Some findings regarding DSP crashes

  • Virus A/B/C:

    I recently obtained a Virus KC that had frequent DSP crashes. When the DSP crashed, the keys would not always create a sound, the BPM indicator would slow down significantly and the sound was either garbled or played back much slower than normal. Initially I suspected the power supply but after making sure it's not that, looking for the root cause led me to make some improvements to the Virus that I'd like to share. Someone tried to replace the commonly failing filter capacitor of the switching voltage regulator but failed, so this is what I first fixed. For those who have issues with their Virus A/B/C and trying a different power supply isn't helping, replace the SMPS filter capacitor - on [1] it's the brown cap right next to the LM2575. In my case, that didn't help with the crashes, though.

    The DSP crashes were most easily reproducible with patches that use a longer sustain, delay or reverb, such as "N-R-G! MS" (A19) or "DryLand MS" (A74) - probably because they use more voices that way. Having the keyboard attached made the DSP crash more likely to occur but since opening the lid of the KC requires the keyboard to be removed, I fed the Virus an endless loop of midi chords instead to reproduce the crash.

    I noticed that the 3.3V and especially the 5V voltage regulators on the PCB were becoming quite hot. I placed them on (bigger) heatsinks to let them work better. However, the 5V regulator seemed to already be damaged because it completely fried after I re-flashed the firmware. The firmware upload would work fine and the writing to flash as well but after the reboot, the outputs only produced a crackling sound. Turns out the 5V regulator suddenly only provided 1.8V and everything was well again after I replaced it. If you're running into problems that show up after updating the firmware or have a crackling noise on the output, checking the 5V voltage regulator on the heatsink should be the first course of action. [1] shows it in the bottom right corner.

    Since the space for heatsinks is rather limited, I also added a 4.7 Ohm, 5W resistor in line with the power switch. It creates around 3V of voltage drop, so the 12V from the power supply are reduced to 9V. This leaves enough for the voltage regulators to work while preventing them from creating unnecessary heat. Why they use 12V supplies is a mystery to me since the Flash only requires 5V to program it.

    What made the biggest difference however was adding some bypass capacitors to both the 3.3V and 5V linear voltage regulators. It appears that Access really didn't pay much attention to their power supply design because adding 100nF caps to the regulator outputs (Pin2-Pin3) kept the DSP running properly for significantly longer.

    A guy repairing his Virus B [2] also added an interesting quote that says to add A 47nF cap in parallel to the LM2575 output cap: "If you look with a scope you’ll see there is ringing at very high frequency short peaks way above specified voltage for the DSP. This helps reduce them substantially."

    So I added a 100nF cap there as well.

    With these measures, the DSP crashes have reduced significantly for me without having to replace the DSP itself. Fully restoring the machine would unfortunately require changing the DSP though, it seems like.

    Virus TI/TI2/Snow/Polar:

    There's a teardown [3] that shows just how hot the 5V linear regulator becomes even on TI/TI2 models [4] since it left burn marks on the ribbon cable which conveniently also blocks air flow. So for my TI2, I'll also add the 5W resistor and the bypass caps the next time I open it to prevent it from taking damage when running for extended periods of time. While Access did change the power supply for the TI/TI2, I'm not really confident in them making sure that the logic power supply isn't noisy and doesn't damage the DSPs in the long run. The fact that we see quite a few TI/TI2/Snow/Polar models having broken DSPs suggests that something indeed is going on. With that, it may make sense for those with working units to add the bypass caps and use a 9V power supply instead of the supplied 12V one.

    [1] https://erichizdepski.files.wo…4/20200412_124211_hdr.jpg

    [2] https://erichizdepski.wordpres…er-access-virus-b-repair/



  • Note: I desoldered the original DSP and soldered in a new DSP - the Virus worked flawlessly again. Since I'm curious like that, I soldered the original DSP back in again to verify. The Virus also worked flawlessly again, much to my surprise. This leaves me to assume that there must've been a cracked solder joint on one of the DSP's legs, as unlikely as it seems. It does somewhat explain though why the crashes didn't immediately occur when the Virus had been off for a while. Not a single crash so far, let's see how it goes. Should the crashes return I'll put the new DSP in again but so far it doesn't appear to be necessary.

  • The original DSP worked for some time but eventually, the sound issues re-appeared and the keyboard would stop working as well. Only a DSP replacement was able to fix it in the long run, unfortunately. Now it runs like new. Quite the strange affair, the whole thing, but since their power supply is so strangely designed, I'm left to assume that the DSP failure has to do with that. It's either that or the DSP itself failing after some time, which I would find rather hard to believe.