Zero through flanging with a single BBD is here

[stm]

Puretube: here is the phase response of my circuit posted above.



As you can see max phase shift = 180º x 6th order = 1080º (attained at infinite frequency).  Flat delay range is up to 360º only, or 5 kHz,  where the phase v/s frequency maintains a linear relationship, as can be seen in the next plot, where I used a linear frequency scale and zommed into the audio region.



Regards,

STM

P.D. if anyone works with MicroCap-7 I can send you the file.

[Chico]

STM:

Thanks for all of your insight here.  I have printed this thread out and am going to study your circuit.

Puretube, I am going to track down those authors and do some research.

Best regards

Tom

[puretube]

somebody called patents sick stuff: well, those files are infectious...

 :wink:

[StephenGiles]

Greetings from Cuenca in Equador, we are sheltering from the rain at the moment taking it easy after a 7 hour drive from Baños yesterday on a road which frequently became unpaved, and visibility was down to almost zero, without any filter whatsover!!!!!!!!! We saw some Inca Indian ladies who have dressed in black ever since a battle with the Spanish 400-500 years ago.
Now all this stuff about BBD-less delay is too much to take when all my electronics kit is 15 hours flying time away. Mark/Mike could I please ask you to email the circuit above to spgstevegiles(at)yahoo.co.uk just in case the thread goes pear shaped before we get back next week. Many thanks.
I had my shoes cleaned by a shoe shine today for 30c - that`s US cents quite a bargain!

[StephenGiles]

Keep this on top!

[Vsat]

Hi Steve,
Saw your request. Sounds like you are seeing some neat places on your SA trip.

Quite a few op amp allpass stages will be required to approximate even a short time delay (say 2 to 5  milliseconds) over an audio bandwidth of say 50 Hz to 10 KHz. Did some rough calculations last year and you are looking at 50 to 100 op amps as a minimum - this could be done with 25 quad op amps.
Haven't tried this one!

As before, you can use far fewer stages and still get interesting results, just not the same as true time-delay TZF, but nice in it's own way.
Regards, Mike

[StephenGiles]

Thanks Mike, I think perhaps a second BBD is a better bet - just think of all the money I`ll save on stripboard!
Stephen

[stm]

I've also done my calculations on implementing a "long" delay based on all-pass stages and agree with the need to use something between 20 to 50 opamps to get delays in the order of the milliseconds.

With the circuit I posted above, you can make a 1 millisecond delay using 16 op-amps, or 4 quads.  It looks as a sixth order delay, but in fact it is two third order cascaded stages.  This makes a difference, because the higher the order, the more delay you get per stage.  In other words, there is a gain when using more stages, but as usual there is a cost to be paid:  component accuracy and tolerance becomes critical, which can limit usability if a large number of series stages is used.

As an example, my circuit above has a 0.5 dB amplitude difference within 0 to 10 kHz frequency.  If I use 4 circuits like the above to get 1 msec delay (i.e. four quad opamp IC's), there will be a maximum amplitude variatin of 2 dB within the band of interest.  Also, this assumes the resistor values have no tolerance, which is not real.

For long delays implemented like this, it would be necessary to use 1% tolerance resistors, and 5% capacitors or better.  If you know how to, just run a montecarlo simulation to such a circuit considering resistor and capacitor tolerance and you will see what I mean.

Finally, and if the readers are still not sleeping with the length of this post, I have still one idea to try to further reduce the number of stages:

Convert my 4 opamp circuit from dual 3rd order to single 7th order, by having three different 2nd order stages plus one single order stage with unity gain recovery.  I estimate this will allow to have a 0 to 5 kHz flat delay of near 330 usec with 4 opamps, so in the end only 12 opamps (3 quads) would be needed per msec, which looks manageable.  The tradeoff is 1% resistors and maybe 2% tolerance capacitors (mica have this tolerance, but are expensive).

Take into account that 1 msec should be more than enough for through zero flanging if your delay line (BBD) con go below this value.

Take care.

