Some thoughts on and analog and digital flangers

[12Bass]

Some months ago I decided to build the A/DA flanger clone using an old SAD1024A which I had lying around.  However, I've taken some liberties with the design, particularly with regard to the low-pass filtering (LPF) before and after the BBD, plus using high-quality op amps and capacitors in the signal path.  Basically, what I've tried to accomplish is to open up the frequency response as much as possible, using a high-quality audio path, while still providing some slight filtering to minimize unwanted aliasing distortion.  The end result is an analog flanger which has a surprisingly open, wide-bandwidth, sound.

So, this led me to thinking about how musicians often talk about how a certain pedal sounds "harsh and digital" or "warm and analog" (e.g. the Boss BF-3 vs. the older BF-2).  For BBD-based chorus and flanger circuits anyway, I'm now thinking that a lot of what people are hearing as "analog warmth" is in fact the rolled-off highs of the LPF for the BBD, and has very little to do with the BBD or the design of the rest of the circuit.  In fact, with the relaxed LPF, and optimal bias, the SAD1024 sounds remarkably "hi-fi" when fed a full-range signal.  As is, once the sample rate is up around 100 kHz or so, the delay path doesn't sound a whole lot different than the bypassed signal, just a tad darker.

Listening to clips of the digital BF-3, what I hear sounds like a flanger which sweeps higher and doesn't have much, if any, low-pass filtering, which allows it to retain wider bandwidth, compared to its older analog BF-2 cousin.  But I'm willing to venture that the old "analog" sound could probably be emulated without much trouble by merely limiting the sweep range and bandwidth.  So, what I'm thinking is that digital technology removes many of the limitations of older analog designs, and it seem that those limitations are what may have been responsible for a lot of what people perceive as the "analog" sound.  Yet, in the end, analog can sound remarkably clear, while digital can sound warm, if the circuits are designed to achieve those goals.

[Mark Hammer]

Look at the control panel here: http://line6.com/tonecore/liquaflange.html  In fact, digital flangers already come with "analog" settings, intended to capture the filtering, and perhaps the signal degradation.

Don't forget that any BBD-based effects is already a "sample-based" circuit.  True, as an analog sample, what a BBD carries from input to output essentially has infinite resolution.  But once you hit 24-bit sample resolution, and 96khz sampling rates, any differences between digital and analog start to melt away when it comes to a sample-based circuit.

Although, now that I think about it, consider that the sampling rate of a time-modulated effect, like a flanger, ranges between both higher (e.g., well over 500khz for an A/DA flanger), and much lower sampling rates (as the circuit approaches the longest delay time within the range), than a 24/96k system does.  As it approaches the shortest delay attainable, one has both infinite resolution of the analog sample AND sampling rates much higher than a comparable digital circuit.  As it sweeps to the longest possible delay time, though, that infinite resolution turns into infinite-plus-signal-degradation-from-leakage, and lower sampling rates.

In effect, if you want to truly mimic analog flangers, you have to vary the signal quality from worse to better.  I wonder if they do that when you flick the button to "analog" mode.

[12Bass]

Look at the control panel here: http://line6.com/tonecore/liquaflange.html  In fact, digital flangers already come with "analog" settings, intended to capture the filtering, and perhaps the signal degradation.

Don't forget that any BBD-based effects is already a "sample-based" circuit.  True, as an analog sample, what a BBD carries from input to output essentially has infinite resolution.  But once you hit 24-bit sample resolution, and 96khz sampling rates, any differences between digital and analog start to melt away when it comes to a sample-based circuit.

Although, now that I think about it, consider that the sampling rate of a time-modulated effect, like a flanger, ranges between both higher (e.g., well over 500khz for an A/DA flanger), and much lower sampling rates (as the circuit approaches the longest delay time within the range), than a 24/96k system does.  As it approaches the shortest delay attainable, one has both infinite resolution of the analog sample AND sampling rates much higher than a comparable digital circuit.  As it sweeps to the longest possible delay time, though, that infinite resolution turns into infinite-plus-signal-degradation-from-leakage, and lower sampling rates.

