Anyone familiar with RME sound cards (repair)?

[living sounds]

I've got a 9632 HDSP that I bought used which had a problem with the analog output. Doesn't matter to me since all I need is the AES input, but unfortunately that stopped working and so did the S/PDIF input, too. I might have accidentally - repairing another converter connected to it - sent too much voltage (something like 24V AC) into it. I replaced 4 capacitors with new ones (the slightly too big panasonic FC ones on the left), two of those looked a little damaged when I bought it. No luck, it didn't cure the problem on the output (noise, sometimes) or the AES input. The ADAT input is working fine, but the card shows nothing coming in from either AES/EBU or S/PDIF.

Is there an obvious part to try and change? Could the small transformers be sufficiently damaged by a very short overvoltage? Sending it in for repairs would cost more than buying another used one, and the trouble is - of course - that there's no way to test it while it's running. RME won't give me schematics either, of course.

 

[living sounds]

Picture 2

 

[living sounds]

Nr. 3

 

[millzners]

I've smoked a lot of parts, here's what I usually start with:

Look with a bright light and a magnifying glass at each component on the board.  When SMD resistors burn up they almost always get a burned crater in them but it can be very subtle and look almost like a tiny indentation or "pin 1" indicator when you reflect light off of them.  Look over every resistor and look for this.

Next I'd look into the input buffers/line drivers on your SPDIF and AES, assuming that's what they use.  If a digital in/out connector gets too much voltage you can bet the first IC in line will smoke.  If nothing else locate them and look at them to see if they have a scorch mark.  If you have a scope or even a meter you could see if they pass signal.

So if I'm sitting there with a part that's fried and I can't find a resistor obviouly burned, and my IO buffers pass signal, I usually end up focusing on one thing that doesn't work like in your case the SPDIF input, applying a signal and following it until I get to the end.  If the end is a receiver or MPU or FPGA or something else with a ton of pins you're probably stuck, so move on.

Hopefully this helps or at least is food for thought.

 

[living sounds]

Thanks!

Nothing looks burned, and IF any overvoltage occured it would have been only for a few milliseconds max anyway. But sufficient to destroy an IC, probably. The incoming signal goes through an SMT cap, then through a small input transformer, a resistor and then into a ds26ls32ACM, a quad differential line receiver.

I have traced the signal with a scope up to the line receiver's input, it is not possible to go further without the card powered up. It's a PCI card, so testing it requires to mount it inside the PC, where room is very sparse, and I don't want to start probing around in the computer while it's running, it's too easy to damage other stuff.

After the line receiver the signal appears to go directly into a huge Spartan XC2S150 FPGA. It's got 208 pins and needs to be programmed.

So it appears my best hope is to replace the line receiver. Apart from the FPGA the only other thing to blame would be the power line feeding the line receiver, which is 5VDC, provided by the PCI bus and probably just filtered a little.

 

[living sounds]

Just checked, the +5V power line connection is fine. So it's replacing the line receiver.

 

[gyraf]

solder a test wire to the line receiver output to test it in-circuit?

 

[living sounds]

solder a test wire to the line receiver output to test it in-circuit?

Did that, the line driver works. Thinking about it some more (I had this card lying around unused for some months) it couldn't have been the problem anyway. The ADAT input works, but the coaxial S/PDIF input using the same socket doesn't. So the problem is somewhere in the part of the circuit processing incoming digital signals other than ADAT...


Is there an easy way to build an S/PDIF or AES ---> ADAT converter? ;-)

 

[MatthisD]

I have traced the signal with a scope up to the line receiver's input, it is not possible to go further without the card powered up. It's a PCI card, so testing it requires to mount it inside the PC, where room is very sparse, and I don't want to start probing around in the computer while it's running, it's too easy to damage other stuff.

Couldn't you use an old motherboard without a case, set up vertically with the PCI card sitting face up with some makeshift support to keep it from moving around?

 

[living sounds]

Couldn't you use an old motherboard without a case, set up vertically with the PCI card sitting face up with some makeshift support to keep it from moving around?

