digital reverb or delay

[johnnyx]

EPE magazine had a project called the "polywotsit", which used a PIC with ADC built in and an external DAC. Several effects are switch-selectable, all implemented in firmware. It uses 8bit converters, not state-of-the-art but probably equivalent to a bucket brigade chip. You might get some ideas from it, link is here

edit;- BTW it's December 2001 edition.

[Devil_H@ck]

ingrast, I'm very interested, please post more info when you have it .

[ingrast]

I will make an initial design in the weekend and post a condensed schematic.

It should be fine to have more input from others in the forum regarding desirable / undesirable features before proceeding to prototyping.

Suggestions welcome.

Rodolfo

[Devil_H@ck]

Well, I know nothing of DSPs, so I don't really know what's possible with them. Maybe it's possible to make this an all-in effects-box?

To solve your lots-of-buttons problem, instead of hooking up one button to one I/O pin, make a resistor chain out of them and use the DSPs ADC to check which button was pressed (asuming a DSP has got an ACD of course).

Here's the idea. You could change the voltage drop to .2 or .1 volt per button to allow up to 49 switches on one I/O pin.

[ingrast]

Here is a preliminary draft design for a hardware based digital reverb processor.
The basic structure is a dual delay chain based on 24 bit A/D converters (18 bits effectively used), 4096 and 1024 x 18 bit FIFO memories, and 24 bit D/A converters (18 bits effectively used).
Delay length is continuously selectable from a variable 3:1 ratio sampling clock generator, and a switch selectable 1:4 FIFO clock (may be jumper selectable for 1:2 ratio). The sampling / shift clock may be further modulated in depth and rate by a low frequency oscillator to generate chorus effects.

The first stage delays the incoming signal from 10 to 32 ms or 64 to 128 ms adjustable as indicated. This signal may be mixed at a selectable level with the "dry" input to generate a "slap-back" effect.

The signal delayed by the first stage is further delayed by a second one (5 to 15 ms) in a feedback loop. Delay varies concurrently with the first stage, since the stereo A/D and D/A converters of which one channel is devoted to each delay stage, have a common sampling clock. This stage in effect adds multiple copies of the incoming signal at multiples of the loop delay and with decaying amplitude. Loop gain is adjustable so as to shorten - lengthen the decay rate. This signal is in turn mixed in an adjustable level with the "dry" and "slap-back" signals to generate the output.

As it usually turns out, the design ended a bit more involved than foreseen, so it will take some extra days to arrive to a workable schematic. The selected low cost A/D and D/A converters generate / expect multiplexed serial digital audio streams that must be deserialized / serialized for writing and reading the FIFO buffers. These chips also are basically conceived for programmability by an external controller, though they default to usable configurations upon reset. So it is necessary to juggle a little with these configurations and clocking - data routing to arrive to a workable design.

Will keep posting details as work progress, and of course will hear any suggestions that may arise.

Rodolfo

[ingrast]

Originally posted by Devil_H@ck
Well, I know nothing of DSPs, so I don't really know what's possible with them. Maybe it's possible to make this an all-in effects-box?

To solve your lots-of-buttons problem, instead of hooking up one button to one I/O pin, make a resistor chain out of them and use the DSPs ADC to check which button was pressed (asuming a DSP has got an ACD of course).

Here's the idea. You could change the voltage drop to .2 or .1 volt per button to allow up to 49 switches on one I/O pin.

Yes, the idea is a full DSP implementation allows for essentially unlimited effects, being the processor speed the only constraint. This includes stereo panning, ping-pong, chorus, reverb, tremolo, flanger etc. etc.

With respect to the multiple switch interfacing solution, it is a clever and compact one with the only limitation it works for only a switch at a time. The one closer to the A/D port effectively blinds out the others. When I mentioned it was easier to interface switches and knobs to hardware based projects, I meant they do their work directly on the circuit, whereas a software approach requires scanning / decoding and processing, all of wich is a chore and takes processor resources.

Rodolfo

[ingrast]

Originally posted by johnnyx
EPE magazine had a project called the "polywotsit", which used a PIC with ADC built in and an external DAC. Several effects are switch-selectable, all implemented in firmware. It uses 8bit converters, not state-of-the-art but probably equivalent to a bucket brigade chip. You might get some ideas from it, link is here

edit;- BTW it's December 2001 edition.

I could not locate the project in the link, but nonetheless you are right in the appreciation. 8 bits are probably rather coarse in S/N for an instrument level effects processor.

Rodolfo

[ingrast]

Originally posted by davesaudio
Yes I assume it (pt2399) is obsolete or obscure

I mention it because it just seems so simple.
( I have no exerience with it )

It and it's obscure BBD chips seem to be available here
http://www.smallbearelec.com/Ordering/ICsDelay.htm

I eagerly await a more substantial design- as yours appears to be


Dave

I was aware of the site, as you may have noticed the chips are NOS (New Old Stock) meaning they are discontinued.

[davesaudio]

All I could find on the polywhatsit
http://homepages.nildram.co.uk/%7Est...olywhatsit.zip
source code only?
anybody got a schematic?

[johnnyx]

I think you have to buy a reprint.. but no matter, I have the article and can scan it. The software (maybe should be firmware) can be downloaded free from the site, as you have seen. The PCBs (1 analogue & 1 digital) are available for only £7.61 for the pair according to the magazine with the article.

If anyone is interested, send me an e-mail from the link on my user info. There is no way to include attachments there, but it will generate a normal e-mail that will accept attachments in a reply. It's fairly large, 8 sides of A4, and includes PCB patterns. I'll see if I can send it uncompressed.

I have two TDA1097, made by Mullard, 1536 stage BBD chips that I never got round to using. They are pre-production prototypes. I have the datasheet too, and I have two TDA1022. I experimented with the latter but found them noisy, almost certainly due to the spaghetti layout I used. My layouts are better now. If you want to experiment with them, send your address in the e-mail. I have no time for it now.