• Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.

Converting Acard AEC-7722 to AEC-7726

trag

Well-known member
I'm moving this discussion to Hacks. The beginning is in Trading Post, here:

Beginnging of Hacking

Following up on my last post in the previous thread:

I wasn't going to mess with it again in what little is left of the weekend, but I tried that last experiment. I connected a Flash-less 7722 (7722 from which I have desoldered the Flash chip) on the SCSI cable and got nothing in Apple System Profiler.

I have found in the past that a flashless 7722 still shows up as a device on the SCSI chain with a message on one of the lines similar to "BadROM16K" or something like that.

So, I tried the first SCSI cable again, and this time the flashless 7722 showed up as expected.

Then I hooked up the 7722 with socketed Flash programmed with 1.69 7726 firmware and connected the 400GB Seagate again. It showed up in ASP, but the drive did not mount on the desktop. It was already initialized earlier and should have a bunch of copied content on it.

I used Drive Setup (or whatever came with OS9) to initialize the drive and it locked up -- which doesn't surprise me, since it's failure to mount suggested issues.

After rebooting the 7722/7726 no longer appeared in ASP.

This strongly suggests to me that I've got intermittent flaws in both SCSI cables. Maybe a broken conductor that makes contact when the cable is wiggles one way and not the other way. Move the cables to switch 7722 units and things may stop working.
 

trag

Well-known member
Okay, I bought new cables. I got an Adaptec 29160 SCSI card and ditched the Atto UL2D. The new cables gave me just as bad results on the UL2D.

Now with the 29160, I seem to be getting consistent results.

Actually, one thing that may be the deciding factor is I was able to find the Adaptec Utility for the 29160 and I set the Termination to On instead of Auto. I actually had a couple of problems initially when I swapped cards, but I quickly made that change in settings.

The UL2D might be fine if I could do the same on it, but I haven't been able to find the utility for the UL2D.

Both the soldered and the socketed PLCC Flash adapters seem to be working properly.

Also evidence suggests that the flasher will flash a blank Flash chip. I'm going to try just erasing one of the TSOP chips and soldering it back in place. That would save the trouble of inserting a byte of 'FF' in every second byte when programming the chip.
 

mg.man

Well-known member
I was able to find the Adaptec Utility for the 29160 and I set the Termination to On instead of Auto
FWIW, I had similar issues when using a Formac SCSI card in a B&W G3... setting the TERM to ON cured it. I may have some ATTO utils... will have a look when I get a moment. BTW, was your 29160 a proper Apple one? I have a stack of "PC" ones, but have never managed to get them to work in a Mac... 😞
 

trag

Well-known member
FWIW, I had similar issues when using a Formac SCSI card in a B&W G3... setting the TERM to ON cured it. I may have some ATTO utils... will have a look when I get a moment. BTW, was your 29160 a proper Apple one? I have a stack of "PC" ones, but have never managed to get them to work in a Mac... 😞

Yes, I found one on an obscure Mac reseller site. I can take some photos if you want to do a component comparison.

https://www.usedmac.com/products/index.php?route=product/search&keyword=29160&category_id=0

Unfortunately, soon after I posted that, things went back to unreliable.

I'm going to switch my test bed to a Beige G3. I've been using a PCC PowerWave logic board and a power supply on the desk, but the PowerWave uses a PCI riser card which puts all the PCI cards horizontal. I've been sort of supporting the far end of the cards, but I suspect that there's just too much ability to wiggle.

I can't believe there's something about the converted 7722 that makes it work one moment, and completely fail to appear on the bus another, and sometimes cause the SCSI card to just appear as generic "PCI" in ASP. I think the connections just aren't reliable.

Might be a need for contact cleaner too. Sigh.

I need a reliable test bed before I can do meaningful testing. Sigh.
 

Dave L

Member
Hi all - have been off the grid for a little while, so just catching up.

I just saw this new listing on eBay: https://www.ebay.com.au/itm/165076746672

If you look closely at the photos, it looks like a 7722 but the firmware sticker is placed over the area on the PCB where it would normally say 7722. The guy is advertising them incorrectly as a 7732 (which is actually a SCSI to SATA adapter).

I think this person on eBay has worked out how to do the conversion - so there is hope for your experimentation @trag
 

trag

Well-known member
@Dave L I think mine are working. At this point, I believe that my SCSI setup (host machine/card/cable) is unreliable for some reason. Actually, I'm pretty sure it's the fact that I'm using a PowerWave which has a PCI riser card which means all the PCI cards are hanging horizontally, since I don't have the board in a case.

It doesn't really make sense to me that things would work perfectly one moment, with an attached 400GB Seagate ATA drive showing up and initializing and accepting several GB of data transfer, and the next boot after changes that jiggle the rig, the drive doesn't show up, or the SCSI card appears as a generic PCI card in ASP, or doesn't appear at all.

You're in South Australia? If you were in the states, I'd say, just let me ship you one or two of the two that I believe are working and if they test okay, we would work something out.

My testing has been delayed because I need to tear down my current test rig and replace it with my Beige G3, but last Thursday, my back started spasming, so I'm not lifting any computers until I'm past this.

