Checkit still running, it's been approximately 24 hours. No errors.

This really has me wondering if it's not a bus issue; this is a small 4-slot backplane and only three cards total.

Does the BIOS have any timings on how long to wait for signals and whatnot, could there be a software problem at the heart of this?

Are you running the quick memory test? or the full memory test? I've never got a full 24 hours out of the quick test -- I always get an "out of memory" error at some point, which I had assumed was checkit running out of space to keep accumulating the log file.

My results definitely correlate with fewer cards = more stable. My bet is still that it's a bus loading issue rather than a bus timing issue. The cards I'm adding shouldn't affect software timing as their onboard peripherals are not used. They are, however, additional loads attached to the bus. I'm anxiously waiting for my bus breakout / bus terminator board to show up so I can start attaching my scope to bus signals and see if they are clean or if there's something weird going on.

Scott

Full memory test.

It finished the full memory test sometime between 2am and 6am this morning with no errors, 99 passes.

System was still responsive, restarted QUICK test with 999 passes, we'll see what happens.

If memory is correct, once I add my ISA floppy disk controller this will start locking up fairly regularly, so that's my next test.

My ISA bus extender test failed miserably with this setup a few months ago, maybe it wasn't a problem with the cable but a problem with the system itself. I assumed the length of the cable (8 inches) screwed up the ISA bus timings somehow, but it could be a load issue as well I suppose. It would regularly fail during the first 64kb of memory tested, and would not make it to a full 640k at all.

the bios has nothing in it regarding timings that I can see that would have such a kind of impact.

Im totally getting different results from you guys, maybe my issues are unrelated. I can barely get a bootup cycle completed without lockup. I think this weekend I'll put the scope on it. Might see if i can find another backplane too 🙁

Sh*t gets wacky when I have more than these three boards in a minimal configuration, so there may be a common thread to all the problems that is yet to be determined.

This is the exact backplane I am using:

http://www.ebay.com/itm/Molex-4-ISA-Backplane … IIAAOSweWVXdPNF

Only it was like $18 instead of $100...

I'm using a Cirrus Logic 5420(?) video card and a home-built XT-IDE adapter as my only other cards.

i have no cards in mine except Sergey's XT IDE CF Lite (no rom in it tho), and a Trident TVGA 9016. The XTIDE rom is embedded inside my BIOS since they work hand in hand.

all other cards removed.

turbo on or off seems to make no difference for me.

I have some ISA dev blank PCBs I can solder some headers onto to get access to the bus easily enough and I have a new german scope that has 8 logic channels so I dont need to break out the Agilent 1670g

My tests have all been with full turbo, for what it's worth.

I would like to get to the bottom of this. The xi8088 would be my daily retro machine if I could rely on it, I missed out on the Heathkit era. Assembling and running my own backplane, processor board, floppy controller, hard drive controller, Tandy sound card, and Adlib sound card would really please me.

I've also gotten a bastardized little ISA wifi interface going on, but it's less exciting than it might sound...

I have tried both with and without Turbo. I haven't noticed a significant difference.

I currently have five boards installed -- CPU, CF, Adlib, VGA, and Floppy. All Sergey's designs. It's pretty stable under those circumstances. It's when I add sixth and/or seventh card that it tends to all go to hell.

didnt have much time tonight, pulled every card out to just xtcf/vga/cpu.

scoped the clock signal just to take a cursory look at it... thats a very rough looking clock cycle for starters!! it is on 14.31mhz frequency too.. spot on to the trimpot, / 3 = 4.77 mhz (which is spot on for xt bus) I guess I'm also surprised by the low voltage of the clock. its what, .864mv, not even 1v

System is still chugging along with memory testing, no failures but did receive the out-of-memory error in FAST mode.

What should the voltage level be for ISA bus clock? TTL-compatible?

From looking at the schematic, it looks like CLK is the output of IC 39, which is a 74ALS04BN on my board. The datasheet says "vOut" is -0.5 to Vcc on that chip.

Why is there such a broad range of values for "vOut"? Maybe this is a stupid comment, but is it related to the "response" time, does it take a few microseconds for the IC to achieve full Vcc output level and it can't because we're switching too rapidly? Would faster logic like AS or S change this?

Ahah, from Sergey's page:

IC U39 74F04, 74ACT04 1
Mouser 595-SN74F04N, 595-SN74ACT04N
Note: This IC should have higher output current (about 20 mA), 74LS04, 74ALS04, or 74AHCT04 can be used, but it might result in stability issues especially on higher CPU frequency.