In effect, if you want to truly mimic analog flangers, you have to vary the signal quality from worse to better.  I wonder if they do that when you flick the button to "analog" mode.

Ha.... figured that this was a thread tailor-made for Mark Hammer.   

The Strymon Orbit supposedly uses dBucket BBD emulation in an attempt to capture the nuances of analog flangers.  Not sure if they included a variable sample rate emulation, though the Brigadier does include a "bucket loss" control which is supposed to emulate the signal degradation which results from long BBD delay times as the sample rates are pushed to the lowest extreme.  With the range of delay times used in a flanger (very short), I'm not sure how much "bucket loss" is likely to occur in this application.  My guess is that most of the sound of analog flangers is related to the LPF, with relatively little contribution from sampling rate and BBD artifacts. 

White paper on Strymon's dBucket: http://www.strymon.net/wp-content/uploads/dBucket_whitepaper.pdf

With my flanger, I can hear what sounds like an increased graininess and a slight loss of detail as the sample rate goes down.  However, given the parallel-multiplex design, even at the lowest delay time, the sample rate is still 70 kHz (35 kHz clock x 2).  Another issue is that the gain in the delay path decreases by about 2 dB at the shortest delay setting relative to the longest delay time.  I've added a trimmer at the output of the SAD1024 to set the balance between dry and delayed signals at 50:50, but it is impossible to keep that mix ratio throughout the sweep due to this gain decrease.  This contributes to a bit of a "hole" as the sweep nears the top of the range.  Optimal BBD bias also varies somewhat, so that probably contributes varying amounts of distortion as well at different points in the sweep.  Still, after all of the modifications, I was somewhat surprised to hear just how subjectively clean the delay path sounds on its own.  If I had to guess, I would think that many would not think that my A/DA is analog.       

[Mark Hammer]

First off, thanks for the link.  The more I hear from this Strymon company, the more impressed I am with them.  These folks think, and think deep.

Second, like I keep saying, the first step towards progress and innovation in audio technology involves describing what you hear (and are not hearing).  Once described, you can then proceed towards explaining, and from there it isn't much farther to predicting, controlling, and implementing.

Third, when fielding single-source signals (that aren't white noise), ALL flangers will seem to crap out.  One thing to set the bandwidth of the circuit wider, and quite another to feed it with a signal that exploits that bandwidth broadly enough.  The best and most archetypal flanger sounds that inspire people are always multi-source mixes that cover much of the spectrum.  After all, you gotta have something to notch for the notches to be audible.

Fourth, leakage from those teeny-tiny caps in BBDs is going to be a function of how long they are obliged to retain their charge.  That, in turn, will depend on what you have to do to make the BBD produce the delay range aimed for.  So, If I try to produce the same delay range with a 512-stage, 1024-stage, and 4096-stage device, the clock rate will need to be slower for the smaller capacity device in order to reach the same destination at the same time.  In that sense, when someone attempts to mimic a BBD digitally, the question needs to be asked about just what sort of time-delay-to-leakage-risk relationship are you talking about.  That's not to pooh-pooh what Strymon has done, though.  I thoroughly doubt whether any musician would find it a useful feature to be able to select between SAD512D and MN3008 emulation on their digital flanger.  Basically, you pick a model of BBD emulation and you go with it.  Still, it bears noting that a "suitably buffered and clocked" MN3005 (suitable for flanging, assuming nothing fries) ought not to have any of the leakage issues of a smaller-capacity chip.  At least nothing of the sort where there would be audibly more leakage at the longest delay than at the shortest.

Finally, I guess it is worth considering that aliasing at the longest delays (and slowest clock rates) is much nmore of an issue for lower-frequency content than for high frequency.  Is my 18khz harmonic from that cymbal "stairstepped"?  Beats the hell out of me, but boy that bass sure sounds grittier.  So, digital emulation of analog flanging also needs to be mindful of how clock rate, aliasing, and frequency, interact.