Due to the room resonances, the spectral components of the diffuse signal are not identical with those of the direct signals, particularly for small rooms. When mixing direct and diffuse signals, how are the readouts from the display to be interpreted?
With pink noise and a psophometric filter (»dBA«), the levels of both the direct and the diffuse components are almost identical, with an error of ±0.2dB or less.
It's all very well to say that, but getting the right data seems to be not as trivial as it looks prima facie. Can I verify that by my own with a do-it-yourself recipe?
Yes. Set the unit's parameters as follows: DryLEV ±0, 1stLEV OFF, 2ndLEV ±0, Density 0, BassGain ±0, RT60Hi LIN, RT60Lo LIN, and 2ndCUT NONE. Feed all relevant inputs simultaneously with pink noise of -20dB. First disable the diffuse signal, and measure the direct level with the psophometer. Then enable the diffuse signal again, but turn off the direct signal instead. Again with the psophometer, check the output level for every combination of room size and RT60.
Jul 2008, updated Feb 2010
While installing new software, I experience problems with XSOFT and RS-232. How can I verify the interaction between my PC, an attached USB-to-RS-232 adapter, and the Yardstick, which is connected to the USB adapter via a null-modem cable?
When clicking the Download program's START button, a small dialog box will pop up and, at the same time, a BREAK signal will be output from the serial port. Do not confirm the dialog box yet. First check if the BREAK from the PC appears at the unit indeed. Check the voltage at pin 3 of your RS-232 port (at the PC or USB adapter end of the null-modem cable), or pin 2 (at the device end). While the BREAK is active, the voltage must be between +5 and +15 volts.
Whenever a Yardstick sees a BREAK while being turned on, it responds by sending another BREAK back to the PC. If you click the dialog box now, the PC verifies that back BREAK. As soon as a back BREAK is recognized, the PC enters the XSOFT protocol. You can verify the back BREAK at pin 3 of the device, or at pin 2 near the PC. Be aware that a back BREAK needs a functional cable link, i.e. both connectors plugged in. For connecting your multimeter, try to remove the hood from one of the connectors.
In contrast to the printed circuit board of competitors' products for 96 and 192 kHz operation, I cannot locate SRC chips (SRC=Sample Rate Converter) inside the Yardstick. To me, it seems that QUANTEC burdens the sample rate conversion job on the DSP, don't you?
No way – there's not any sample rate conversion at all. If, as it's the case for the 2492 and 2496, we explicitly allow 192 kHz operation, the entire algorithm operates at 192 kHz then, and all delay cells are quadrupled compared to 48 kHz. Or, from the other side: if a 2492 Yardstick with a 2492_QRS_22x4_SIMPLE plug-in is operated at just 48 kHz, a lackadaisical DSP has to wait 75% of its time for the next sample.
Jul 2008, updated Jan 2010
To me it appears that its predecessors sound more mellow and with more bass than the 249x Yardsticks. Do I hear too much into things?
You are right, and I'll give you the low down.
Besides the basic RT60 setting, there are the RT60Lo and RT60Hi parameters for modifying the reverberation time gradually towards lower and higher frequencies. One should expect that both parameters modify nothing but the reverberation time, and there should be almost no audible effect when e.g. fed with stationary pink noise. This exactly is the behavior of the new 249x series.
With the older 2402 Yardstick and its DSP limitations, the result is a bit different. Besides the influence on reverberation time, there is always a frequency-dependent boost or cut in the reverberation level. If e.g. RT60Lo will be set to 2.5, the bass will be a bit louder, too, just as the bass potentiometer of a Baxandall filter network would be operated in parallel. In the same manner, the treble level will be reduced when setting RT60Hi to e.g. 0.25.
The frequency characteristic is so simple that it could be mimicked with an equalizer in the effect-send or return path of your console. In fact, it's identical to a Baxandall equalizer with 6dB/octave.
Now back to the 249x series: for high frequency components, you may reduce their levels with the good old Bandwidth filters (renamed to 2nd CUT recently). For low frequency components, there is a new parameter pair called Bass Gain and Bass Edge – available since Rev. 2.0. This allows to archive bass gains and bass edges in the presets.
Jul 2008, updated Feb 2010
With headphones, an output correlation of 50% does a wonderful job. But when monitoring with loudspeakers, we will run into the problem that the left speaker also hits the right ear, and the right speaker also hits the left ear. Is there a way to counterbalance that?
Available since Rev. 2.0, there is a new parameter correlation, which unbalances, or detunes, the device's output matrix in order to optionally emphasize or deemphasize the lateral or the center signal. This multi-step parameter is available for each output pair separately. With an appropriately allotted emphasis for the lateral signal, crosstalk between speakers may be compensated, at least to a certain degree.
Why are there no analog inputs and outputs anymore? – The world is analog, and many recording studios still swear on analog audio technology.
AD/DA converters age at a much higher rate than outboard signal processors. This means that separate AD/DA converters typically are a more economical approach.
Digital audio inputs and outputs are an almost perfect interface: full 24 bits in and out – no noise, no distortion, and, if required, double or quadruple sample rates.
Our units are designed for an operating life of 15 to 20 years. Analog inputs and outputs, especially A/D and D/A converter chips, are changing from year to year. What today means the spearhead of technology has to be paid for with astronomical prizes – but may turn out to be outdated in just a few years. Here is an example: our 2493 builds on (currently still) high-performance and thus expensive converters. But what if a new trend will further manifest, which favors IIR instead of FIR anti-alias filters (because of the IIR's non-existent pre-echoes)? What about the 2493 with its time-tested linear-phase FIR filters – will these filters turn out to become a problem in a year or two? From our point of view, it doesn't make a lot of sense to burden an outboard unit with possibly doubtful converters, until its end of life.
Digital audio ports are much easier to keep up with converter advances. So we recommend: use your digital I/O Yardstick with external AD/DA converters, and replace them every few years with a then state-of-the-art model. Then the cost/performance ratio will be your decision, not ours...
You seem to feel uncomfortable with our reasoning? Have a quick read into FAQ 053, we're sure that we will convince you that way...
Aug 2008, updated Jan 2011
Is it realistic to consider rebuilding an existing unit, e.g. from AES3 to Analog?
All 249x units are designed with modular subassemblies. So it may look straightforward to exchange one or more modules. The main problem lies in the then obsolete and useless front panel and rear hood, which are costly to be recycled or disposed. Conclusion: never say never...
A very important and huge advantage with the M7 is that its diffuse field can perfectly blend with prerecorded ambiences. For example, I have a live recording which has a lot of live Early Reflections, but because of too much audience, the decay is shorter than it should be. With the M7, I can match the Size very precisely to the recording's space, and I can add diffuse field only to make the reverb tail longer, not to mention that its reverb tail really lives, not static as the recorded real reverb.
When I used the 2402, this was my main problem with it. As soon as I wanted to add longer tail to a prerecorded material, the sound field started to be a little more crowded and too spatial.
Due to the somewhat restricted DSP power in 1997, the 2402 had one room size only: 100,000 cubic meters, which is extraordinary large. Such a room will never match smaller rooms like clubs.
Thirty years ago, one would have solved your problem with a plate. A plate has an incredible density, right away from the start. There are almost no early reflections. In this respect, a plate closely resembles a small reverberation chamber. And possibly resembles the M7 diffuse field generator output.
The corresponding QUANTEC approach is easy and straightforward:
Pick a 249x Yardstick room which has been set one size step smaller than the room you want to correct. All early reflections emerging from this smaller room will be masked by the pronounced live initial reflections. While the live echoes have exhausted, the Yardstick has "warmed up", which means it is dense enough and ready to step in. You may fine-tune the switchover timing with appropriate postdelay.
As the sizes of the initial-reflection real room, and the late reverb Yardstick room, are of the same magnitude, their resonances will blend nicely. Now you can add Yardstick reverb tails at will, without being bothered by much-too-late early reflections from a totally wrong room size, way too large to be melted.
Oct 2008, updated Jan 2010
I have serious problems with the user interface. The big switches and rotary wheel are ok, but to use only those is quite time consuming. The mini switches are really not very practical if you happen to have big hands, plus you need to be really on top of the unit to see what you are doing.
You seem to have overlooked Parameter Follow Me mode. When enabled, the left display follows every mini key pressure immediately, showing the current parameter name and value with larger characters. There is also a large status bar for the selected parameter. You can turn the wheel to modify the parameter.
Confirm that Parameter Follow Me (System Setup / Preset) has been turned ON. With the big switches and the rotary wheel, step down the menu Edit Scratch A until you control an arbitrary parameter's value from the rotary wheel. Instead of allocating a new parameter by going up, sideways, then down again, just hit the appropriate mini switch.
If I install a new plug-in on an already used bank, what happens to the presets of the previous installation?
It depends on the type of plug-in. If you simply install a newer version of a functionally identical plug-in, your presets will be maintained. If the parameters of the new version have been modified or expanded, all previous parameter versions will be recognized by the new software, and appropriately converted on-the-fly. New, previously non-existent parameters will be initialized to uncommitted values, which try to minimize changes of the previous presets' sonic characteristic.
If you install an older version of the same plug-in, and both preset structures are identical, your presets will be maintained. If caveats found, the older software assumes an unknown, non-related effect, and the preset area will be erased. If the over-installation has been done inadvertently, the old presets can be salvaged as long as the over-installed plug-in has not yet been manually selected, confirmed, and launched. It's only an actively running plug-in, which irrevocably erases or converts the presets.
Example: Whenever a DLY plug-in will be installed on a bank which previously hosted a QRS plug-in, all presets will be lost. Functionally different plug-ins do all structure their preset areas at their own discretion, and are blissfully ignorant of any foreign heritage.
Incidentally, the installation tools' default setting always targets at an unused bank.
How can I recognize that a plug-in will definitely not damage my presets?
Installing only newer versions of otherwise functionally identical plug-ins is a safe bet. Functionally identical means that the identification strings of both plug-ins match in every aspect but the version number. Exceptions from this rule are feasible, but will be documented explicitly.
A plug-in can be identified from the 3rd and 4th lines of the plug-in boot loader. This same text will be re-used in the filename of the installation tool. The filename always starts with the device type, followed by the full name of the plug-in, and ends with a version number. Blanks, '.', or other special characters will be replaced by '_'. The boot loader may show the following text:
The corresponding installation tool will then be called:
What's the use of all that SIMPLE, MEDIUM, and COMPLX stuff, anyhow?
For any DSP algorithm, extensive feature lists or higher sample rates can be traded against each other. The QRS algorithm's Yardstick 1.x versions have been designed under the assumption to be fully operational even at 192kHz. On the other hand this means that, at 48kHz, the DSP is active only one quarter, and otherwise waiting in idle the rest of the time.
For the user, such an approach is straightforward: any sample rate from 38 to 216kHz is acceptable. The unit digests everything, and the operating surface remains consistent.
For an ambitious algorithm designer, such an approach may be frustrating. How much more can be realized with the otherwise uselessly wasted double or quadruple DSP power. Before continuing, the designer must come to a momentous decision: If I double complexity, my algorithm cannot be operated at 192kHz anymore; but at least will reach 96kHz. If I need four times the DSP power, 48kHz is the end of the line. Losing a few customers when resorting from 192kHz to 96kHz may not really hurt your company, but losing some 10 to 20% when sacrificing 96kHz will do. Moreover, such a move will alienate all the rest.
On the other hand, the entire customer base will be preserved if various complexities of an algorithm will be offered. Sometimes it doesn't make sense to stuff all bells and whistles into one plug-in. Quite often it may be realistic to trade-in some of the less important features for a result, which can be operated at double rate, too. Many a high-end customer may even favorite a feature-reduced version. Motto: why should I need Baxandall filters for my $3000 preamp, anyway...
Exactly this is the idea of SIMPLE, MEDIUM, and COMPLX. For a SIMPLE plug-in, all unnecessary ballast will be avoided; this is sort of a minimum version, but fully operational up to 216kHz. The COMPLX plug-in will be stuffed with all bells and whistles, even if just of occasional use; it's the flagship. And, right in between, a MEDIUM plug-in may be positioned, which provides some potential to play with, but compromises only marginally when it comes to sonic quality.
Quite often, the following relations can be assumed: COMPLX:48kHz, MEDIUM:48..96kHz, SIMPLE:48..192kHz; but don't take that for granted. To make it clear, all plug-ins have their maximum sample rate labeled as x1, x2, and x4.
The most universal and most commonplace plug-in is definitely the SIMPLE version. If the required design target can be achieved with it, any future plug-in should be limited to this singular version. This reduces maintenance efforts for the software staff, and minimizes gray zones and learning curves for the end user.
Oct 2008, updated Aug 2011
If you want to learn about the sonic implications of the SIMPLE, MEDIUM, and COMPLX QRS algorithm's room models, we'd recommend reading FAQ 069 ...
Why do QUANTEC Room Simulations sound more impressive and more realistic than real rooms?
First a formal objection on the part of the Reverberation Czar! - It goes without saying that natural rooms still sound better than our simulation models, first and foremost because of their outstanding spatiality. However, this only holds true as long as the listener is physically positioned inside the room, and thus gets flooded by the omnidirectional sound waves.
As soon as one tries to capture the room impression with microphones, all 3D magic will vanish.
Every sound transducer the acoustic event has to come about on its way to our ears is a bottleneck, which strips off a certain amount of the original room response's spatiality and liveliness. This is by no means a devaluation of our state-of-the-art microphones. But fact is that the microphone bundles all the omnidirectionally-arriving signals on its diaphragm, and then tunnels the precious payload through an electrical mono channel. Even if you would totally shut-off one ear while at the church, that does, on no account, sound monaural.
