A simple tweak to delay circuits/builds

[Mark Hammer]

I'm a fan of adding some gentle treble-trimming to the feedback/repeats path of delays, whether analog or digital.  I do this by adding a cap to ground after the feedback pot.

Well, a local guy was selling his attempted (unfinished) PT2399 delay build, plus a non-functioning MXR Commande Flanger for $25, so I figured what the heck.  I finished up his unfinished PT2399 delay (which seems to be identical to a Rebote 2.5, and thought I'd try out something.  Instead of JUST the switch-selectable treble-trim in the feedback path, I thought I'd use a 3-way toggle and insert a "grit" function by using the other toggle position to insert a pair of Schottky diodes (1N5817) to ground to clip the feedback signal.

Not especially elegant, but adds some bite and grit to repeats that goes a little beyond the grit occurring without any such clipping when aiming for many repeats at longest delay.

The Schottkys are essential, since they have the lowest forward voltage (mine were around 160mv).  If one uses germanium or silicon types, it is likely that the amplitude of the feedback signal will not be sufficient to clip.  Note that repeats, in general, tend to decline in amplitude.  So if one needs to use very low forward-voltage diodes in the first place, how would 3 and 4th repeats end up being high-enough amplitude to result in clipping, unless one dimed the repeats control?  Easy.  The tonal qualities of the first repeat are retained for subsequent repeats, and the first repeat is generally more than sufficient to produce the intended grit.

Not earth-shattering, but a simple and cheap mod to add some variety to delay sounds.

[Ripthorn]

Keeley does this in the mag echo as well. I'm a fan of doing things with delay signals, be it filtering, clipping, general destruction, love it all.

[Mark Hammer]

Keeley does this in the mag echo as well.

I did not know that.  He's a smart fella! 

[FingerBlisters]

Could someone smarter than me explain off which pin this is occurring? Or draw a rough schematic to help those more visually inclined?

I guess you're using an on/on that adds the diodes in when flipped?

[niektb]

Keeley does this in the mag echo as well.

I did not know that.  He's a smart fella! 

https://www.pedalpcb.com/docs/MagnetronDelay.pdf

[11-90-an]

Could someone smarter than me explain off which pin this is occurring? Or draw a rough schematic to help those more visually inclined?

I guess you're using an on/on that adds the diodes in when flipped?

Depends on the certain pedal, but for example the Magnetron Delay that niektb posted above...
Schematic (without power supply):



delayed output is taken from pin 12, then filtered by an internal opamp in the pt2399, which is accessed with pin 13 and 14. After that it has some other components then the signal is sent in 2 ways, the "level" pot and the "feedback" pot. The feedback is what we are interested in right now. as you can see in the schematic, there are 2 diodes, D1, and D2, which clip the signal. Other than those diodes you can add some filtering, etc., to condition the sounds of the repeats...

Hope that makes sense...

[bluebunny]

I thought "delay" in the thread title was a verb, and this was going to be advice on curtailing my build addiction... 

[Mark Hammer]

Could someone smarter than me explain off which pin this is occurring? Or draw a rough schematic to help those more visually inclined?

I guess you're using an on/on that adds the diodes in when flipped?

Depends on the certain pedal, but for example the Magnetron Delay that niektb posted above...
Schematic (without power supply):



delayed output is taken from pin 12, then filtered by an internal opamp in the pt2399, which is accessed with pin 13 and 14. After that it has some other components then the signal is sent in 2 ways, the "level" pot and the "feedback" pot. The feedback is what we are interested in right now. as you can see in the schematic, there are 2 diodes, D1, and D2, which clip the signal. Other than those diodes you can add some filtering, etc., to condition the sounds of the repeats...

Hope that makes sense...

Thanks.  Yes that's how to do it.  Personally, I'd recommend Schottky diodes like the 1N5817, instead of the silicon 1N4148 shown.  The forward voltage of the silicon units may be too high to result in audible clipping.  Any clipping one hears may well be a result of trying to push the PT2399 too hard, rather than from the diodes.

[jubal81]

Could someone smarter than me explain off which pin this is occurring? Or draw a rough schematic to help those more visually inclined?

I guess you're using an on/on that adds the diodes in when flipped?

Depends on the certain pedal, but for example the Magnetron Delay that niektb posted above...
Schematic (without power supply):



delayed output is taken from pin 12, then filtered by an internal opamp in the pt2399, which is accessed with pin 13 and 14. After that it has some other components then the signal is sent in 2 ways, the "level" pot and the "feedback" pot. The feedback is what we are interested in right now. as you can see in the schematic, there are 2 diodes, D1, and D2, which clip the signal. Other than those diodes you can add some filtering, etc., to condition the sounds of the repeats...

Hope that makes sense...