So it would be faster in teh long run to just send them to you, if you weren't so expensively far away. :)
 

trag

Well-known member
Just to update where I am.

When I ordered five (5) of the 7722, I received two different revisions of the board.

I received two Revision 1.5 boards, which has a position for a PLCC Flash and has an W29EE512 on board.

I received three Revision 1.8 boards, which have positions for both a 16 bit wide TSOP Flash and the more common 8 bit PLCC flash. The TSOP position is populated with an SST39VF04 (IIRC). I may have that wrong, working from memory. It's a 16 bit wide, 3.3V flash with 4 Mbit capacity.

As stated above, I believe the two boards that came with PLCC Flash chips are working properly and that my test rig is unreliable.

Switching to a more reliable test rig (Beige G3) is delayed while I wait for my back to stop spasming.

The last test I did was to solder wires on to the Vcc pins of the 7722's which came with TSOP flash chips. I soldered the wires onto the Vcc pin for the PLCC Flash position. Measurements were consistently 4.88V while powered on.

So the PLCC position is getting 5V. I installed a PLCC socket on one of the boards, but again, am waiting to test until I set up a reliable test rig.

This suggests that the Rev. 1.8 boards (they come with 3.3V TSOP flash, but have position for 5V PLCC flash) would work if one simply installed an SST29EE512 in the PLCC position.

This may be easier because desoldering, cleaning, programming and resoldering the TSOP chips is challenging because of the tiny pin pitch.

The board shown in the Ebay auction that @Dave L linked to above is a Revision 1.5 board, which has only the PLCC position for Flash.
 

trag

Well-known member
I finally swapped out my test rig. I stopped using the bare PowerWave board because the PCI slots are on a riser, putting the PCI cards in an unsupported horizontal stack when the logic board is not cased.

On my Beige G3 I'm getting consistent results, finally.

My first try last night, actually had me worried that things were still indeterminate, as at first the machine would not boot with the converted 7722 attached. However, I tried swapping SCSI cables (different brand/make) and that resolved the issue.

A little disappointing, as both SCSI cables were brand new with terminators. In the German thread about this conversion the original poster repeatedly states that the 7726 is very finicky about SCSI cable quality, and he seems to be correct.

Anyway, the two 7722s on which I've installed PLCC sockets are working great. One, where I think I resoldered the chip directly seems to have a corrupted ROM, but I thought it was working before. Two other cards haven't been modified yet.

So I have one rev. 1.5 board and one rev. 1.8 board working.

1) Next steps are to desolder the flash on the corrupted system, reprogram the flash, and probably reinstall with a socket.

(not sequential)

2) Erase one of the SST39VF400 TSOP chips and see if the Acard flasher will flash a blank chip. That would save me trying #3.

3) Load Strawberry Perl on my Windows laptop and write a little routine to insert a byte of 'ff' between every byte of code on the 7726 firmware. This is necessary to use the SST39VF400 chips that are on the Rev. 1.8 boards. The chips are 16 bits wide, but only 8 bits of the chip are connected/used by the 7726. So when the firmware is programmed on teh chip, from teh chip's point of view, every high byte is blank.

Or I could just buy a supply of W29EE512 chips in PLCC package. Seems inelegant though.

One last curiosity. After the conversion, the SCSI ID seems to be stuck at 8+. In other words regardless of the status of the jumper for SCSI ID 8, all SCSI IDs behave as if that jumper is in place. @Michael_b do you remember if you saw anything like this? If you used yours only on 50 pin interface you wouldn't have noticed it, as the 50 pin interface ignores/lacks the ID8 bit.
 
Last edited:

Michael_b

Well-known member
One last curiosity. After the conversion, the SCSI ID seems to be stuck at 8+. In other words regardless of the status of the jumper for SCSI ID 8, all SCSI IDs behave as if that jumper is in place. @Michael_b do you remember if you saw anything like this? If you used yours only on 50 pin interface you wouldn't have noticed it, as the 50 pin interface ignores/lacks the ID8 bit.
I only used mine on the 50 pin interface on the Beige G3/SE/30 so I can't comment on that. Glad you've sorted out the inconsistencies, though.
 

trag

Well-known member
This is all great news - well done! I'm keen to buy one whenever you're ready to part with one.

I've got a little more testing to go on these before I'm ready to part with one. If I had all five working, I'd be ready, but with only one each of the rev. 1.5 and rev. 1.8 working, I want to hold these two until I get the issues with the others resolved.

I did some more work on it tonight. I took the W29EE512 (PLCC) chip out of the socket on the working rev. 1.8 board and erased the W29E512. Then I put the chip back in the socket and hooked it up.

With the SCSI-Bridge utility ver. 2.05 and the 1.73q firmware, I was able to update the board with a blank flash chip. Testing after that showed it working just as before.

I already had the SST39VF400 chips (TSOP) removed from the three rev. 1.8 boards. So I erased the three TSOP chips.

Then I soldered two of the TSOP chips onto the two Rev. 1.8 boards on which I have not installed a PLCC socket.