Interesting. How'd I miss that?

keenerb wrote:

Ahah, from Sergey's page: […]
Show full quote

Ahah, from Sergey's page:

IC U39 74F04, 74ACT04 1
Mouser 595-SN74F04N, 595-SN74ACT04N
Note: This IC should have higher output current (about 20 mA), 74LS04, 74ALS04, or 74AHCT04 can be used, but it might result in stability issues especially on higher CPU frequency.

hmm. I have a 74F04 in U39.

interesting he specifies 74F04 or 74ACT04, F is TTL, ACT is CMOS.
silkscreen on the board says 74LS04...

I dont have any 04 replacements around to see if that changes the clock pulse shape.

He also has F/ACT specified for U40 on the board, but ALS/AHCT specified in his bill of materials.

I wonder if the IC document on his webpage is incorrect. You could try swapping U39 and U40 and see what happens if they're different types. I have ALS for U39 and F for U40, I'll swap those two ICs and see what happens for me this evening.

I have 74HCT04 installed into U39 and U40, which seems like a lousy choice. I have 74F04 sitting in a tube that I ordered for this project and never used. Weird.

I'm going to add some 74ALS04 to my next order, as well as an 82C54 (my current non-CMOS 8254 is easily the hottest chip on the board). One thing that may account in differences in our builds is that there's actually a fair bit of variability that Sergey allows in his bill of materials. For example, 8259 vs 82C59, 8088 vs 80C88, etc.

BloodyCactus, my first thought was that your clock pulses look like some signal reflection going on. I've done a little bit (well, very little) of RF work and when signals aren't properly terminated, sometimes they start looking crappy like that. Googling I came across this post: https://codingcipher.wordpress.com/category/h … ware-interface/. His third picture looks a lot like yours, though that's the one he describes as looking "nice".

All my chips are 82cNN's and all are Harris (aka intersil). I have ebay'd a couple of new ones as I realised my two 82c59 and the 82c37 are the slowest speed grades you can get so I've got 2x 82x59-12mhz versions on order (mine is 5mhz) and the 82c37-10mhz which will also replace the 5mhz one..

Hm I never considered that. My 8237 is a -5 part as well. I'll check the spares I ordered for my second xi8088 unit.

Perhaps this is why the DMA test is failing under CheckIT. Running at 4.77mhz should show if that's an issue or not. My suspicion though from looking over the schematic is that the slower DMA controller will simply mean slower DMA transfer?

All I see on ebay for a -10 part is KS82C37A; I assume that should be compatible?

I'm unclear what speed my NEC D8259AC uses, it has no speed prefix. Data sheet doesn't make any references to clock speed or anything like that, so I'm not sure if that's significant.

The position that I've taken is that if I can't get it to work reliably at 4.77 Mhz, then I'm not going to worry about Turbo speeds. To that end, I really haven't picked up faster DMA controllers, interrupt controllers, etc.

BloodyCactus, where did you pick up the 82C59 12 Mhz? I see digikey has them for about $13 each. I was hoping for cheaper.

avnet has then for around $8. Shipping is $9.

I don't think DMA speed is going to be a problem, *SIGH* BUUUUT I went ahead and ordered two anyway.

I had some spare 74F04's so I swapped out U39 on my system. I'll let Ultima run overnight, it's the most consistent indicator of potential issues for me. Running at 8mhz currently

keenerb - the KS82C37A-10 is what I ordered. Your 8259ac is a 5mhz part.

8259A-8 is a 2 MHz part.
8259A is a 5 MHz part.
8259A-2 is an 8 MHz part.

SMBaker - I got two 82c59's from ebay.

The turbo, does not affect the bus speed. The bus is still 4.77mhz, its just internally its clocked at oscillator/3 speed (so for me with 24mhz oscillator, thats 8mhz). (I've verified on my scope).

BloodyCactus wrote:

The turbo, does not affect the bus speed. The bus is still 4.77mhz, its just internally its clocked at oscillator/3 speed (so for me with 24mhz oscillator, thats 8mhz). (I've verified on my scope).

That doesn't seem right. According to the schematic it seems that bus speed SHOULD be tied to CPU speed; my EGA card also won't work at CPU speed 10mhz but will at 8mhz, and I assumed that was due to ISA bus speed difference.

*edit*

Ahh, yes. It is 4.77mhz according to schematic, the X1 input on the 8284 is tied directly to the 14.318 oscillator. My mistake. Since the DMA clock is tied to the 8284 I expect there's really no need for > 5mhz part there. Same with the PIT?