Thanks.  Yes that's how to do it.  Personally, I'd recommend Schottky diodes like the 1N5817, instead of the silicon 1N4148 shown.  The forward voltage of the silicon units may be too high to result in audible clipping.  Any clipping one hears may well be a result of trying to push the PT2399 too hard, rather than from the diodes.

I'd advise against the 1N5817. They are EXTREMELY leaky and will have a big effect on the filter. Bat41s have a similar FV but don't leak much at all.

[Mark Hammer]

Fair enough.  I have some of those I think.  I'll try them out when I have a moment.

[Paul Marossy]

I've been using a DemonFX "Tape Echo" clone of the Keeley Magnetic Echo and thought it sounded really nice, so I dug up a schematic out of curiosity. I've added modulation to PT2399 delays before, and Keeley's implementation of that is pretty standard. But, I was surprised to see the 1N4148s in the feedback section. It seems to me that the intent is to have them act as clamping diodes, not to introduce distortion. You can't get runaway/forever repeats on this circuit, so I'm not sure if that is part of the reason why or not. I just know that I like it. 

[Rob Strand]

I've been using a DemonFX "Tape Echo" clone of the Keeley Magnetic Echo and thought it sounded really nice, so I dug up a schematic out of curiosity. I've added modulation to PT2399 delays before, and Keeley's implementation of that is pretty standard. But, I was surprised to see the 1N4148s in the feedback section. It seems to me that the intent is to have them act as clamping diodes, not to introduce distortion. You can't get runaway/forever repeats on this circuit, so I'm not sure if that is part of the reason why or not. I just know that I like it. 

Normally diodes in that position just limit the signal level if it does happen to go nuts.

What makes it grow is the gain around the loop is greater than 1.  For example if the circuit has a gain of 2 around the loop and initial input is 1V, the gain around the loop is 2 so the effective input level for the feedback level is 2V.  The 2V now feeds in and comes back as 4V.  The values aren't important, what's important is the increased level.  If the gain is 1/2 then you get 1V, 1/2V, 1/4V..  which continues to decrease.

I just took the pedalPCB schematic
https://tagboardeffects.blogspot.com/2019/09/keeley-magnetic-echo.html
https://www.pedalpcb.com/docs/Magnetron-PedalPCB.pdf

Start at output then progress around the loop.

output_filter_gain = R17/R13 = 15k/10k = 1.5

Next we hit R16 = 10k.  That forms a divider with:
- the level pot, 100k
- R10 = 10k, the loading on R16 depends on the setting of the level pot
- the feedback pot, 100k
- R8+R9, the loading on R16 depends on the setting of the feedback pot

Lets say with the feedback pot is on full and the level pot is on min.
Both give these setting largest gain around the loop.

The load on the R16 is R = 100k // 100k // 11k = 9.02k
So the *DC* gain of the divider formed via R16 is
divider_gain = 9.02k / (10k + 9.02k) = 0.472

The gain from R9 through the to the output of the PT2399 at R7 is about
input_filter_gain = R7 / (R8 + R9) = 10k / (10k + 1k) = 0.909

So the gain around the loop is:

fb_loop_gain = output_filter_gain * divider_gain * input_filter_gain * PT2399_gain
                  = 1.5 * 0.472 * 0.909 * PT2399_gain
                  = 0.643 * PT2399_gain

The PT2399 gain is around 1 but according to the datasheet it can be -0.5dB to 2.5dB, but it's
not clear what conditions those are under.   2.5dB is a gain of x1.33.

For PT2399_gain = 1,  fb_loop_gain = 0.642
For PT2399_gain = 1.33,  fb_loop_gain = 0.85

So even under the higher gain the loop gain is less than 1 so there's no blow-up of the feedback.
If we advance the delay level the level (divider_gain) will be even lower (by quite a bit as that's a addition
10k load on the divider).

So a big approximation I've made is calculating the DC gain.   The cap C5 provides some HF boost and that's counteracted by the roll-off of the filters.   I'm sure the net effect will be C5 provides a gain higher than
the DC gain I've calculated - a spice sim is required to see the fine details.   Nonetheless the loop
gain is looking low-ish and that's probably why the feedback doesn't blow up.

I think diodes limit the feedback level to prevent the PT2399 overloading with high signals.

[Paul Marossy]

Normally diodes in that position just limit the signal level if it does happen to go nuts.

That would make sense although I haven't really seen it anywhere else other than Keeley's circuit. I haven't seen every PT2399 delay/echo circuit there is to look at however, just thinking of the most popular ones from memory.

What makes it grow is the gain around the loop is greater than 1.  For example if the circuit has a gain of 2 around the loop and initial input is 1V, the gain around the loop is 2 so the effective input level for the feedback level is 2V.  The 2V now feeds in and comes back as 4V.  The values aren't important, what's important is the increased level.  If the gain is 1/2 then you get 1V, 1/2V, 1/4V..  which continues to decrease.

Yeah, that makes sense. Kind of a snowball effect.