As the omnidirectional output of our room acoustic models is fed straight into the audio signal path, instead of being tunneled through a microphone's diaphragm, our models clearly sound more live than a natural room captured with microphones. What is still unavoidable, even for us, is the bundling through the diaphragm of the playback transducer, currently the final acoustic difference remaining between a QUANTEC Room Simulation model and a person's physical presence inside a real room.
From the number of sound transducers involved, three quality classes may be derived:
- Class 0 – no sound transducer:
The listener is physically positioned inside a real room; flooded by the omnidirectional sound waves.
- Class 1 – one sound transducer:
The listener harks attentively to the QUANTEC Room Simulation model through loudspeakers or headphones.
- Class 2 – two sound transducers:
The listener follows a recording where room acoustics have been captured by microphones. Only those components of the natural room acoustics not yet stripped-off by the multiplication of recording and playback transducers will remain.
While QUANTEC Room Simulation with its parameterized models belongs to class 1, the nowadays extremely popular convolution-type reverberation is clearly at class 2 candidate. Reason is that the convolution chain still comprises of two sound transducers: first the microphones to capture the room's “fingerprint”, then later the playback transducers to monitor the convolution result.
Strange: if the room fingerprints would be directly deducted (i.e. “stolen”) from our Room Simulation models, a convolution-type reverberator still would not ascend to a class 1 sonic quality. With just about two "fully L" and "fully R" room fingerprints tapped, we'd end up with the same problem a real strero room microphone would end up with; there's no help to recover the totally jagged "crosstalk domain", which is paramount for acoustic space depth impression (see FAQ 041). It's not until a large number of different panpot settings are being sampled separately (both "intensity" and "time delay"), and all of them recovered during the mix (which would be technically unrealistic), that a convolution reverb fed with QUANTEC-IRs could pit against the breathtaking transparency of our original room models.
Still unachievable for any size of fingerprint library would be our parameterized models' unrivaled ease of ad-hoc characterization.
Nov 2008, updated Jan 2011
Out of curiosity, I've installed an older 1.x software on a unit already converted to 2.x. Now all my 2.x plug-ins have vanished, including my valuable preset collection. Too bad...
Keep cool – that's what has happened: The Yardstick's bank 0 of a 2.x installation is being used by the SWITCH boot loader, while in version 1.x, the (one and only) effect software has been settled there. If, after an existing 2.x installation, a 1.x software is being installed, it will overwrite the 2.x boot loader. Because of the missing boot loader, the „bridge“ to the 2.x plug-ins and their preset libraries has gone. But they still exist, safely captured further down the Flash memory. It's just the „access“ which needs to be fixed.
How can the access be re-enabled?
Download the current SWITCH code from our website, and re-install it to the boot sector again. Additional benefit: the 1.x software, installed by you out of curiosity, will not be overwritten initially, but will be offered by SWITCH to be relocated to a bank further down the Flash memory. Before copying, SWITCH will compile a list of all banks, regardless of occupied or free. Now it's up to you to enter a preferred target bank, wait a few minutes while the installation is being relocated, and then proceed with installing SWITCH until finished. With the next reboot, not only all your 2.x plug-ins are being advertised again, without any change. As an additional bonus, your stray 1.x installation will be offered for booting, too.
When doing some rework on older sessions, I sometimes wish to access earlier versions like 1.6, 1.7, or 1.8 - including the original presets of that time. Would it be possible to smuggle various 1.x versions into the unit, by repeatedly overwriting the SWITCH boot loader. Then relocate each one to another bank, in order to have all of them at my disposal when reopening a session?
I've just finished overwriting my former 2.1 installation with version 2.3. Now I found that only parts of my preset collection can be used. Amidst my presets there are lots of voids yawning - I can read EMPTY all over the place. Looks like I've lost parts of my presets due to compatibility problems...
Nope. Those “knocked out teeth” are all those former factory presets, which have been haphazardly interspersed with the Local user preset archive up to version 2.1 – just like a rope of pearls. All older factory presets, and many new ones, have been relocated to library directories like Dialog-Lib and Music-Lib. All users may refill those residual EMPTY voids with their own presets now.
Can I access older user presets located in a foreign bank?
Yes – that's generally possible. Older presets inherited from previous software versions will be converted to the new preset format on the fly. If the new preset format contains parameters not yet available in the older version, similar parameters will be converted (if possible). If all else fails, new parameters will be set to their default values. The idea is to make the updated preset sound as similar as possible to the one from the older software. In spite of all automatism, we strongly recommend to check the updated presets thoroughly.
Can I access factory presets located in a foreign bank?
Yes, but not before they have been copied to the foreign Local archive.
Bullshit! - My Yardstick does not fit between the rails of my rack. Aren't those QUANTEC people in a position to adhere to the 19” specs?
According to IEC60297, the clearance between the 19” rack rails is required to be 450 mm minimum. Our Yardsticks external width is 448.2 mm maximum – so where's the problem?
Mar 2009 - updated Oct 2011
Is erasing of a bank mandatory before installing a new plug-in?
No. But if a bank has not explicitly been cleared, the previous Setup and all local User Presets remain in that bank. When updating to a newer version of the same plug-in, it often makes sense not to erase that bank. Launched for the first time, a newer plug-in tries to adopt the remaining data, then start to build on that heritage. Before, for whatever reason, installing an older version, a bank should always be cleared in advance!
What about controlling my Yardstick from multiple positions?
It's possible to operate a Yardstick from both front panel and web browser window in parallel. It may haptically be useful to select a parameter from the browser, but modify it by turning the front panel wheel. Or, to select a Preset archive from the browser, but stepping through the Presets from the wheel.
The only critical function is the bargraph display. Bargraphs only work on front panel and one browser concurrently. Opening multiple browser windows to control the same Yardstick from different PCs would, at the current stage of development, disarrange the bargraphs.
For jump cuts in movie post production, I regularly need a frame-accurate switchover of the room acoustics. With the Yardstick I run into two problems:
1. While turning Preset selection, it takes me 2-3 seconds until the Preset has been loaded and will be audible ("Selecting, Initializing...").
2. If old and new Presets are located near opposite edges of the library, while still underway to the target, I'd risk some Presets being loaded temporarily.
Is there a better practice?
The simplest way would be loading old and new room situation to Scratch A and B. When the jump cut occurs, simply hit the A/B button.
How long would that take?
When switching between small rooms, there's only the gap for the intentional fading down and up again. The larger the room size, the longer the transition will take, which, by surprise, will still sound organically. To name the numbers: 1E0=>5ms, 1E1=>15ms, 1E2=>40ms up to 1E6=>almost 1sec (always take the room size you are switching to).
And what about repeated jump cuts in rapid succession?
Enter Setup and look for the menu item Preset Load (Operations Guide 3.1 - p. 100), then select Enter. While in Enter mode, a Preset will no longer automatically load as soon as the wheel slows to a standstill, but will wait for a confirmation from the ENTER key. This allows for switchover at the push of a button, with the abovementioned initialization delays. A proven trick is to copy the required room sequence to adjacent Preset slots in the user area, and then retrieve them sequentially.
Competitive products always provide plenty of different algorithms: Room, Hall, Plate, Space - exactly what I need in daily routine. As far as I understand, a Yardstick has a single algorithm only, which seems a bit narrow-gauged to me, isn't it?
No way. You're right that a Yardstick does not provide more than that single QRS algorithm, but that's a powerful, all-inclusive solution. Its universalism is hidden within its parameters. Just by well-directed tuning of the various parameters, you can end up inside a bread bin, or a giant cathedral with multiple side naves and transepts. Other than competitive devices, which recommend an optimized reverb for a specific type of sound (e.g. Voice, Guitar, Drums), the QRS algorithm is being defined by the room the musicians are playing in, independent of the construction of their instruments.
If I'd squeeze the entire mix through reverb, or a room for your sake, I'd totally loose transparency and comprehensibility.
That's true for reverb, but not for QUANTEC Room Simulation. The QRS algorithm applies diffuse components individually for each instrument, according to its momentary style of playing - fully automatically, without any intervention. Just the same as with a real room that never needs to be tailored to a specific instrument. The trick lies simply in an exact simulation of the physical laws effective inside a room. In particular it's the attack behavior, i.e. the gradual effervescence of the room. The longer a tone will be maintained, the further it will climb up the attack slope, and the louder it will tilt over into the reverb tail. With this "launch pad" trick, long tones lead to strong, short tones to gentle reverberation - all of its own volition. All-important: each one individually, i.e. without any interdependency of the envelopes. (»Hallelujah diagrams« - see links below).
Let's construct an extreme example: if we concurrently feed a room with a flute and a spoken voice, the low-bandwidth, sostenuto flute tones will play around the room much louder, haunting and longer observable, while the short, wide-bandwidth transients of the spoken consonants just tip the room. To clarify the consequence of this approach even further, let's simply exchange the roles for a moment. Sixteenth-note sequences from the flute just tip the room, while a sustained "Aum" will considerably pump up the room response. This is one of the secrets why transparency will be so strikingly maintained with the QUANTEC approach.
To get back to your question: it's the QRS algorithm's ability to do all things right with just a single parameter setting. With multiple dedicated algorithms, you'd need multiple reverb units, one for each instrument class. More than once it will take you hours of fiddling about with the various units, until all the individual components will end up coordinated in an acceptable manner.
Another QUANTEC premium: all the instruments are always playing together within the same room, and don't live their parallel lives in different rooms.
Competitive products always provide dozens of adjustable Initial Reflections; but only two with the Yardstick. Is QUANTEC running out of reflections?
Less is often more, especially in this situation. QUANTEC has recognized very early, that, within a real room, each initial reflection is being emitted from the same structure that also generates the reverberation tail. In its simplest case, a room consists of six boundaries; left, right, front, rear, top, and bottom. If we'd allow a sound engineer to add arbitrary early reflections, which, because of the freedom of choice, can no longer be consistently co-created by the late-reflection structures, would do more harm than good to a coherent sonic image. Physics simply do not allow separately created echo patterns to seamlessly merge with the subsequent room response - disruptions and side actions are inevitable. In this context, we were told of a singer who once put this phenomenon into words. »With this device, I had the impression for the first time that there was no more annoyances between my voice and the room.«
What exactly, from the QUANTEC viewpoint, does the established competition do wrong?
To preserve complexity, i.e. to minimize DSP power, most competitors split room response into two partial algorithms
- A multi-tap delay line with a fistful of taps, which, independent of the contents it is fed with, repetitively reiterates the same distinctive early reflection pattern. This approach follows the idea of a raytracing model, where a few initial echoes (»early reflections«) are derived from the geometric circumstances of that room. The temporal and panoramic distribution of the delay line echoes is told to characterize the desired room.
- Separated from the direct signal by suitable Predelay, follows a continuum with as much density as possible. Allegedly, any room behavior according to size, individual reflections, and room resonances can be completely ignored. What irritates us most: the one and only design target for this continuum is to increase its density beyond limits. This does mean nothing else than, as far as DSP resources can deliver, to end in musically dead "White Noise" - for any type and any size of room.
QUANTEC's conclusion: an evermore reiterated pattern is an annoyance, sooner or later. A similar effect is known from cheap loudspeakers, which continuously repeat their impulse response with each and every transient coming in. Second: an "acoustic harassing fire" in contradiction to the room would completely destroy the interaction between tone duration and reverberation level, which is so stunning with the QRS algorithm. Net result: the QRS algorithm would still supply clean reverberation; but its legendary transparency will be lost and gone.
But why are there still two adjustable initial reflections, after all?
What has been outlined above is basically required for musically inspiring rooms. Movie and radio drama dialogs are a different story. What's needed there over and over again are standard acoustics, painted with a pretty wide brush. Most situations are "nonmusical" rooms with pronounced boundary surfaces in the near field, e.g. a hallway, a car, or a staircase. For such near field simulations, the reverberation tail is generally second-rank, and the emphasis of a simulation is shifting towards early reflections, indeed. There, and only there, one single, representative early reflection for left and right each does make sense.
Compared to the competition, QUANTEC is showing off with a highly idiosyncratic approach for yonks. Such an esoterically prancing idea looks pretty untrustworthy to me, as it suggests that QUANTEC follows other physical laws than the rest of the universe. Seems like pure marketing wishy-washy, isn't it?
It goes without saying that QUANTEC is bound to operate within the limits of physical laws, as anybody else. Major difference and unique selling proposition is our way to approach common room-acoustic phenomena via mechanisms and models defined by the air fill of the room. In other words: we focus on the transportation medium of the sonic energy, and include those frequency-selective resonators within a room. This is contrary to the competition, which follows delay line patterns deducted by a raytracing perspective. Translated to a wind instrument, the competition concentrates on the wooden or brass instrument body, while we turn our attention to the vibrating air column within the instrument.
Still frivolous. As resonance behavior in the frequency domain may be represented by impulse response in the time domain, and vice versa, simply by Fourier transforming one into the other. Plain-talking: both approaches lead to absolutely identical results; they are two faces of the same coin.
Your statement is valid for unlimited resources only, which means abundant DSP power. Whenever one is bound to economize DSP power, because one has a meager 1% compared to that of a real room, one should thoroughly evaluate the important details to devote the scarce resources to, and what uncalled-for knick-knack would go unheard anyway. Point is rather: what can be omitted without harm, and which one of the contrary approaches may lead to music-esthetically pleasing results faster and easier.
Well, based on a resonance approach, QUANTEC claim that they're in a position to extract more useful results from limited resources than the competition. Which means they fish out those better 1%, simulate only that, then spoof the auditory system music-esthetically more pleasing than the competition.