[Vsat]

Thanks for sharing these ideas stm!  A few comments to make about TZF and tape flanging in general. I  suspect two of the main reasons why tape flanging sounds as good as it does are 1) both channels (reel-reels) have more bandwidth than a typical flanger (good tape machine gets up to  15 to 17 KHz with fairly flat response) and 2) both channels are fairly well matched. I suspect TZF is closer to the ideal (max notch depth maintained across the entire audio spectrum) than a typical single delay flanger since a) the insertion loss of a BBD varies several dB over the spec'd range of clock rates and b) sinX/X droop will also cause the response to drop as signal freqs approach the Nyquist limit. This causes incomplete cancellation over some freq ranges when the direct signal is combined with the delayed signal. It would be possible  (although messy) to compenste for these effects (fortunately flangers usually sound good even if you ignore these effects, particularly with regen cranked up). With a dual-BBD TZF approach, these effects are still present, but since both channels don't vary much in clock rate, the channels remain more or less matched, with resulting good cancellation over a wide range of frequencies  (if you are using long BBD's and a high sample rate). To get a 40 dB notch depth across the audio spectrum, over the entire range of delay times requires that the freq response of the  delay be flat to a small fraction of a dB... something quite difficult to do. The flanger s I've looked at with a sweep generator fall far short of this, particularly when a wide range of delay times is involved - but they still sound good. It would be very neat to try your circuit stm! Also try  a voltage-controlled variable delay version of a similar network using OTA's to combine with the fixed op amp version of the same network - then you will have a BBD-less TZF (but the noise from the OTA's will be very bad if many stages are involved).
Regards, Mike

[stm]

Hi, here are my comments to last Mike's post:

1) I agree tape reels have better bandwidth (10-15 kHz) but on a guitar system, the guitar speakers have a vary steep high frequency rolloff above 5 kHz (4th order, or 24 dB/octave)

2) A signal 20 dB stronger than other, will likely mask the lower one (i.e. not be heard by a human ear).  Taking this into account, it is reasonable to think that maybe just 20-30 dB of notch depth are enough to obtain a deep flanging effect.

3) Reel tape reproduction can be very accurate in terms of phase if two analog recordings are cloned at the same time, which is good to obtain deep notches when mixed.  However, I'm pretty sure there will be at least one or two dB's in amplitude difference between both playbacks due to differences in tape media and the many calibration adjustments inside the tape unit. I haven't heard of any adjustment done to calibrate this, so I rather think the through-zero flanging sound we like is good despite there is some amplitude mismatch, which finally traduces in less deep notches.  In fact, a 1 dB amplitude difference will limit notch depth to 20 dB approx.

My point is that maybe we are looking too much at numbers and figures instead of just building the actual thing and listening to judge it.  Let your ears be the judge!

Regards,

STM

[Vsat]

stm,
Agree that with audio it's a good idea to listen to what your ears say!

I often use a white/pink noise source for testing phasers, flangers etc as it gives a very good idea of the coloration which is imposed on a sound. Recently, in work with a time-delay TZF have found that I really liked upping the bandwidth to beyond 10 KHz to get some extra snap to the sound - but with a guitar speaker this would be of no additional benefit. Another thing I preferred was to have a max time delay difference on the order of about 7 - 10 mS... this permits a deep growlish sound with closely-spaced notches, more so than which a delay of just a few mS. A max delay difference of 100 mS really doesn't add that much to the  effect, and adds a "latency" which  would be objectionable to most users in a live situation.
Regards, Mike

[stm]

Vsat:  I see your point.  Very interesting what you mention about the pink noise generator usage!

So you say delays in the order of 7 to 10 msec sound good.  I don't see it very feasible to get that much with analog circuitry due to part number count and tight tolerance required.  However, my intention of implementing a short analog delay was to get the through zero flanger effect when the flanger just does its minimum delay, i.e. to adjust the analog delay to coincide with the flanger's minimum delay.  I suspect that in this case it may sound as through-zero and the the BBD signal will foldover, which I think would sound similar to real through zero.  Maybe this can be simulated playing numerically with a WAV file to validate the concept.  Gotta see if I can implement something just to see where we are standing.

See ya!

[Vsat]

stm,
Regardless of whether it sounds like regular TZF or not, it will be neat sounding... and thus worth trying! 10 mS delay difference is about the most you'd really want, before getting into noticeable delays - 2.5 to 5 mS would still give most of the thru-zero swoosh.
Regards, Mike

[stm]

Boss Dimension Chorus (DC-3) implements something that could be thought of as a dual tape playback system.  It uses two MN3007 BBD's whose clocks are modulated 180º out of phase, effectively crossing delays twice per LFO cycle,.  The two BBD outputs are mixed with the original signal.  Also, there is some low frequency reduction on the BBD'ed signals and some bass boost on the dry path to compensate for overall bass loss.

My point is that a Boss DC-3 could be tweaked to get the through zero flanging effect bypassing the frequency shaping networks and not using the dry signal in the mix.