The PT2399 gain is around 1 but according to the datasheet it can be -0.5dB to 2.5dB, but it's
not clear what conditions those are under.   2.5dB is a gain of x1.33.

For PT2399_gain = 1,  fb_loop_gain = 0.642
For PT2399_gain = 1.33,  fb_loop_gain = 0.85

So even under the higher gain the loop gain is less than 1 so there's no blow-up of the feedback.
If we advance the delay level the level (divider_gain) will be even lower (by quite a bit as that's a addition
10k load on the divider).

So a big approximation I've made is calculating the DC gain. The cap C5 provides some HF boost and that's counteracted by the roll-off of the filters.   I'm sure the net effect will be C5 provides a gain higher than the DC gain I've calculated - a spice sim is required to see the fine details. Nonetheless the loop
gain is looking low-ish and that's probably why the feedback doesn't blow up.

Sounds reasonable to me.

I think diodes limit the feedback level to prevent the PT2399 overloading with high signals.

That seems to be the case but for kicks I tried adding these diodes to a circuit I built that's been on the market since at least 2009 and it didn't seem to do anything at all. Still got the runaway repeats with ever increasing noise. It's like every time it repeats it adds a little more noise. Maybe there would need to be more done to the circuit if that is implemented? The two circuits are pretty different with the way the delay and repeats are implemented. In any case, Keeley's design is clever I think in this regard. Doesn't really have enough repeats to do the kind of "ambient" stuff I like to do but still is very nice sounding. It repeats about 5-6 times with repeat at full clockwise vs. literal almost forever repeats.

[Mark Hammer]

I put a simple one-pole lowpass filter on the feedback loop after the pot.  Pick the right corner frequency and it progressively removes top end with each repeat.  That does several useful things:
1) Moves repeats to the perceptual background so the sound doesn't seem so cluttered.
2) Allows you to dime the repeats without getting runaway feedback.
3) Helps to reduce cumulative grime that comes from regenerating analog or digital signals through an analog feedback path.
4) Sounds more like the way real echoes sound in the real world, where nearby imperfect reflecting surfaces damp the tone.

[niektb]

I happen to own a EQD Space Spiral (okay okay I built a clone) and I can get it to go completely nuts with all kinds of distortion and mangling

Here is a simple demo  (no other effects enabled, just dry sound from my smartphone built-in mic)

PedalPCB calls it the Dark Rift Delay. Maybe someone else can tell whether it's the PT2399 or the diodes clipping?

[Rob Strand]

I put a simple one-pole lowpass filter on the feedback loop after the pot.  Pick the right corner frequency and it progressively removes top end with each repeat. 

Maybe that's why the Keeley circuit messes with the input low-pass filter.   Notice the filter cap that normally goes to ground is connected across the resistor in the feedback path.   That boost the highs in the feedback path and will offset the loss of highs on each repeat.

When you emulate room reflections typically the high-frequency reflections are attenuated more than the lows, so the normal connection with the loss of highs on each repeat emulates a physical process.   A men's room has less attenuation of highs because of the reflective surfaces.  When you have long delays the delayed copies are more a deliberate effect rather than emulating natural room effects - so this is perhaps where the Keeley connection come in.

[Mark Hammer]

When you emulate room reflections typically the high-frequency reflections are attenuated more than the lows, so the normal connection with the loss of highs on each repeat emulates a physical process.   A men's room has less attenuation of highs because of the reflective surfaces.  When you have long delays the delayed copies are more a deliberate effect rather than emulating natural room effects - so this is perhaps where the Keeley connection come in.

The other aspect is that, even IF one imposes the same additional progressive LPF on long repeats, because they are spaced further apart, one doesn't necessarily notice the progressive loss of highs as much as when those repeats come 40msec apart.

[Paul Marossy]

I put a simple one-pole lowpass filter on the feedback loop after the pot.  Pick the right corner frequency and it progressively removes top end with each repeat. 

Maybe that's why the Keeley circuit messes with the input low-pass filter.   Notice the filter cap that normally goes to ground is connected across the resistor in the feedback path.   That boost the highs in the feedback path and will offset the loss of highs on each repeat.

Is this different than a "tone control" on something like the Wampler analog echo? I imagine it is more like a fixed "tone control" tuned to a specific frequency? I did notice that an adjustable tone control affects how much runaway repeats you can get on a Wampler analog echo type of circuit. The diodes however seem to have no effect if you add them into a circuit like that... but the filtering the between the Keeley and Wampler circuits are also quite different.

[vortex]

The Ibanez EM-5 Echomachine has back to back 1N4148 diodes in the delay line.

I always thought that was noteworthy... 

[iainpunk]

The Ibanez EM-5 Echomachine has back to back 1N4148 diodes in the delay line.

I always thought that was noteworthy... 

in a soft clipping opamp configuration!

cheers