The signal voltage for me is good - 5v
The duration is OK ?

Guessing that's only the DRQ2 signal. could you add the DACK2 line to the second input of the scope ?
http://www.hardwarebook.info/ISA#Single_Transfer_Mode

also not sure about the timing, but >8ms feels a bit high to me.

yes second scope on dack2

yep, like you said before, there's just no response to the dma request. dack2 should go low, but for some reason doesn't.

dataino.it wrote on 2021-05-30, 17:39:

The signal voltage for me is good - 5v
The duration is OK ?

8ms is pretty long, no single DMA transaction should take anywhere near this long. When you said it's a short pulse I though it was some 10 or so nanoseconds. I think it might be timing out now.
I see you have some sort of logic analyzer? 2MHz is not much but maybe it will do. Connect CPU pins HOLD and HLDA to it, in addition to DRQ2. Let's see if the CPU even gets HOLD signal from the '206 - probably not because a DMA cycle that long would interfere with DRAM refresh and the system would crash. But it won't hurt to check.

It's a very weird situation where the card sends DRQ but gets no DACK, it can happen if DMA is not enabled. But it should be, the system expects floppy data transfer. So maybe one of the data or address lines of the '206 is not connected to the chipset, so it's not being programmed correctly for DMA operation. Check if pins XA0-XA9 on the '206 are all connected to chipset. Also check XD0-XD7 but I think a problem here would be detected as CMOS issue, so I think it's more likely an address line problem.

Is this the right sequence of signals?
DRQ2 (Slot -> 206), HRQ (206 -> 283), HOLD (283 -> CPU), HLDA (CPU -> 283), HLDA1 (283 -> 206), DACK2 (206 -> Slot)

Deunan wrote on 2021-05-30, 19:36:

8ms is pretty long, no single DMA transaction should take anywhere near this long. When you said it's a short pulse I though it […]
Show full quote

dataino.it wrote on 2021-05-30, 17:39:

The signal voltage for me is good - 5v
The duration is OK ?

8ms is pretty long, no single DMA transaction should take anywhere near this long. When you said it's a short pulse I though it was some 10 or so nanoseconds. I think it might be timing out now.
I see you have some sort of logic analyzer? 2MHz is not much but maybe it will do. Connect CPU pins HOLD and HLDA to it, in addition to DRQ2. Let's see if the CPU even gets HOLD signal from the '206 - probably not because a DMA cycle that long would interfere with DRAM refresh and the system would crash. But it won't hurt to check.

It's a very weird situation where the card sends DRQ but gets no DACK, it can happen if DMA is not enabled. But it should be, the system expects floppy data transfer. So maybe one of the data or address lines of the '206 is not connected to the chipset, so it's not being programmed correctly for DMA operation. Check if pins XA0-XA9 on the '206 are all connected to chipset. Also check XD0-XD7 but I think a problem here would be detected as CMOS issue, so I think it's more likely an address line problem.

XD0-XD7 -->OK connected to 74LS245
XA0-XA9-->OK connected to chipset

I changed Bios with MrBios
V007B317 OPTI317 OPTi 386SX SxPW/SxPI 82C281 / 82C283 82C283 - 386 SxPI single-chip 386SX non-cache
System OK but problem persist

tomorrow tray to investigate CPU signal

snufkin wrote on 2021-05-30, 20:08:

Is this the right sequence of signals?
DRQ2 (Slot -> 206), HRQ (206 -> 283), HOLD (283 -> CPU), HLDA (CPU -> 283), HLDA1 (283 -> 206), DACK2 (206 -> Slot)

This depends on the mobo and chipset. In general yes, although it would also be possible to connect the '206 directly to CPU. And it might be so in this case since this is an SX systems, I doubt it has VLB slots. So ISA and logic inside the '206 should be the only things requesting the bus. But on the other hand there is also system RAM refresh, that can be done via a pseudo-DMA and it would be controlled by the chipset.

It is a good idea to test where HRQ from '206 is connected, to CPU HOLD pin directly or to the '283 OPTi chipset. It has to be one of those two. I wanted to wait for the CPU HOLD/HLDA test resulst to suggest that next - since the XA lines all seem OK. Then maybe it is HRQ line - that can also be tested with some other DMA, like SoundBlaster sample playback. If SB doesn't work either then it's a safe bet the problem is with HRQ line or even the chipset itself.