here is pretty much same idea as yours

had used this and encounter issues described above. there is exists termination resistors, but on bridge card, not on extension card. also i know of one case of using same connections as yours on mb with soldered it8888 chip, but issues is same. shorter cables will help but in my case they was as short as voodoo sli cable (about 5cm/2 inch).

This is very cool! Thanks. It is very useful, specially if you combine Core 2 Extreme and fully functional ISA in one system, like i've done here.

This article is in Russian, so, please, use Google Translate, i will translate it manually later.

In 2012-2013 I made similar construction and modified it a few times because I could not get a stable signal.
This is the last version:


It is true: when we use chipsets with native ISA like i440 or VIA133 the signal on ISA lines will be much more usable than on ITE8888 bridge. Most unstable cards (as I found) are Terratec EWS64XL and GUS GF1, especially EWS64. This card won't work anytime. When I use i440 chipset I got two, sometimes three cards working stable together, not more. If ITE8888 is used there is only one sound card working without problems. Of course I mean "full" cards which consist of digital sound and midi parts and use IRQs and DMAs. The "only-midi" cards as Goldfinch or SW60XG can work much more stable.
I did use two SCSI 50-pin cables and female ISA connector on ABLY riser. I did plug into this slot a standard ISA riser about 4 slots. If I try to use short cables it doesn't help well. I did think about backplane for 8-10 slots but my fail with a few cards killed my enthusiasm.

anthony wrote:

here is pretty much same idea as yours
had used this and encounter issues described above. there is exists termination resistors, but on bridge card, not on extension card. also i know of one case of using same connections as yours on mb with soldered it8888 chip, but issues is same. shorter cables will help but in my case they was as short as voodoo sli cable (about 5cm/2 inch).

Uhh.. I don't think termination resistors would do much, so close to the bridge chip. They're for "termination" so that means they should be on the far end. Maybe that could cause issues?
Then, on the other hand, all cards on the bus can be a signal source so I'm not even sure if there is a "far end" on the ISA bus.

PARUS wrote:

In 2012-2013 I made similar construction and modified it a few times because I could not get a stable signal.

It is true: when we use chipsets with native ISA like i440 or VIA133 the signal on ISA lines will be much more usable than on ITE8888 bridge. Most unstable cards (as I found) are Terratec EWS64XL and GUS GF1, especially EWS64. This card won't work anytime. When I use i440 chipset I got two, sometimes three cards working stable together, not more. If ITE8888 is used there is only one sound card working without problems. Of course I mean "full" cards which consist of digital sound and midi parts and use IRQs and DMAs. The "only-midi" cards as Goldfinch or SW60XG can work much more stable.
I did use two SCSI 50-pin cables and female ISA connector on ABLY riser. I did plug into this slot a standard ISA riser about 4 slots. If I try to use short cables it doesn't help well. I did think about backplane for 8-10 slots but my fail with a few cards killed my enthusiasm.

Cool to see more people have been experimenting with this 😀

You're using some seriously long cables there... I would suspect those could cause trouble. But if shorter cables didn't help then it must be something else. Do these bridge chips run at some super-high clock speed, maybe?
And what exactly do you mean by "three cards working stable together": having all three play sound at the same time, or having three plugged in with one card playing? Or does even detection fail already with more than 3 cards?

White wrote:

This is very cool! Thanks. It is very useful, specially if you combine Core 2 Extreme and fully functional ISA in one system, like i've done here.

This article is in Russian, so, please, use Google Translate, i will translate it manually later.

Wow. That's a crazy machine you built there! 😳 Very impressive.

If you decide to try a backplane with the ISA bridge, please report back on how it works! 😀 As anthony and PARUS have said, the PCI-ISA bridge chips seem to cause trouble when the bus becomes too long.

jwt27, I mean three plugged cards. It's no matter how many of them are playing at the same time. It's no matter even how many of them have been initialized. It is enough to plug three and get unstable work each of them. The amperage drop occurs for three and more cards and too much electromagnetic interference through cable. Almost all sound cards are detecting properly except EWS64 but correct sound is not present. They can be detected even five together but won't work properly. EWS64 does not detect even if I plug it single. The ITE8888 bridge was not overclocked, it soldered on industrial motherboard with i865G/ICH5 chipset. This mb is Itox G7S620-N-G.

If I take for experiment mb with native ISA I see three cards work properly (included GUS). Not ever, it is necessary to make a good shield. The EWS64 won't work ever.

Hmmm. I will have to experiment some more with this and see if I can replicate this problem on the backplane. So far I haven't really tried to use more than 4 cards at once (MDA + YMF719 + SCC-1 + SW60XG). I don't have the EWS64 or GUS cards so I'm unable to check those.
Do you think it's caused by interference on the signal lines or a lack of power? Do you have an oscilloscope maybe to check for skewed or reflected signals, or noise on the power rails?