Exactly. Approaching room acoustics from the frequency domain, instead of geometry, shifts the importance of various physical phenomena essential for the simulation of a certain target room acoustic behavior. Main target for QUANTEC are musically ambitious room acoustics. More than 30 years ago, QUANTEC was the first to recognize that the silver bullet for simulation of musical rooms lies in modeling the engaging and subsiding of resonators. Those who index a room's boundary surfaces, then raytrace geometry, then deduct tapped delay lines from that data, will inevitably dissipate their energies and DSP power on musically meaningless sidelines.
My reverb unit offers 4 algorithms - a Yardstick just one. How come?
My Chinese guitar from the DIY store has 4 pickups - an Ovation just one. How come?
With regard to price: why are those Yardsticks considerably more expensive than the competition?
Just have a quick listen - and you'll put this question aside.
Most competitive units feature a Modulation parameter. Has QUANTEC diligently missed the need for modulation?
Yes, as modulation is of no significance for Room Simulation.
Without random modulation, all reverberation tails would be equal.
True for reverberation, but totally ineffectual for Room Simulation. Here's our competitors' problem: in front of the actual reverberation tail is a cluster of so called initial reflections, which are told to characterize a room according to a raytracing concept. Such a cluster has one major problem - with each and every transient fed into the system, the cluster reiterates the same response over and over again. This kind of "fricative", which sticks out of any plosive like p, t, k, will be fatiguing, and will turn out to become boaring or annoying before long. Quite similar to a cheap loudspeaker, which cannot cope with fast transients, and thus will replay over and over again its critical onset pattern: kprwt, kprwt, kprwt.
Our competitors circumvent that problem by randomly moving the initial reflection delay line taps slightly. The goal is that the fricative will mutate quickly, thus prevent the hearing from associating subsequent fricatives with preceding fricatives. Basically, the problem seems to be solved, but unluckily this trick has serious adverse effects due to the inevitable Doppler effect. Moving the taps would result in frequency fluctuations, which detune the incoming signal. The reverb commemorates Chorus effect - it whines, and is no longer "piano proof".
But such a chorus reverb may sound ridiculously beautiful, e.g. with a guitar solo.
As a creative means of design, chorus reverb certainly has its value. But in a real room, frequency modulation won't happen, so there is no such parameter in the context of Room Simulation.
Just a minute! - There's a paper by David Griesinger (»Lexicon«) where he demonstrates that, due to airflow in large rooms, slight frequency deviations can definitely be detected.
Extremely thrilling reading for HVAC engineers. But where's the music-esthetical benefit? - Show me one conductor or choirmaster, who pilgrimages to the caretaker to ask him to crank up heating or ventilation - in order to make the room sound more pleasing.
It's sort of weird, isn't it?
And what's QUANTEC's trick to suppress the undesired regularities within the algorithm?
No trick at all. With Room Simulation, this problem simply doesn't occur, as we don't add separate initial reflections. As documented elsewhere, loudness and character of each reverberation tail depends tone by tone on the duration of the stimulus. This will be mixed with the previous history, i.e. longer tones played recently, whose tails remain louder in the mix. Based on individual tone durations, it's quite possible that originally staggered tones will decay with synchronous tails (»Hallelujah diagrams« - see links below).
The individual composition of each reverb tail according to tone duration and previous history leads to continuously changing interferences between one another. Such interference results in frequency-dependent accentuation and cancelation, which is beyond association for human hearing, thus sounds very lifelike and musically pleasing. But without the cumbersome Doppler effect now, which, if noticed too late, could lead to aggravating surprise during subsequent processing steps.
How do I use a Yardstick in the context of a 5.1 or 7.1 production? - Part 1 - Original Recording
Similar to a mixing console, a Yardstick doesn't process encoded material in the sense of x.1. It builds upon the unencoded raw version of the audible material. The effects send path operates out of the console as usual; the effects return will, together with all the direct signals, be encoded to x.1 in a subsequent processing step.
There's some danger that through the effects send bundle, certain instruments may enter the device twice, with arbitrary phasing. Example: if the violin microphones overhear the cello section playing behind the violins, the cello sound would come in over two effects send channels. If I'd mix both strings channels to feed the effects send path, I would risk unacceptable comb filter effects on the cello reverberation.
Refrain from mixing in advance; benefit from using all 8 effect send paths. Fortunately, the 2498 Yardstick offers an unparalleled property: all 8 paths into the simulated room are completely phase-insensitive. In other words: if one would exchange polarity on one of the inputs, there would be no change in sound at all.
Conclusion: those cancellations you're afraid of would need some care only if the "small surround" 2496 Yardstick with its 2 inputs would be used. If you'd use the "large surround" 2498 Yardstick with its 8 input channels instead, none of such cancellations would develop at all.
Note: in a surround production, if one would move along a secondary stereo downmix version either, all cancellation problems should have been solved there already - independent of a Yardstick. When working on a 5.1 surround production, such a controlled free-rider downmix is an ideal effects send for the "small surround" 2496 Yardstick with its 2 inputs and 6 outputs.
How do I use a Yardstick in the context of a 5.1 or 7.1 production? - Part 2 - Post Production
Completed x.1 productions, which should be re-opened and augmented with Room Simulation, need to be unassembled (»decoded«) before they can be fed into a simulated room. Fortunately, the 2498 Yardstick offers an unparalleled property: all 8 paths into the simulated room are completely phase-insensitive. In other words: if one would exchange polarity on one of the inputs, there would be no change in sound at all.
To exploit this feature, assign each single channel of your 5.1 or 7.1 production to one of the 8 virtual driving speakers. Not just the 5 or 7 satellite channels, but all 6 or 8. Regardless of the incoming phasing on the 6 or 8 channels - especially picky would be the crossover frequency between subwoofer and satellites - nothing like cancellations would develop at all.
How do I use a Yardstick in the context of a 5.1 or 7.1 production? - Part 3 - Stereo-to-Surround
Here we take advantage of another inherent convenience of the QRS algorithm: an added simulated room will routinely blend with an already recorded room or reverberation. Contrary to competitors' reverb units, such a double reverberation will not degrade transparency at all, or marginally only.
For this type of projects, both 2496 and 2498 Yardsticks are likewise applicable. To integrate otherwise idle DSP resources of the 2498's #3 to #8 input channels, please use it in the 2->8 mode, optimized for this purpose (instead of 8->8).
In web forums one can routinely read that 'the QUANTEC' is an outstanding primary reverb. Actually, what's a primary reverb?
Many music productions deploy different reverb units and reverb plug-ins, tweaked to specific instrument classes in an optimum way. This requirement for optimum tweaking is the result from a limited range of artifact-free operation of low to medium-cost reverb algorithms, which may be useful for vocals only, or solo instruments only, or percussion only, etc. Argumentum e contrario, one has to be prepared for inacceptable side effects for every wrong instrument class. Because of this persistent "requirement for chasing the instruments", automation is indispensable here.
With this strategy, the musicians are playing in separate rooms. It's the job of the primary reverb, a device with outstanding quality, to re-collect them and put them back into a single room. Problem is that a primary reverb generally compromises transparency of the cumulative sound image, and thus will often be used reluctantly.
A QUANTEC Room Simulator meets highest standards at this critical position. Most of its Music Library presets flawlessly process all instrument classes simultaneously, i.e. the entire mix. Side effects such as coloration, clatter, intransparency, or mushy sound are almost nonexistent with a QUANTEC-based primary "reverb".
Another peculiarity of primary reverberation is that it is an always-on effect, which means that its settings are modified infrequently. To put it into perspective: automation is seldom required; similar to monitoring speakers. This may explain why, despite unavailable automation, even the very first 1982 Room Simulator generation is still in daily operation worldwide.
2496 or 2498 have more outputs than I need - what to do with surplus outputs?
Simply ignore them and leave unconnected. Please avoid mixing surplus outputs to used ones. Due to the 50% correlation output matrix, any cross connection would cancel out 50% of the reflections, which would result in both thinned out and 3dB lower level reverberation. On the other hand, this behavior may come in handy when mixing L and R channels, as the resulting mono signal has lower diffuse energy.
2498 has more inputs than I need - what to do with surplus inputs?
Contrary to outputs, you should drive all input pairs in parallel from your less than 8-channel input signal. Each input channel pair contributes 25% towards total number of a room's normal modes and reflections. If one or more inputs remain idle, the resulting reverberation will by unnecessarily thin and a bit lower in volume.
Moreover: each input pair of a QRS COMPLEX plug-in behaves like a QRS SIMPLE plug-in. So only a close concert of all inputs will end up in that wonderful COMPLEX sound
When driven from one stereo pair only, it's recommended to feed the 2498 through its XLR input. Such operation will not only forward the input signal pair to all 8 virtual speakers, but compensates for any level fluctuations, too.
What exactly is the low-frequency edge for Room Simulation?
It should be noted that a room can only produce reverberation on frequencies (»spectral lines«) where there are room resonances (»modes«) available. For any gaps in between, there's never any reverberation. The low-frequency edge is determined by the lowest 'standing wave' that can manifest itself within a given room. This depends very much on the size of a room. For large rooms, the edge is a few Hz, while for boxes and cabinets, room resonances can manifest only above a few hundred Hz. For DC, which would correspond to a constant air pressure, all QUANTEC Room Simulations deliver a zero, which means that DC components are being suppressed.
For all those users who'd like to have a hell of a party with their supernova, there is an additional goodie hidden in the Effect Setup menu. With the Subsonic switch, the DC zero will be converted to a pole. In reality, this would mean that a given air pressure in a room would continuously increase over the course of a few minutes, until the room would burst and collapse. In simulation, a DC offset would increase and increase until full level is crossed, and output overflow occurs. In other words: the Subsonic setting would annul the law of energy conservation. Useful for such kind of effects is the AES/EBU port's ability to transfer unlimited DC, which is generally not the case for analog AD/DA.
The chief attraction is that 0Hz has changed to a pole (»resonance«) now instead of a zero (»cancellation«), which means that there is no lower frequency edge any more. Even below 1Hz there is still a lot of cannonade and boom, which resembles no less than a respectable earthquake. Made-to-measure for those addicts of D-BOX motion code and Action Seats.
More than once one can read an indication that for the 2498, the phasing of the 8 input channels is told to be irrelevant. Already difficult to understand in theory, I cannot imagine such a pipe dream to work in practice?
Acoustic quantum mechanics.
Heisenberg uncertainty principle.
Weren't we just in audio engineering yet?
For simplicity I'll call it "QUANTEC physics". I'm talking about phenomena to become relevant whenever the frequency resolution of an acoustic phenomenon will become more filigree than the frequency resolution of the sense of hearing. Room simulation is such a situation, especially for large rooms, where the distances of room modes ("room resonances", "spectral lines", "standing waves") are often closer than 0.1Hz to one another. Considering that, from a narrow band viewpoint, both phase and amplitude behavior radically change from line to line (40dB are not uncommon), one can only speak about wild, completely irregular random fluctuations. To get hold of meaningful figures at all, a statistical physical definition of the room is an option. Simply use third octave band noise instead of sine waves, and your frequency response will be flat and clean.
Unlike a crystal-stabilized sine wave oscillator, the bell-shaped spectral width of a human voice or a musical instrument will stimulate an entire bundle of adjacent room modes simultaneously. The exact hits will change permanently along the way, as the bowed or blown instrument's pitch will fluctuate continuously.
Short digression: imagine you have two uncorrelated noise generators identical in construction. If you mix both outputs, the output noise signal increases by 3dB. What do you think you hear if you'd activate the 180° key on generator A or B (or both)? - No change at all - all variants sound absolutely the same!
Exactly herein lies the trick for multiple phase-insensitive effect send inputs. Instead of mixing the 8 input signals additively before feeding them into the reverberation chamber, they're being forwarded disjoint. Not before the room itself, the various input signals stimulate their individual and not exactly predictable room modes. Which signal eventually hits which room mode at which location remains unknown, and can only be quantified statistically. And once again: unaffected by imaginary (180° or whatever) keys at the inputs.
All Yardstick I/Os are AES/EBU. Does that mean that I cannot operate a Yardstick in my S/PDIF studio?
Not at all.
S/PDIF Channel Status bits are being recognized properly. As most sound cards deliver wrong sampling frequency indication beyond x1, there's an optional fallback mode based on an internal sampling frequency counter.
For adapter cables and plugs, there's an S/PDIF section in the manual with exact schematics.
Contrary to a much hyped competitive unit, my choir stands amidst the room here, not in front of it. Why does QUANTEC perform better than the other guys?
Buy our no-hype product and enjoy it.
What's the stunning transparency of QUANTEC Room Simulation based upon?
Psychoacoustics - the cocktail party effect in this case.
The cocktail party effect, discovered in 1953, describes the ability of the human auditory system to focus its listening attention on a single talker, among a mixture of conversations and background noises, ignoring other conversations.
Photo: Henning Hraban Ramm / pixelio.de
How pronounced is this effect?
Within such selective perception, the hearing reaches noise suppression from 9 to 15 dB, i.e., the acoustic source, on which humans concentrate, subjectively seems to be many times louder than ambient noise of the same level. This even holds, if the ambient noise is louder than the information by a similar amount (»negative S/N«). On the other hand, a microphone recording would capture mainly background noise, as any selective perception does not apply.
How is this related to the acoustic transparency of a room?
To come to the point: within a room, a side effect of the cocktail party effect would result in a greatly reduced perception of spatiality: the information sounds dry and with minimum reverberation. A microphone positioned at the same place would capture blurred, slushy, and over-reverberating information instead.
Once again, QUANTEC seems to be ahead of the pack...
That's by no means our merit! - It was surprise to us, too, that even within the scope of the cocktail party effect, perfect Room Simulation behaves just like a real room, and provides the human hearing with the necessary clues. If those clues are unavailable, as it's the case with our competitors' reverberation devices, the human auditory system reacts just like the microphone: the sound is unintelligible, slushy, and with far too much reverberation.
