Slapback Echo

[ElectricDruid]

I saw a promo recently for the EHX Slapback Echo, and was inspired to build one. Thinking that the PT2399 does a perfectly reasonable job for the sort of short echos required, I used that to keep the schematic as simple as possible. So here it is:




Pretty simple: two chips, two pots and one switch. It's a one-trick pony, but if it's a nice trick, it might be fun to play with. Anyway, I'm not paying $80 for one!!

Now I'm working on a PCB layout for a 1590B.

[GibsonGM]

Very nice, Tom, thanks for posting it!   

[StephenGiles]

One could substitute a 1K5 resistor with a 4K7 pot for the delay switch bit.

[ElectricDruid]

Of course. It's easy to add a delay time control or even a feedback control, but then it's just another PT2399 delay. So I've been fighting the temptation to add anything and kept it strictly to exactly what EHX have on theirs!

[Mark Hammer]

I have a little 9v battery-operated Danelectro Hodad amplifier.  It provides tremolo and "reverb".  The reverb is simply an on-off function with fixed delay time, mix and feedback level, using a surface-mount PT2399.  Never content to leave well enough alone, I added a delay-time pot to it.  Tom's is the "deluxe" version, with a 3-way toggle, blend and volume control.  Phew!  Too many knobs! 

[Ben N]

See, Corned Beef "Reverb", Danelectro. Probably the same circuit as the Hodad amp. I remember being disappointed to discover that it was just a slapback, then kind of tickled--hey, it's a slapback! Those little super-cheap PT2399 delay modules on Aliexpress that Pink Jimi Photon riffed on a while back would work just fine with a minimum of work (although I don't doubt that Tom put a bit more effort into his).

[antonis]

I think our Portuguese friend is favourably inclined towards mains supply & rectified/filtered +V..
(34 Hz corner frequency of R1/C1 LPF can't classify +V as "hum-free" supply..)

P.S.1
Of course, TL072 high PSR should take care of supply ripple and 1.3 Hz corner frequency of R2//R3/C5 VBias LPF should also make non-inverting inputs immune but, IMHO, a ten times bigger R1C1 time constant should be much more effective while maintaing lower than VBias charging up time..

P.S.2
Of course, I strain the gnat but do like a lot to search for minor blemishes in Tom's really flawless designs..!!

[ElectricDruid]

I think our Portuguese friend is favourably inclined towards mains supply & rectified/filtered +V..
(34 Hz corner frequency of R1/C1 LPF can't classify +V as "hum-free" supply..)

I'm not Portuguese! I only live here. I was British originally. Now I'm only Brit-ish.
Otherwise yes!

IMHO, a ten times bigger R1C1 time constant should be much more effective while maintaing lower than VBias charging up time..

Perhaps so. Please feel free to adjust to taste.
Partly it's that I haven't tested the current draw, so I don't know what the volt drop across R1 looks like. Consequently, I left R1 as a pretty cautious value initially. I often put 68R in that spot, which lowers the cutoff a little more. And if you can get 470uF/35V caps that are nice and compact then they'd obviously be an improvement!

P.S.2
Of course, I strain the gnat but do like a lot to search for minor blemishes in Tom's really flawless designs..!!

LOL! That's high praise coming from you, Antonis! Thanks.

[Mark Hammer]

See, Corned Beef "Reverb", Danelectro. Probably the same circuit as the Hodad amp. I remember being disappointed to discover that it was just a slapback, then kind of tickled--hey, it's a slapback! Those little super-cheap PT2399 delay modules on Aliexpress that Pink Jimi Photon riffed on a while back would work just fine with a minimum of work (although I don't doubt that Tom put a bit more effort into his).

Well, given that they were really intended to cater to the karaoke market, it's small wonder that 2399s do a decent job on slapback.

[ElectricDruid]

It turns out there was one other thing that Antonis missed! 

In my original schematic I'd put a 470KC (e.g. Rev Log) pot for the Boost control, thinking that would give the best gain response. Well, intuition is a fabulous thing...when it works! I've been wrong so often that I knew I needed to check, and when I did, I found that actually a linear pot gives the best response.
I plotted a graph with the gain created for each of the typical pot responses (so +dB's of gain against degrees of pot rotation). Having done that, I wondered if adding parallel tapering resistors would be better or worse. Turns out it's worse. I tried double-the-pot-value with an equal double-the-pot-value fixed resistor in parallel and plotted that too, for both log and lin pots. Here's the graph:



I've updated the schematic in the first post, so it's corrected now. If you still see "470KC" for the boost pot, empty your browser's cache.

UPDATE: My "Log pot" curve was a bit over-curved. I've adjusted it to be a theoretical copy of the "10% value at 50% rotation" that many audio log pots use, and that actually puts the log pot response slightly closer to the ideal than the linear. There's not a lot in though  - I've updated the graph above, so you can see.

