ADA flanger: Why mix & output before removing the clock noise?

[flo]

In: http://moosapotamus.net/IDEAS/ADAflanger/ADAflangerSCH.gif
The delayed signal of the mixer is taken from IC1c before the clock frequency is filtered out by IC1d.
The delayed signal is then mixed with the clean signal for the flange effect and then send to the output by IC2b.

Why is it done like this?
To me it makes more sense to first remove the clock and then mix & output it.

--------------------------------------------------
For more info on the ADA flanger see also:
http://www.adadepot.com/adagear/gearpages/stompboxes/ADA-Flanger.htm

[StephenGiles]

I'm afraid the answer is that because it is. The minimal filtering is done in the feedback section, which is probably why it sounds so good.

[flo]

... probably why it sounds so good.

Could very well be the case.
The EH Electric Mistress also has very poor clock filtering and I like it's sound.
Boss BF2, Pearl FG01 and Ibanez 301 on the other hand first filter and then mix & output & feedback. The Boss BF2 and Pearl FG01 I have but I don't like them so much. I don't know how the Ibanez 301 sounds.

[George Giblet]

The Boss CH-1 Chorus uses very high filter frequencies and a lot of people hate it - some even speculated that is was an evil digital effect (it's not).

[Mark Hammer]

1) Where analog delays aim for longest feasible delay time from the fewest BBD chips, flangers aim for the shortest possible delay, which means that the clock frequencies used are well out of the range of what is audible.  Indeed, less lowpass filtering is required for flangers than for analog delays for that very reason.

2) Flangers sound "best" with widest bandwidth.

3) The A/DA flanger is best known for having one of the widest sweeps, particularly at the "high" end, where the clocking is very fast.  Although certainly guitar amp speakers would work at cross-purposes, the removal of lowpass filtering would make it easier to hear any notches produced at those highest points in the sweep.  Or rather, more lowpass filtering would make that extra sweep rather moot.

4) There is the speed with which the signal is clocked through the BBD, and there is the extent to which the signal is recirculated through the BBD.  Those little storage cells are at their best behaviour when they don't have to hold onto an analog sample for very long (i.e., at faster clock rates), but they aren't perfect, and every cycle through the BBD creates a little more imperfection.  Consequently, while the delay output is minimally filtered, the recycled signal is lowpass filtered before being sent back to the start of the line.

[flo]

Thanks! Now I understand the design decision that was made.

[12Bass]

Sorry to bump an old thread....

Mark's comments above have been helpful, and mirrored my experience playing with the regeneration circuit in my A/DA flanger clone.  What I found was that changing the low pass filter feeding back into the BBD significantly altered the personality of the flanger.  Although a more relaxed filter (smaller capacitors) opened up the regeneration on the high harmonics, it also seemed to lose much of the A/DA flanger's trademark "swoosh".  This makes me think that the regeneration circuit may be a prime contributor to the characteristic sound of different flangers.  My speculation is that the lack of low pass filtering might also explain why newer digital flangers like the Boss BF-3 might have a more "digital" sound... which is actually not a digital artifact so much as little to no sculpting of the regeneration path which leads to a more "pingy" tone.  Other comments regarding a steeper low pass filter for making the flanger less metallic may lead to further experimentation....

Speaking of bandwidth... if my hunch is correct, the SAD1024A A/DA clone has two 0.01 μF capacitors between the input and ground (C10, C34), for a total of 0.02 μF, and making for an ~ 8 kHz low pass filter going into the BBD.  I've removed one of these, increasing bandwidth to 16 kHz, and have gained a perceived increase in fidelity.  Given that the sampling rate used doesn't go below 34 kHz, I'm not sure why the bandwidth of the circuit has been so restricted.  It seems to me that the ultra-short delay times, and resultingly high notches, that this circuit is capable of are somewhat negated by choosing such a low cutoff for the high pass filter.  Earlier, I noticed that the MN3007 version only uses one 0.01 μF on the input, which may help to explain why some have reported that it has higher fidelity.

