Atari TT's LM7905CT getting too hot

[mikro]

I have had a TT for some time but I haven't really used it for longer periods of time. Just installed the Thunder/Storm, then Lightining, then NOVA/Mach64 and all seemed fine. Sometime during those times I have also recapped the PSU with the Exxos kit + replaced the CNY17-3 opto-isolator with a PC111, everything as per https://www.exxosforum.co.uk/forum/viewt ... ?f=9&t=300.

Today I have started testing it a bit more - everything is fine but after some time I have noticed that the TT simply restarts. And usually shortly after the first restart it does again and again, in much shorter intervals. I found possible culprit: LM7905CT is getting ridiculously hot. And since its functions seems to be "employ internal current limiting safe area protection and thermal shutdown for protection against virtually all overload conditions" I'd guess that something is either indeed overloading or just something is broken which makes the circuit so hot.

As a test, I stripped the PSU of all shielding and manually pointed the PSU fan directly at the regulator. That seemed to help, I could run the TT for 1+ hour without a single restart.

Any ideas what to do with this? I tried to reverse the PSU fan (in its original position), i.e. sucking air in (instead of out) but that doesn't seem to have any effect.

The PSU is PSM-5341 so it should be the same as in MegaSTE. I'm performing more tests (like stripping Thunder/Storm etc hardware in case they are the source of the overload) but I'm guessing that others like @czietz are using TTs and their original PSUs day and night so this hardware shouldn't be an issue in TT.

[czietz]

Thunder, Storm and Lightning do not use the -5V rail that is provided by the LM7905. In fact, it's not even available either at the TT RAM connector (where Thunder and Storm TT are) nor at the VME connector -- so we can rule out the Lightning VME and also the Nova. I would have to check the schematic to see which component is actually using it.

You could check (with the power off) if there is something accidentally shorting the -5V rail.

PS: And yes, I'm using the original PSU (recapped) in my TT and MegaSTE without issues.

[mikro]

Intermediate result so far - when I disconnected the Storm, the computer seems to be much happier, no reboots so far (about 30 minutes). Mach64 (but not the NOVA adaptor) has been disconnected the whole time, just to be on the safe side.

[czietz]

In that case the hot LM7905 would be a red herring, though, because it is not involved with the Storm TT in any way.

Bad RAM? You can populate just one of the SIMM slots on the Storm (marked "Slot A"). Does it run stable then? What happens if you let YAARTTT test the TT-RAM?

[mikro]

... and we can safely rule the Storm out, it has rebooted shortly after I posted the last update.

Incidentally, I am running YAARTTT as a check whether it has restarted, *when* it was running memory errors were always zero.

Going to look at the -5V rail, thanks for the tip.

[czietz]

Assuming I didn't overlook something in the schematic, the -5V rail only connects to the audio DAC and to the TT high video circuit. Particularly the latter (being ECL logic) will draw significant current even in non-faulty state.

[mikro]

I guess you don't have your TT open so you could check whether your PSU's regulator is getting so hot as well? It's so hot you can't even touch it for longer than half a second.

[czietz]

The TT is open but the PSU is not. I can't promise I'll have time tomorrow. If I do, I'll check.

[mikro]

Thank you for the effort, though. I'll try to investigate further in the meantime.

What can I say with a certainty, it definitely seems that the lesser devices / CPU activity, the better. Just staring at empty desktop makes the computer more "stable" (but the regulator is still super hot). With all hardware connected & YAARTTT running, it would reboot within 2-3 minutes.

[Maximilian]

After carefully checking all the voltages under load, a voltage meter with high/low setting would be interesting.
(to check if there is a dip)
That would at least prove it is a voltage related issue and where to focus your efforts.