exxos's Atari, Amiga & retro forum

I've been looking into PCB fabs in relation to quality and the bottom line is, it's not looking good.

The problems I've been seeing have been reported from jlcpcb users on other sites. But it's not just them. There's not much evidence but the trend seems be be a sharp decline in PCB quality on several common PCB suppliers. There's even a thread on startdot from 2020 which is concerning as I've only recently seen the issues.

I'm still waiting a reply from the fab which made the dfb1x boards. I've had no major problems with so far. But they are like x10 cost of JLC already. I want to get them to make the next batch of boards to see how they hold up.

My concern now, is if they have gone bad as reported on startdot, then my only option is to go to a uk fab. I've used one for the 2,000 30pin simms I got made last year and no apparent problems. But of course uk fab and uk assembly is ramping up costs. But I just can't use fab houses I don't trust. That list seems to get longer every year..

Though to myself, I need a new song.. Couldn't think of one. Then thought, I wonder if I could do a song about the ST536 chaos.. So had a go ! 8-)

Made some good progress today!

Firstly, AI found the bug on why CRC was failing on reboot. Basically what I did was, I had TOS check the TTram block was available before copying ROM to it.. This was doing a WRITE, which bascially trashed the first word in the rom shadow.. hence CRC fail after reboot as the first word was now wrong :lol: oopies :P

Secondly.. After some discussion with AI, we rewrote BLTFIX and integrated it into TOS. Basically if blitter is available and we are blitting within STram, then the blitter is used. If we are trying to blit into TTram, then softblit (CPU) routines are used. Realistically though, the CPU is not far behind most blit operations anyway. The blitter shines during block copies like the blitting test in GB6 as it doesn't have the overhead which the CPU has. So I wanted to keep blitters strengths running where possible. Then "everything else" goes to the CPU.

So it may not look much but it actually works!



Thirdly, AI also found the long-standing mystery of what BLTFIX (and my new port) slowdown was. So @agranlund will be pleased to know, that the issue was found in TOS itself, the cache logic was wrong! So the cache was been incorrectly set and not correctly flushed, which basically caused chaos and resulted in those weird slowdowns which have plagued everything since day one.....


And we now have a nice speed bump in scores :)

There could possibly be more optimisations to come very soon ;)

@exxos that really cool!
Now the big question is.. what are you going to put in the "Random unknown space" option :)

agranlund wrote: 06 Apr 2026 21:04 @exxos that really cool!

Yeah, TOS306 needs rebuilding now, its actually turning off the caches at random :P

Now the big question is.. what are you going to put in the "Random unknown space" option :)

Probably will end up being a option to turn off all the stuff I added as other stuff is likely broken now :P

I was toying with the idea ages ago to put a "boot rom" menu there. Mostly when using flashy-clock, a quick way to change boot rom. But I think it be more trouble than its worth..

Possible it could turn into a ST-ram cache on/off eventually :shrug:

Or just leave it there to annoy people....

exxos wrote: 05 Apr 2026 19:05 Though to myself, I need a new song.. Couldn't think of one. Then thought, I wonder if I could do a song about the ST536 chaos.. So had a go ! 8-)

I like this one! :)

New rev board.. Can you believe I will get 5 made at JLC again ! BUT, I am choosing different options this time. I still plan to use UK fabs for production boards. I still have a lot of parts in my JLC account to use up yet.. So hopefully I won't end up regretting this decision.. (probably will !) But I really need to make sure the board actually works before committing to a production run anyway.. Which kind of lands me back at JLC again anyway :roll:


Thing is, at least I know what to look for with manufacturing defects / bad vias, BEFORE I start soldering the things up this time.....

exxos wrote: 07 Apr 2026 18:30 New rev board.

Big arsed through-hole osc?

BW

Badwolf wrote: 08 Apr 2026 09:47 Big arsed through-hole osc?

Sort of.. Its going to be a adjustable clock for overclocking experiments ;) it was actually on the current rev as well.

Knocked up the test OSC board. *should* adjust from about 95mhz to 130mhz.