Schematic of 0.5 msec Analog Delay for Through-Zero Flanging

[stm]

Hi all!

I settled to get 0.5 msec delay with the maximum possible bandwidth and using at most two quad opamps, intended for through-zero flanging without using a second BBD's and its associated frequency limitations, restricted dynamic range and noise problems. The circuit shown below is the result:



The circuit is an analog time-continuous allpass network; it generates a 0.5 msec +/- 1% time delay from 0 to 6 kHz; frequencies above 6 kHz suffer no attenuation, however the delay falls down as frequency increases. With the values shown gain remains constant within +/- 0.2 dB from 0 to 20 kHz. The following curves illustrate this:



Component tolerance is critical , no only for the flat delay response, but to maintain the allpass or 0 dB attenuation characteristic across the audio band.  1% metalfilm resistors and 2 to 5% silvered mica capacitors must be used (hence the small capacitor values chosen).  Also, as a consequence of the op-amp saving topology used to implement the 2nd order all pass sections, every fourth stage in the circuit provides 26 dB of gain in order to maintain overall gain at 0 dB, so low noise opamps are recommended to avoid noise and hiss.

I think this circuit is "advanced" in terms of building complexity due to the large number of components and exact value dependence for proper operation. If there is enough interest, I would sell assembled and fully tested PCBs, so they can be used as building blocks and add-on boards for existing flanger designs. Just let me know.

I am posting this circuit for personal DIY use only. It is OK if you want to host it, as long as you credit me and retain all the copyright information.

As usual, I eagerly await the comments from our DIY community!  

[Vsat]

Wow.... very nice, stm!! Will you be building this soon?
Regards, Mike

[stm]

Mike,

My current priorities are:

1) 3-band universal tone stack (circuit already protoboarded & tested, just need to build it seriously, i.e. in a painted box with knobs and stompswitch!)

2) Dual MN3005 delay (status same as above)

3) MN3009 flanger/chorus/vibrato (haven't started with the prototyping yet).  Here I will try the analog delay line, perhaps with a SPDT switch to tuck it in the dry signal path!

If there is demand for kits on this I would consider devoting to the analog delay first.  For now I want to finish what I've started waayy looong agooo!  :roll:  :oops:  :cry:

[Chico]

STM

Very cool indeed.  In all honesty, it will take me a while to digest this circuit and all the other ideas that were discussed in the previous thread on this subject, but I am plowing through it.

I have just started a new flanger project myself and plan on testing out this concept in the near future.  My design is still very much a work in progress, so it may be a while, but I will let you know how it turns out.

Best regards and thanks for the build tips.

Tom

[stm]

Chico, good to see this stuff *could* be useful.

I've spent so many hours on this design, and really hope it is useful to somebody!

By the way, I found in digikey some 2% metallized poliester caps, which are cheaper and more compact than silvered mica.  In this case you can multiply by 10 the cap values (to use something more normal) and divide "frequency determining resistors" by 10.

Note that second order sections would require to change only the two upper resistors (105k and 309k), because the others on the + input just form a voltage divider and can be left as they are. Also, in the fourth opamp of each row, there is need just to scale the 215k resistor if the correspondig cap is changed.

Regards  

[puretube]

GREAT!

1 remark allowed: instead of buying expensive caps,
one could go out and buy a couple of handfulls of cheap ones
and measure them out: this can lead to <1% accuracy...

(look for capacity-meters/comparators elsewhere)

[Mark Hammer]

I salute your diligence and dedication, sir.

Mike/Vsat and I were discussing time-produced vs allpass produced "stagger" yesterday, and one of the things that occurred to me is that with time-produced stagger (i.e., the so-called "dry" signal is staggered/delayed by a small fixed amount so that the swept-delay signal has the opportunity to arrive at the mixing stage *before* the staggered signal once in a while), the amount of equivalent "phase delay" varies across the spectrum.

I'll try and express it more clearly.  If I delay the fixed signal by 1msec, the cancellation that results when it is combined with a swept signal does not occur instantaneously at all frequencies.  If I have a 10khz signal as part of the swept path, quite a few cycles of that signal will actually reach the mixer stage BEFORE the 1msec-delayed fixed signal gets there.  Naturally, the number of cycles that pass prior to any cancellation occurring (remember, it has to be the same waveform in anti-phase versions at the mixing junction for any cancellation to occur) will be fewer for lower frequencies, and more for higher frequencies.

