[sdiy] Mixed-signal problems (dsPIC digital delay project)
yo at vacoloco.net
yo at vacoloco.net
Thu Jan 26 17:04:05 CET 2017
I have to admit I've never been a fan of star grounds, principally
because you end up making the return paths long and overly complex so
you get an increase in ground resistance and hence a larger differential
between the two, and often noise too.
I assume this is a two layer board? If you can, I'd suggest rolling a
four layer board, not much more cost but *SO* much better for noise. Top
layer is signal, 2nd layer is a solid common ground, 3rd layer is power,
again using big tracks/planes and the lower layer is signal again. I'd
also ground fill top and bottom layer.
You might also look at adding some inductors before the grounds hit the
common 'star' earth to try and stop any noise coming along the ground
that way. Similarly look at inductors on the input to the regulators
(REG1 and REG2), and even go as far as adding one for each of the rails
that supplies the opamp (in this case the ground here is effectively
your -ve supply rail.
Not sure that helps.
On 2017-01-26 15:34, Tom Wiltshire wrote:
> Hi All,
> I'd like some advice. I recently did a PCB for my DigiDelay project:
> This is a commercial project in as much as I'm selling chips and PCBs.
> If my asking for help for that bothers you, please stop reading here
> and accept my apologies.
> The pedal is good, but it has more hiss than I'd like. A certain
> amount is going to be inevitable because the circuit uses the dsPIC's
> on-chip 12-bit ADC for audio, and the delay processing is only 16-bit.
> But that's not the problem I'm seeing.
> The problem I've got is high frequency noise on the 9V power supply
> rails. This looks like spikes at the 64KHz sample rate (the audio is
> sampled every other sample, the interleaved samples are used for the
> control pots). The spikes are causing an oscillation at around 2.7MHz.
> This comes out as basic white noise on the audio.
> There's an image of the top of the PCB here:
> I thought I'd done as much as I could to reduce interference between
> the digital and the analog sides of the circuit: The PCB divides the
> two areas completely (analog on the left, digital on the right), both
> have separate ground planes. The grounds only meet at a single point
> back at the power supply input. The dsPIC's analog and digital 3.3V
> supplies are generated by separate regulators. This is all supposed to
> help, right?!
> It seems to me there must be some simple thing I can do to remove such
> a high frequency. The audio from the dsPIC is already heavily filtered
> once it's back in the analog world. This helps (there's less noise
> after the filters than before them), but it's not eliminating it, I
> guess since the noise is getting in through the op-amps' power pins.
> Any pointers or things I could try would be appreciated.
> I'd like to get this properly sorted since I'd like to do a rack
> version with delay time modulation like the old Ibanex DM2000.
> Synth-diy mailing list
> Synth-diy at synth-diy.org
More information about the Synth-diy