RME is a really great company making high quality audio interfaces and their decent software policy helps to keep them usable for a very long time. I appreciate that very much and so do people all over the internet. Sadly, the same people are asking how to solve dead, noisy or blown mic pre-amps in their RME Fireface 800. Problem seems to be connecting and disconnecting microphones while +48V phantom power is switched on. Don't ever do that, whatever equipment you use. Components might fail and this is what happens regularly to Fireface preamps. Here is how I solved this.
There are some useful threads to be found on the web. I recommend this one at gearspace.com where a user is helping to pinpoint the cause of pre-amp noise, although in my case the solution to the problem was a little different.
What's that noise?
I acquired my FF 800 a while ago with a noisy channel 10. It's been like that a long time the former owner said, and he did not know how and when the issue started. Symptoms of the problem were crackle and sizzle on this channel, increasing as gain was raised. Switching phantom power on/off made no difference. Shorting XLR input pin 2 and 3 made no difference either. Noise was always present, no matter if I used a front line/mic input or the input from the back.
Opening up is straight forward. Rack ears and some smaller screws have to be removed, then you can lift off the top. In the back a ground wire is connected. Unplug it and lay the top housing aside. Now connect the interface to your computer and switch on the the whole system. Don't touch anything in the interface by accident to prevent deaths of components. Don't touch anything in the power supply to the right to prevent death of yourself. Take a multimeter and make sure the power lines are ok. The op amps need +/-15V. If you have a look at the image above showing channel 10, all yellow dots are Ground, plus voltage is orange, minus is light green. All input channels look more or less the same, you'll find your spots. In the image you see a THAT 1510 input stage op amp to the right. Check pin 5 against pin 7 and pin 4, accordingly. However, most op amps in this device are 4580, you'll find the first one when you follow the 1510 output lines. As long as there is only one channel noisy you don't need to care, stick to the channel circuit first.
Since power was ok I started tracing the noise with my oscilloscope. It was present on 1510 output pin 6 and on the negative input (pin 2). So I suspected one of the 4 General Semiconductor (now Vishay) diodes marked S407. I took the multimeter again and switched to diode check mode. The diodes were ok. Alright, shut down your system and pull all plugs.
Disassembling and further checks
Since there is neither a service manual nor schematics available (I asked kindly in person at a fair) I took the image above with my camera, opened it in Photoshop, erased the caps, started tracing the lines with my multimeter and sketched the results in the image.
To be able to trace I had to disassemble the whole device. You should do that, too, it makes handling much easier. Start with the clock and digital I/O board. Disconnect the ribbon wire by pulling the board's plug, remove the nuts on the word clock connectors and grab the small board. Now remove 3 screws from the LED board in the front, disconnect it, too and lay it aside.
Remove the pot knobs by simply pulling them, remove a lot of screws from the XLR inputs and the main circuit board, one screw that ties a regulator to a heatsink in the power supply and two on the left and right side of the housing. Now lift off the mainboard together with the front panel. If you want to remove the frontpanel completely you might need to remove all XLR connectors first, at least so it seems. The panel is stuck even if you center all potentiometers and try to twist it. Well, I did not bother.
Measuring is not that easy since the device is partly fullpacked with really tiny components. You can move the caps a little aside to reach all spots.
First I checked the caps for shorts (bottom side). All ok. Then I documented Ground connection wherever I could find it (yellow, probe in XLR pin 1). Next I moved the probe to XLR pin 2 and traced the hot (+) line through the circuit to op-amp pin 3 (red). Then I did the same with the cold line (-, dark green).
Eventually I found a 4.7Ω resistor that was dead and this made a lot of sense. The 1510 got no signal on this line except for the presumeably very little power supply noise that's not being blocked by the neighbored diodes. While the hot side got the normal signal and since the op-amp is increasing the differential voltage of both lines you hear a lot of noise.
Get a grip on that resistor and fixing
How to solve issues with these tiny beasts? Usually I use a heat gun to remove SMD components. To protect the plastic and components nearby I placed aluminum foil around the area I wanted to heat up. It took less than a minute to get rid of the faulty resistor with around 390°C. The solder did not melt at lower temperatures, unfortunately. You can always test the right temperature in less critical parts of a circuit. I did this on one of the Ground connections of the 10K-pot. Someone from RME told me the layers in their multilayer PCBs have become quite thin and complex throughout the years, so careful work is a must or unrepairable damage may happen. Fortunately, a FF800 is not as complex as latest products.
I found a good quality 4R7 (measures on spot, size 0805, probably 125mW) on an outdated TV receiver board and took that as a replacement. You might consider taking a larger 1206 sized 250mW resistor to be safe but then you should replace the one on the positive line also.
After cleaning with a solder sucker I added a little leaded solder to the two soldering pads because of the lower melting temperature. I put the "new" resistor in place and heated up thoroughly. Another multimeter check convinced me of proper connection.
Then I reassembled the boards and made a first check with the whole system. Tadaaa – the noise was gone and channel 10 now sounds exactly like any other channel. I reassembled the rest of the housing and put the knobs back on. The complete repair took about 3 hours.
Hope this helps someone. Feel free to leave a comment.