DL240 Quantel HAL Express & Dylan Restoration Part 4 [Long & Technical!]

www.pcbway.com - The full featured PCB prototype service!
www.peakelec.c... - Electronic Test Equipment for the Hobbyist, Student, Maker and Professional
In this long and technical video i look deep inside the Quantel Dylan 'Chatterdisk' video disk array how it works and offer some insights into why it doesn't!
Discord:
/ discord
Patreon:
/ dexterslab2013
Mastodon:
mastodon.socia...
Twitter:
/ dexterstechlab

[@tekvax01]

The long-awaited dyan drive crate! I remember once replacing a couple of the failed SCSI Seagate drives, with spares that had been sent to us from Quantel. One of the replacement drives had its main 12-volt and 5-volt capacitors installed backwards. I replaced the drives, restarted everything, and made sure the RAID rebuild process had begun correctly. I then left for lunch, and returned an hour later, to see the entire equipment room filled with smoke from the failed caps! Of course, the entire system had now crashed, and the rebuild process had failed as well. I ended up talking with CS tech support for the rest of the afternoon so that the data could be recovered. That made for an interesting couple of days!

[@heinrichhein2605]

Would it be remotly possible to get the designs to an interview, so they can maybe tell a view story designing such beasts. Just would be unbelivable!

[@DextersTechLab]

We are working on that very idea! There is also a proposed documentary on the impact the Paintbox had.

[@frankowalker4662]

@@DextersTechLab That would be so cool.

[@juliansmith1888]

@@DextersTechLab Perhaps interview Dave Chapman who worked on Doctor Who, and the chap who did matte paintings for Who at the BBC on a Paintbox.

[@heinrichhein2605]

@@DextersTechLab You are amazing!

[@BeauTardy]

"Spinning up the Dylan drives" was a thing back in 1993. :)

[@frankowalker4662]

The way it distributes the Thunk could get a bit confusing, but it does make sense. It's great you are in touch with the designers. It's also a shame you can't get Dylan running.

[@DextersTechLab]

Thank you for the comments Franko, it is a shame but i need to get on with other projects. I will probably re-visit Dylan at a later date.

[@JensChristianRestemeier]

I think there were some ISDN modem cards for the German telephone network that were using Transputer chips - IIRC old cards were quite popular for playing with the chip, as you could upload your own firmware.

[@DextersTechLab]

Yes i think they started to get used for embedded applications rather than as a complete computer solution.

[@merseyviking]

Fantastic stuff Mark! When you said to grab a brew and settle down, I was already one step ahead of you :) It'd be a bit of a pain (and probably too late now), but you could try swapping out chip 12 for chip 11 to see where it stops.

[@DextersTechLab]

Hey John, we have already ruled out a problem with the FAS236 ICs on that board, even without it plugged in it still stops at 12! It's a challenging problem for sure!

[@merseyviking]

@@DextersTechLab Yeah, figured you would have, especially given the hardware designers being on hand. It's frustrating for me, it must be unbearable for you!

[@thomasguilder9288]

I hope it‘s not an issue with one of the cpld‘s. I guess if the addressing is splitted over four chips as there were no big enough ones it might be that drive 0 to 12 are receiving their cs from one cpld as the remaining chip selects come from another one… might be interesting to trace the cs lines to verify this theory… as the cpld‘s are socketed it wouldn‘t be too hard… Can you attach a logic analyser to the data bus on the transputer? Even without actual code the steps configuring the 20 scsi controllers should look quite repetitove, also with the stuff written to the cplds for choosing the proper cs lines. Perhaps this gives some insight to when and why it stops executing, perhaps it just waits for an ack that never comes

[@DextersTechLab]

All excellent points! One issue with the logic analyser approach on the transputer is it uses a multiplexed bus, this is how it managed to be 32 bit and be in such a small package. So the bus switches between an address bus and a data bus. It can be done, it's just more work and i had already spent close to 6 months working on HAL & the Dylan.

[@tekvax01]

I'm guessing that no other Dyan drive crates to compare with are available? can you swap around the CPLDs so they feed different SCSI controller CS locations? I'm guessing you have already checked the traces from the CPLDs to the SCSI control CS pins for continuity

[@DextersTechLab]

hi Dan! Yes we have checked all the CS, they all trace back to the pins on the CPLDs. We have also swapped what we call the 'wing' cards over (the ones above and below the PSU as you see it in the video). It turns out the connectors are wired identical. That made no difference. Interesting idea on swapping CPLDs, i'd have to check their part numbers and see if they are identical.

[@tekvax01]

Scoping out the memory modules, might be a good idea too… do you have access to a logic analyser?

[@DextersTechLab]

I do yes but the SMD parts with memory in it's tightly packaged ZIPs and the Transputer in its PGA package make it extremely hard to use a logic analyser and get meaningful info. The Transputer also has a multiplexed address and data bus which further complicate things.

[@BeauTardy]

My memory is fuzzy but it seems to me that the Dylan only worked with the Ramcorder, not the image library. Not sure if that is relevant to your issue here.

[@DextersTechLab]

Yes, Dylan only stores video, any still pictures, cutouts stencils etc are stored on the System or Shared disks.