Feb 2010, updated Aug 2011, Nov 2011, Mar 2013
Whenever I try to adjust a parameter from the mouse wheel, the slider jumps right to the top or bottom stop.
Open your mouse driver configuration. Within various mouse wheel properties, you'll find something like advance one screen page per mouse wheel step. This option needs to be disabled.
Whenever I'm recording the impulse response of any QUANTEC Room Simulation, for putting it through my convolution plug-in, its spatiality collapses altogether. To name a number, the 60 to 150 feet depth of a sacred building has completely vanished into thin air. No idea of what's going wrong. Maybe some copy protection QUANTEC has hidden intelligently?
It's neither you doing something wrong, nor there's a copy protection. It simply doesn't work.
QUANTEC really cannot make a claim for themselves, that proven mathematical and physical methods like convolution and Fourier transforms ironically fail, as soon as they're applied to their Room Simulation algorithm.
Nobody did insist on that. It goes without saying that convolution works; you do hear flawless reverberation, don't you? It's just it's spacial depth that you're missing; which has gone flat somewhere in the course of your manipulation.
Flawless indeed. But, where exactly, the spatiality has fallen by the wayside?
Right from the start - when feeding the unit.
In plain language: first I feed the left input with a click, and record the impulse response on both outputs. Then click on the right, and again record both outputs. What's wrong with it?
For now, you've recorded no more than two, let's call them "labyrinths": one for 100% left, and one for 100% right. Now hurry to proceed with sampling center, slight left, slight right, and all the rest of it.
Just wait! - I've two labyrinths - one for left, and one for right. If I'd feed the unit with a center signal, i.e. mono into both labyrinths at once, the output always delivers the sum of both labyrinths. Generally, this should hold for any panpot settings, right?
In the context of Room Simulation, we don't deal with two delayline-based labyrinths, but with hundreds of thousands of resonators distributed throughout the room. From the perfect coordination of all those resonators - and only from there - results that stunning transparency.
Feeding the unit with a sine wave test signal and very fine frequency steps (<<1Hz) would stimulate a non-specified subset of those resonators, to a more or less extent. Depending on the phase and amplitude conditions at the two inputs, quite a few resonators may not even respond at all. Moreover, one cannot estimate if a specific resonator will respond to the left, right, or only to a specific phase or amplitude relationship between the two input channels. In other words, as jagged and bumpy as the amplitude and phase behavior of a single labyrinth, is crosstalk between the two labyrinths with room simulation. The operative point here is that those hundreds of thousands resonators jump wildly with even the slightest frequency drift, while your two labyrinths stubbornly deliver their vector sum, regardless of the frequency - definitely a bit of a yawn.
I must admit that my approach would indeed force a majority of those resonators into lockstep. This may paralyze time-of-arrival stereophony, but what puzzles me is that intensity stereophony does collapse likewise. Did I overlook another important detail?
Due to the complex crosstalks within a room, one may realistically imagine one resonator at one specific room position, which would respond to either left or right channel, but not both. With mono, genuine Room Simulation may deliver a resonance gap here, while your convolution clone may still deliver the sum, e.g. a peak - as it does with any other resonator. Be aware that you haven't captured those singularities while taking the fingerprints. Moreover, just 1 Hz higher, both approaches might match again, and 2 Hz higher, some completely unexpected behavior could occur.
In short: while having sampled the two room fingerprints, you've completely disregarded the "crosstalk domain".
Stepping aside just ¼ of a certain pipe's length would change its individual room IR signature drastically: what have been peaks may be gaps then. - Photo: Timothy K. Hamilton.
Finally, it looks to me that there's no feasible way to counterfeit the QUANTEC Room Simulation algorithm by means of a convolution plug-in?
Sure enough, there is one dedicated configuration where a convolution clone would be 100% identical.
Would you mind to tell me?
The idea is to just lever out the uncapturable "Crosstalk Domain". Take care of feeding your Yardstick always with a mono signal. Just feed both left and right input with the click, record both outputs' IRs, and then off to convolution!
With this trick, both the Yardstick and its convoluted IR clone do really sound exactly the same?
Absolutely - both are as flat as pancakes now.
Feb 2010, updated Aug 2011
What if I'm on tour and one of the parameter switches breaks off?
Solution 1: Access the lost parameter from the menu, then adjust it with the rotary knob.
Solution 2: Go to "System Setup / Preset / Overview Parameter Line Up" and remap the broken parameter to an unused toggle switch (all parameters with a * are currently unused).
Why do certain competitors' units sound more wide-screen than the Yardstick?
That's because they ignore the need for a proper output correlation of 50%. Just play around with our "Correlation" parameter - so you may adjust the tradeoff between "wide screen" and "center hole" according to your own preferences. Please note that the sonic results for loudspeakers and headsets differ remarkably.
Why do outboard effect units still sound so much better than software plug-ins?
Software dudes wont catch up in just a few years. They might have the CPU power but not 30 years experience with the algorithms.
Why shouldn't they catch up 5 or 10 years? - There are quite a few nearly perfect "guidelines" to learn from, or just for copying.
Just have a look at those Korean piano manufacturers...
While working with QRS and XL over the years, I've been compiling quite an impressing library, which I'd like to reuse for the Yardstick. What do I have to take care of?
Any of the QRS parameters are available on the Yardsticks, too. In addition, the Yardsticks provide some new parameters, which have been pre-assigned for previous devices. Two examples are the edges of the frequency-dependent reverberation time, which have been pre-assigned to TBA Hz and TBA kHz before.
Having said this, neutral standard assignments of the excess parameters are are assigned as follows:
- Bass Edge: don't care
- Bass Gain: ±0dB
- Low Edge: TBD
- High Edge: TBD
- Density: TBD
- DrySrc: 12
- 1stSrc: 21
- 1stDly±: 0ms
- 2nd Cut: 8kHz
- 2nd Corr: 0.0
"TBD" values will be added when available!
Why not sell Yardsticks as RTAS plug-ins for ProTools?
1st problem: the QRS algorithm requires oodles of long audio delay lines. For 1E6 COMPLX, all internal delay lines sum up to almost one minute.
2nd problem: in addition to the 24 bit input and output wordlengths, additional 12 bits of overflow headroom (within the algorithm) have to be accounted for, due to those irregularly distributed room resonances. Finally, there are 8 excess bits for frequency-dependent RT60. All this requires the DSP56321 DSPs on the HD1 cards to operate in double-precision, which cuts the already marginal RAM memory in half from the very first.
Let's consider the widespread Digidesign HD1-PCI(e) card, which can be purchased for around €5000 (VAT not included) over here. The sum of all its delay lines would amount to respectable 82 seconds at 48 kHz (theoretically; disregarding any code memory). As already mentioned, we have to do all calculations in double-precision, which cuts the total delay in half, so there are 41 seconds left. It would take a more thorough analysis to verify that all those fragmented memory morsels can still be efficiently grouped for the QRS algorithm.
Even if the entire HD1 resources would be allocated exclusively to a Yardstick plug-in, this would only allow a single instance of 1E5 (or two instances of 1E4). Both examples represent only one half of what a real Yardstick delivers with its on-board means.
Some would argue that you could install other (3rd party) functions on your HD1 card; as long as they are frugal enough to exist with that memory shortage. What's far beyond the card's capacity would be installing that fictitious Yardstick RTAS plug-in twice, i.e. in multiple instances. Well, a handful of living rooms may well be achievable, but something like a bunch of cathedrals on the HD1 would be a pipe dream...
Jul 2010, updated Oct 2011
A sound effect unit that cannot be integrated within my automation system is absolutely no option to me. Period.
Too bad for those Cappuccino Americans...
...and too bad for Quantec as a self-confessed plug-in technophobe!
Wait a sec! Don't jump to conclusions, please. It's our utmost concern to cooperate with external equipment, but always strictly independent of manufacturer or operating system. That's why we refuse controlling our devices from a hardware plug-in. We're using a manufacturer-independent web browser solution instead.
Within the scope of automation, a clumsy web browser would be pretty useless...
Wrong! Our Yardstick automation builds on the fact that every available audio editor is capable of recording both PCM and MIDI tracks. To benefit from this for automation, set up a new MIDI track specifically for your Yardstick.
Whenever you click a MIDI button (displayed on your web browser), or hit a certain front panel key from the Hotkey menu, the Yardstick sends out a MIDI SYSEX snapshot of all current settings to your audio editor, which will record that snapshot. To be precise: the current active parameter set from Scratch A (or B) will be dropped to the cue point 'now' on that MIDI track.
Whenever one of those cue points is being crossed during a later playback, your editor sends back the deposited SYSEX to your Yardstick. This overwrites its Scratch A (or B) with exactly the same settings that have been disposed to that cue point before. All Yardstick displays skip to the SYSEX values, and the recall will be immediately audible.
This approach should not be mixed up with Presets, or a more or less cumbersome Preset library. It's no more than taking anonymous snapshots, and reconstruct them at a later time. That's completely independent of a currently installed Preset library. Generally, you could go to a different studio, then load your session with all PCM and MIDI tracks to their local editor. With every cue point being crossed, all parameters of the foreign Yardstick will skip to your session values. Contradicting presets of the foreign Yardstick will never be overwritten, just simply ignored.
What can be done if I hit the MIDI key untimely?
Within your audio editor, simply shift the SYSEX object (visible on your MIDI track) forward or backward a few frames - until it fits.
How many of those cue points can I drop on the timing axis?
As many as you need; for gliding parameter values, you could drop several cue points per second, if you like.
Why there is nobody who told me about that?
RTFM - it's documented in the operations guide version 3.1 or subsequent.
Most competing units deactivate their front panel operating elements whenever I try to control the unit from remote (e.g. from a plug-in). Does a Yardstick likewise lock out its front panel while the unit is being controlled by the web browser?
No. In contrast to the competition, all operation elements remain active, regardless of their combination. What's more, all those operating elements (e.g. hierarchical menu, toggle switches, or web browser sliders) remain in lockstep all the time. With minor restrictions, a Yardstick can even be concurrently operated from multiple web browsers on different PCs.
Why don't you ship printed booklets with the Yardsticks?
For the Yardstick series, about three firmware updates are being released each year. With every new firmware update, another operations guide version is being published. Therefore, a printed booklet would be out of date after just 2 months on average.
And if I unconditionally need one, really?
Simply forward the current operations guide's PDF link to you local print shop. No more than two hours later, and at a very reasonable price, you may pick up your spiral or adhesive binding booklet there.
We recommend you to print the full-color version, but you're free to order black-and-white and a yummy pizza instead.
With competitive units, I'm free to split the reverberation algorithm in two so-called "engines", i.e. two separate mono in / stereo out devices. Not so with Quantec. Why?
The room simulation algorithm can deliver its one-of-a-kind quality only if there is unrestricted access to the "crosstalk domain". That's only possible when fed from at least two inputs. More about the crosstalk domain can be learned from FAQ 041, which also outlines why convoluted Quantec-IRs sound flat and dull.
If I've got it right, there is no such restriction with the 2498 and its 8 inputs and outputs. So I can use the 2498 as four independent 2in/2out engines, can't I?
Not yet with the current plug-ins. But we're working on a special 2498 multi-engine plug-in that provides you with up to four totally separate rooms. They all can be individually driven and parametrized.
My dealer has sent me a 2498 for evaluation. How can I hook up that pretty complex unit for testing with simple stereo in / stereo out ?
Connect your AES/EBU stereo "Echo Send" signal to the 2498's rear panel XLR input. For your AES/EBU stereo "Echo Return", use the XLR connector labeled "Out 1+2" from the snake cable included in the shipment. Keep any other of the snake's XLR connectors disconnected. Any unit with the original facory settings will then automatically configure to stereo in / stereo out.
For S/PDIF, the hookup will be more complicated. Please consult the S/PDIF section in the operations guide.
What else should I keep an eye on?
Please double check that you do not duplicate the Dry signal by mistake. If the Dry signal concurrently passes through both the console and the Yardstick's Dry path, ugly interferences (»comb filter effects«) will be inavoidable. As all factory presets carry ideal amounts of dry components, you are recommended to temporarily turn off any direct signals through your console. If you need it the other way round in future regular operation, you may opt for globally disabling all Dry path settings from the Yardstick setup menu.
How about downloading new presets or new software to the Yardstick...
... uploading - the Yardstick is being UPloaded!
Really - why?
Imagine sitting in front of your PC while pulling MP3 clips or movies from a distant server. You're doing what?
I'm waiting for those Downloads - no doubt!
Imagine sitting in front of your PC while pulling a Yardstick plug-in from the Quantec web server. You're doing what?
Imagine sitting in front of your PC. You're just done with a few HTML updates for your Internet home page, and now start transferring your stuff to your web hosting company's web server. Upload or download?
Upload, whatever else?
Imagine sitting in front of your PC. You're just done with downloading the Quantec stuff, and now start transferring your new plug-in to your Yardstick's internal web server for installation. Upload or download?
Ok, we are finished for today. Regardless of size or weight or distance: the only important thing is the direction of the data transfer. A transfer from PC to web server is always an upload, even if the web server is as tiny as a Yardstick.
Whenever I use a 2493 plug-in with a sample rate of 192 kHz, the -80dB LEDs of the ADC bargraphs light up. If I listen into the ADCs digital output, the audible noise floor is almost as low as with 96 or 48 kHz. What's going on?
You're running into a problem of almost every delta-sigma ADC chip on the market today, regardless of consumer or professional. The lighting of the LEDs is caused by a severe noise peak, short before half the sample rate. With 192 kHz, the noise floor between 72 and 96 kHz skyrockets by some 40 to 60 dB.
