Hi Scott, nice to see you join, sad to hear you too are having issues now 🙁

Following the time proven engineering practice of "If something broke, undo the last thing you did, even if it couldn't possibly matter", I removed an 8255 IO board that I had installed. The Xi 8088 is no longer having epic freakouts during Checkit seem to have stopped. I've been running checkit continuously for about 4 hours. One pass logged an error in the 9000-9FFFF block.

I'm thinking maybe in my case there are two issues going on: 1) the same issue with the second SRAM chip that has been plaguing others, and 2) instability caused by the IO card that I had installed. The IO card shouldn't be able to effect a memory test, my only guess is that maybe there's a bus loading issue going on, and the bus just started getting wonky with a couple additional ICs plugged in. I'm going to have to consult the schematic and see what logic is being used in the Xi 8088's bus interface. Maybe try swapping a few chips with a different logic family and see if it has an impact.

Of course, my "success" could just be me misinterpreting randomness, and the problem could have nothing to do with the IO card.

Scott

That's good to hear. As long as I stay below the 512kb boundary it seems to be fairly reliable for me, all things considered...

keenerb wrote:

It was a great relief that the project involved no SMD parts. […]
Show full quote

BloodyCactus wrote:

since its all through hole, piece of cake. now if it were smd birdfeed, yes 😉 birdfeed before sockets!

It was a great relief that the project involved no SMD parts.

When I put my 1mb RAM board together from lo-tech I was most disheartened when I opened the box and this appeared:

Managed to get it working though. After using up about an entire spool of desoldering braid, anyway...

Don't solder each Pin 😉
Get Braid to remove solder.
Use a Flux Pen.

https://www.youtube.com/watch?v=lauw0bSe-Cw

I recommend the Drag Soldering Technique.
https://www.youtube.com/watch?v=wUyetZ5RtPs

The legs look separated enough to solder individually given fine enough solder and good magnification. Drag soldering is usually best for high density QFP and TSSOP chips. I'd probably solder pins one and five to secure the package, then solder the remaining odd pins, then solder the evens.

smbaker wrote:

Following the time proven engineering practice of "If something broke, undo the last thing you did, even if it couldn't possibly […]
Show full quote

Following the time proven engineering practice of "If something broke, undo the last thing you did, even if it couldn't possibly matter", I removed an 8255 IO board that I had installed. The Xi 8088 is no longer having epic freakouts during Checkit seem to have stopped. I've been running checkit continuously for about 4 hours. One pass logged an error in the 9000-9FFFF block.

I'm thinking maybe in my case there are two issues going on: 1) the same issue with the second SRAM chip that has been plaguing others, and 2) instability caused by the IO card that I had installed. The IO card shouldn't be able to effect a memory test, my only guess is that maybe there's a bus loading issue going on, and the bus just started getting wonky with a couple additional ICs plugged in. I'm going to have to consult the schematic and see what logic is being used in the Xi 8088's bus interface. Maybe try swapping a few chips with a different logic family and see if it has an impact.

Of course, my "success" could just be me misinterpreting randomness, and the problem could have nothing to do with the IO card.

Scott

Now that I think about it, I was using my xi to test an Isa bus extender ribbon cable. With the bus extender installed but no cards on extension I saw many issues with memory errors during system boot. It seems quite similar to your issues in retrospect.

keenerb wrote:

That's good to hear. As long as I stay below the 512kb boundary it seems to be fairly reliable for me, all things considered...

I left Checkit running overnight. No errors this morning, but the machine is firmly locked up.

What are you using for U27, U28, U36, U37, and U42? These are the ICs responsible for chip selecting the second bank of memory. I'm using Sergey's recommended parts (ALS family) for U36 and U37, but used HCT logic that I had on hand for U27, U28, and U42 under the assumption it wouldn't really make a difference. HCT is slower than ALS. I'm thinking of ordering ALS logic for those chips to see if it makes any difference (or AHCT logic, which is also faster than HCT).

