VOGONS


Trying to upgrade a board.

Topic actions

Reply 20 of 22, by Moogle!

User metadata
Rank Member
Rank
Member
Horun wrote on 2020-08-19, 22:52:
Moogle! wrote on 2020-08-19, 18:25:

Minor update. Got the DS1000 up off the board, but I somehow managed to break three of its legs off. Got a socket down and tried a compatible chip from another board (one of those bigger, blockier types) with 40NS delay. It tries to boot with that, but it's gets stuck with 0A on the POST card. I then went and found another one (again, another big blocky one) that was also 50NS like the original Dallas chip, and it does boot with that. Neither works at 33Mhz though.

I think you need to slow/add-wait time, not speed up/reduce. The DS1000-40 is roughly 20% less wait time by TAP5 than DS1000-50. By increasing from 25Mhz to 33 you are increasing speed by 30%. If you could find a DS1000-60 or -75 it would nearly balance that speed increase by adding more wait time. One thing: all of the Dallas DS1000 -20 thru -100 still have 2nS as nominal delay on Taps 1 and 2. So if some critical part was clocked/signaled off Tap1 or Tap2 any of those could be getting signals too quickly no matter which DS1000 you used at 33Mhz. Think of it like a typical VLB controller and you increase board clock to 40Mhz from 33Mhz, you need to add wait states either on board to VLB if it has them or on the VLB card itself (thru jumpers). Yes am simplifying things but going to a DS-40 is going wrong direction.

I have been wondering on this as well, and indeed, the 40NS delay chip came from an old 386. But the other 50NS delay chip that I got came from a 286. It's all too confusing. D:

Reply 21 of 22, by Moogle!

User metadata
Rank Member
Rank
Member

Well, I stuck what I believe to be a 100NS delay line in that socket, and it still won't boot at 33Mhz. While I haven't ruled that out as being the sole cause of this, given the flaky behaviour at 25Mhz with a 40NS delay line chip, I believe there is more to this issue, and I already have a new theory. On the board, as you can see in the picture, is a single large scale semi conductor. A CHiPs and Technologies 82c206 Integrated Peripheral Controller. (Basically all the main, important chips from a PC/AT). We discussed previously that the ISA bus (which I believe this chip is connected to) probably runs at 8 or 8.33 Mhz, being divded by three from 25. So, assuming that is true, at 33Mhz the chip, this will be operating at 11Mhz. (I do have a counter I could test this, but I am too lazy to pull it out, but I might do so later) However, the datasheet for the IPC says it's only good until 10Mhz. I do have a Siemens chip that is pin compatible and good up to 16Mhz, but I cannot find my chip puller and I have had bad luck in the past trying to jimmy it out with paperclips and such. (Those damn sockets just crumble after a number of years).

Bed for now though, worked 12 hours.

Reply 22 of 22, by Moogle!

User metadata
Rank Member
Rank
Member

Another update, for those who might be interested. I got the IPC chips swapped, but that still didn't cure the problem, though I did notice one of the LEDS on the POST card that wasn't lighting would occasionally come on now. So I looked again, and lo and behold, there is an honest to god discrete 8254 (programmable interval timer) on the board.

You know, like the IBM PCs had, and 286s and up tended to not have.

Anyway, like the IPC, it's also only good up to 10Mhz. I don't have one faster than that, but I did have an 82c54 on my XI8088. So I decided to just piggyback that onto the one on the board, and to my shock, it sort of works. The board still won't post, but numbers are flashing on the post card now before hanging. I have ordered what I hope to be a genuine 82c54-12 (the fastest one made to my knowledge) to see if that cures the problem.