Vintage Synth Repair

>
> I have a Harmonizer displaying a symptom which I reckon is specific enough to diagnose the component failing, so here goes.
> 
> The Harmonizer model in question has 2 main circuit paths - delay and pitch shift. The delay side of things works fine - it's based around a bank of RAM chips and everything is good.
> 
> The pitch shift uses 6 FIFO chips - 40105 - low capacity memory where the data can be clocked in and out at different rates. I believe it gets written at 41kHz and read at anything for 10kHz to >82kHz - giving the pitch shift.
> 
> Something has gone in the pitch shift circuitry and I'm getting a very noisy output - sounds like digital noise to me. Here's the thing though - as I crank the pitch up, the noise gradually disappears, until the point where it hits 2.000 times the pitch, when the noise disappears completely.
> 
> I've put my money (literally) on the fact that one of the FIFO chips has gone, and have ordered 6 replacements at GBP1.50 each. The thing that's bugging me though is I can't figure out why it's doing what it's doing - at twice the readout speed it will have to be going twice through the buffer so I would have thought the problem would get worse. If it became noise free at half the speed I could understand it, as it would have to be ditching some data, and so if it was ignoring the data from a duff chip then I could understand it cleaning up the signal.
> 
> Another possibility is that the chips are wired in parallel (they are only 4 bit) and I've lost the 4 least significant bits on 1 of the 2 fifo queues - I know they're wired as 2 parallel queues, and maybe each bank of 3 chips are themselves wired parallel to give 12 bits? The manual talks about an algorithm for reading from alternate queues to improve sound quality. At double speed maybe it's only reading from the 1 queue? This is complete guesswork.
> 
> Anyone have any ideas?
> 
> If the new chips fix the problem then this is all academic, but if they don't fix the problem then I'll be very interested in opinions.
> 
> Susie
>

>
> I have a Harmonizer displaying a symptom which I reckon is specific enough to diagnose the component failing, so here goes.
> 
> The Harmonizer model in question has 2 main circuit paths - delay and pitch shift. The delay side of things works fine - it's based around a bank of RAM chips and everything is good.
> 
> The pitch shift uses 6 FIFO chips - 40105 - low capacity memory where the data can be clocked in and out at different rates. I believe it gets written at 41kHz and read at anything for 10kHz to >82kHz - giving the pitch shift.
> 
> Something has gone in the pitch shift circuitry and I'm getting a very noisy output - sounds like digital noise to me. Here's the thing though - as I crank the pitch up, the noise gradually disappears, until the point where it hits 2.000 times the pitch, when the noise disappears completely.
> 
> I've put my money (literally) on the fact that one of the FIFO chips has gone, and have ordered 6 replacements at GBP1.50 each. The thing that's bugging me though is I can't figure out why it's doing what it's doing - at twice the readout speed it will have to be going twice through the buffer so I would have thought the problem would get worse. If it became noise free at half the speed I could understand it, as it would have to be ditching some data, and so if it was ignoring the data from a duff chip then I could understand it cleaning up the signal.
> 
> Another possibility is that the chips are wired in parallel (they are only 4 bit) and I've lost the 4 least significant bits on 1 of the 2 fifo queues - I know they're wired as 2 parallel queues, and maybe each bank of 3 chips are themselves wired parallel to give 12 bits? The manual talks about an algorithm for reading from alternate queues to improve sound quality. At double speed maybe it's only reading from the 1 queue? This is complete guesswork.
> 
> Anyone have any ideas?
> 
> If the new chips fix the problem then this is all academic, but if they don't fix the problem then I'll be very interested in opinions.
> 
> Susie
>

>
> does the noise go up in pitch as it disappears? sounds to me like a device that works this way would have tracking filters to remove clock noise from its output at the lower readout rates; I bet they're gone awry. they're possibly just analogue VCFs driven from a voltage derived from the clock speed....
> 
> duncan. 
> 
> --- In vintagesynthrepair@yahoogroups.com, "Susie" <gardens@> wrote:
> >
> > I have a Harmonizer displaying a symptom which I reckon is specific enough to diagnose the component failing, so here goes.
> > 
> > The Harmonizer model in question has 2 main circuit paths - delay and pitch shift. The delay side of things works fine - it's based around a bank of RAM chips and everything is good.
> > 
> > The pitch shift uses 6 FIFO chips - 40105 - low capacity memory where the data can be clocked in and out at different rates. I believe it gets written at 41kHz and read at anything for 10kHz to >82kHz - giving the pitch shift.
> > 
> > Something has gone in the pitch shift circuitry and I'm getting a very noisy output - sounds like digital noise to me. Here's the thing though - as I crank the pitch up, the noise gradually disappears, until the point where it hits 2.000 times the pitch, when the noise disappears completely.
> > 
> > I've put my money (literally) on the fact that one of the FIFO chips has gone, and have ordered 6 replacements at GBP1.50 each. The thing that's bugging me though is I can't figure out why it's doing what it's doing - at twice the readout speed it will have to be going twice through the buffer so I would have thought the problem would get worse. If it became noise free at half the speed I could understand it, as it would have to be ditching some data, and so if it was ignoring the data from a duff chip then I could understand it cleaning up the signal.
> > 
> > Another possibility is that the chips are wired in parallel (they are only 4 bit) and I've lost the 4 least significant bits on 1 of the 2 fifo queues - I know they're wired as 2 parallel queues, and maybe each bank of 3 chips are themselves wired parallel to give 12 bits? The manual talks about an algorithm for reading from alternate queues to improve sound quality. At double speed maybe it's only reading from the 1 queue? This is complete guesswork.
> > 
> > Anyone have any ideas?
> > 
> > If the new chips fix the problem then this is all academic, but if they don't fix the problem then I'll be very interested in opinions.
> > 
> > Susie
> >
>