[@Spritetm]

If you're willing to hack the PCB, what you could do is effectively swap two SCSI controllers by cutting the CS PCB trace to them and adding a bodge wire to cross-connect them. (So e.g. route CS for chip1 to the CS pin for chip 12 and vice versa.) If the problem switches along with it, it's on the side of the SCSI controller; if it's on the side of the common control logic, the problem will stay the same. Edit: Ah, you said below you managed to do the same by swapping the 'wings' cards. That would accomplish the same. On a more software-wise approach, perhaps it's an option to grab the Transputer firmware from the HAL, disassemble it to figure out how it talks to the SCSI controllers, then write a custom payload that allows you to poke those chips manually? Combined with a scope to see what the actual signals look like if the Transputer 'orders' them, it may give a wealth of insight.

[@DextersTechLab]

Hey, i did try something similar to your suggestion. I lifted the CS pin for number 6 to see if it would hang earlier, it didn't. It just continued as before stopping at CS12. Very odd indeed. Yes we also swapped the wings over and nothing changed in that case too.

[@worldother2080]

se podra emular el programa de quantel hal en tu pc con tarjeta grafica, me encantaria probar el programa :(

[@akspa0]

Just to clarify - is it that there's an absence of the chip select signal, or is it not being pulled low? As for dumb ideas, has the connecting cable been continuity tested to ensure proper connectivity with the HAL? Since you mentioned replacing some of the surface-mount caps, did you see any electrolyte damage in that area?

[@DextersTechLab]

hey, thank you for the comments! All the chip select lines have been continuity tested and check out, the chip selects only connect to the CPLDs and then onto the transputer via a separate bus. The connections from HAL to Dylan have been checked also, if they were not working we wouldn't have been able to complete the initial boot process. There was no electrolyte damage thankfully, changing the caps actually didn't change anything.

[@pbpraha789]

Love this series :) Did you try swapping disks / controllers to see if the problem also moved? e.g swap disk 12 with disk 2 etc

[@DextersTechLab]

Hey, thanks for the suggestion. In this state, swapping drives, even removing them makes no difference. We are very sure the fault, whatever it is will be located on the main board (the large board to the right as you see it in the video)

[@pbpraha789]

@@DextersTechLab Oh, that's unfortunate. I hope you are able to fix it and enjoy it working again. Also, thank you for uploading "Painting with Light", it was this very program that inspired 14yo me to write a painting program on my Commodore Plus 4, sans mouse/tablet/mass storage. Granny was very impressed when I plugged it into her TV. If it wasn't for that show I'd likely not be the software engineer I am today :) Thanks for the memories and good luck :)

[@DextersTechLab]

Thank you and what a great story. Maybe one day you might get to try a Paintbox, i do try and get them out for people to see and use.

[@Darkstar2342]

Very interesting video. A couple of ideas on what you could still try (but I'm sure these have all been suggested to you before) a) When you disconnect the drive+cable from chip #12 does it still fail? b) It could still be bad memory (maybe only the part of the memory responsible for chip #12 is bad). ZIP memory was very fragile and sensitive, I have seen lots of bad chips on the Amiga (ZIP memory was used extensively there). You can test the memory with e.g. the RetroChipTester Pro but you'd have to unsolder them all c) what's the connector PL4 for? Maybe there's some serial debugging port or something similar on there? d) What happens when you disconnect the whole "upper" daughter PCB from the main PCB? Maybe the system will boot further with a reduced number of disks available? (I have no idea if there were units sold that had less than 20 drives, but even if it later errors out, it might be worth trying) e) any other of these diagnostic commands could be helpful? INISYS, COMMS, FIFOS, ...? f) have you dumped the code and analyzed what it does? That might lead to some clues. The instruction set of these transputers is reasonably well-documented, even if not a lot of tooling seems to exist...

[@DextersTechLab]

Thanks Darkstar, i appreciate the comments. A: yes B: i would think this unlikely, the FAS236 SCSI controller will only access that ram when it's asked to read or write data to the disk. This hasn't happened yet, the transputer is only trying to configure the registers. We have also swapped the two smaller boards (the boards are mirrors but the connectors are the same) and it will still stop at 12, in fact even with the board disconnected it stops at 12.

[@Darkstar2342]

@@DextersTechLab for B, I actually meant the transputer's RAM, sorry, should have made that more clear. If there's a broken bit in parts of the RAM, the code could simply crash or enter an endless loop or something

[@mikemines2931]

I have six pieces of Peak test gear and I've never regretted buying any of them plus it's a bonus buying British. No I've no connection, I believe in credit where its due.

[@DAVIDGREGORYKERR]

Can you use 4x 5TiB SSDs as I think you can get a SCSI LVD RAID adapter.

[@DAVIDGREGORYKERR]

If anyone living in Oregon looking PCBs what about Osh Park PCB Fabrication Plant.

[@FrozenHaxor]

Generally not worth it, you'll get 100x better price from chinese fabs with quicker turnaround.

[@markdjdeenix6846]

Desolder all the ram and test It’s the only thing I can think of

[@markdjdeenix6846]

I only say this because the ram looks similar to the Amiga 3000 ram and it is 30 years old and fails to this day

[@DextersTechLab]

Thank you, yes it is a possibility but generally faults like that i would expect to put a CPU eventually into a tight program loop or a halt condition which doesn't seem to be the case.