In a sense, you not only have variation in the distribution of notches with flanging, but differential ONSET of cancellation because of the time differences.  Not having done the proper experiments, I obviously can not force the issue and say phase-delay will absolutely NOT yield the same sort of effect as true sample/time-based stagger, but the fact that musical effects are used with constantly varying, rather than steady state input signals means one has to set aside all those assumptions derived from textbooks and steady state signals, and consider real-world signals.

My gut sense is that true time-based stagger will produce cancellations in a different manner than phase-delay produced stagger, because of such differences in cancellation onset.  Again, that may just be a different feeling TZF and not necessarily a worse, better, or un-TZF-ey TZF.

A reasonable inference?  You tell me.

[stm]

Mark,

I'm pretty sure a time delay is a time delay independent of how you produced it.  In this respect, a signal through a BBD, or written in tape and read some time after, or an all-pass filter are equivalent, but IF and only IF the delay is the same across all the bandwidth of interest.

In this respect, BBD's and tape delays usually have LIMITED BANDWIDTH (above a certain point amplitude drops off), whereas an allpass filter has LIMITED DELAY from a certain frequency and above.

Having said that, the key for the different methods to be comparable is that they have similar or better bandwidths for flat amplitude and flat delay WITHIN THE AUDIO BAND OF INTEREST.

This poses the question of which is your audio band of interest.  In this respect, the circuit I proposed has flat delay up to 6 kHz, which I believe is reasonable for through zero flanging.  A typical flanger using BBD's may have frequency limitation around this point.  I agree tape flanging has better bandwidth, but also typical music has wider bandwidth than a guitar.

** I think the only way to tell is trying the actual thing **

By the way, the delay I propose can be reduced to 0.4 msec to increase flat delay bandwidth to 7.5 kHz if necessary.

Finally, regarding the fact that actual through zero flanging is not homogeneous in terms of which frequencies cancel first, I agree, since always one of the delayed signals has some slightly varying delay, that introduces slight pitch shifting.  This is independent of how you created the delay (IMO), and believe is integral part of the TZF effect.

Hope this makes sense.

Take care.

[Vsat]

stm,
Got the simulation running (finally... a resistor that had one end unconnected) and it works just as you say....NICE NICE NICE!
Regards, Mike

[stm]

Mike, good!

I still have one last card under my sleeve...

As I told you, each row of four opamps form one 4th order allpass + one 3rd order allpass with gain.  I can also use the very same topology (only with different component values) to obtain a single 7th order allpass stage.  The advantage is that the higher order would allow better bandwith (perhaps 8 kHz) at 0.5 msec.  The disadvantage is that component tolerance becomes more critical.

I have yet to cook the values for the 7th order to evaluate feasibility.  In other words, a working circuit in the simulator with perfect and stable values doesn't guarantee a real world working circuit.  So far I managed to obtain the poles of the transfer function. Converting them into the circuit is currently giving me a bad time, but I won't give up yet!

Regards,

STM

[MR COFFEE]

Folks,
When we are trying to design LOW NOISE circuitry, the impedance needs to go down, or the the thermal noise of the resistors gets every bit as intense as BBD noise.

As puretube noted, we can forget BUYING 1-2% capacitors, if we simply SELECT our caps from a pool (pile) to get ones that match accurately. And even with different values involved, we can MEASURE our results (of an actually built unit) and tweak it into tolerance - a luxury unavailable on the production line, but quite available to the DIYer.

For all the op amps involved, may I be so bold as to suggest scaling values by at least 20 times downward to get the NOISE out  ...I really, really, REALLY have this thing about NOISE!!! ...UGHHHHH....
uh, sorry about that, folks. I must need to take my medication again... ;-)

BTW, many tweaked circuits for such purposes have an alignment proceedure and trimpots to compensate for overall device tolerances, i.e., adjust pot1 until phase delay at X hz. = 90 degrees, using a Lissajous (oh hell, I know that ain't spelled right) figure on a 'scope.

This is SUCH a cool thread. You folks are AWESOME..

[stm]

Well, as Duke Nukem used to say "Damn, I'm good"  

( :oops: sorry for that, just couldn't help it.)

I DID IT!

I really improved the above circuit just changing component values, so now it is implemented as two seventh order 0.05º linear phase allpass filters.  In plain english these means the following improvements to the above design:

1) I get 0.5 msec delay with over 8 kHz flat delay bandwidth

2) Using 1% resistors and 2% caps, a montecarlo analysis shows amplitude variation remains within +/- 1dB and delay stays within +/- 2%

3) I reduced capacitor value usage to just two different values (1nF and 10 nF) so getting 2% caps is easier

4) I reduced attenuation introduced by the second order stages from 52 dB (for the complete filter) to less than 12 dB, so there should be no noise problems now.

