ggg's pt-80 compandor

[jbm222]

i'm going to build a pt-80 delay from generalguitargadgets.com, but i forgot to order the compander chip.  Will the pedal work without it?  I know I'd have to probably add another opamp or two (i've got plenty of 4558s and 072s laying around), but how important is the compression/expansion to the over all sound of the pedal?  If it's extremely important, can a build a simple compandor using common components (meaning no OTAs or LED/LDR kind of stuff)?

[Mark Hammer]

Other projects using digital delay chips or BBDs but not a compandor, have been around for a while.  In principle, it is possible to make a delay-based pedal without companding.  Plenty of companies have done it.  The problems you will face in the absence of companding, however, are:

a) The need to determine what the optimal input level is.  Having a compressor in the front makes it pretty much a no-brainer, since the compressor assures that even large transients won't exceed what the delay chip can handle (or at least not often, and not by much).

b) The need to use more filtering or aim for lower bandwidth.  The nice thing about companding is that the noise emanating from the delay chip is reduced enough that one can often tolerate a somewhat higher corner frequency.  That's not to say that greater bandwidth always makes the best-sounding delay, but you can always choose to trim the treble if you have it - you can't bring it back if you've gotten rid of it already.

Try and score yourself a 570 or 571 if you can (NE or SA prefix).  It will simplify life considerably.  You CAN make your own compandor, but the 570/71, along with Scott's component choices, will circumvent having to match the amount of re-expansion to the amount of compression, and a host of other things.  This is one of those cases where it just isn't worth it to roll your own.

[jbm222]

ok thanks for the advice.  I think i'm going to throw it on a breadboard tommorrow without the companding, and if it doesn't sound as good as I want i'll go ahead and order a compander.

I was also wandering... in the future, i might want to add another delay chip for longer delay times.  If i do this, would it work ok just to set up another chip right after the first one, fixed at max delay time and without the lowpass filtering, and switch it in and out of the circuit for long/short delays?

OR.... if i decide to do this, would it be better to get the other chip now and work out a design with a dual pot or something that simultaneously changes the time on both chips?

I realize now that i probably should've gone with a design that uses external memory and clock so that I could just add RAM to get longer delays, but I already have most components for this delay now, plus the sound samples i've heard sound really awesome... way better than the digital delay on the processor i sold.

[Mark Hammer]

Well now that scenario makes for an interesting discussion.

The PT2399 has its own onboard clock that is adjusted via a single resistance value on pin 6.  Because the clock is internal, that means that two cascaded chips can't share the same clock/frequency.  I suppose in a perfect world they might, but few dual-ganged pots will be that perfect.  In contrast, one of the advantages of analog chips (BBD) is that they can all be ganged to the same master clock.

Why is the one better than the other?  As Ton Barmentloo (Puretube) has noted on multiple occasions, multiple clocks on the same board run the risk of heterodyning.  "What does my sexual orientation have to do with this?" you ask.  Nothing.  It's about having the clocks on your board behave like renegade ring modulators and forcing their way into the audio output.  As you probably know, ring modulators produce the sum and difference of two signals.  When it comes to high frequency clocks, the audibility of those products may well be minimal, but still needs to be thought about in advance to make that determination, because sometimes they CAN be a problem.

For example, if one chip is being clocked at, let's say, 150khz, and the other is being clocked at 154khz, their sum (304khz) is obviously well out of human (and likely dog) hearing range.  But their difference, ahhhh, now THERE'S the rub - 4khz.  That's gonna be really irritating to hear, and difficult to suppress without having to simultaneously take out some of the treble you hope to retain.

You suggested having one set to max delay.  In the datasheet for the PT2399, it indicates the internal clock at that delay time is set to about 2mhz.  If the other chip is set to less delay, then clearly the sum and difference of a 2mhz and 3mhz clock are not going to be audible.  As you approach the max delay on the adjustable chip, however, there is risk of the two clocks getting close enough that a difference signal may become audible.  Besides, even though the master clock on each chip is pretty darn high, as digital chips, things get divided down to provide coordination of the various proccedures required for sampling, shifting the samples along the RAM, etc.  The real villain may not be the master clocks, but rather those divided-down clock signals that DO start to have values where sum and difference can create audio demons.

As Ton has also pointed out, it's not like the heterodyning can't be successfully managed - after all there are plenty of audiophile multi-clock systems out there to prove otherwise - but if you're designing your own board from scratch, there is clearly a challenge ahead of you.  One which I doubt either you or I are up to.

The suggestion for changing to the RAM-assisted big brother - the PT2395 - is probably the smart way to go, not only because the larger RAM will yield a potential 800msec of delay, but because there is the one clock to be concerned with.  It will still benefit from companding, but it won't need much MORE than companding and the same degree of lowpass filtering as the PT-80.  

If it matters, I threw together Dean Hazelwanter's old Holtek HT8955A-based echo unit and it worked fine.  The HT8955A is a now-obsolete chip that also provides up to 800msec delay with an external 256k RAM chip, although Dean's circuit/layout for it is not as purposeful in reducing noise as Scott's is for the PT-80.  The noise level without companding or extensive filtering is...okay...but not at all anything you could gig with professionally.  Being able to squeeze out another half second of delay time, though, is a real blessing and allows you to enter another realm of thinking about use of delays.

[jbm222]

wow, now that does make for an interesting discusion.  I never thought of that problem with multi-clock systems.  I don't know if would actually be a problem with this circuit though.

The data sheet only lists info for frequencies down to 2.0M, but says that resistor values up to 50k can be used.  The pt-80 uses some value close to this to get a 600ms delay.  I'm not sure what the clock freq would be here.  For just a quick and dirty extrapolation, lets say it's about 1.5M.  I could make that be the delay time for the fixed delay chip.  I could make that 1k in series with the 50k pot a trim pot, and adjust it so that at max delay time, there are no audible artifacts.  All it would take would be a difference of about .01M or 10k for it to be almost completely cut by the filtering.  make it 20k and theres no way anyone is going to here it.

So I should be able to run two chips simultaneously with a 600ms delay on one and an ALMOST 600ms delay on the other, giving me almost 1.2s.

But the thing is... at 600ms, this THD becomes noticible.  Noticible distortion, into another chip, which at max delay settings would also produce noticible distortion may lead to intolerable distortion.  This is why I thought maybe a dual pot would be better, because at lower settings--say up to 800-900ms, the distortion would not be nearly as noticible.  Taking your point into consideration, I could probably make sure there was a big enough difference in the fixed series resistance that it never became an issue.  Dual pot set to zero, and a 1k and a 1.1k or similar values should do the trick.

However, assuming I could get one of these two methods to work I would have to redesign the filters since the design Scott uses the pt2399's built in opamps for some of the filtering, both pre and post.  He put a lot of work into carefully designing these filters.


The switching is probably not an issue with either design... I might end up with some popping and other random transients, but typically I wouldn't switch delay times mid solo or anything so it shouldn't be an issue.  Either way I could probably get way with just a SPDT toggle, but would use a DPDT just to be safe since they only cost about 10c more.


I think what I'm probably going to do though, is just build a pt80 to specs and sell it on ebay or something if I'm not happy with it.