[antonis]

It turns out there was one other thing that Antonis missed! 

Shame on him..!!!

(I think an S taper pot should bring 150^o rotation close enough to +10dB..)

[ElectricDruid]

There's one last thing to mention as well. I copied all the basic parameters from the EHX version - so 45, 65, and 100msecs delays, and 0dB to +20dB gain in the boost. That's pretty much all there is to it.

There's one thing that I don't know though, which is where EHX set the filters. We can have an educated guess though. If they used a MN3208 2048-stage delay to create the echo, they'd have needed a 10KHz clock rate to get out to 100msecs, and the filters would be around 3.3KHz (datasheet recommends 1/3rd of the sample rate). The original unit has a reputation for being pretty dark, so this is quite likely. Of course, it's *possible* they used a longer BBD in the modern version. 4096 stages would give them another octave on the filters at least, so 6.6KHz.

I ignored this and set the filters at 7KHz, so fairly bright by comparison. I might adjust that, but I need to build it and hear it first. Mark H says that a bit of filtering on echoes helps your brain separate them from the main sound, which seems like a good theory to me. So it might be better to lower the filter a bit more to help keep the slapback separate from the dry signal. We'll see.

[Frank_NH]

I can confirm that the original EHX Slapback Echo is rather dark sounding.  I had one briefly and didn't like the dark tone, so I'm anxious to try ElectricDruid's circuit (thanks for sharing!).

Another thing I didn't like was the tiny enclosure they adopted for the EHX unit.  My preference is for larger pedals that feel solid under my foot, like a 1590BB or 125B - something I can do with a DIY circuit. 

[bean]

There's one last thing to mention as well. I copied all the basic parameters from the EHX version - so 45, 65, and 100msecs delays, and 0dB to +20dB gain in the boost. That's pretty much all there is to it.

There's one thing that I don't know though, which is where EHX set the filters. We can have an educated guess though. If they used a MN3208 2048-stage delay to create the echo, they'd have needed a 10KHz clock rate to get out to 100msecs, and the filters would be around 3.3KHz (datasheet recommends 1/3rd of the sample rate). The original unit has a reputation for being pretty dark, so this is quite likely. Of course, it's *possible* they used a longer BBD in the modern version. 4096 stages would give them another octave on the filters at least, so 6.6KHz.

I ignored this and set the filters at 7KHz, so fairly bright by comparison. I might adjust that, but I need to build it and hear it first. Mark H says that a bit of filtering on echoes helps your brain separate them from the main sound, which seems like a good theory to me. So it might be better to lower the filter a bit more to help keep the slapback separate from the dry signal. We'll see.

I honestly don't know if the PT2399 is that consistent as far fixed resistance to VCO vs real world delay time. It's possible that different devices might produce inconsistent results. IMO, a 10kB pot would probably be just as good to eliminate that if it is a problem.

You could reuse the switch idea to select between different values of parallel caps @ C14 for bright/dark tones. Just my $.02, hope you don't mind the input.

[ElectricDruid]

Good question, Bean!

The datasheet gives numbers for the relationship between resistance and delay time, but it doesn't actually specify any tolerance for any of it. Could be +/-50% for all we know!

I've got a decent handful of PT2399's so I can do some tests.

[drdn0]

 I'd love to try and squeeze this into a 1590B with my sort of janky universal PCB layout - happy for me to give it a try?

[ElectricDruid]

I'd love to try and squeeze this into a 1590B with my sort of janky universal PCB layout - happy for me to give it a try?

Absolutely. Go for it.

I've had a go too, and I've got some boards on the way from China, so we'll see if they're any good. It'd be a rare time I'm happy with the first spin, but you never know!

[Kipper4]

Cool I like it.

[ElectricDruid]

I honestly don't know if the PT2399 is that consistent as far fixed resistance to VCO vs real world delay time. It's possible that different devices might produce inconsistent results.

Ok, I've got my boards back and soldered one up (and it works!) and I've tried a few PT2399's to see how much they vary. I've got two different batches, one set I got from Banzai a few years back, and some that I bought recently direct from Princeton's distributor in Europe.
The older ones give 40ms, 60ms, and 100ms. The newer ones give 45ms, 65ms, 105ms. There's next to zero variation within the batches.

It looks like what varies is the shortest-possible delay that the chip can do, and it looks like it varies by around 5msecs. While that might make a difference for Little Angel chorus applications, I don't think it's really significant here.

The equation I worked out from the datasheet figures for the delay time vs resistance is:

Delay msecs = (11.46 * Resistance KΩ) + 29.70

The part of this in brackets doesn't seem to alter, so the response of the clock to resistance seems to be fairly solid. What changes is that constant stuck on the end.

[ElectricDruid]

Just a later note to say that I actually got around to finishing this off and boxing it up. Details are over on my site:

https://electricdruid.net/slapback-echo/