>
> hmm...
> The behaviour tonight was somewhat different. The noise no longer disappears, but does go up in pitch. I traced the signals coming in to the fifo array and all the clock signals are clean, however there appears to be a lot of noise coming in from somewhere on the various lines that enable the fifo circuitry - I'm still trying to figure out where the noise starts.
> 
> Would I be right in thinking that if the input of a nand gate is noisy then the source of the noise is on the input side and not the output side? ie is it possible for noise to leak backwards through the output of a nand gate to the input? I grounded the inputs and the noise instantly disappeared, but maybe all I was doing was disabling the fifo array.
>

>
> I need to give this a rethink as I'm starting to understand a bit more on how the fifo array works, and I think the noise I'm seeing is of a low amplitude and not the same noise I'm hearing on the output. Rather than 2 fifo arrays, the 6 chips are connected as a single 12 bit 32 word fifo containing 2 sets of data multiplexed down the single fifo. The queue is only long enough to hold a fraction of a waveform, so I think a lot of the logic of 'filling in the gaps' is handled by the cpu feeding the data into the array. The output of the fifo goes straight to the DAC, which I know works as I swapped it with the delay line DAC. The output of the DAC is then 2 multiplexed audio signals, and its the demultiplexing of this signal which is now receiving my attention.
> 
> btw, strictly speaking, this thread is off topic, as it's not a keyboard instrument, but as you can play the harmonizer with a keyboard CV then I'm hoping you'll forgive me.
> Susie 
> 
> --- In vintagesynthrepair@yahoogroups.com, "Susie" <gardens@> wrote:
> >
> > hmm...
> > The behaviour tonight was somewhat different. The noise no longer disappears, but does go up in pitch. I traced the signals coming in to the fifo array and all the clock signals are clean, however there appears to be a lot of noise coming in from somewhere on the various lines that enable the fifo circuitry - I'm still trying to figure out where the noise starts.
> > 
> > Would I be right in thinking that if the input of a nand gate is noisy then the source of the noise is on the input side and not the output side? ie is it possible for noise to leak backwards through the output of a nand gate to the input? I grounded the inputs and the noise instantly disappeared, but maybe all I was doing was disabling the fifo array.
> >
>

>
> I think I said I'd put my money on a duff fifo chip - I win! I'd lost the middle 4 bits of the 11 bit word, which I realised last night when I was looking at the demultiplexed audio and seeing it stratified on the scope - a sure sign of missing bits.
> 
> Yet another cheap buy fixed up. That's a Linndrum and a Harmonizer resurrected in the same week. Next up a Binson Echorec - real old school! 
> 
> --- In vintagesynthrepair@yahoogroups.com, "Susie" <gardens@> wrote:
> >
> > I need to give this a rethink as I'm starting to understand a bit more on how the fifo array works, and I think the noise I'm seeing is of a low amplitude and not the same noise I'm hearing on the output. Rather than 2 fifo arrays, the 6 chips are connected as a single 12 bit 32 word fifo containing 2 sets of data multiplexed down the single fifo. The queue is only long enough to hold a fraction of a waveform, so I think a lot of the logic of 'filling in the gaps' is handled by the cpu feeding the data into the array. The output of the fifo goes straight to the DAC, which I know works as I swapped it with the delay line DAC. The output of the DAC is then 2 multiplexed audio signals, and its the demultiplexing of this signal which is now receiving my attention.
> > 
> > btw, strictly speaking, this thread is off topic, as it's not a keyboard instrument, but as you can play the harmonizer with a keyboard CV then I'm hoping you'll forgive me.
> > Susie 
> > 
> > --- In vintagesynthrepair@yahoogroups.com, "Susie" <gardens@> wrote:
> > >
> > > hmm...
> > > The behaviour tonight was somewhat different. The noise no longer disappears, but does go up in pitch. I traced the signals coming in to the fifo array and all the clock signals are clean, however there appears to be a lot of noise coming in from somewhere on the various lines that enable the fifo circuitry - I'm still trying to figure out where the noise starts.
> > > 
> > > Would I be right in thinking that if the input of a nand gate is noisy then the source of the noise is on the input side and not the output side? ie is it possible for noise to leak backwards through the output of a nand gate to the input? I grounded the inputs and the noise instantly disappeared, but maybe all I was doing was disabling the fifo array.
> > >
> >
>

>
> nice work, & thanks for the update.
> 
> see if you can get the binson sorted in time for the 5th anniversary of syd's death next month, & then play some slide guitar through it with a zippo. that'd be a nice tribute.... :-)
> 
> d.