If such a problem is supposed to exist for almost any AD converter, I wonder why people don't care about it...
You've already confirmed it's lack of audibility, didn't you? This does not mean that the people at Quantec do silently ignore it. There's a snag, undoubtedly, which is disclosed by the bargraphs. In addition, some mirroring frequencies from the sampling process may well be mirrored down to audible frequencies by intermodulation. As long as AD converters with such a deficiency are commonplace, Quantec recommends to dispense with 192 kHz altogether.
With 96 kHz, the problem doesn't exist, right? Not a tithe of it?
No. The origin of the problem is caused by the basic funcionality of delta-sigma converters. With delta-sigma conversion, audio signals are sampled with just 1 to 4 bits of resolution, but with a very high sample rate of typically 6.144 MHz. With a subsequent downsampling to 48, 96, or 192 kHz, that stroke of genius called "noise shaping" shifts all quantization errors upwards, far beyond the acoustic range, so all that smut will no longer be audible. Please be aware that the smut still exists - just consult your dog...
Once again: why at 192 kHz, and why not at 96 kHz?
The ultimate cause originates from the primary oversampling rate of a constant 6.144 MHz. This rate is predetermined by the response of the internal switched-capacitor filters, i.e. technology, and cannot be scaled upwards any more. With 96 kHz, we have an oversampling rate of 64x - just enough to shift all the smut beyond 50 kHz, from where we let it go skyrocketing. The interesting thing is that for a sample rate of 96 kHz, residuals above 50 kHz are beyond Nyquist, i.e. half the sample rate, which means that they are completely removed by the antialiasing filters. This holds even more for 48 kHz.
With a sample rate of 192 kHz, and an oversampling rate of 32x, the results are completely different. The same frequency of 50 kHz, i.e. the edge from where the smut starts skyrocketing, is now far below Fs/2, and thus fits comfortably within the given conversion bandwidth.
Here are a few practical examples:
A very popular quad-channel audio ADC for mixing consoles and multichannel outboard ADCs is Burr-Brown's (now Texas Instruments) PCM4104, which, according to its data sheet, excels with a very low RMS wideband noise floor of - 116 dB FS. Here are the narrowband noise diagrams for each of the three sample rates.
It's obvious for both 48 and 96 kHz that the noise floor is pretty flat. For 192 kHz, the frequency-dependent noise floor starts skyrocketing beyond 50 kHz. For each 10 KHz step from there, the smut energy increases by some 20 dB!
Concentrating on that 20 kHz wide band between 80 and 100 kHz, which is near Fs/2, it's evident that, although comparable in size with the useful bandwidth (aka. "audio band"), this area on the frequency axis roars with a hundred or thousand times the noise energy.
Are all Burr-Brown audio ADC chips just trash?
Nope, any audio ADC from any manufacturer shows such kind of side effect. It's the overtorqued 192 kHz sample rate that should be scheduled for disposal.
This is Crystal's 192 kHz "pro-audio flagship" CS5381 for comparison. Mind the diagram's linear frequency axis that allows an easier recognition of the four virtual 24-kHz-wide "bins".
How and when may this problem be solved?
If future switched-capacitor structures would accurately perform under double the current oversampling rates, we'd get rid of the problem at 192 kHz - just as such noise is meaningless for 96 kHz today. Next is ARDA's recently announced AT1201, the first of a new, but power-hungry breed of audio ADCs with sampling rates up to 432 kHz.
What I fear is a sudden increase in those unawakened users, who seize the opportunity to once again double their house sync to ridiculous 384 kHz. It's exactly them who need to continue coping with those noise peaks, which will be an octave higher anyhow.
Dec 2010, updated Jan 2011
A Quantec competitor like Bricasti succeded, with no more than clever activity in the Gearslutz forum, to kindle an unprecedented worldwide hype for their product almost overnight. Wouldn't it be high time for Quantec, to likewise start stirring the rumor pot by stimulating their certainly respectable user community with the achievements of viral marketing?
Nope, that won't work for Quantec. Quantec is not a lifestyle product, which would gain more and more attraction with every "friend" jumping on the bandwagon. It simply doesn't go with Quantec to have quickies in bedroom studios.
Quantec counts on the expert knowledge, ambition, and experience of professional individuals, who are able to implement ambitious projects, score accurate, and according to acknowledged standards. Every single user accounts "his Quantec" to his individual wealth of experience, complemented by interdisciplinary skills aquired over the years. People rate their Quantec (often acquired with considerable financial means) as their competitive advantage, which should not be carelessly sold below cost - in a drive for recognition or whatever. Gentlemen don't kiss and tell.
But there are still a few scattered Quantec users to jump in once in a while.
They always act defensive, never evangelistic. It happens whenever simple-hearted trend-surfers shoot their mouth off with smattering, which cannot remain unresolved without further discussion. If, in the course of a forum thread, somebody parades Quantec users as village idiots, just because they try to get to the bottom of separate initial reflections, or modulation effects, or just find them music-aesthetically misplaced, they will fight back - not just verbally, but also substantiated with practical examples. Without defending themselves, they would, in the medium term, jeopardize their own special status among the competition - be it impartial or just self-opinionated.
How to update a Yardstick with new software?
No matter if your Yardstick is within reach, or locked inside a machine room down the corridor: an update will be finished in less than 2 minutes, while you remain seated in front of your PC.
- Visit www.quantec.com with your favorite web browser and download new software.
- Visit your Yardstick's homepage, enter "Software Management", then upload the file to one of the free banks.
- Select the new bank from the browser, and, voilà - your Yardstick will restart with the new SW version.
Done!!! - No screwdriver, no EPROM, no cable.
Do I need to feed the direct signal through the Yardstick, or may I do it the usual way, i.e. passing DRY through the mixer?
To minimize potential grey zones, we clearly recommend using the Yardstick's direct signal pass-through for all digital I/O Yardsticks. Doing so would allow you to simply tick off the item, as everything may be optimized automatically.
And for analog I/O Yardsticks?
With analog I/O, the subject is more complex. With AD/DA converters, a latency in the order of 1.5 ms is inavoidable. Sending dry through the console would cause a potentially audible time scatter between "Dry" and "Wet", i.e. wet lags dry by 1.5 ms. Specifically for a room simulation of small rooms, the temporal mapping of direct signal and attack slope may be damaged now.
Feeding my precious dry signal through the Yardstick's AD/DA converters would cause severe headache to me. My first concern is a rise of the noise floor, my second is opening another door for unexpected side effects when recording with 96 or 192 kHz. In other words: I tenaciously try to get rid of any burdens in the signal path, so I don't want to take any risk with another unneccesary AD/DA pass-through for my dry signal.
This problem cannot be solved uncompromisingly for purely analog signal paths. If you would agree to a digital I/O Yardstick, the DRY signal would be handed over with bit transparency, i.e. without any loss, even with 96 or 192 kHz. Even if you insist on feeding the dry signal through your console, the temporal scatter with a digital I/O Yardstick would be no more than 35 samples (valid for all sample rates).
Are all-digital consoles connected to digital I/O outboard stuff generally a safe bet when dealing with latency problems?
Yardsticks, at least, are a safe bet, as their DSP algorithms are capable of operating on 96 and 192 kHz directly.
Does this mean that there are certain outboard devices that still cause latency troubles, or do I read too much into things?
That's right. Be careful with outboard gear that cannot handle sample rates of x2 or x4 directly (e.g. because of lack of DSP power). If such units promise 96 or 192 kHz operation, you shoud expect an internal sample rate converter (»SRC«), at least for the algorithm path. Although noise floors and accuracies of modern SRC chips are excellent, their latencies are comparable to AD/DA conversion chains. Once again you'd risk to damage the timing relation between the direct and diffuse signals.
Take special care with outboard gear that splits its signal paths internally. Some units provide an internal 96 or 192 kHz direct path, but only 48 kHz through the algorithm. 48 kHz samples may be forwarded to both paths, but higher sampling rates may be split. Whenever the SRC needs to jump in, with its inherent latency, even the timing relations internally to the device are going to be scattered by a few ms, silently and without warning.
Jan 2011, updated Aug 2011
Why can such inherently simple circuits like D/A converters sound so different in practice?
Speakers may sound vastly different, too. What exactly your question is targeting on?
It's puzzling me that certain recordings perform very well with a given D/A converter, while with a certain other D/A, I observe an unnatural harsh sound, and other unexpected artifacts. Such audible differences do exist, although the specs of both converters are comparable; even if both specs are above reproach.
You mean that much discussed problem with clock regeneration? - Jitter, phase noise, and so on?
No, the reason that causes my problem seems to be of a different nature.
Since more then 30 years, we at QUANTEC are addressing those unavoidable sampling intermodulation hazards by getting to the root of the problem. We are suppressing potential aliasing directly within our converters.
As an advocate of gray zone minimization, we avoid to shift the responsibility for certain intermodulation risks to downstream devices, like D/A converters or speakers, which, in certain circumstances, may be unable to cope with.
In other words: instead of a blind faith that downstream equipment will do it right, we at QUANTEC assume a priori that it does not. Nevertheless, a sound engineer requires the chain to always function properly.
Time has proved this strategy as a viable, real-world solution with the QRS (1982) and QRS/XL (1987) converters. Up to now, one can read in community discussion boards that those converters are judged to be "outstanding for their time".
That sounds very promising. By the way: regarding the converters mentioned, I concur with the community.
So let's proceed to FAQ 058.
********** Work in progress - currently all audio clips are missing **********
Why is it of such eminent importance that an A/D or D/A antialias filter's attenuation at half the sampling rate is more than 30 dB, and not just the industry-standard of 6 dB?
The easiest way to discover the physical phenomenon behind this problem may be demonstrated with purpose-built sound clip FAQ#58_1 (see below).
A gliding sinewave from 20 to 25 kHz has been recorded with 48 kHz, as usual. To scale the critical frequencies near half the sampling rate, 24 kHz in this case, down from ultrasound to the human audio band when played back, each recorded 48k sample will be stuffed with 7 trailing zero samples (and level increased by 8). Net result is that the recorded material is played back with 1/8 of its pitch and speed, i.e. 3 octaves lower, similar to an 8 times slower magnetic tape.
Please listen to sound clip FAQ#58_1 (see link below).
What exactly should I listen for?
You'll hear a gliding, ascending sine wave; nothing spectacular at first glance. If, however, its frequency approaches the reach of half the sampling rate (3 kHz in this experiment), you'll notice that the ascending sine wave more an more clearly will run into a second tone, descending from the top.
Surprisingly, there are two sine waves now: the externally-fed ascending sine wave (0->Fs), and its synchronously descending mirror image (Fs->0) (»Alias«). At exactly Fs/2 (3 kHz), both tones meet each other in unison. Frequency difference will be zero then, so only a single tone will remain, possibly with a slow beat.
Let's note: You feed the A/D converter with a single tone, but it responds with two.
Since all this happens in the ultrasonic range, nobody would care about it; with the exception of my studio doggy "Dolby". So why should I ever have a problem?
A well-known term in analog audio is Intermodulation. This term describes the disturbing effect of creating additional tones, whenever two tones are applied to a non-linear transfer function. Those tones are equal to the sum and difference of the individual frequencies. From F1 and F2 thus arises F2+F1 and F2-F1. As low intermodulation distortion represents an important quality criterion, there are standardized test requirements.
Through the principle of sampling, or more precisely, by an inevitable ambiguity associated with the process of capturing the samples, when feeding the system with a tone F1, a physically inevitable mirror tone F2 with F2=Fs-F1 will result. This is the downward moving sine wave you've just heard.
Below is a chart with sampling points from a 48kHz recording, intentionally constructed to match two frequencies, i.e. both 23 and 25 kHz. This causes an ambiguity in the sampling process to cause mirroring of the real tones, and, based on this, is ultimately responsible for any difference tones in the audio band.
Here, the two distinct tones "traditional" to an intermodulation process need not to be provided explicitely. It's enough to provide a single stimulus, which will automatically be joined, more or less intensively, by its inevitably emerging "mirror partner". From these two, now really-existing components, intermodulation products can arise, too, and likewise with their sum and difference frequencies.
A numerical example: if, at 48k sampling rate, an A-to-D converter is fed with a sinewave tone F1 of 23 kHz, its "mirror partner" F2 responds with 25 kHz. If there are any nonlinearities in the subsequent signal path, as described above for traditional intermodulation, secondary tones of F2+F1 and F2-F1 will show up. The sum of F2+F1 is irrelevant, since it is always 48 kHz, thus inevitably falls into the first zero. Within the audio band, and therefore acoustically relevant, however, lies the 2 kHz differencial tone F2-F1. As long as there are no non-linear transfer functions in the signal path, differencial tones would only be created theoretically.
Here, a potential gray area becomes apparent, as that highly explosive ultrasonic double tone is passed down the signal chain as kind of a cuckoo's egg. We are likely to agree, that an energy-rich mirror tone outside the harmonic spectrum has no business within the audio payload, even if it's in the ultrasonic range.
If I understand you correctly, caused by sort of "phantom tones" in the ultrasonic range, I'm getting issues in the acoustic range, right?
Initially, it's just the risk of a problem in the acoustic range. Picture yourself on the domes of your monitor system's tweeters, which need to radiate both 23 and 25 kHz concurrently with large volume and low distortion - certain doubts regarding their linearity are advisable for obvious reasons. In particular, one needs to clarify how your tweeters can put away, with acceptable intermodulation rejection, much more complex sound clusters, such as triangle and mirror-triangle.
Please listen to sound clip FAQ#58_2 (see link below).