[Vsat]

stm,
Yes - you should be able to do TZF by modifying a Dimension "C". You can hear TZF going on in some of the multiphase choruses such as the Solina or Juno (particularly noticeable with white noise).

I was getting a bit off track earlier.
With the method suggested by Chico, using your network, the approx. short time delay won't have a very noticeable effect until the BBD gets down to a similarly short delay... so it should sound like normal flanging when the BBD is delaying by say 1 - 10 mS, with closely-spaced notches and lots of low-freq rumble, then when the BBD delay time closely approaches the network delay time, the notches will spread apart much more rapidly, as in normal TZF. Should sound nice - probably better than the Eventide!
Regards, Mike

[Mark Hammer]

Happy Monday.

There are a number of ideas kicked around here which require clarification so they don't end up unintentionally misleading those with less background knowledge.

1) When VSAT (Mr. Pike to some of us) refers to maximum delay of 7-10msec, that refers to the modulated delay signal, not the time-staggered fixed-delay signal used to rovide room for the modulated signal to "move ahead" in time.  Seven to 10 msec is a commonly-occurring maximum delay time in flangers.  It would seem that the ideal max-delay time, though, is one which allows a good ratio of sweep.  Because they use bare-bones (but cost-effective) clocking circuits, a great many flangers have only a limited ratio of less than 20:1.   (I suggest reading the interview with Steven St. Croix in DEVICE at hammer.ampage.org if you haven't already done so).  Although it doesn't go through zero, one of the reasons why folks go ga-ga for the A/DA flanger is because it sweeps to such a short minimum delay, that it has a much higher sweep ratio than many other commercial flangers from that era.
Of course the key player in that ratio is the minimum delay time, not the maximum.  So, a flanger with a 20:1 sweep of 500usec to 10msec will likely sound more dramatic than one with a sweep of 1-20msec.  As near as I can tell, the reason why this is true is that flangers become more dramatic-sounding when there is a big contrast between a largely unaffected-by-notches state and a completely-infected-by-notches state.  The shorter the minimum delay, the fewer the audiblenotches at that point in the sweep.  If the minimum delay is already long enough to produce notches (and even 500msec would be if you had more than 5khz usable bandwidth in your signal path; 1msec especially so), then the contrast created by the sweep is reduced.  At the longest delays in the sweep, as well, there comes a point where the number of notches does not change dramatically, and where the listener's attention changes over from focussing on the notches/timbre to focussing on the delay-difference and the pitch change (i.e., the chorus zone).

2) I have no idea, and I know of no theoretical basis for determining, the optimum amount of ahead-time needed for TZF.  That is, if one can arrange for the time-swept signal to arrive at the mixing point before the fixed-delay time-staggered signal does, how much time staggering is needed to do that in an aurally satisfying or dramatic way?  Stated another way, when the swept signal passes "through zero", how far ahead in time does it need to go to create the disorienting swirl we associate with TZF?  My gut sense is that once you pass a millisecond or two of time staggering, nothing is gained.  It is the momentary "phase confusion" between the harmonics of the two signals that yields the delightful psychoacoustic confusion, and 1-2msec is likely enough to produce that.  Maybe, just MAYBE, 2-4msec if one is dealing with a restricted bandwidth signal path (e.g., 6-8khz total bandwidth).
To make matters even more complicated, one of the classic features of tape-based flanging is the "bounce" that occurs from jiggling tape transport mechanisms as the direction of sweep is reversed with the thumb.  This is so built into our template of what "Sky Pilot" and "Big Hurt" flanging sounds like that companies like Eventide built in electronic simulations of that sweep anomaly into their modulation circuits.  (Steve Giles has worked with it a lot and can provide further comment on what it does or doesn't add).

3) Will a Dim C be amenable to TZF?  It'd take a bit of work, perhaps more than is alluded to (some of us can find our way around that baby easier than others!  :wink: ).  Given that there is only one master LFO, the two MN3101's it has (for its pair of complementary MN3007's) would probably need to be staggered in their ranges.  Stock, they each use a 100pf cap to set their sweep ranges.  I'm wondering if one wouldn't need to offset them a bit by using discrepant timing caps.  The other thing is that , if I've interpreted Mike correctly, one would use just the delay signals and ignore the real time signal.  The DIM C uses a pair of counter-swept delayed signals to produce a wobble-less chorus.  In this application, one of those delays subs for the fixed time-staggered signal needed for TZF.