[Mark Hammer]

Glad my ramblings could be of use.

Not sure that ANY flanger could be accused of "fidelity".   So, I gather you mean a discernibly crisper sound due to wider bandwidth.

When one questions the design philosophy behind any vintage pedal, it is always wise to consider the historical context.  This is a pedal that was produced when:

a) high-quality cable was not widely used;

b) the absolute best one could hope for in a readily-available op-amp was probably a TL072 or maybe an NE5532, and those cost noticeably more than others;

c) the price differential for most lower-noise components (e.g., metal film resistors) was substantial;

d) players either used PAF-style humbuckers which HAD no top end to lose in filtering, or used noisy single-coils;

e) there wasn't a universe of competitors (especially cheap stuff from China and Korea), so you sort of had the freedom to design what you wanted, rather than worry about having as much or as little of this or that as your closest competitor.

[12Bass]

Sure, I'll agree that anything based off of BBDs is not likely to meet the definition of a 'high fidelity device'.  Still, I was somewhat surprised when I listened (using my K702s) to the soloed BBD line on my A/DA clone.  I've used an OPA1642 (input and regeneration) and a pair of OPA2211As for the audio path, and the sound of the BBD output is admittedly a bit distorted, but clearer than I expected, especially after I relaxed the filtering.  So, perhaps it is a matter of semantics; however, I think that the flanger sounds clearer than it did originally (using LM324s, stock capacitors).

BTW, over the past few days I've been reading a bunch of your old posts and found that they have answered a number of my BBD-related questions.  This forum is a great resource!

[StephenGiles]

Sure, I'll agree that anything based off of BBDs is not likely to meet the definition of a 'high fidelity device'.  Still, I was somewhat surprised when I listened (using my K702s) to the soloed BBD line on my A/DA clone.  I've used an OPA1642 (input and regeneration) and a pair of OPA2211As for the audio path, and the sound of the BBD output is admittedly a bit distorted, but clearer than I expected, especially after I relaxed the filtering.  So, perhaps it is a matter of semantics; however, I think that the flanger sounds clearer than it did originally (using LM324s, stock capacitors).

BTW, over the past few days I've been reading a bunch of your old posts and found that they have answered a number of my BBD-related questions.  This forum is a great resource!

Surely using LM324 in the audio path is like using 4 x 741. I don't have the ADA circuit with me but I thought they used MC3304s, about which I remember Mike Irwin telling me something as to why.

[12Bass]

Surely using LM324 in the audio path is like using 4 x 741. I don't have the ADA circuit with me but I thought they used MC3304s, about which I remember Mike Irwin telling me something as to why.

Hey Stephen,

Think the chips were MC3403.  

I started out with the LM324s because they were in the moosapotamus A/DA clone BOM list, and because I wanted to be sure the flanger was working before making modifications to the circuit.  Although they use more current, the new TI op amps are leagues better, IMO.

From reading through past posts, I seem to recall you saying something about how you found that the A/DA would get lost in the mix when it swept through the upper octaves.  My hunch is that it may have been partly to do with the low pass filter on the BBD, and partly due to the extremely short delays that the A/DA can produce which put the frequencies affected up past 10 kHz where there's relatively little spectral content.  

FWIW, I did some more experimentation with the low pass filter on the input of the SAD1024A (the two 0.01 μF caps connected to ground).  First I tried removing one, then both, to see what would happen to the sound coming from the BBD line.  In doing so, I was able to confirm that the Reticon chip is quite capable of reproducing high frequencies, provided they are not filtered out before they get there.  For my taste, I'd rather put a bit less low pass filtering on the input, and a little more on the output (if necessary).  As it is, I'm still not hearing any objectionable noise or aliasing, plus the flanger has a clearer, more extended, sweep into the upper ranges, so it doesn't get lost up high like it did before.        