Note: Make sure that, at least for this Sample, you're listening with full bandwidth. The following test may fail if the audio band is trimmed to e.g. 10 kHz, just because your PC's sound system is unable to cope. Frequencies of up to 25 kHz need to be reproduced virtually undamped.
What exactly should I listen for?
With a sampling rate of 48k, an ADC with -6dB antialiasing filters is fed with a sine wave F1 slowly gliding from 20 to 25 kHz. Filtering of a mere -6dB results in an only slightly attenuated mirror tone F2. Despite full scale level, you shouldn't hear anything. If you'd hear sliding whistles instead, these would be objectionable F2-F1 difference tones, caused by an intermodulation problem of your PC, or your monitor speakers.
Sound clip FAQ#58_3 serves as a crosscheck (see link below).
Unlike the previous example, the ultrasonic difference tone is not output directly here, but, just before recording, distorted by a slightly curved transfer function, which roughly corresponds to a THD of 10%.
What exactly should I listen for?
This distortion causes that the two ultrasonic tones F1 and F2 cannot slip-through unnoticed any more. Any cheating will be disguised unmistakably by a strong differential tone F2-F1, caused by that curved transfer function. Because the tones are right within the audio range, it is impossible to remove them at a later step.
Can I estimate, or measure, this kind of problem in advance?
If your monitoring speakers would specify intermodulation behavior at near ultasonic frequencies, you could look there for some hints. Since tweeter domes are mechanical systems, I would be reluctant if there's just a single-frequency spec. However, if results from a whole bunch of intermodulation tests at various frequency pairs would be available, you may estimate intermodulation suppression from the set of data, and, as a consequence, the attenuation of mirroring products from the sampling process.
Then everything should be fine with my outrageously expensive studio monitors, right?
Obviously everything's alright! - Same holds for your dentist, who has forked out some twenty thousand €€€ for his stereo system. Or even those audio performance tests at a high-end professional paper's editorial office; even there, the problem will go unnoticed, as with their precious equipment, disturbing difference tones simply won't occur.
But the grey area remains. Try to put yourself in one of your more common customers. You probably don't even believe that a €20-class tweeter dome can put away our fictitious triangle caleidoscope the same way your studio monitors do - or the high-end bolides at your dentist.
I can easily imagine those completely overwhelmed low-cost domes beating around the bush. This makes me think of the following idea: could some maybe hidden intermodulation chirping be causally involved whenever one and the same CD, with different D/A converters, sometimes sounds clean, sometimes veiled?
Or that 96k sound recordings usually sound more transparent than 48k, because the (analog) audio energy in the vicinity of Fs/2 = 48 kHz is already too weak for causing energy-rich, and thus annoying differential tone artifacts?
Perhaps, a theoretically required Nyquist sampling rate of 48 kHz would be sufficient anyway, if you could only get that basically easy-to-understand intermodulation problem under control?
Simply try it! - With the 2493 Yardstick's AD/DA converters, you could easily switch from industry-standard antialiasing filters (-6 dB @ Fs/2 - flat) to carefully optimized Quantec filters (-30 dB @ Fs/2 - steep) - and back again. Simply put your 2493 in BYPASS mode, and ... hear for yourself.
Please listen to sound clip FAQ#58_4 (see link below).
What exactly should I listen for?
With a sampling rate of 48 kHz, a sine wave tone F1 slowly gliding from 20 to 25 kHz is fed into an ADC with -30dB antialias filters; which results in an already very low level mirror tone F2. Once again, the ultrasonic difference tone is not output directly, but, just before recording, distorted by a slightly curved transfer function, which roughly corresponds to a THD of 10%. Even here, an audio band differential tone F2-F1 is generated by the two ultrasonic tones F1 and F2. But as F2 has already been strongly attenuated by the optimized filter edge, a differential tone is no longer perceptible.
Final question: When and by whom have these relationships actually been discovered? Given the currently still overwhelming supply of A/D and D/A chips with "built-in -6dB problem" - and a huge number of onboard and outboard converters built with these chips - the moment of enlightenment can probably be not too far in history ...
It was 1978 when the QRS designer (and coincidentally writer of these lines) has worked on A/D and D/A concepts for the QRS. As a young guy, he could easily perceive tones up to 18 kHz. Moreover, as the sampling rate of the QRS was just 20 kHz, the problems already manifested with higher-pitch audible tones then, and not in the near ultrasound as today.
In this historic scenario, a certain intermodulation distortion was not at all a mandatory prerequisite for folding down the problem into the audio range. The gliding, descending sine wave was directly audible, similar to the FAQ#58_1 sound clip.
The following sound clips are provided solely in wav format. A lossy encoding like e.g. MP3 is no longer suitable for such challenging material. Normally insignificant artifacts of data reduction, especially masking effects, would cause completely unexpected side effects when used on such borderlines.
All sound clips to FAQ 058 in preparation!
Jan 2011, translated Jul 2011
Recently there's some buzz about "piano reverb". You may already have guessed it: similar to the QUANTEC QRS algorithm, piano reverb is based on a resonance instead of a time-delay model. Are there any more similarities?
Yes and no. Despite significant differences in detail, one could say that the conspecific relations between QUANTEC and piano reverb are somehow closer than e.g. between QUANTEC and LEXICON. While at least the early LEXICON algorithms have tried to map the entire room acoustics on a single loop, both QRS and piano reverb have always relied on massive parallelism.
Piano reverb is based on the centuries-old principle of freely vibrating, not artificially (e.g. felt) damped resonant strings. The most famous representative is the harp, but the principles can be found at a variety of other stringed instruments as well.
In piano reverb, many narrowband resonators based on tense strings are available, which, due to their high Q-factor, show a pretty sluggish attack and decay behavior. Each of these well over 100 resonators can be excited in its natural harmonic series of resonances. In their initial phase, they produce some kind of a gentle attack slope, and, when the stimulation stops, change over to a decay similar to a reverberation tail: Hal-le-lu-jah!
The reverberation time of piano reverb is largely determined by the piano's soundboard. The more energy the soundboard detracts from the strings, also dependent on frequency, the shorter their reverberation times, i.e. RT60. A special feature of algorithmic piano reverb is its precise, in steps of e.g. half tones controllable reverberation time, which is said to be reminiscent of a one-third octave equalizer. However, it seems to me that the advocates of piano reverb have overlooked overtone excitations ...
Probably piano reverb stands or falls with the intonation of the strings, right?
Right. But beware: a common mistake would be, for example, to equate the number of attainable "room modes" with the number of piano keys available. Moreover, a piano provides pairs and triplets of strings for the higher keys, and each string cannot only be excited on its fundamental, but also on its 2nd .. nth harmonic.
This means that the standard piano tuning is extremely unfavorable for piano reverb. Not only because all higher octaves were already redundant, but also because any existing pairs and triplets would uselessly vibrate in unison. So you should avoid any intonation in which the valuable higher strings would be wasted for spectral lines, that are already "done" by the harmonics of some lower-tuned strings.
So what about some fundamental differences between the QRS algorithm and piano reverb?
Although our QRS room models are also based on an optimized interaction of a multitude of resonators, the tuning does fundamentally differ from that of piano strings. The "intonation" of the QRS matches much better an important quality criterion for real rooms, which can be described as "number of resonance lines per Hz of bandwidth."
Another fundamental difference lies in those characteristic, extensive propagation delay times within the QRS algorithm. While piano reverb kills all temporal relationships within a tone sequence, the QRS algorithm provides a multitude of "built-in samplers", which can record trills or similar progressions, and replay them in the course of the reverb tail as wildly swirling sequences.
Up to now, the issue of "echo density" was still left out. How about that?
The extremely short length of piano strings of no more than 2 meters results in extraordinary high echo densities, which would be typical for very small rooms. Even the piano wire braiding makes no spring reverb out of these strings, because the turns are almost "shorted" through the heart strings.
Such extreme echo densities are by no means a challenge for QUANTEC room models. Just work with similar sonic propagation delays to match the piano strings.
Ironically, to mimic the extremely short propagation times of piano strings seems to be of limited use, since such ill-practices would cut e.g. the QRS algorithm's "trill bonus". Musical quality rooms concurrently deliver high mode density and long walking distances and a high echo density - piano reverb cannot compete with that.
Looks to me that exactly this balancing act can be managed with QUANTEC room models in an incomparably uncompromising way.
Jan 2011, translated Jul 2011
Apparently, with the exception of the 8-channel 2498, there is no "Predelay" parameter. How could QUANTEC overlook basic reverberation ingredients?
What you're looking for has been labeled "Postdelay" for all Yardsticks with stereo inputs. The idea is that a classic predelay would be redundant in surround mode, since between the various surround outputs, a time delay variation needs to be explicitly defined anyway. If you want extra pre-delay, just pack it onto the already existing post-delays.
But I need no surround.
For Yardsticks with stereo outputs, the approach is even simpler: it doesn't matter whether you place the extra two-channel delay in front or behind the Yardstick. To keep presets compatible between our different devices, we have maintained "Postdelay" (as a concept and functionality) even for pure stereo versions. In practical operation ("chain of linear four-poles"), both predelay or postdelay would do the trick: reverberation starts later.
And why does the 2498, and only this one, provide an explict "Predelay", in addition to "Postdelay"?
Because the 2498 provides the unique feature of true staggering of depth inside the simulated room. In contrast to getting out a surround signal, depth-staggering is based on a time-delayed putting in of individual stereo pairs into the reverb chamber. Apart from different input material, these stereo feed-ins differ in bandwidth, room mode distribution, and, you guessed it, predelay.
Maybe the following mnemonic can help readjusting your perspective:
Predelay is for staggering of depth - postdelay spans the surround-tent
Feb 2011, translated Jul 2011
Why does QUANTEC not tolerate publishing of IR collections, even if they are not for profit?
Because convoluted room simulation IRs by far cannot keep up with the quality of the real device (see FAQ 041).
You have to make concessions once in a while - as long as it's free ...
Our problem can be explained when looking at the distributed market volume of devices. From competitors' units in the $500 class, many more original equipment has been brought among the people than with QUANTEC. At affordable prices for everyone, the quality standard applies to the original device, which can be acquired or borrowed easily. Consequently, interest in dubious IR collections, targeted to replace the original, is limited.
And with QUANTEC?
QUANTEC units are relatively expensive and therefore rare, so the majority of reported "expert knowledge" couldn't come from first hand experience. Particularly in connection with our products, we regularly face forum posts, where it is claimed that "a Yardstick does not provide much more transparency or space than a cheap MXR or Behringer device."
Over the course of the thread, you will probably notice a puzzled real QUANTEC user joining in, asking if the talk is about the now outdated 2402 Yardstick from 1997, or about a Yardstick from the current 249x series. Typically, the originator will admit sheepishly to have "downloaded Yardstick IRs from xyz, and tested in a convolution plug-in."
If people compare our quality devices with an inferior plagiarism taken from an unknown supplier, they must end up with an out of thin air, in practice totally baseless misjudgment. The only bridge to the original are just two (of several hundred) broken out of context QUANTEC IRs, which are touted as "95% original" in some circles. We do not want to deny that, at least in the reverb tail, one ounce of QUANTEC may flash up. But accounting for no more than 1 percent of all IRs, namely those of 100% left and 100% right, requires to accept that the legendary QUANTEC spaciousness inevitably collapses altogether.
If our original device user had not joined that thread, such nonsense would have been communicated to the public for years - what a loss of image to our brand ...
Hence our credo: QUANTEC is only inside if you can read it on the enclosure.
Mar 2011, translated Jul 2011
In Internet forums, people regularly praise Bricasti's M7 with superlatives like "the highest-performing audio signal processor in the known part of the universe". Can we really leave such claims unchallenged?
Of course not - although Bricasti uses six dual-core Blackfin DSPs that are clocked at 600 MHz. All together, at first glance, gigantic 14,4 GigaMACs/sec. Anyhow ...
What is it? - Don't hesitate trying to make the best of such a super galactic advantage of the competiton! - As you have surely read already: "The M7 has more power than a battleship!"
It's, once again, this familiar response cocktail of hype, half-knowledge, pseudo-arguments, and marketing bluff. I'm currently not even sure, if I should entrust you with anything. In these days, just because of a leak in the publication of inconvenient truths, you risk being credited with sexual misconduct, or dubious gas deals ...
Sorry! - I'll keep it for myself! - Promise!
Thank you! - So, here we go: In a city somewhere in Massachusetts, carefully sealed off from the rest of the world, lies a skeleton in the closet, hushed up by mighty marketing strategists, and hosts of loyal zombies, thoroughly distributed throughout all social networks around the world ...
A skeleton in the closet? - That sounds exciting to me ...
Hold on tight now: the BlackFin DSP engine, incorporated six times into this alleged "battleship", is a 16-bit processor! - That's the dishonesty behind such impressive benchmark figures. Fast but narrow-chested, the "PMPO of digital signal processing". Each one of these processors offers four brisk hardware multipliers, but all they can do is 16x16 bit! - All right, still a battleship, but one that is moved forward by six outboard engines whining full rev.
For heaven's sake! - 16 bit audio? Isn't that a dusty, ancient technology from the '80s? Well, such kind of antiques have long been extinct for professional audio applications, alone because of the now-standard digital inputs and outputs with 24 bits.
Not necessarily, because 16-bit DSPs are meanwhile ridiculously fast. The key to make it work for the most part is called "double precision". Which means a software routine to split the 32-bit job on multiple 16-bit jobs. Down the speed, up the torque: a sort of reduction gear optimized for the task to be managed. And, even in this context, the law of conservation of energy applies ...
As a consequence, we now have half the performance, i.e. 7,2 GigaMACs/sec. That would still be gigantic!