Haven't got one here. Unfortunately, I think this repair is not something I can do anyway. Without schematics, without precise test gear etc. I've checked the AES output, it has the same digital noise as the analog output, very loud and only audible when audio is played back. It goes on and on at 44,1 khz, but at 88,2 khz the whole playback stops after 2 seconds or so. Only the analog input and the ADAT input are working as they should. There is something wrong with the clocking circuitry I think.

The funny thing is that the problem with the noise was supposed to be there when I bought the card, but it worked fine here and it only started after a few weeks or so.

 

[MatthisD]

I've looked through the datasheets for the AD1852 and the eval board version. It now seems logical to me that your analog output and AES S/PDIF inputs are both problematic. The DAC looks like it takes the external digital ins from a digital input receiver (the ds26ls32ACM like you mentioned maybe) to its SDATA port (pin 27) and obviously the DAC handles analog outputs too. You said the analog output does function but with noise so the DAC chip must be receiving digital output signal but not digital input signal or else it is but the DAC is not working properly so it seems replacing it could be worth a go but the following seems more likely given the symptoms,

AD1852 eval board datasheet;
"Switch SPDIF/EXT (S3) controls CPLD (U4) and U11,
which is used to switch signals between the SPDIF input (J1)
and the direct input, via the 10-pin header J2, EXT DATA"

'EXT DATA' I presume is used in the 9632 for clock sync to another RME card. I doubt that you have it set like below but it would correlate with non-functional AES and digital noise as it would have no clock source.

from the 9632 manual:
"SPDIF In / Internal has to be selected in the Settings dialog. Additionally Pref. Sync
Ref has to be set to SPDIF In for this internal connection to work properly. Please note that
the external SPDIF- or AES input can no longer be used."

If your card is not set like above then the component for selecting ext.clock source could be faulty causing it to act as if clocking from another RME card and disabling AES input, see if you can find a part equivalent to 74HC00 as below. I can see a 14pin chip to the right but can't read the markings. In the RME manual, page 85 the block diagram shows the equivalent switching labelled "input select".

from the AD1852 eval board datasheet;
U11 (74HC00) is used to select the correct master clock
source for the DAC when it is switched between the SPDIF
receiver (U2) and the Ext Data Interface (J2). A discrete logic
gate is used for this function, because of the excessive jitter
that modulates high-frequency clock signals when they are
handled by PLDs.

 

[living sounds]

Thanks alot! This does make sense. I've got a replacement DAC chip somewhere, but I doubt it's the culprit because the noise is present on the digital output, too (also the headphone out, which just gets it from the DAC, I guess).

There are two logic gates CT4046AM and 74HCT04D.


Here are hi-res pictures of the 9632 I found on the web:

http://www.head-fi.org/t/566627/best-opamps-for-my-rme-9632-and-m-patch-2-mods-need-help

Looks like they switched to tantalum caps from the electrolytic SMT caps later, but it's otherwise identical.

Is it reasonable to assume that the 74HCTD4D does the digital input switching, the HCT4046AM the clocking/reclocking and the two ADG451BR switches the switching between hi/+4db/-10db etc. for analog input and output respectively?

I've ordered replacement 74HCT04D and will try those first.

 

[MatthisD]

I think that 74HCT04D must be equivalent to U11 in the eval board diagram.

ADG451BR could be used for the equivalent switches S1,2,3,4 in the eval board diagram.

"The ADG451/ADG452/ADG453 contain four independent,
single-pole/single-throw (SPST) switches"

That would be worth replacing before anything else maybe as it switches between AES/spdif in and EXT DATA (RME calling that; SPDIF In/Internal)

What would be interesting to do is to plug a second 9632 into the spdif internal and set your one to clock from it and see what happens. You wouldn't happen to know anyone with 9632 that you could borrow?

 

[living sounds]

What would be interesting to do is to plug a second 9632 into the spdif internal and set your one to clock from it and see what happens. You wouldn't happen to know anyone with 9632 that you could borrow?

The internal S/PDIF input seems to be in parallel with the external one, so it wouldn't make any difference, I think.

 

[MatthisD]

You mean you've tested it or according to datasheet or manual? they shouldn't be in parallel should they, doesn't the switch select to either receive AES/spdif in or a spdif clock source from another RME.