[stm]

Mark, your points are interesting. By the way, I did mention the possibility of tweaking a DC-3 and also leaving out its dry signal.

I will try to come up with a higher order filter capable of producing more delay and higher bandwith with the minimum opamps possible.

According to what's been said here, I will settle my next design for:

a) 7 kHz bandwidth
b) 500 usec fixed delay
c) 1% metalfilm resistors
d) 2% mica capacitors
e) 8 opamps max

Will post when I have this solved.  Bear with me, as this can take a few days since I'm quite overpassed here at the office.  Also, I need to borrow the "magic" book that has detailed info on these delay networks from my former boss.

Regards,

STM

[Vsat]

Mark (and list),
I'll make a few comments/clarifications. With normal flanging, the notch positions and spacing are determined by the time delay between the direct signal and the delayed signal. eg. a 10 mS delay will have the lowest freq notch located at 50 Hz (one cycle of 50 Hz has a period of 20 mS, and delaying the 50 Hz signal by 10 mS will put it 180 degrees out of phase, enabling cancellation when the two signals are summed). With TZF, the same situation applies, only that it's the time delay DIFFERENCE between the fixed and variable delays that counts. For example, with a fixed delay of 20 mS, and a delay that varies from 10 mS to 30 mS, the thru-zero point will be reached when the variable delay is at 20 mS... 20 mS-20 mS = zero mS. At the extremes, the delay  (relative to the fixed 20 mS delay) will be 20 mS-10 mS=(+)10 mS, and 20 mS-30 mS=(-)10 mS. If the LFO sweep is stopped at either + or - 10 mS, the sound will be identical to that of a single-delay flanger set to 10 mS. If you mix in the dry signal with the TZF output then the fixed 20 mS delay will be noticeable... but TZF does not use the dry signal.

Now, TZF does not normally use regen... it depends on good notching for it's effect (it sounds awful with regen as the individual delay times vary very little compared to an ordinary flanger). So for best effect attention should be paid to maintaining consistent and deep notches across the audio spectrum. With BBDs, there is a problem in that the insertion loss (attenuation of signal within BBD) varies with clock rate. Using regen with a flanger tends to hide this effect. With TZF, the BBD need only be swept over a fairly narrow range of clock rates as compared to normal flanging. So the insertion loss will have less variation, and notch depth will be consistent from one end of the passband to the other.

Having a large fixed delay (say 100 mS) and a variable delay (say 90 mS to 110 mS) is advantageous since the variable delay only changes by 10 percent in either direction. With a shorter fixed delay (say 11 mS) and a variable delay of say 1 mS to 21 mS, the variable delay is operating over a 21:1 range, and insertion loss will have greater variation. This is largely a moot point since a musician will NOT want to be playing through a delay on the order of 100 mS, and flanging will still sound good even if the notches aren't 40 dB deep. Just something to keep in mind if you are trying to maximize notch depth... also a VCA tied to the LFO can be used to compensate for insertion loss variation.

I tend to like 10 mS as the max "delay difference" in TZF as a compromise which allows for that rumbly "bottom of sweep" sound (notches every 100 Hz) and without presenting excessive "throughput delay" to the performer. 5 mS would give even less delay but would sound like a flanger with a max delay of 5 mS... less rumble than with 10 mS delay.

As an aside, to produce a first notch at 20 KHz (period = 50 uS) would require a delay of 50 uS/2 = 25 uS. To produce a first notch at 20 Hz (period = 50 mS) requires a delay of 50 mS/2=25 mS. This gives a "flange ratio" of 50 mS/50 uS = 1000:1 to sweep the audio spectrum.  For a guitar speaker with upper cutoff of about 5 KHz, a first notch at 5 KHz (period =200 uS) will be achieved with a delay of 200uS/2 = 100 uS... as in stm's design.  The design proposed by Chico and stm uses a fixed 200 uS delay (approximated with an allpass chain) and thus enters the thru-zero region only as the variable delay approaches 200 uS... for the rest of the sweep, the 200 uS fixed delay is less significant and it behaves more like an ordinary flanger. This design however gives it the ability to go thru-zero near the top of the sweep.
Regards, Mike

[puretube]

one of the nicest, itching threads in a long time...  

[StephenGiles]

Well after a normal September afternoon of 102F in the shade that makes good reading Mike. We fly back to Europe tomorrow, hopefully with no delays. The 3rd engine of our MD 11 wouldn't start in Quito so held up for 3 hours - plane already 4 hours late!!!!!!!!!
Stephen