[Mark Hammer]

Of course, with flangers, the clock rate needed to achieve a given minimum delay not only represents what must be contended with (or not) as far as keeping the clock aliasing inaudible.  It also represents the sampling rate and corresponding bandwidth possible on that basis.  The same 1024 stages clocked more slowly for a chorus not only require lowpass filtering targetted lower in the spectrum to keep that clock whine inaudible, but force the user/designer to aim for lower bandwidth even in the absence of concern about clock whine simply because sampling at a lower rate only provides faithful rendering of an input signal up to a certain maximum frequency.

Flipping this around, flangers can aim for greater bandwidth at the high end if they also aim for faster sampling.

Yes, the chips were MC3403s and the same chips were also used by Morley/Tel-Ray for their flangers too.

[StephenGiles]

Yes you're right, various 3403s in the circuits I found, but I don't remember what I used - possibly MC3403. There's a photo floating around somewhere.

[12Bass]

If I'm reading correctly, when using parallel-multiplex mode, with a minimum clock of 34.8 kHz, the sampling rate (fs) is effectively 69.6 kHz, and the Nyquist frequency is equal to f clock, or 34.8 kHz.  Thus, the low pass filter going into the BBD should be a minimum of 34.8 kHz (fs/2), or 23.2 kHz if we are going for a more conservative fs/3.  Both figures put the input low pass filter above the audible range, so I'm not too surprised that I haven't heard any audible aliasing.  Hmmm... does it matter, with regard to aliasing, that the sampling rate in each half of the BBD is only 34.8 kHz?  What I've surmised from this is that this particular flanger design doesn't require anything more than a 23 kHz low pass filter on the input.  Now I'm left wondering why the specified low pass is so much lower?

Any suggestions on capacitors for making the input low pass 23 kHz?          

Although I suspect that many guitarists might not find this information too useful, perhaps this might be helpful someone who is interested in a flanger with more extended bandwidth.  As a bassist with Q-tuner and Lane Poor pickups and a fairly high fidelity rig, I appreciate hearing a broader frequency response from my signal processors.            

[StephenGiles]

Of course, with flangers, the clock rate needed to achieve a given minimum delay not only represents what must be contended with (or not) as far as keeping the clock aliasing inaudible.  It also represents the sampling rate and corresponding bandwidth possible on that basis.  The same 1024 stages clocked more slowly for a chorus not only require lowpass filtering targetted lower in the spectrum to keep that clock whine inaudible, but force the user/designer to aim for lower bandwidth even in the absence of concern about clock whine simply because sampling at a lower rate only provides faithful rendering of an input signal up to a certain maximum frequency.

Flipping this around, flangers can aim for greater bandwidth at the high end if they also aim for faster sampling.

Yes, the chips were MC3403s and the same chips were also used by Morley/Tel-Ray for their flangers too.

Same designer?

[TELEFUNKON]

Same designer?

Cloner?

[StephenGiles]

Ask David Tarnowski of ADA, I believe he designed the ADA Flangers.

[12Bass]

Just wanted to revise my earlier post where I said that the removal of the feedback filtering had a negative impact on the A/DA's trademark "swoosh".  From further experimentation, I've found that removing the regeneration low-pass filter actually increases the flanging effect, particularly at high frequencies/extremely short delay times.  It's the bee's knees if you're after that "high regeneration going into hyperspace" sort of sound.  Also, the clock noise is mostly canceled by the trimpot on the SAD1024A's output... and is out of audible range, so there's no great need for severe clock filtering in this design, IMO.  

FWIW, I think I prefer the sound with a 0.047 μF capacitor coupling the signal back into the BBD.  This raises the high-pass on the feedback circuit by an octave and, while it is somewhat thinner, has a clearer, more resonant "swoosh".  Probably better with bass this way, as it doesn't affect the fundamentals as much.  

 

[Mark Hammer]

If you think of it, classic tape-hub flangning had no noise-control circuitry involved at all, unless there was some sort of EQ-ing of the mixed tape signals.

The magic in flanging is always highly dependent on the opportunity to produce audible notches across the entire spectrum.  So yeah, no big surprise that lifting lowpass filtering (with whatever noise risk is entailed in doing that) would help.