Decimator/bit crusher effect

[Dr.EM]

I'm trying a simplified version on breadboard, just the free running ADC connected up as shown with the op-amp input and DAC output. The op-amp uses a higher supply and 1/2 supply is used as artificial ground. It's not going very well, here are my observations:

When I switch it on, there is a lot of noise on the power rails (supplied by 317 regs). Around 2v-pp. It is high frequency and is present even if I remove power to the ADC, so is somehow caused by the op-amp? It has a bypass cap.

The same noise exists at the output. It becomes much larger when I connect my mixer to the output.

The ADC outputs seem to do what they should, though at the wrong times. With no sound input, they have a lot of activity (probably because the noise which is also present at its input?). They do change with a signal applied, I have attached the noise at the output followed by the noise with an input signal and you should be able to hear it a little.

The input signal is audable at the output even with the ADC outputs disconnected!? It is quite clear if I attach the ADC analogue ground to the 1/2 supply artificial ground. It is less clear with it at actual ground, but I can hear more of the digital signal (as in the attached mp3). I really don't know where it should go, why are there seperate analogue and digital grounds? How can they possibly be seperate? I think this is the cause of the problems.

Vref2 is unconnected. Should it go somewhere?

Any ideas?

 

[Nigel Goodwin]

When I switch it on, there is a lot of noise on the power rails (supplied by 317 regs). Around 2v-pp. It is high frequency and is present even if I remove power to the ADC, so is somehow caused by the op-amp? It has a bypass cap.

Sounds like the 317 regs are oscillating?, what capacitors do you have round them?.

 

[Dr.EM]

I followed this:

http://www.techlib.com/electronics/regulators.html

But actually used 4.7uF for the outputs as they were convenient and they share a 0.1uF (close together). I haven't checked properly, but think they were fine without the circuitry connected. Its somehow making them unstable.

Just looking through that ADC datasheet again it looks like vref2 should go somewhere, to 1/2 the supply in my case? Still don't understand this seperate analogue and digital grounds, but it seems to make a point of them. Are they supposed to be totally isolated, or is this more to do with the grounding topology? I guess I could use the circuit at the bottom of that page to isolate them if that is what it wants, but I can't see how that would work?

 

[Nigel Goodwin]

If you've got 2V ripple on the supply lines there's something SERIOUSLY wrong, and it's most likely the regulators oscillating - opamps wouldn't take enough current to drop the supply two volts. Have you checked what's going in the regulators?.

 

[Dr.EM]

Ok, I've checked it again and they appear fine now!? However, the artificial ground has a high frequency sine wave of about 1vpp imposed on it (perhaps I was checking the regs outputs in relation to this rather than actual 0v, causing the off readings. I've got 4 voltages floating around and its kind of messy). At its input, via the 2 10k resistors, its fine. The noise is only after the buffer, so is something to do with the op-amp. It is there even with the ADC power removed. It would make sense that this is upsetting the circuit as it needs to be stable. I'm assuming its the op-amp oscillating, but don't know how to stop it.

EDIT - when I remove its output connection from all other points in the circuit it is fine. Something must be loading it wierdly, and since it does it without the ADC power connected, it must be itself (the 2.5v reference for the summing amp, output buffer and input amp)

EDIT Again - I fixed the vref2 to 1/2 supply (the currently noisy line) and analogue ground to actual 0v. I'm getting some very promising patterns out of the ADC by just connecting the R2 ladder directly to my scope. Thing is, they dissapear when I connect them to the op-amp summing amp instead . At the output of that I get a lot of noise and a faint signal (as per the mp3)

---Don't worry. I'm trying it with normal 084 op-amps and the "using +/-10v input" diagram from the datasheet. Having much better luck already . Will just require a more obscure supply, but anyhow.

 

[Dr.EM]

Ok, well it seems to working as expected. I have the /WR fed by a 555 with HPF and schmitt for the short pulses. I wanted to use the LM331, but it isn't doing what I expected. I assumed it would be a supply voltage level square wave, but i'm getting like a 100mv fuzzy wave whose amplitude changes over the range too :/ . Not very good, I don't think its right at all since it says "logic output".

Anyhow, it does work with the 555 system. It goes a bit haywire every so often, I think it starts picking up mains radiation since I have a long wire from the pulses source to the ADC as they are on seperate breadboards. At least I hope thats all it is, it goes back to normal when switched off then on. Perhaps /INTR should go somewhere even with /WR fed by a seperate source?

Removing the LSBs does quite little, so I remove the ones from the 250k upwards. That gives the effect I expected/wanted more. Just need to do it with the logic switched instead of pulling wires out .

I have attached a short demo (I hope it's ok to do this, its well under the limit?). This would work much better MIDI controlled, so an accurate sample rate can be moved directly to, rather than sweeping to it via a pot (currently a trimmer!) and guessing the pitches. First you hear bit rates changing in drums, then bit depth in drums, then some bit rate with a keyboard (needs to be midi controlled for this in particular). Only have any hope of getting MIDI control if I can get that LM331 working (then use MIDI-CV convertor). Anyone used one before? Thanks for the help with this people, I probably will build it into a unit, though it isn't quite the tone I wanted

EDIT - Ok, it won't attach.

Hmm, a nice idea here. Mabye I could use the audio out of a keyboard, clip it into a square wave and put it through the pulse generator. Then you could just play the decimator with the normal audio out of the keyboard