Then I attempted to flash them. And failed. Every time, the flashing utility just spins forever. It's willing to perform the update, but my best guess is that the utility is sending the Erase Chip command, and it's not getting the proper STATUS bits back from teh flash chip in response.

Maybe the flashing utility doesn't support the SST39VF400? Those were OEM boards. Perhaps they were never meant to be updated.

Another possibility is that my soldering on both boards is bad. TSOP has awfully fine pitch pins. But I redid the soldering twice on both boards and got the same result every time. Examined under brightly lit magnification.

I also tried touching up the soldering on the non-socketed Rev. 1.5 board that isn't working, but that didn't help.

One interesting thing I found is that the SCSI IDs are proper when there's no firmware on the board. After usable firmware is added, then that +8 thing starts happening. I wonder if there's some slight wiring difference between a 7722 and the 7726. I wish I had a netlist, schematic or actual 7726 to examine.

Next:

1) Pull the TSOP flash back off of the two Rev. 1.8 boards, and try programming them directly with the 1.73q firmware, or one of them with 1.73Q and the other with 1.69q, perhaps. If that works, then try updating and see if that works or not. My expectation is that programming directly will work, and updating still won't work. I'm still debating whether it would be smarter to just throw away the TSOP chips and buy some extra PLCC flash chips for these boards.

2) Pull the PLCC chip off of the non-socketed Rev. 1.5 board and probably install a socket. I think it will save time in the long run.

3) I have a small collection (8) of SST29E512 chips, which I think I desoldered from Adaptec 2940UW boards, 20 years ago. I'm going to try one or more of those in the boards with sockets. The SST and the Windbond 29EE512 may use the same programming codes -- fingers crossed. If they do, then I can convert up to eight of the Rev. 1.8 boards without messing with TSOP and without ordering extra chips.

4) Test the working converted boards with optical drives. I think I did this and found that they still work, but I can't remember for certain.
 

trag

Well-known member
Made a deceptively large amount of progress this weekend. Feels like tiny progress is why it's deceptive. A lot of it was on other cards.

Last weekend I desoldered a bunch of chips off of cards.

This weekend, I soldered PLCC sockets onto several cards (many of them unrelated to the 7722/7726.

I also programmed one SST39VF400 (TSOP flash from ver. 1.8 boards) with the 1.73q firmware and soldered it back on a board. This is a bigger accomplishment than it sounds, because first I had to write a routine to insert a byte of 'ff' in between every other byte of the 1.73q firmware. This is necessary, because the SST39VF400 is 16 bits wide, but the 7722/7726 only uses the lower byte. So only the lower bytes can contain the firmware. If one just programs the stock firmware onto the chip, every other byte would be lost in the unused upper bytes.

At first it didn't work, but I put it under magnification, found what I think was residue from cleaning. At least, it came off with a pick, with no heat, so I don't think it was stray solder.

After that the ver. 1.8 with TSOP flash chip worked great. I didn't try flashing it with the utility to determine if it's reflashable. I'm not sure why one would want to since 1.73 seems to be the latest available. But mainly, if it corrupted the firmware, I didn't want to have to desolder, clean up, and resolder the chip again. After I get a second one working with TSOP flash, I'll test in-system flashing.

I also put a socket on another ver. 1.5 board and tested it with a normal PLCC flash and as expected that worked.

I tried the SST29EE512 PLCC flash in the sockets of a couple of the units, but I could not get them recognized. Makes me wonder if the firmware checks the chip IDs.

However, I had trouble on my chip programmer with the SST29EE512 chips, where if they were all the way at the bottom of the socket they wouldn't work. They had to be inserted so the top of chip was just level with the top of socket. So the issue may be a contact issue, not a chip compatibility issue.

I'm going to try some contact cleaner on them are retry.

Status:

2 ver. 1.5 7722 with sockets installed that work as 7726 when PLCC Winbond 29EE512 chips with 1.73q (or 1.69q) firmware are installed. Seems to be reflashable.

1 ver. 1.8 7722 with PLCC socket installed over removed TSOP flash. Works as 7726 when PLCC Winbond 29EE512 chips with 1.73q (or 1.69q) firmware are installed. Seems to be reflashable.

1 ver. 1.8 7722 with TSOP SST39VF400 chip removed, programmed with 1.73q firmware and reinstalled. Works as 7726. Haven't tested in-system flashing.

1 ver. 1.8 7722 with TSOP SST39VF400 chip removed. Awaiting installation of reprogrammed TSOP chip.

SST PLCC flash waiting to be retried after contacts are more thoroughly cleaned.

@Dave L Getting closer. There's not a lot of experimenting left to do.
 

trag

Well-known member
Dead in the water. My test rig died on me, again. I guess every machine in the house needs recapping at this point. Or something. Electronics hate me. Sigh.

BTW, how many folks want one of these? In the original thread I jokingly suggested selling them for $125, but I'm thinking $60 plus shipping is probably reasonable.

By the time it's all paid it costs about $25 to get the unmodified cards. Then with time and soldering materials and such, another $35 seems reasonable to me. Plus I sold a bunch of 7720Us for $39 each about 15 years ago. So, with inflation...

These have a wide SCSI interface on them. You'll need an adapter to use them on a narrow bus.
 
Top