No, not half! It wasn't meant that literally with the energy conservation law. In "double percision" with 32x32, the performance collapses to 1/10, because a respectable number of additional operations turns out to be unavoidable.
What you say now so easy! - Is there a traceable source for this?
Natch! - Just google Analog Devices' "Engineer To Engineer Note EN-186" titled: "Extended-Precision Fixed-Point Arithmetic on the Blackfin Processor Platform" - you'll find that approach described in detail there. Whereby there are definitely also critical voices to this approach:
I really think that application note EE-186 is wishful thinking on ADI's part in order to try to sell its Blackfin processor to audio folks. The Blackfin looks like a great 16-bit DSP, but for high resolution audio work I think a better choice would be a DSP that has native 32-bit arithmetic.
Could you briefly summarize the paper for the reader?
Using an example with decimal numbers may outline the procedure. Consider the following example, and take each decimal digit (2, 3, 4, and 5) as a replacement for a 16-bit word.
Determine partial products:
3 x 5 = 15
3 x 4 = 12
2 x 5 = 10
2 x 4 = 8
Shift partial products to the correct decimal position:
Add up partial products:
Result 23 x 45 = 1035
Calculate 23 x 45 with a multiplier that can only handle single-digit decimal numbers ...
In that way, at 32-bit fixed point requirement and 32x32 MACs emulated, the M7's impressive six DSPs effectively boil down to just 0.6 DSPs. Thus we arrive at 1440 MegaMACs/sec; admittedly still a respectable net result.
Are 32-bit fixed point algorithms at all sufficient for professional audio?
Apart from some special cases, such as parametric EQs with IIR filters, 32-bit fixed point easily fits most audio algorithms, such as chorus, flanger, delay, compression, and FIR filters.
Would 32 bits and 1440 MegaMACs/sec be sufficient for the QUANTEC room model?
Close, but no cigar! - This surprisingly multi-faceted topic is being covered in depth and in all respects within FAQ 064.
Now I understand how the DSP power of the M7 collapses dramatically with increasing dynamic requirements and more accurate coefficients. Perhaps it would be possible to trade these dependencies against each other in specific cases. Is there any further documentation?
Here's a diagram, in which the dramatic depletion of the remaining DSP audio performance of the M7 (red) with increased requirements of dynamic range and accuracy becomes evident. Note in comparison the characteristic constant performance (blue) of the 249x floating-point DSP. Exact figures may be found in FAQ 064.
Mind you, here compete six DSPs (red) against a single DSP (blue). Once the requirements of an audio algorithm are higher than the mainstream (the two gray bars with black lettering), our rock solid floating-point DSP does not need to hide behind the "6-engined battleship" - quite the opposite.
As reasonable choice for high-end audio algorithms, it seems wise to actually accept no else than floating-point processing ...
Actually, yes! - The exceptionally high dynamic requirements of the QUANTEC room model are the main reason why the Yardstick was equipped with a high-performance floating-point DSP. The IEEE 754 32-bit floating-point format offers, at reasonable cost, a huge audio dynamic range, that would not be achievable in fixed point with word lengths of less than 256 bits. All that in a single machine cycle, thus highly efficient.
If 32 bit would be sufficient for an algorithm, then the six 16-bit-processors of the M7 still deliver respectable 1440 MegaMACs/sec, despite being thwarted by double-precision overhead; that's after all 2.4 times the throughput compared to the Yardstick.
If however 32-bit double-precision is no longer sufficient, as it's the case with the QUANTEC room model, realizing a triple-precision fixed-point format does not make much sense any more regarding the M7 architecture. Who, instead of working through the algoritm, nearly 96% of his time is occupied with himself, will be gradually running out of steam, even with a formidable sixpack of DSPs.
In such scenarios, the Yardstick's more seasonable floating-point architecture can play out its strengths convincingly: better throughput compared to the M7, easier programming, extremely small size, and last not least a much lower energy consumption, which renders active cooling and its associated »vacuum cleaner noise« superfluous.
Oct 2011, translated Dec 2011
Compared to the competition, where is the Yardstick in energy demand?
Photo: Björn Schwarz / pixelio.de
Why does the M7 stand out so fiercely when used for 2in/6out?
It's because for surround mode, three M7 units need to be cascaded.
What are the values shown based upon: real or apparent power?
Unfortunately, in the course of our investigations, we couldn't find out whether our American colleagues aimed at active power or apparent power in their declarations. It's puzzling for us that the Americans use the unit Watt (active power: equipment heat, electricity meter speed), while in Europe, for the a.c. power grid the unit VA (apparent power: cable heat, power loading of the grid) is exclusively used.
All QUANTEC backpanel legends and all technical documentation invariably refer to apparent power, i.e. VA. Taking into account the measured power factor of 0.59, the following table completes the picture:
It's clearly seen that power information in Watt can show power consumption a lot more favorably. Regarding the American devices, we'd be grateful for valuable information from readers.
For a power factor of 1.0, both data values would merge. Such high values, however, would require a power-factor correction circuit, which, for powers of less than 75 watts, is neither required nor customary.
Now it's also clear why those guys at Bricasti need a fan ...
Mar 2011, image added Jul 2011
Would it be realistic to port the QRS room model code to Bricasti's M7?
No! - The dynamics of the M7's built-in 16-bit processors are not nearly enough.
But it should be possible, with reasonable effort, to use a software for operating the 6 BlackFins in a "double precision" mode. With 32 bits of word length, this would make available still 10% of the original 16-bit throughput (see Analog Devices EN-186).
That's correct. But, as you already mentioned, you would burn 90% of the computing power. Any idea how to dig all that heat away from your brick?
1440 MegaMACs/sec at 32x32 should be more than sufficient, right?
The bottleneck is not in the throughput, but that the 32-bit fixed point format is not sufficient for the QRS room model. For the correct processing of the imposing resonance magnifications caused by "room modes", which arise by necessity at many places within the simulated room, an overall headroom of at least 12 bits is necessary.
If the audio inputs and outputs have a word length of 24 bits, as it is common today, the internal requirement of the QRS algorithm would be at least 36 bits. Add another 8 bits for the frequency-dependent reverberation time, and you'll end up with at least 44 bits of dynamic range for the QRS room model.
Then I offer Triple-Precision: 436 MegaMACs/sec at 48x48.
Unfortunately again off the mark, this time in throughput!
DSP machine cycles
QRS room model
for BlackFin Extended Precision
M7 16x16 native
M7 32x32 ext.prec.
M7 48x48 ext.prec.
M7 64x64 ext.prec.
249x Floating Point
Since, obviously, a tricked-out BlackFin engine with "double precision" at best creates a 32x32 fixed-point version of the QRS room model, the algorithm probably needs to be slimmed down. At which places restrictions are to be expected?
I can only speak for the QRS room model, as I have no insight into the Bricasti algorithms. In any event, with 32 bits internally in the algorithm and 24 bits at the inputs and outputs, a headroom of just 8 bits would be totally inadequate.
Where exactly would you make cuts to the QRS room model for still squeezing it into 32 bits?
The first thing you need, is to say goodbye to reverberation times continuously gliding on the frequency axis, which has always been a basic requirement for QUANTEC. Instead, as it is customary with many competitors, split the audio signal with a three-band configurable crossover filter, then tweak its 3 bands in terms of RT60 as needed. Every natural space shudders such manipulations, but for now, this trick would take us from 44 to 36 bits. To get rid of the still overshooting 4 bits, the following strategies could be considered:
Reducing the word length of the audio interface from 24 to 20 bits (possibly fully acceptable for analog I/O)
Reducing of the maximum reverberation times to 1/16 for each size. For example at 1E4 from 20 to 1.25 sec
Another option would be a combination of both strategies, e.g. 22 bits and 5 sec.
All the above compromises are unacceptable to me - what remains?
Perhaps mixed-precision - a few approaches are compared here:
DSP machine cycles
QRS room model
for BlackFin Extended Precision
M7 16x16 native
M7 32x16 ext.prec.
M7 48x32 ext.prec.
M7 64x48 ext.prec.
249x Floating Point
Whoops: M7 green and ok at 48x32; so it's not yet lost?
No! - But in fact, the much-hyped "King of audio DSPs" has slipped aside an embarrassment just by a hair's breadth: the M7 is rather not an order of magnitude more powerful than any other audio signal processor on the market, but positions itself for the QRS room model, despite its concentrated charge of 6 signal processors, only in a single, fairly exotic operation mode a humble 14% above the Yardstick. In any other mode, it fails at the QRS room model. For the popular 48x48 mode, used in audio technology especially for IIR filters and equalizers, the Yardstick would be even 38% faster than the M7.
After all, we now know that there is one trick, how the Bricasti M7 (with 6 DSPs) can have an edge over the Yardstick 249x (with only a single DSP) - albeit just curtly. You have to admit that a mixed-mode with 48x32 is something rather contrived, exotic, and computationally cumbersome ...
And compared to the Yardstick?
For the QRS room model, the Yardstick DSP provides sufficient 600 MegaMACs/sec, and does so without contortions, with a luxurious floating-point arithmetic unit, which fits much more organically to the entire spectrum of signal types occuring in the professional audio technology.
Now in plain English: could QUANTEC, in a more or less distant future, consider a port of the QRS room model on the Bricasti M7 for realistic?
No! - In contrast to a Yardstick, there is no way of foisting "3rd party" plug-ins on a Bricasti M7. So it would require interweaving of both Quantec silverware and Bricasti silverware inside their EPROMs, which, for reasons of confidentiality, would be absolutely unthinkable for both parties.
Oct 2011, translated Dec 2011
Why are there no QUANTEC devices at SAE (»School of Audio Engineering«)?
Have you ever seen a driving school that trains on Maserati?
Apr 2011, translated Jul 2011
A real reverberation chamber with driving speakers and capturing microphones - that's it, right?
Not necessarily! - Reverberation chambers are usually much too small, and Mother Nature cannot be deceived with regard to room size. There are a number of shortcomings with reverberation chambers, that cannot be compensated with ugly tricks like pre or post-delay.
With the usual two speaker feeds for the reverberation chamber, you'll get only two room mode spectra, nowadays often referred as »IRs«: a spectrum originating from the left speaker, and a spectrum originating from the right speaker. An orchestra, like a church organ, consists of a large number of sound sources distributed across the room; each of them stimulates the room in a highly individual way. So to speak, every instrument and every organ pipe comes with an own, very personal IR. Such behavior cannot be mapped easily with 2 feeds and 2 IRs, and not even with a few more. For my part, a few hundreds of IRs may do it, but don't overlook the need for a corresponding number of echo send paths ...
Reverberation chambers usually have a relatively small volume. Anyway, they're incomparably smaller than e.g. churches. In comparison to large rooms, small rooms show a much lower mode density (=density of room modes on the frequency axis).
Photo: Henry Mühlpfordt / flickr.com
Since a room can only form reverberation at points of resonance, smaller rooms show a lot more "holes" in the frequency spectrum of the reverb. One can clearly hear that as a more or less metallic coloration, especially with narrow-band sounds like choral music. Reverb with insufficient self-resonant lines on a soprano aria may cause physical pain in your ear ...
That's it then, right?
No, one more step. Picture yourself on a church gallery playing a trill on the organ. What does a trill? It spreads through the room. Let's now trace one of those signal paths that hit the front of the sanctuary, bounce back, and eventually return to your ear after several 100 milliseconds. You've guessed it, still warbling, why not? It's the interaction of a variety of return paths with different lengths that blurs and neutralizes time factor, and turns the trills into a continuous reverb tail. But the molecules inside the reverb tail continue warbling. When listening carefully to the reverb tail of a large church, you can occasionally hear distinct audible trills rising up from the crowd; that's part of the game.
For small rooms like a reverberation chamber, due to much too short signal paths, the individual trill components can no longer be sequencially aligned, because some sort of "sampling buffer" is missing for capturing the trill. Inside small rooms, whether real or digital, a spreading of musical sequences cannot happen, so spatial depth is not possible. Result: in contrast to a large real room, the spacious impression will be drastically reduced for conventional reverberation chamber sizes.
May 2011, illustrated Aug 2011
In Internet forums, people occasionally complain about "paying exorbitant prices" for "the Yardstick" compared to the Bricasti M7. What do you, as a manufacturer, think about the origins of such statements?
With an audio tool such as the Yardstick, the primary requirement is realistic sound, to an extent that it cannot be distinguished from a real room in any way, even by a trained ear. In addition, as various journalists have outlined in their reviews, you can "buy competence" with a Yardstick, to put your studio a great leap ahead of the competition (that alone is "priceless"). Ultimately: when asking for the price, one cannot afford such device anyway - am I right?
Talking about sound: we invite every customer to challenge the Yardstick against any competitors in an A/B contest, before finalizing their purchase decision. Up to now, we've won almost every shootout brilliantly. The reason is that, provided real practical experience from first hand available, we score flawlessly ...
Which means that the Yardstick is still considerably more expensive than the competition?
No, no way! - We can explain such prejudice only by the fact, that many an annotator not just misses the view (or the patience) for the "big picture", but particularly that those "wannabe spindoctors" cannot substantiate any first-hand practical experience. As with any professional investment, it's about the price-performance ratio of the total solution, not a low entry price for merely a partial function. It's notably the Yardstick's high utility, which is explicitely praised in all reviews from the trade press.
What do you mean by "big picture"? - Where would I benefit, or have savings from?
It's about the standard accessories. Unlike the competition, a Yardstick has its remote controls virtually "built into the base unit". Moreover, Yardstick and remote controls aren't connected with a point-to-point connection to the base unit, as it is commonplace. In our approach, right away after unpacking and plugging into the nearest Ethernet switch, you'll be given Yardstick remote controls throughout your entire premises, even several at once. In the near future, we'll present a similar concept for providing all workplaces with Totel Recall and Dynamic Automation.