5) A single row (4 amps) can produce a respectable 0.4 msec delay with 5 kHz bandwidth, a great improvement over my very first delay circuit posted that achieved 200 usec with 5 kHz bandwidth and 4 opamps.

When I have time I will arrange the schematic and characteristic curves for posting.

Take care!

[Mark Hammer]

Damn, you ARE good!  (And is it me or do many schematics start to FEEL like side-scrolling games?    )

As an aside, the benchmark through-zero sounds that many people are familiar with (and make them think "Geez, I wish I could do that") tend to be fairly broadband and cover a huge chunk of the audible spectrum.  No big surprise since tape-hub flanging was traditionally applied to multiple tracks during post-production (think "Axis, Bold as Love").  In discussion the other day, Mike/Vsat concurred that any attempts he's done at TZF (or just about any of his experiments in modulated comb filtering, whether mega-phasers or flangers) sound fantastic with white noise, pretty darn good with "music" (i.e., multiple sources mixed down to mono), and not that much more inspiring than regular stuff when applied to single instrument sources.  Of course, when you play a distorted power chord into an overdriven amp, the distortion introduced delivers up a lot of the bandwidth coverage needed to exploit these "electronic comb-overs", and you start to move into the zone of what will sound best with TZF.

There is a kind of balance between signal and device bandwidth to be considered for pleasing TZF.  I'm not going to speculate on what the required bandwidth is or will be, but a wide-bandwidth device fed a limited bandwidth (or rather narrow bandwidth) signal is unlikely to sound jawdropping.  Conversely, a full-spectrum coverage signal (let's go crazy and say white noise) fed into a limited bandwidth device will also make you go "That's nice, but less than I was expecting".  What one needs is a broad-spectrum signal fed to a wide bandwidth device.  Again, what that spectrum and bandwidth needs to be at a minimum is certainly outside the scope of my knowledge or experience.

Once again, I salute your dogged persistence in pursuing the ideal, and your skill in doing so.  Steve Giles, if he is reading this thread, is probably thinking "Yes, yes, dear the Andes are very nice.  Can we go home now?  I have some, er "stuff" that needs doing." :wink:

[puretube]

this co-incides with the "That Lady" phenomenon:
guitar thru "TZF" depends in its "felt intensity"
largely on the harmonic content at the input of the box.

No wonder most soundclips on the website of
"a currently marketed" TZ-flanger manufacturer
are using distorted guitar sounds...

[Vsat]

stm,
Attenuation  now only 12 dB! - that puts it way ahead of the previous design...great work.

Mark, puretube - other flangers/phasers have regen as an additional parameter for "strengthening" the sound. TZF doesn't have this, so has to rely on the quality of it's notches (and bandwidth). At least a couple of people have suggested to me that internal noise in a phaser/flanger may not be entirely bad.... it may even be essential in the case of tape flanging... it can help the ear to "delineate the moving comb" with otherwise spectrally-sparse input material.
Regards, Mike

[Peter Snowberg]

FANTASTIC work STM!!!!

12dB.... Nice!

[StephenGiles]

Hi Folks - just arrived home in England after a 9 hour flight from Bonaire to Amsterdam, 1 hour to get through security!!! then straight on to our connecting flight back to London Heathrow, just 40 minutes. Despite a short 2 hour sleep after unpacking, all of which has absolutely nothing to do with thru zero flanging, I'm fighting to keep my eyes open so not really taking much of all this in yet, but fantastic work stm. Incidentally, I always hide a few important things when we go away for our annual holiday and having found my wallet, my LAN connector for my laptop and a few others, I just couldn't remember where I put my Stripboard ADA Flanger - bloody important -  a chap can't go on without that! I put it somewhere safe really to keep it out of the way of our cleaning lady in case she goes berserk with the vacuum cleaner!
Stephen

[Ed Rembold]

Very interesting thread!

STM,
I'd like to learn more about the filter topology you used,
Does it have a name?
or design program?

Thanks Ed R.
(Mike I. and Sean C.  nice to see you guys here-  been a long time)

[Vsat]

Hi Ed,
Nice to see your here!
Regards, Mike

[StephenGiles]

stm - any low noise FET opamp? Someone's bound to ask this - what would happen if the 10k resistances are modulated by a LFO?
Stephen