Request: PT2399 delay circuit analysis

[tatou]

After about 15 builds from kits and PCBs, I'm branching out into hacking/modding/designing my own circuits, and learning more about the basic building blocks. AMZ and ElectroSmash have been good resources, and I've also appreciated the circuit analysis in some MBP projects (like the Slow Loris).

Anyway, right now I'm trying to understand simple PT2399 delays. I started by redrawing two low-parts-count delay schematics to have similar layouts: MBP's 2015 Cave Dweller and Tonepad's Rebote v1 (which appears to just be the manufacturer's example circuit, plus the voltage regular and input buffer?). My plan was to compare and learn from the "diff" between these two schematics.

However, I'm kinda lost.

My drawings of the Cave Dweller and Rebote schematics are attached. Can you fine folks help dissect these circuits for a relative n00b like myself?

Specific questions:

• I'm guessing the JFETs are input buffers to retain tone. However, neither of them look like Jack Orman's example JFET buffer here. How do these two buffers work?
• My understanding is that the PT2399 is notorious for distorting repeats at long delay times (I build a Zero Point Micro II and its delays are too crunchy for my taste). So you have to do a lot of filtering. Where are the filters in these circuits? Is the LED on pin 7 in the Cave Dweller meant to help with the distortion?
• Most everything coming off pins 9-16 on the PT2399 looks wildly different between the two circuits to me. So are they using the LPFs and OAs on the chip differently?
• How does the mix pot in the Cave Dweller work? Is there a corresponding (fixed) portion of the Rebote?

I'd appreciate any help understanding these things. Thanks!

[midwayfair]

The parts we mess around with on the PT2399 are the analog half of the chip, which are inverting op amp stages that are used for high- and low-pass filtering. You can see them drawn out looking like regular op amps in this: http://www.madbeanpedals.com/projects/ZeroPoint/ZeroPointMini.pdf

You need a way of mixing the dry and wet signals, because most people don't want their delay 100% wet. The tools to do this properly don't really exist on the PT2399 itself because you're using the op amps on it for filtering. So you need a "wrapper" for the delay circuit, which is at a minimum an input buffer; an output buffer lets you also do a mix pot very easily and still use both op amps in the PT for filtering.

The Cave Dweller 2015 is probably as simple as you can possibly get. It uses the input device as both a buffer (the source output) and amplifier for the dry path. (cf my Hamlet delay.) The drain output is the dry path, and since it's inverted, it just gets fed into an inverting stage of the PT and now it's back in phase.

The dry path is in phase with the input, but then it goes through a single inverting op amp, so it needs to be inverted. It goes back into the PT at the same point as the dry path via a mix pot that either grounds a 4.7K mixing resistor connected to the wiper of the mix pot (so no delay signal can get in) or connects the delay output (lug 3) to that mixing resistor (full delay signal).

The LED on pin 7 is a current limiter, which prevents the signal from exceeding ~1.7V peak to peak inside a chip that runs on 5V (so, max of under 2.5V peak to peak before clipping). The internals of the PT2399 aren't entirely known, but there's a thread on DIYstompboxes linked in the build doc for the Hamlet delay on my website if you want to know as much as I know about it. Note that the LED has nothing to do with the distortion on the repeats at longer times. That's more to do with the chip's memory and digital programming.

Why's the CD look so different from the Rebote? Because the Rebote has no mix pot. So it doesn't split the signal and just runs the dry signal right into the PT2399, and you get control over the number of repeats and the delay time only.

There is no exactly corresponding spot for a mix pot in the Rebote 1.0, but there probably is some way to achieve some sort of level control for it. If you look at the Rebote 2.0 and the million subsequent variations of it, the input amplifier is replaced with an inverting amp stage, which outputs to the PT and the mixer stage; the second half of the op amp mixes the dry and wet signals together.

[tatou]

OK, I think I learned a couple things here.

THING ONE: I guess I don't understand how buffers work. I haven't been able to google out super-useful information, but this page suggests that if the output comes from the emitter of a BJT transistor, the signal is not inverted; if it comes from the collector (aka "transistor switch"), the signal is inverted. I assume that FETs have the same properties, such that G/S/D correspond to B/E/C?