Yes! I'm agree. I started ISA-VGA Trident video, Creative Goldfinch, Yamaha SW60XG, Sound Blaster AWE64 together and they are working well with i440 chipset. But two of these sound cards are MIDI-sequencers as your SCC-1 and SW60XG too. They use small MIDI commands and do not use massive data transfer of "live" digital samples. Therefore they work fine. It is of great importance.
I think that interference much matters because if I make foil shield it gives me good effect. Lack of power on signal lines (not main power!) maybe causes too because two cards work better than three and one card - better than two. No, I haven't oscilloscope. I even don't know how to use it 😀

I very wanna that you would make usable decision! I would be very glad.

If shielding foil around the cable helps, maybe the issues you're seeing could be attributed to EM interference? In that case perhaps interleaved wiring (like in 80-wire IDE cables) could help too. I don't know if these cables and connectors exist for such a large number of pins though.
And still that doesn't explain why using more cards causes more trouble. Low signal strength seems unlikely to me, since everything is high-impedance TTL all the address/data pins should be buffered by line driver chips on all the cards. The signal voltage would have to drop from 5 to below about 2.5 V to cause any trouble.

I have some data-intensive cards like network cards, I/O controllers and VGA cards. Will see if I can get them to work on the backplane.

Would you happen to have a schematic for your header PCBs? I've decided to add very similar headers to my GUS clone and would actually like to incorporate your design into my board.
I could rework all that from the gerber files of course, but a schematic would be a bit nicer to work with.

Never mind - found a suitable gerber viewer for now 😀

The schematic is pretty simple: http://i.imgur.com/PJvPbuj.png
I made this in DesignSpark since the free version of Eagle only allows very small boards. I could upload the original .sch and .brd files if it's any help.

In short: two pin headers, 64 and 36 pins, connected straight through. Top side is connected to the far pins, bottom (with the power rails) connected to the near side. Two pins left unused (tied to ground) on the 64-pin cable, my first cards had these on the inner side (near the 36-pin header), but I moved these to the other side (near the slot brackets) to make space for shrouded headers.

As far as I know, you can use pretty much any motherboard with ISA slots as backplane a as long as you don't power it on.

In theory. Because ISA BUS runs in parallel, though I wouldn't recommend doing that.

ISA carries power pins too, so connecting a second board would still power it on anyway.

Ultra cool!

I just wanted to say that I think this is awesome... even if I don't have a use for it. Well, this would come in handy for boards I have with only a single ISA slot. I have a huge Dell server, here somewhere that has, IIRC, a single ISA slot and plenty of room where I could mount this.

Thanks for posting this, OP!!!

And what about signal stability through cable during three and more cards plugged together?

jwt27 wrote:

And still that doesn't explain why using more cards causes more trouble. Low signal strength seems unlikely to me, since everything is high-impedance TTL all the address/data pins should be buffered by line driver chips on all the cards. The signal voltage would have to drop from 5 to below about 2.5 V to cause any trouble.

If all else fails maybe add those drivers to the backplane? In another project I worked on we had a similar problem getting a super unstable CPU bus (one running at 8MHz natively, the other being sped up to 48MHz) and we had to refresh the adress lines twice in 2 places (project consisted of 4 PCBs sandwiched on top of each other) connecting the CPU bus almost 1:1

One use I see for this is portables with ISA slots. One could make a card with a huge connector and mount the backplane in a case where a cable connects it to the portable. You could then power the backplane externally and add more cards in, say, a T3200 or those Compaq boxes that had ISA slots... At least, in theory.

Cable length may be OK, because expansion chassis for systems like that used a long cable which basically carried the ISA BUS signals in and out of the portable to said chassis over 3ft long wires which would imply that either the problem isn't really there all that much or else it can certainly be eliminated.

Hello,

i must say this is a awesome ISA Extender Project. Thumbs Up for jwt27 😀
But i have a Question about Powering the External ISA Backplane:

When you use your Ribbon Cable to provide Connection from your ISA Mainboard to the Backplane *without* adding any Power to the Back-Plane directly, the Backplane become Power from the Mainboard so we need no Aux Power to the Backplane.

But:

When Adding external Power like a additional AT Power Supply AND plug the ISA Ribbon Cable Extender with your mainboard we have this Situation:
Power the Mainboard AND Power from the Aux Power Input from the Backplane isn't a good Idea and i think we will blow up something, isn't it ?

I didn't use passive ISA Backplanes. Have a passive AT ISA BUS Extender Board some Circuits to cut the Power from the Master Mainboard to the Slave ISA Bus Extender when external Power is giving from the Aux Power Input from the Backplane ?

I have a very nice Schneider Euro PC II XT 8 Bit ISA Bus Keyboard Computer and would like to add an external ISA Extender, but this will bang my head about the two Power-Sources, first from the ISA Bus and second with the Aux Power Supply from the passive ISA Bus Backplane to blow up the Computer or more with two Power Outputs, ...

Greetings

jwt27 wrote:

Step 6:
Install and test. Your backplane will likely have AT power connectors. It'll be powered by your mainboard, so there's no need to connect these.

Yes there is a reason/need to connect those.
The traces on a motherboard have current limits.
If you try to power all those cards through the motherboard's power distribution there is a good chance you will burn open the traces on the motherboard.
.