by In this series of “When I find something interesting during a repair and remember to make notes” posts, the aim is to get information out there so that all the web crawlers, archive sites and search engine caches may preserve it for ever. In a world of Discords and Facebooks, it’s important to make sure knowledge isn’t kept in a walled garden.
I recently recapped a CD32, and it all went very well indeed. The customer had caught it early, there was no devastating leakage as there sometimes is, and no actual faults present – just a bit of preemptive maintenance. They also included a brand new TerribleFire TF328 card, which because it came from the official UK seller Supaduper, included the riser, and a MicroSD-IDE adaptor and cable with a 32GB card. Great service indeed, all plug and play. Notably, the customer had never used this card before, they’d waited for it to arrive and then shipped it all off to us.
I’m doing the usual tests – everything here gets at least 16 hours under load which usually means sitting down for a few levels of a game, and then some looping demos, as well as all the functional tests. And I noticed something a bit weird…. as I was playing a CD, it seemed to be getting quieter. Ever so gradually just fading out, I kept having to turn my amp up to hear it and it ended up with my amp at 50% which would normally blow your eardrums, and more noise than audio. Oh! Well, at least it proves the value of a good soak test. It was late at night so I switched off and went to bed.
The next day, it started out fine and then again, over the course of about 5-6 minutes, it went quiet. Reboot, and it goes quiet almost immediately – the longer it’s powered off for, the longer it stays loud. Heat related perhaps? Caps? I double checked my work but could see no fault. I started stripping things down to isolate the fault, and the first step was to take the TF328 and riser off. Aha! It’s loud again. Reinstall it, goes quiet.
So it’s the TF328 at fault, right? No.
I have a TF328 in my test box, with riser, and it lives in that box for exactly this scenario, where I need to check a CD32 expansion slot. I plug it in using the adaptor and card supplied…. quiet! So either they’re both faulty, or it’s not the card. And notably I used to use my test card all the time and never had this issue. At some point through all this I plugged in the TF328 without the IDE cable connected and was astounded to hear the audio coming through loud and clear. It’s the MicroSD-IDE adaptor!
I rechecked all my work as at this point I’ve got two whole CD32 setups with TF cards next to each other, swapping parts over. And sure enough, the common part is whichever machine has the MicroSD-IDE adaptor connected has an audio issue. A digital device affecting analogue playback.
Why?
Let’s delve into the schematics. And as easy as it is to type this out, it was a couple of days of puzzling. The CD32 has a “Mute” circuit, which seems to serve the purpose of muting the RCA output (from both CD audio and Paula) when you press the reset button. Not sure why, no other Amigas do, but without it I suspect there’s a “pop” from the speakers. This takes an input _RESET and uses that on a transistor – U342 – to affect the audio output levels. I won’t pretend to understand everything that’s going off, it looks a bit chaotic, but that’s the gist. I’ve highlighted _RESET on the PCB layout (from amigapcb.org) and R945E is a nice place to measure this on the front edge of the board.
_RESET also surfaces on the IDE port. The specification suggests drives could know when to reset because the host has, although in practise they never need to. In fact some drives back in the 90s used to spin down when you reset them, so it was common to have to cut the signal – it’s pin 1 of the cable.
The theory is that this specific MicroSD-IDE adaptor for some reason isn’t playing nicely with the reset line, and it’s pulling it down. Who knows why, it should only be an input to the IDE device but these are the joys of random Chinese adaptors. Measuring _RESET on the CD32 board with it connected is 4 volts, and with the adaptor disconnected is 5v. Using a *different* SD-IDE, or a CF-IDE adaptor, it’s also 5V. Remember this is an active low signal, whilst 4V isn’t enough to reset the Amiga it is enough for that transistor to slowly start to mute the output.
So the cause is, either a manufacturing fault or some weird design decisions on that specific model of MicroSD-IDE adaptor. The fix? Simple cut pin 1 of the cable like it’s 1995 all over again – or try a different adaptor. In this case MicroSD was affected but SD and CF were not, but I’ve also since heard about full size SD adaptors affecting it. I don’t think the TerribleFire card is the issue at all, it’s just passing through the reset signal exactly as it should.
