PT2399 noise problem. Have you encountered it?

[merlinb]

I am messing about with a PT2399 and I am finding that when the delay resistor exceeds about 20k I am getting a lot of quantization noise. It makes the repeat sound a bit like a 'chuff'. Has anyone else encountered this?

[~arph]

Oh yes, it's notorious to get noisy on long delay times, the filtering on the output is really important as well as the filtering in the feedback path (if any)
There are a bunch of threads here on PT2399 noise reduction, none with clear final examples though.

I run four PT2399's in series so imagine the noise in there. I did find that a lot of the noise in my circuit actually came from the mixing circuit and not from the PT2399's

[Mark Hammer]

Forum member Alain Parent asked me to give his Rebote 2.5 build a listen over the weekend because there was some "distortion" that he couldn't seem to get rid of.  And what you describe is precisely what I heard; almost like a "zipper" sound at longer delays.

Given how many PT2399s there are in the world, obviously there has to BE some sort of solution to this or else none of the pedals that use them would sell as many as they do.

In the case of the Rebote, I gather the simple solution is to a) keep the max delay at around 350msec or less, and b) up the value of those .01uf caps in the filters when in the longer delay-time zone.

[R.G.]

You're right mark. The PT2399 gives a very good compromise on noise and delay, but when the delay gets out to about 400-500mS, it runs out of delay stages, the sampling rate goes down, and the internals can no longer spread the noise and imperfections out to high enough frequency. It gets down into audio. At that point, the only options are to filter the resulting audio to even lower cutoffs to lose some of the noise.

It is also true that there may be internal issues which make this more zipper-eque. In the case where the noise is especially bad, I'd say to try another PT2399. It may (or may not) be different. Fortunately, this is cheap, and the worst case is that you spent $5 to learn that you can't fix it. Zipper noise is unlikely to be coming from the rest of the circuit.

[frequencycentral]

The PT is a very seductive piece of shit.     Wish there was something comparable but slightly more upmarket.

[merlinb]

Well I have managed to improve it by increasing the integrating caps between pins 9-10, 11-12, to 100nF. There's still a bit of chuff, so maybe I'll just limit the maximum delay to 200ms or something...

[Mark Hammer]

The PT is a very seductive piece of sh*t.     Wish there was something comparable but slightly more upmarket.

With BBDs, we can simply daisy chain more stages and run the clock a little faster to squeeze equivalent delay with higher quality.  The last time I looked at the Tayda site, they had PT2399s for 48 cents each.  Set aside the rather mind-boggling idea that a complete digital delay on a chip is available for the price of a quad op-amp (and less than a knob).  Is it possible to gang these things to a common clock and run several in series to get that elusive longer delay without the zipper noise?

[frequencycentral]

No clock input.  

Using a single delay time pot with two PT's in parallel causes horrible noisy issues - not sure why, as it seems like it *should* work. A dual gang would work ok I guess. Maybe the answer for multiple PT's would be to have a VCR on each chip, like the BC560 in Echobase for example, with all the VCR's controlled by a single pot.

[Mark Hammer]

That assumes that every clock on every chip is made to tight tight specs, and having each chip governed by some attempt at precise equal-resistance-but-independent clock control will have all individual clocks synced.  The possibilities for heterodyning strike me as way too risky.  Keep in mind that while the individual clocks may not be within the audio spectrum, difference products of very close digital clocks may well be.  And of course, if one moves into 3 or more cascaded chips, the prospects for difference products being audible increases substantially.  Moreover, even if same speed is attainable, syncing is not a given.

Or am I being too pessimistic? 

[frequencycentral]

No, you're correct of course. But it's the only way given the limitations of the PT to even get close. But surely worth a try anyway. Like I said, seductive.......but you'd be into matching PT's and matching VCR's.......

[R.G.]

That assumes that every clock on every chip is made to tight tight specs, and having each chip governed by some attempt at precise equal-resistance-but-independent clock control will have all individual clocks synced.  The possibilities for heterodyning strike me as way too risky.  Keep in mind that while the individual clocks may not be within the audio spectrum, difference products of very close digital clocks may well be.  And of course, if one moves into 3 or more cascaded chips, the prospects for difference products being audible increases substantially.  Moreover, even if same speed is attainable, syncing is not a given.

Or am I being too pessimistic? 

No,  you're being entirely too realistic. Exactly those problems happen. I ... um... have some experience with that.   

I did get a four-PT2399 chain running in synch, with a little effort. Got the fabled tap-tempo happening too, which was actually the easier part of the whole thing. Discarded it in favor of another cheaper and less technically demanding approach. The PT2399 is a good compromise for the task it does; get too far outside the set of assumptions and the best-compromise point shifts on you. It's another illustration of creeping featurism pushing a design outside the optimum set of parameters.

[Mark Hammer]

I guess we need to remember that BBDs, though limited with respect to some aspects of their performance, are essentially open-ended devices, where something like a PT2399 is explicitly a special-purpose chip and is really more of a closed system.  Maybe not a one-trick pony, but a pony nonetheless with a finite repertoire of tricks.

If they weren't so tantalizingly cheap, it probably wouldn't bug us quite so much.

[Seven64]

at less than 50 cents a pop someone should just see if they can chain 5 or so of them togeather on a breadboard and see what happens 

[~arph]

There is allready two people in this thread that did four...

[merlinb]

Maybe the answer for multiple PT's would be to have a VCR on each chip, like the BC560 in Echobase for example, with all the VCR's controlled by a single pot.

A much easier method would be to use a current mirror. Any number of PTs could then be controlled from a single source (hfe matching helpful but not essential). Still an ugly way to do things since the audio gets A-D converted every time it goes through a chip  

Discarded it in favor of another cheaper and less technically demanding approach

What approach? A big BBD?.. PIC?..

[Mark Hammer]

The PT-80 project from Scott Swartz also uses a PT2399, but includes companding, which the rebote does not.  Does anybody have any experience with what companding can do for that zipper noise at longest delays?  Maybe we're just expecting too much from a chip that only uses simple lowpass filters in its "defense of the signal".

[R.G.]

A much easier method would be to use a current mirror. Any number of PTs could then be controlled from a single source (hfe matching helpful but not essential). Still an ugly way to do things since the audio gets A-D converted every time it goes through a chip

Yep. This is yet another way to get quantization noise. It helps a lot if the clocks are running in lockstep so the sample is at least at the same place on each recovered waveform. Otherwise you get heterodyning as a consequence of there the clocks sample the previous delay's reconstituted waveform; the only cure for that is to lowpass heavily between each delay, and soon you have only very dark mush left.

[Seven64]

There is allready two people in this thread that did four...

PWND.  i guess i didn't understand that for some reason.  ill go back to my corner now.

[slacker]

Any chance of a sound clip? I'm having a hard time figuring out what sort of noise is being talked about, and whether it's just what's normal for the PT2399 or something else.

Alternatively does this clip suffer from what's being talked about, this is about as clean as I managed to get a PT2399 at longer times. http://www.eskimo.plus.com/fxstuff/13452.mp3 This is with a 47k pot giving about 700ms and makes a perfectly usable pedal, although I guess that depends on what you're trying to achieve.

[Mark Hammer]

No, that's considerably cleaner than what I heard on Alain Parent's built.  Although to be fair, those "punched" chords avoid the worst-case scenario where the strings are decaying and the quantization noise exceeds the string signal.