I took out my 9652 and it has four 4046AM chips located close to the four adat ports.

 

[living sounds]

I checked it with the multimeter, yes, they're not in parallel. But I also tried to sync an S/PIF signal to the internal input, didn't work. I'll replace the 74HCT tomorrow.

 

[MatthisD]

Hopefuly thats it then, keep us updated.

 

[living sounds]

So, I've replaced the 74HCT04D. Sending in an external S/PDIF signal I got sync for a second, and then it went away. There are intermittent single sample crackles on one channel now, the other is silent.

I've noticed checking with the scope that the DS26LS32ACM puts out a signal of much higher amplitude than goes in. Is that how it should be?

I could try replacing the hct4046am next...

Just checked my mails, the guy I bought the card from originally had the noise problem on the output a few times in 4 years with his old computer, and always was able to get rid of it with a reboot. In his new system the problem was there all the time.
Somewhat similar here, since the card worked well the first few months.

There were blown caps (I replaced), so there definitely was too much voltage in some places for some time. There are two zeners on the board which are working fine apparently.

There are are several SMT regulators on the board, too, could this be a problem with the power supply?

I found the DAC chip, but I doubt replacing it would do any good, and with all those legs and the solder unfriendly position on the board...

 

[MatthisD]

Was that the only 74HCT04D on the board?

There should be one of these;
"U8 (ADM811TART) is a RESET generator that provides
a debounced 200 ms reset signal from the push button (S3)
or on power up. The reset is active if the 5 V supply drops
below 3 V."

If you can find a chip that looks like it:
http://ie.farnell.com/analog-devices/adm811tartz-reel7/supervisor-reset-3-08v-sot143/dp/2074891

If you can find it could you solder a wire to it and measure the voltage at this reset generator output, rebooting your system to make sure it is functioning. There should a resistor of about 100k from one of its legs to ground which I think drops the voltage after power-off enabling it to reset once it goes below 3v. Doing this will also tell you that the digital 5v supply is alive.

If you refer to the block diagram I posted the components should be arranged similarly in the RME.
The four spst switches, u10/u11 74HCT04D, reset function generator (if its there), the DIR...
can you identify U4 as it uses a 74HCT04D for clocking it seems, depending on whether or not thats the case with the RME as well.

"U4 (M4-64/32) is a Vantis CPLD and has been programmed
to provides input signal MUX selection, LED buffering and
switch decoding for the different interface modes. The output
interface mode of the DIR (U2) must be compatible with the
input to the AD1852 (U3) and this is selected at the same
time as the mode for the AD1852 is selected, with switch
Interface Mode (S5). S5 is decoded to drive the DIR mode
pins (M0-3) and AD1852 mode pins (IDPM0) and (IDMP1).
The source code in included in Appendix A.

"Note: Because of
excessive jitter degradation in CPLDs, a separate IC (U11) is
used to select the MCLK input to the AD1852."

I would have thought the hct4046am are related to the ADAT ports, one per port perhaps.

 

[Andy Peters]

I've noticed checking with the scope that the DS26LS32ACM puts out a signal of much higher amplitude than goes in. Is that how it should be?

That's a standard RS422 differential receiver chip. The output is TTL, so its minimum output level for a logic high will be above 2.4 V. The data sheet says that "typical" output level for logic high is 4.2 V, which is unsurprising for a 5 V device, especially if lightly loaded.

Now the input is differential and the input sensitivity is 200 mV, meaning the difference between the two inputs has to exceed 200 mV for a valid logic level to be recognized. If the + input voltage is greater than the - input voltage by at least 200 mV, then the output will be a TTL logic high. If the + input voltage is less than the - input voltage by at least 200 mV, then the output will be a TTL logic low.

Note that the inputs can deal with signal levels up to +/- 7 V (to withstand common-mode differences); in practice, the system uses the complementary 26LS31 differential driver, which provides complementary TTL-level outputs. The 26LS32's input sensitivity makes it possible for one to use a very long line between driver and receiver and still discriminate valid logic levels.  So it's quite reasonable to see the 26LS32 output voltage being greater than the input.

-a

PS: haven't used the 26LS31/32 pair in ages. We're all LVDS, all the time these days.