I'm also thinking I should hook up a scope and watch my 5V supply. If there are power supply issues, then it could manifest as sorts of weirdness.

Scott

https://goo.gl/photos/SjjVhrAqoENdLvBD7

SHould be able to read the chip labels pretty clearly.

everyone of my chips is a 74F except for 1, maybe I should switch to 74LS...

My system def. still has problems but I've been able to spend a long time in Champions of Krynn and Ultima III without crashing...

I pulled mine apart, added some capacitors to the prototyping area of the backplane to filter Vcc, and left it running all night with my oscilloscope monitoring the 5V line. It ran all night without crashing. At least part of that time Checkit was running memory tests, but eventually stopped with an "out of memory" error (I assume Checkit is only able to hold a finite amount of logging data?)

Anyhow, I can't say whether it is the capacitors that helped, just the act of tearing it down and reassembling it, or simple randomness at play.

I'm going to let it run the comprehensive memory test for a while today, then I'll start re-inserting IO cards to see whether or not issues reappear.

Scott

Let's hope for the best.

My results are not quite conclusive. With the 8255 board removed, the Xi 8088 stands a good chance of running through the night without crashing, but it's not a 100% success rate. With the 8255 board installed, the Xi 8088 fails within an hour or two, sometimes logging a memory issue in checkit, sometimes just locking up with no error message.

I'm not using the recommended logic family (F or ACT) for the bus drivers. I'm using HCT. I'm going to order some F, and maybe some ACT too to try them. It still feels to me like this is a bus loading or a bus noise issue.

I also designed myself a bus breakout board / bus terminator that I should be able to play with in a few weeks. I'd like to get my oscilloscope on some of the bus signals to see if they're getting loaded or reflecting, or something else going wrong, particularly as I add additional boards to the system.

Scott

That's similar to my observations.

I also feel that more boards = more likely to lock up. My best results seem to be using a four-slot ISA backplane with xi8088, video card, and floppy drive adapter.

When I have a lot of ISA cards installed I recall occasionally failing th eBIOS boot memory test...

I replaced my 74HCT244 and 74HCT245 chips with 74F logic, and replaced my 74HCT139 and 74HCT02 with 74ALS logic, to bring my board closer to the recommended logic families. I then reinstalled the two problematic IO cards, and it promptly locked up again while doing a Checkit memory test.

Next I'm going to try replacing some of the vintage parts (8284, 8288, 8259, etc) and see if that makes a difference.

My bus terminator / bus breakout board should be here in about a week, then I can try terminating the bus and putting the 'scope on the bus lines to see if there is an obvious problem.

I have reassembled my minimal system (xi8088, VGA, xt-ide card) on a four-slot ISA backplane.

I'm going to let champions of krynn run for a while, it's IO/memory/CPU intensive so should give a pretty good burn-in.

I have noticed that certain circumstances cause my xi8088 boot-up music to sound off, like it's a higher pitch or something. Anyone else noticed that? I think it's related to the backplane...

There was maybe once or twice where it outright crashed while playing the startup tone, but other than that the pitch has been consistent.

I finished replacing all of the vintage ICs. It still fails checkit if the bus is packed full of cards.

If you run checkit and do a "system board" test, does it fail your DMA controller? I'm trying to determine if my DMA controller is bad, or if it's simply different from what checkit is expecting.

It locked up after about six hours running CoK demo.

Checkit reports failure for DMA, don't know why.

keenerb wrote:

Checkit reports failure for DMA, don't know why.

I'm thinking this might not be a problem. Channel 0 is traditionally used for DRAM refresh, and since we don't have DRAM to refresh, maybe that's what Checkit is upset about.

I pulled the second SRAM chip to see if it was causing any issue. The board seems to be happy booting with only 512K, but it still locks up as before.

My machine was still running Checkit this morning after about twelve hours. No errors reported that I could see. I'll see if it's still running when I get home...