If so, from the CaveDweller design, it makes sense how we're using both buffer signals. (1) The inverted (drain) signal is the "dry" one, and gets inverted by LPF2 (pins 13-14 in the original CaveDweller schematic but not my redraw) and mixed with the delays via the MIX pot for output. (2) The non-inverted (source) signal gets fed into LPF1 (16) which is delayed, inverted, spit out of OP2 (pin 12), then mixed with the already-inverted dry signal into LPF2... but also fed back into the LP1 (16) via the FBK pot for more repeats.

THING TWO: I made a mistake in my redraw of the CaveDweller schematic. I assumed LPF1 and LPF2 were interchangeable opamps based just on the PT data sheet, so I swapped them in the drawing to look more parallel with the Rebote schematic. However... that ValveWizard doc you pointed to suggests that the input of LPF1 (pin 16) gets inverted for output (15) **AND** is what goes into the delay circuit and output on OP2 (12)

If that's the case, then Rebote is a little harder to follow, but what I think is happening is that only the inverted (drain) signal ever gets used, both for "dry" and repeats. The "dry" signal gets fed into LPF1 (16) and deinverted for output (15)... while its delayed signal comes out of OP2 already deinverted (12), so it has to be dampened and re-inverted by OP1 (13-14), and finally merged with the "dry" signal back into LPF1, etc. (The rat's nest is confusing to me, and while I suspect the polarized caps have something to do with keeping the inverted/deinverted signals going in the right direction, I'm still confused.)

Anway... is this any closer?

For anyone following along at home, the docs Jon mentioned are here:

• Hamlet Delay build doc
• ValveWizard's PT2399 notes
• DIYStompbox thread on PT2399

PS: Hey Jon, I see that you've been a fawmer since 2013! I'm Burr, the guy who started FAWM. Pedal-building is a recent development... thanks so much for the help!

PPS: I actually investigated your Hamlet circuit a little bit over the weekend as well, but after scratching my head too many times decided to focus on Rebote & CaveDweller since they are simpler.

[midwayfair]

Wow! Great to meet you, and thanks for starting FAWM!

A buffer is literally any active stage you put between one thing and another. It can be doing something else, like how the drain outputs are also amplifiers, but the guitar is buffered from the rest of the circuit by Q1.

I'm on mobile so I'll try to come back to some other questions.

[tatou]

Follow-up.

So here's another very simple PT2399 delay, the Dirty Punk Repeater, which I drew based on the perf/PCB layout here. Schematic attached, again meant to be "parallel" to the previous two.

If I understand correctly, the input goes straight into LPF2 (inverted) and then split into (1) the Q1 output buffer and (2) LPF1, where it gets delayed out of pin 12 and looped back to (1) the Q1 buffer and (2) LPF1. Super simple.

My questions about this relative to the previous two are:

• Is each repeat out of phase with the previous one? If so, is that (probably) on purpose?
• Compared to CD and Rebote, there are very few res/caps on the analog side of the PT2399. What are these doing?

[midwayfair]

I'll try to suss out the repeats on that.

As far as the caps go:

C5 and C6 (100nF caps) are some standard thing that's not actually part of the analog circuitry. I think Merlin Blencowe (merlinb) might have explained what they do at some point, but god knows where and when. Some people report getting lower noise when they change them.

The easy ones are:

C4: This forms a mild low-pass filter by shorting R5 for all frequencies above a certain point. The gain stage is just over 1x, but for frequencies above a certain point, it's no gain. It's not doing a whole lot, eh?

C2 is a decoupling capacitor, to remove DC before going to the base of Q1; it also forms a high-pass filter with the repeats pot value. The DC bias for Q1 is provided through R6, which is attached to one of the pins of the PT2399 without a capacitor.

C8 is also a couple capacitor, this time for the input to the chip.

Note there's no input buffer! Your input impedance is the series resistance plus feedback resistance of the first LPF.

Actually, I can't answer the phase question. I'm not totally sure what happens internally in the chip. There are some things about the PT that are mysterious because they aren't in the datasheet. I think you migth need to scope it.

[tatou]

Thanks, Jon. This is all very helpful.