Show Posts
1
Hardware / Re: SymbiFace3 is near your Enterprise...
« on: 2020.February.19. 23:59:38 »
Sadly, yes! Some pics here https://enterpriseforever.com/maintenance/glitch-on-rd/
My problem is in the WIZ W5300 network chip - it has an internal FIFO with a auto-incrementing data pointer. So to read a received frame, for example, you just do lots of reads, and the pointer auto-increments to the next byte. I interfaced it using I/O, so it happened when the I/O address was decoded and there is a /RD cycle. The main glitch happens when /RD is low
so the W5300, being a modern fast device, sees 2 read cycles and increments the FIFO pointer twice. Not every cycle, just occasionally.
I have seen the glitch on a EP with no expansion attached - you just need to look at /RD on a scope that can trigger on a short high pulse (basic Rigol scope in my case). My early EP64 seems worse than my later EP128s.
Having got to the bottom of this, Zozo pointed out that a few EXDOS cards have a factory-fitted bodge on the back. When I looked, it was a R and a C on the /RD line going to the EPROM, and it successfully removed the glitch! I could also see it at the WD1770 but it works...so must happen at a slightly different point in the cycle. Or the WD1770 is too slow to notice the glitch. Or a double read doesn't matter to it.
@gflorez sorry, I have accidentally hijacked your thread
B.
My problem is in the WIZ W5300 network chip - it has an internal FIFO with a auto-incrementing data pointer. So to read a received frame, for example, you just do lots of reads, and the pointer auto-increments to the next byte. I interfaced it using I/O, so it happened when the I/O address was decoded and there is a /RD cycle. The main glitch happens when /RD is low
so the W5300, being a modern fast device, sees 2 read cycles and increments the FIFO pointer twice. Not every cycle, just occasionally.I have seen the glitch on a EP with no expansion attached - you just need to look at /RD on a scope that can trigger on a short high pulse (basic Rigol scope in my case). My early EP64 seems worse than my later EP128s.
Having got to the bottom of this, Zozo pointed out that a few EXDOS cards have a factory-fitted bodge on the back. When I looked, it was a R and a C on the /RD line going to the EPROM, and it successfully removed the glitch! I could also see it at the WD1770 but it works...so must happen at a slightly different point in the cycle. Or the WD1770 is too slow to notice the glitch. Or a double read doesn't matter to it.
@gflorez sorry, I have accidentally hijacked your thread
B.
Then I remember the factory glitch mod that some EXDOS cards have...
Something went very-very wrong at the companies crash. As he wrote it: "one of the saddest periods of my life and I do not like to think about it."
