Falcon refuses to boot without diag cart

[stephen_usher]

:dualthumbup:

[rubber_jonnie]

:girldance: :girldance: :girldance:

[kodak80]

Great to hear that the Falcon is now back and working again.

:cheer: :cheer: :cheer: :cheer:

[Badwolf]

Lovely old job, mikro!

BW

[mikro]

I verified it from numerous angles: first, it boots every time now. After a minute, 10 minutes, one hour, always. Second, it boots in CT60 mode (that didn't happen at all before). And third, it boots from IDE, I could run DSP demos, FPU demos, memory / disk io hungry demos... all fine. And as a bonus, my Falcon now looks like taken from production line. So... wow.

Actually, this had an interesting aftermath... after some time (when I decided to put the PCB back into the case) it seemed like the symptoms are back but in a strange fashion -- the board would either work (for hours, running BadMood, MP3 playback, Gembench) or wouldn't boot properly -- sometimes I'd get the infamous "colour noise" (i.e. random pixels/lines), sometimes I'd see a corrupted Atari logo, sometimes I'd see a corrupted message from HD Driver, sometimes it would freeze in XBoot, sometimes even reaching the desktop (and freeze as soon as I entered some menu).

This has been going for about two weeks (to be honest, I tried to diagnose it only 2-3 times during that time) and I couldn't get any closer even to a reproducible scenario. And yes, the diag cart would still be rock solid.

Today didn't seem any more optimistic -- I again tried knocking on the ICs, on the PCB, gently bending the board and nope. If the Falcon worked then it worked, if it didn't there were those random symptoms (changing their form even after a gently reset button push) and then suddenly it would miraculously fix by itself. The opposite scenario (to break it when it worked) was surprisingly much, much harder than going from a broken to working state. I really had to knock on the PCB like crazy (I wouldn't expect even a working Falcon to operate properly with all those vibrations!)

I was working with the following thoughts:

• The PCB is somehow cracked somewhere inside and I'm f**ked (possible but I didn't believe I was so careless)
• While recapping, some capacitor (or other component) got loose, lost connection, had a damaged trace/via from my soldering work (I resoldered the caps around the place where I thought the crashes happen most frequently - under the PSU and around the Combel, didn't help ... was open to the possibility to check all caps and their connections)
• There is a small piece of solder / metal somewhere from my soldering work (very possible but ... how to find it?)

I was really close to giving up (at least for today) and then it occurred to me... all those crashes (except the earliest one) are somehow related to drawing ... Atari logo, boot text, XBoot, desktop menus, gembench... it rarely crashed when the desktop was still. Thinking about it a bit more I realised that drawing itself couldn't do any damage (it's writing to ST RAM after all, just like anything else) but what could do some serious damage? The Blitter!

So around midnight I was like OK, let's give it one more chance, let's test how keen it is to crash during some Blitter-intensive work (Gembench's drawing tests, my SillyVenture 2019 invitro's end part, ...) and indeed, I was able to crash it much faster while knocking on the Combel during a blitting code.

So... even if this were true, what now? In a pretty hopeless mood I took my (only) magnifying glass, to check on the Combel (it was really just a box checked in my head, no expectations whatsoever)... looking at it, looking... and then I see a small blob of solder on one of its tiny legs! Actually in two places! Well, to make the unnecessarily long story short ;), using the classic soldering iron, the glass and a razor I somehow managed to get rid of it and not destroy the Combel.

Connected everything again, ran "Joy" demo (blitter code from start to end), were smashing with a toothbrush on the Combel and so far so good. So fingers crossed, I'm not going to make any bold claims anymore! ;-) Please note that this also explains the diag cart (at least partially) because it uses its own drawing routines for text (i.e. no VDI, no Blitter).

Anyway, another lesson learned the hard way... when using the desoldering station (blowing air out), protect the neighbouring ICs! (I hear the laugh coming from the back seats ;-))

[exxos]

:bravo: :girldance: :girldance: :girldance:

[JezC]

@mikro Great news! Always good to see another Falcon saved!

:thumbup:

[Badwolf]

So... even if this were true, what now? In a pretty hopeless mood I took my (only) magnifying glass, to check on the Combel (it was really just a box checked in my head, no expectations whatsoever)... looking at it, looking... and then I see a small blob of solder on one of its tiny legs! Actually in two places!

Great bit of diagnosis, Mikro!

Well done in getting her flying again. Properly this time. :D

BW