You'll save thousands of $$, as any additional costs for ICONs, LARCs, or M10s are completely eliminated (possibly even multiple times). Installations at each workstation (eg. software or cables) are unnecessary, because you benefit from an already installed Ethernet infrastructure (including laptops, netbooks, and tablets with WiFi/WLAN), available throughout your studio or broadcast center anyway. Complemented by any of the ubiquitous Internet browsers, often installed first when a new workstation is being set up.
Oct 2011, updated Mar 2012
Exactly how many presets are available for the Yardstick?
For the plug-in with the designation 'QRS' in version 3.1, there are 70 manufacturers presets. But for future plug-ins with other functions or effects, this could be quite different.
All Yardsticks with 2 input channels offer 15 banks, while the 2498 offers 63. In any one bank, any plug-in functionality (same or different) can be installed, each of which brings its own presets.
The 'QRS' plug-in version 3.1 holds 250 user presets per bank. Consequently, a fully-equipped Yardstick, with 15 QRS instances installed, can hold 3750 presets, while configured with 63 instances, a 2498 can hold 15750 presets accordingly.
If, by browser and ethernet, all 250 presets from a bank (en bloc or individually) are being exported to a PC, the number of presets is unlimited.
Can you name the presets at will?
Yes, with both "cranking" on the front panel, and - much more comfortable - on the web browser.
How many characters do I have for the name?
Such a low number of characters tends to get messy, especially on the PC ...
The bottleneck is the quite limited 1U-sized front panel display. We want more important items to be displayed than lengthy names. On the PC, we recommend to create an own directory for each archive and each project.
You may have overlooked that when written to memory, each customer preset will automatically be stamped with date and accurate to the second time. Consider this time stamp, which is displayed prominently on the front panel display, as part of the preset name. In other words: without any need for typing a name, a customer preset will automatically get a time stamp when stored. This makes your preset unmistakable, even if all presets would carry the same name. With the help of your studio diary, you can assign a "stray" preset to a particular session again, even after months.
Incidentally, the file names on the PC are very much longer, because, in addition to the time stamp, we tackle a whole range of environment variables in the file name. This includes the customer-assigned device name, the name of the plug-in the preset comes from, and a few other things.
What about the room mode density of the QUANTEC room model?
This of course depends very much on the assigned room size, and also the complexity of the room model used. But it's comparable in magnitude to real rooms.
Would you please allow me to substantiate your question with the help of a numerical example?
As is well known, a room can only form reverberation on frequencies where room resonances (»modes«) are available. The less of the useful signal "slips through the cracks", the lower the coloration of the reverb sound.
A feedback delay line of 1 sec has a natural frequency density of 1 Hz, so it can form reverberation at e.g. 98, 99, 100, 101, 102 Hz. For a delay loop with just 0.1 Hz, the density would be 10 Hz, e.g. 80, 90, 100, 110, 120 Hz. Imagine now operating 10 loops in parallel with an average (but not exact) of 0.1 sec, specified in a clever way so that the natural frequencies of each loop fill the gaps of the others, then again an average density of 1 Hz will result.
What I'm aiming for: to estimate the natural resonance density of a reverberation device in a first approximation, one could simply add up all delay lines built into the algorithm, and then form the inverse.
Incidentally, in the days when RAM was still really expensive, a cursory glance on its memory chips was usually enough to estimate the mode density of a reverb.
So I just open the screws e.g. of a vintage QRS, find memory chips with 64 kWords, which deliver a bit more than 3 seconds with a sampling rate of 20 kHz. So the natural resonance density of the vintage QRS is about 0.3 Hz.
Exactly! - But beware, this only applies to the maximum room size of 106m3. The smaller the room, the greater the gaps on the frequency axis, and the more metallic coloration will be in the reverb sound.
What exactly needs to be fixed in a bathroom to get rid of its coloration?
Increase the natural resonant density or, in other words, add more delay lines. Based on the loop calculations above: a bathroom of 3 m cannot deliver delay times of more than 10 ms, but you're still free to increase the number of paralleled delay lines, and try to bend them away from the uniform 10 ms whenever you can. Ultimately, this leads to the well-known reverberation chambers that attempt, with crooked walls and ceilings, to stuff as many independent delay lines into a given space as possible.
Photo: Henry Mühlpfordt / flickr.com
I should perhaps mention that such a strategy does reduce coloration indeed, but, even with pre or post delay, cannot pull a church from a small echo chamber. Delays of a mere 10 ms simply don't "reach" far enough for a convincing impression of auditary spaciousness. In addition, such agile small echo chambers lack the sedate complacency associated with large churches.
But now to the facts!
For the COMPLEX QRS room model with a room size of 106 , the sum of all nested delay is almost 50 seconds - hence the self-resonance density calculated by the inverse is 0.02 Hz. This is 15 times denser than the vintage QRS. You can clearly hear the improvement when listening to the 010 piano sound clip, which compares the two diffuse signals directly.
A natural resonance density of 0.02 Hz is undisputedly sensational; but what about smaller rooms?
For smaller room sizes, the natural resonance density scales, to a first approximation, with the cube root of the rooms' volumes. Or, slightly oversimplified, with edge length.
Hence a deconstruction from 106 to 104 would scale to about one fifth. Consequently from 0.02 Hz to 0.1 Hz, which would be, coloration-wise, still a very good value.
So far, only the computationally intensive COMPLEX room model was mentioned. I guess the natural resonance density may reduce with MEDIUM or SIMPLE?
Indeed! - For MEDIUM, the self-resonant density is cut in half; and again for SIMPLE. For 104 we would end up in 0.2 Hz for MEDIUM and 0.4 Hz for SIMPLE.
In comparison, the 1982 vintage QRS provides some 1.2 Hz at 104, so to follow our notation, it could be relabeled as "⅓ SIMPLE". This also explains our reference back to the 1982 original algorithm, which lives on, concerning basic structure and sound character, but delivers drastically improved self-resonant density (and also many new parameters).
Are there any minimum or guideline figures in the literature?
For Schröder and Kutruff, the early pioneers of digital reverb (University of Goettingen, c. 1960), a target value of 1 Hz has been constituted. For the vintage QRS, this corresponds to a room size setting of 1E4, for example a concert hall (10000m3).
In comparison, the current SIMPLE room model reaches the target natural frequency density of 1 Hz already in a large classroom (500m3), while our flagship COMPLEX reaches the 1 Hz target already in a broom closet (8m3). There will be few real spaces that can shine with such values ...
Within this context I'd like to note that the sizing of the room models cannot exclusively be derived from simple scaling. Just one of many additional effects to add complexity is the influence of the respective diffusors on room acoustics. Just take a simple "room design component" such as a bathtub, and project this diffusion object in correct scale from your bathroom to a cathedral; you know what I'm talking about ...
Did those tricks from the echo chamber tool box contribute to the drastically reduced coloration of the new Yardsticks? Contrary to the vintage QRS, even small rooms sound aesthetically clean, i.e. do not show any metallic discoloration.
It's the same!
Jun 2011, updated Aug 2011
If you want to learn about the basic ideas of SIMPLE, MEDIUM, and COMPLX plug-ins, we'd recommend reading FAQ 012 ...
User A: I'm trying to integrate a just unwrapped Yardstick with my LAN. Unfortunately, nothing works at all. What next?
User B: I've substituted my previous PC against a more powerful unit, which now runs under <my new operating system>. Within my modified LAN setup and with my new <browser type, high version number>, I can't access my Yardstick any more. Do I need to configure anything special for <browser type, high version number>?
A Yardstick is totally independent of browser type, and of course independent of PC, Mac, Linux, or whatever. To clarify: a Yardstick does not install any software on your computer - everything takes place within a standard web browser, just like surfing the Internet. Actually, any internet-enabled browser, out of the box, should successfully interact with the Yardstick's built-in web server ...
Please work through the following checklist to pinpoint your problem:
Can you reach other web sites from your browser? If the HTTP protocol works properly on other web sites, then at least the routing within your LAN should be ok. Moreover, a problem with your PC's internal firewall would be unlikely then.
Verify that the green LED near the Ethernet connector turns on when plugging an Ethernet cable into your Yardstick. Does the device on the other end of the cable show a similar behavior? Then your patch cable should be working properly.
If not: is there a point-to-point link between PC and Yardstick? Which means no LAN with star connections to an Ethernet switch? In this case, you don't need a standard, but a crossover patch cable between PC and Yardstick!
Double check that, with the IP number used, you're really addressing your Yardstick, and not some completely unrelated device. Maybe you're running into a void with that IP number used ...
Can you ping your Yardstick from your PC using that IP number? Each ping packet triggers the ETH LED on the front panel, so it would flash once per second.
Confirm that a few ping responses temporarily fail to appear while turning your Yardstick off and on again? This would be a final confirmation that, within the scope of the network, everything plays together flawlessly.
If nothing out of this works at all, then go to the Yardstick setup, following SystemSetup/Ethernet/ShowEthernet, and look for its current IP. You need to locate an IP number which matches your local LAN, either configured manually, or automatically from your DHCP server. If you read all zeros, the Yardstick might be waiting for an allocation from the DHCP server; which fails to appear for some reason. Alternatively, you may configure the IP address manually. Please verify that a manually configured IP address does not conflict with your DHCP server's IP pool.
A problem that occasionally occurs with older switches is incorrect rate negotiation. The Ethernet interface on the Yardstick is 10BASE-T, not 100BASE-T. Solution: insert another Ethernet switch in between, which may better cope with erroneous rate negotiation algorithms. Or even replace your main switch with a newer model, or a different brand. Quite often, inserting an old Ethernet hub from your vintage collection may help, too.
I don't have a LAN, it's just a Point-to-Point (»P2P«) link with an Ethernet patch cable.
For a standalone point-to-point link between PC and Yardstick, you could use arbitrary IP numbers, but still a few rules need to be accounted for:
You need to configure both ends manually, as there is no DHCP server to assist you.
You may ignore Gateway IPs altogether, as there is no way out either.
Use same Netmask on both ends, e.g. 255.255.0.0 (or even disable with 0.0.0.0).
Use two adjacent numbers for the IP addresses, e.g. 169.254.81.192 on your PC, and 169.254.81.193 on the Yardstick (in this case, enter 169.254.81.193 in your browser to access your Yardstick).
Remember this is just one example for a working point-to-point link (with no connection to the outside world, of course). There are millions of IP settings that may work the same way.
Sorry - still doesn't work.
If you could confirm everything up to now, but it still does not work, there might indeed be a problem related to your browser, or its configuration.
There might be an HTTP proxy pre-configured in your web browser that possibly cannot reach your internal LAN. A Yardstick is basically a web server like any other, only that it's not located in the depths of the Internet, but stands right next to you. Solution: by entering an exception rule for your local LAN's IP space, you need to create a bypass around the proxy.
If you need telephone or email assistance, first of all please clearly state, if it's a LAN or a P2P-Link.
How do I specify at which sampling rate the 2493 AD/DA converters are going to process an analog input signal?
There's no way for manual intervention. The AD/DA converters always operate at the maximum sampling rate a currently active plug-in can still cope with. For example, if the active plug-in is of type SIMPLE, the AD/DA converters are automatically configured for 192 kHz.
To my ears, 192 kHz sounds horrible. What can I do for a SIMPLE plug-in to be forced to 96 kHz?
You cannot do anything! - Steer clear of the SIMPLE plug-in category, and switch to the MEDIUM category instead.
What about my creative freedom as a sound engineer?
Whenever a MEDIUM category plug-in forces the AD/DA converters to 96 kHz, the user will benefit from the double-sized DSP window per sample, which means that the algorithm complexity can be doubled. Ultimately this will increase sound quality considerably. In the case of a QRS plug-in of type MEDIUM, both room mode and reflection density will be doubled; with the result that a MEDIUM reverb tail sounds smoother and with even less metallic coloration than a SIMPLE reverb tail. Still another doubling of quality comes with COMPLEX. A lot more about room mode density, in particular compared to real spaces, can be found in FAQ 069.
From what I understood, I may control the AD/DA sampling rate by selecting the appropriate plug-in category ...
Exactly! In the context of COMPLEX type plug-ins, which accordingly force the AD/DAs to 48 kHz, there's surprizingly at least one single option within the analog-specific setup menu:
Please enter System Setup -> Analog -> Anti-Aliasing Filter
You may select whether the antialias filters (cascade of both AD and DA filters) adhere to the industry standard of attenuating 12 dB at 24 kHz, or whether a special QUANTEC optimization is active, which attenuates an effective 60 dB at 24 kHz. For more details, please consult FAQ 058.
Aug 2011, updated Mar 2013
How are RTC, time server, time zone, and DST interrelated?
The 249x timebase is its built-in, battery-backed RTC (»Real Time Clock«). It is advantageously set to local winter time.
Because each clock has a certain inaccuracy, clocks need to be put forward or backward from time to time. This correction can be automated by referencing a time server. Those public time servers are available worldwide, so they do not administer local times, but a time that's valid all over the globe: UTC (»Universal Time Coordinated«).
To recover local time from UTC, the decoding routine needs to be aware of which time zone you are in. From this information, the 249x RTC generates local winter time. Final step is consulting the calendar, whether summer (»Daylight Saving Time«) or winter time (»Standard Time«) needs to be presented. For summer, the value stored in the DST offset is being accounted in the result.
By the way: it is considered impolite if all your PCs, servers, and Yardsticks retrieve their times directly from the public time server. Please consider configuring a local submaster on one of your 24/7 servers, which can be accessed from any piece of equipment within your facility.