Repacing PCM1702 DAC with PCM1794 for old piano?

[frmdstryr]

A family member brought me a Yamaha CMVP-207 piano that was not working to try and repair. It would play sound from the audio in port but nothing from the piano keys. The full schematic can be found in the service manual but I've attached the relevant part below. 

I traced the issue down to the two PCM1702U DAC's which are no longer made. I looked on ebay and found some... but I think they are fake because while it actually played sound from the keys (unlike before), the audio quality was horrible. When debugging I noticed the -5V regulator was overloaded (the -5V net would only reach -0.72V) making the pcb hot. I couldn't find any shorts so I removed the two ebay chips and all the digital signals and +5V and -5V are present.

After some l searching for alternatives, I found https://www.ti.com/product/PCM1794A which I think maybe could be used with a small converter pcb that adds a 3.3v regulator, configures the PCM1794A via hardware pins, and re-routes the traces as needed. Would this work?

If the PCM1794A is put in MONO mode with Standard 24-bit format (shown in section 8.1.3 in the PCM1784 datasheet) will it work with the same digital signals the PCM1702 used?

I'm also confused how the different analog outputs would be converted. The PCM1702 takes +5 and -5 inputs with one L output, while the PCM1794A has a L+ and L- output but only takes a 5+ supply.  Could I just cut the AGND connection on the "+" of the output Op amp (UPC4570G2) then run the  L+ and L- of the PCM1794A and to the "-" and "+" of the op amp?

Is there another part (or approach) that would work better or can the PCM1794A be used? Any help would be appreciated!

[frmdstryr]

Not that anyone cares, but I got the PCM1794 to work by simply shifting the WCLK/LRCK signal by 4 bits so the MSB of the 20 bit data signal aligns with the start that the Standard 24bit interface expects.

[MathWizard]

Well it's good you got it working. But didn't the 1702 output positive and negative voltages, but the 1704 only outputs positive voltages? So how does that end up working the same.

I haven't tinkered with digital stuff for a while (in fact I've taken a break from all physical electroncis lately, but I've been doing some book work at least) how did you shift the input by 4 bits? I don't see a bunch of input pins, so you can't just move the input signals up or down to different pins.

[ledtester]

I haven't tinkered with digital stuff for a while (in fact I've taken a break from all physical electroncis lately, but I've been doing some book work at least) how did you shift the input by 4 bits? I don't see a bunch of input pins, so you can't just move the input signals up or down to different pins.

Let me see if I've got this...

The data enters the 1702 serially and it works a lot like a 595 shift register... there are CLOCK, DATA and LATCH_ENABLE signals.

Since the 1702 is a 20 bit device the data stream being sent to it is a sequence of 20 bits words -- it sees 20 bits of clocked in data followed by the LATCH_ENABLE signal and this sequence repeats over and over again.

However, the 1794 has a 24bit internal shift register. By delaying the LATCH_ENABLE by 4 clocks the 1794 will see a 24 bit value in which the the upper 20 bits matches the 20 bit word that was sent to the 1702 and the least significant 4 bits will be the most significant 4 bits of the next word. However, since they are LSB bits they won't affect the output that much.

[MathWizard]

Yeah I'm not that familiar with ADC/DAC's in real life, and less so with DAC's. OK one of these days I'll have to try some of it in LTS with R2R networks, and maybe try what you talk about. But yeah I'll remember such things are possible, if I ever run into such a problem.

The last things I was working on with DAC's was a guitar FX processor. I scoped some of the 16-bit words, but I never got around to decoding them or trying it with my logic analyzer.