Trying to optimize PT2399 reverbs before microprocessors take over their job

[Synthlurker]

This is one of those projects that I might possibly have no time to complete myself, but thought that I might share some findings that will make it easier to design some good sounding reverbs with good ol’ PT2399.

First of all, the Belton Brick is apparently soon 20+ years old! There are a few different designs floating around, and seems that “Modulo Belton” is the cleanest “trace” with approximately stock values:

It is possible that the original Belton brick used 4,99k, 3,24k, and 1,82k for its delay timings as per the patent sheet, but it is also likely that they also just used E12 values as in the Modulo Belton schematic. Other possible difference is the CD40106 (or 7414) LFO speed, which is 0,47Hz in Modulo Belton, but might be faster/slower in real Belton bricks.

Scraping through the different already used and hypothetical delay values (and calculating using the formula Delay time (ms) = 11,46 * pin 6 resistor kΩ + 29,70), I was able to create this table:

Most PT2399 reverbs (as most reverbs use Belton bricks) use 1,333 ratio (or 0,75, whichever seems clearer), which does sound “good”.

Equinox II and clones have a big 1,618^2 ratio, although doubling D2 delay length results in approximately the common 1,333 ratio.

There is a possibility that the Hamuro reverb is only conceptual, but anyways it is the only reverb with approximately golden ratio delays.

T60 also has size dual pot (like in BTDR-3H brick) to adjust the ratios, but this will result in less mathematically-perfect reverb delay taps. But it works if it sounds good enough (and I think it does!)

I’m thinking about trying to make Belton brick-like delay, but with silver ratio 1,414 (= sqrt(2))… It will hopefully be a bit more natural. Or maybe I could try adding a fourth PT2399 with longer delay to improve the sound…

PT2399 can be supplied with up to 5,5V safely, which will improve headroom and therefore reduce distortion. Alternatively possible distortion of the delay network could be reduced by adding a green 3mm LED from all the pin 7s of PT2399s to ground, altough as the audio passes first through only one PT2399, adding LED to only that one could be enough.

The filters could probably be adjusted a bit, but I’m skipping this here as tuning the multibeedback lowpass filters it is rather complex topic.

There is possibility to get rid of the opamp and oscillator chip and therefore make smaller reverb “brick” by using deadastronaut’s ‘after chorus’ modulation trick of 470n capacitor + 1M resistor from pin 8 to pin 6 for at least one PT2399 (preferably one of the longer delays as modulating shorter delays is less audible). The now free PT2399 opamp could be used for the feedback circuitry!

Also I have a suspicion that the feedback circuit feedback resistor (R19 in Modulo Belton) is varied between the different flavours of BTDR-2 bricks (short/medium/long)… changing this instead of using external feedback circuit would simplify feedback reverb circuitry (like Grindcustom fx Tenebrion reverb). 10k pot should be good enough for this.

But as I mentioned in the start, making reverbs like this is starting to become unnecessary complicated as even cheap microprocessors (for ex. Raspberry Pico 2) have sufficiently good internal ADCs (12 bits resolution, low noise etc.) and enough processing power for proper floating-point calculations and >1s of clean delay without external parts.

[K.ostas]

A couple of links I’ve come across you might find useful:

• Jack Orman suggests to push PT2399 at 6V (in that case, for a better chorus effect).
• Brian Thornock’s T60 DIY reverb pedal with three four (as you correctly mention, my recollection was wrong) PT2399s.
• Rod Elliott has an interesting article including a suggestion of how to simplify the stock (datasheet) PT2399 circuit.

[Synthlurker]

As en example of improving ANY PT2399 reverb, here is modded version Ectoverb from Bluff Chill Devices (maybe call this the “Plasma mod” lol). I’ve built the original some time ago, but it was VERY noisy due to way too big resistors used (this one should not be noisy anymore as resistors are scaled to more reasonable values). The main point I want to show with this are the basic filters/mods one can adjust or add into practically any PT2399 reverb (or delay).

Logically, this would also need some kind of input buffer and output wet/dry blend but I’m not including that because it isn’t part of the reverb circuit itself. I have no idea what feedback resistor (R3 and R6) values would be good for this, anything above 10k shouldn’t oscillate (consider series 10k + 10k pot). As this does not have MFB filter at all, the schematic is easier to read than most PT2399 schems, and all filters are simple 6dB filters. Also, this has practically the same delay time ratio as Equinox II (approx 1,618^2 or 2*1,333)!

And here is the same as simple paint drawing ignoring all filters. My point is that for the cleanest, most natural reverb, there needs to be some modulation to get less metallic sound (LFO modulation is another alternative, but the simple pin 6-8 cap+resistor circuit which reacts with input level feels more interesting to me) and pin 7 LED limiter/clipper to prevent the PT2399 from clipping internally (PT2399s internal clipping sounds horrible compared to this).

Utilizing one of the PT2399 internal opamps (pins 13 and 14) for feedback mixing circuit likewise in the Belton brick could work for this circuit too. I’m not going to redraw the circuit to show that, so here is once again a paint drawing to visualize the circuit: