Delay -> Reverb ideas I had

[RaceDriver205]

I was just thinking the other day (I get plenty of time to think on the daily journey to university), about delays in series:
In theory, every time you add a delay in series, you actually multiply the number of delay times by 2.
E.g.-
Guitar into Delay 1 (200ms) -> 0ms + 200ms delays
Delay 1 into Delay 2 (100ms) -> 0ms + 100ms + 200ms + 300ms delays
Delay 2 into Delay 3 (50ms) -> 0ms + 50ms + 100ms + 150ms + 200ms + 250ms + 300ms + 350ms
And so on.

The point im getting at, is that by the time you're at 4 delays, you actually have a total of 16 different delay times - i.e. a reverb! Does this sound about right?

[RaceDriver205]

Of course, I would imagine that the outputs of each delay would have to be fed back to the inputs of every other delay to make it work best.

[scaesic]

you'd need more than 4 to sound like  aproperly diffused reverb, and some filtering.

[RaceDriver205]

How many do you reckon?

[Seljer]

might be of some help: http://www.harmony-central.com/Effects/Articles/Reverb/

[dirk]

I think this is what your looking for: http://www.soundonsound.com/sos/may04/articles/synthsecrets.htm
Some more stuff on delay's: http://www.soundonsound.com/sos/jun04/articles/synthsecrets.htm

[SeanCostello]

In a real room, the echo density increases as the square of time, so for anything but super tiny rooms, you will need more than 16 reflections. For a good hall simulation, you probably need an echo density on the order of 10,000 reflections per second.

However, you can create a decent reverb with 4 delay lines, if you put them in the proper recursive network. Google "feedback delay networks" for some hints.

Sean Costello

[A.S.P.]

don`t use whole-numbered multiples of delaytimes

[mattpocket]

I guess you wouldnt use whole numbered multiples of delay times because after a while they'd come back to when they start right? and over lap other ones and stuff... so making them silly numbers would reduce liklihood of this?

[SeanCostello]

If you use delay lines that are related to each other by prime number ratios, theoretically you reduce the risk of echoes overlapping with each other. The ratios should ideally be large; in digital systems, the prime numbers used generally relate to the number of samples in each delay line (i.e. each delay line contains a large prime number worth of samples).

However, this only works OK at best in practice. Even if all the delay lines are prime numbers, you can have bad sounding resonances in the system. My guess is that this is due to delay lengths "almost" overlapping, which ends up producing a very similar effect to overlapping delays.

In addition, with time-invariant delay lines, you inevitably will have an audible artifact that occurs at a rate equal to the sum of all the delay lengths. For larger systems, with lots of delay lines, this is less of a problem, but for systems with only a few delay lines, you can hear a beating sound, even if everything is tuned perfectly.

Sean Costello

[RaceDriver205]

Wow, that complicated? Screw it then.

[mattpocket]

Wow, that complicated? Screw it then.

Haha, a man after my own heart!

[Processaurus]

Have you seen those huge blue yamaha ud stomp delay pedals?  They have 8  separate delay lines you can program to do different things, like this.

[mattpocket]

The Line6 delay modeller looks nice too!

[Mark Hammer]

The MN3011 used 6 taps that were not multiples or divisors of each other.  This avoids the unpleasant resonances.  Unfortunately, while such a chip moves in the direction of mimicking those aspects of reverb, I have never seen a design that went the extra mile to make it sound more realistic by having differential filtering of the various taps.  In the real world, the longer delays have more high end absorbed, compared to the initial reflections.  The preponderance of designs simply mix the 6 taps to one op-amp with slightly different value mixing resistors to gve more amplitude to the early taps/reflections.  What they should do is to maybe mix 2 consecutive taps to their own op-amp with successively more lowpass filtering applied to the longer-delay taps.

One way of achieving the sort of reverb-with-single-tap-chips arrangement that RaceDriver was aiming for would be to harness one master clock via a counter chip, like a 4017, and divide the master clock down by odd numbers to feed several BBDs of different capacity.  Keep in mind that BBDs are typically multiples of each other, so dividing the clock by 2 or 4 or any even number still gets you multiples.  To move in the direction of an MN3011 (still the easiest way to mimic reverb in the analog mode) you wold need to divide a master clock by 3, 5, 7, etc.

For instance, say we had a master clock running at 100khz (That assumes two complementary 50khz clock pulses once flip-flopping takes place).  Feed that to a 2048-stage MN3208 or BL3208, and we get 10.24msec of delay time.  Divide 100khz by 3, feed that to a 1024-stage MN/BL3207, and you get 15.36msec of delay time.  You get 25.6msec of delay if you divide that 100khz clock by 5 and feed another MN/BL3207.  Feeding the divide-by-3 pulse to a 256-stage MN3209 gets you 3.84msec of delay which you can stick between the BL3208 and BL3207 stages.  You get the general idea.  The gist is to use a single master clock to extract different delay times out of different capacity chips which can be tacked onto each other in series to get a variety of non-harmonically related  delay times that are then summed like those of a multi-tap chip.

So what's the catch?

1) Though there is one master clock, there are still multiple clock signals prancing around the circuit, so the risk of heterodying is still there.  You'd need to keep those clock line traces mercifully short.  This is likely NOT a good case for vero or perfing.

2) The filtering needed for minimizing clock noise and aliasing would depend on the divided clock frequency hitting each BBD.

So, it CAN be done, but clearly the MN3011 is the easiest way to attempt it.  A pair of them in series, or perhaps a 2nd one fed by one of the midway taps on a primary MN3011 would move even closer towards mimicking the diffuse sound of actual reverb in an analog environment.