Fastlane Z3 / SCSI problems

[amigasith]

Cheers, folks! I recently got a very nice Fastlane Z3 for my A4000 Desktop, which is now giving me serious headaches I managed to get it to work reliably after trying lots of different things, but now I have a situation where I need your fellow advice...

So here it is: The Fastlane works really well if I leave the desktop case open. I can run file copy activity, etc. for hours and the A4000 is absolutely rock stable. No lock-ups, no file system errors, no gurus, nothing.

As soon as I close the case and run the same workload, it takes maybe half an hour and strange things start to happen: PFS3 throws errors like "Wrong dirblock id" and randomly a lot of other error requesters. The "scsiconfig" tool from Phase 5 says that the drive's RDB has a checksum error, etc. It looks like the OS cannot access the drive reliably anymore. I don't get lock-ups or gurus, however. All the issues are related to the file system and / or the SCSI disk.

When I let everything cool down for maybe half an hour, everything returns back to normal: scsiconfig is able to read the RDB without a checksum error, PFS3 does not throw error requesters anymore, etc.

While leaving the case open, I used a hot air blower to heat up the different parts: First the Fastlane itself, then the A4000 mainboard, then the drive, then the PSU, but nothing happened. I simply cannot reproduce the strange behavior when the case is open.

-> Could it be something else than temperature?

Another important thing to mention is that the A4000 does not have any problems like this when using the onboard IDE with a real hard disk. I can leave the case closed and run all kinds of funky stuff for hours and everything runs absolutely rocks solid. Therefore, I think it must be related to the Fastlane and / or the SCSI drive somehow.

-> Could it also be related to the SCSI hard disk? It's an old IBM DCAS-32160 from June '97.

While writing this, I come to think that maybe a hard disk issue could be the root cause of all of this. Can an old drive behave like this? I mean working okay when cold, and returning bad data when heated up?

Any tip is greatly appreciated

[mark_k]

Are you able to put the SCSI HD in an external case (with a fan)? See whether you still see errors with it connected that way. The idea being, the drive would hopefully stay cooler like that.

[amigasith]

Hi mark_k, thanks for replying! When I finished my 1st post, this was exactly my thought, too

Although I don't have an external case, I did a similar experiment: I let the SCSI cable and a molex power connector hang out of the Amiga's case and connected the drive that way. Then I closed the case and started a small Amiga DOS script that does some file archiving in a loop. After maybe 20 minutes or so, the Fastlane / SCSI drive started acting weird again: The "Wrong dirblock id" error requester popped up again and scsiconfig complained about the RDB having a checksum error.

-> I think I can rule out the SCSI drive from the equation...

[Kin Hell]

Your A4000 obviously has a Buster 11 fitted, otherwise the Z3 Fastlane would not work. If you're running an 060 CPU, you need V8.5 ROM on the Z3. If your Buster 11 is socketed, strip the Amiga & get the motherboard on a flat, firm & Anit-Static suface. Push down firmly on the Buster 11 Chip to make sure it is fully seated. Do all the chips in a socket just to be sure. Also make sure your CPU card is fully seated along the entire length of the CPU Slot connector before you shoe-horn it all back into the case. Make sure all the socketed chips on the Z3 are batted down too & check the RAM on the Z3 if fitted. Does it stay stable with closed case & no RAM on the card....?

There is way too much space between the Motherboard to chassis standoff's to do this in the chassis. You could cause serious damage doing so. The CPU slot connector suffers "gap-age" in the mid area when fitting the CPU card to the mobo whilst in the chassis. By removing the front drive bay, you can jam your right fingertips between the underside edge of the mobo & chassis whilst stabbing the fingers on your left hand pushing the two together. This eliminates the gap-age in the middle of the slot connector.

I'd still check the CPU slot connector even if your Buster 11 is surface mounted. (not many were)

Other factors could be failing PSU, Mobo needs Re-Capping, restricted Air-flow... At least do the above to eliminate these possibilities. Heat will cause sockets & connectors to move, When failing caps get overly warm, they will do weird things too.

[amigasith]

Hi Kin Hell, also a big "thanks" to you for replying, cool Before I start explaining what I tried from your suggestions, here is a bit more information regarding my setup:

- As you already guessed, I have a Buster 11. It is socketed.

- The A4000 mainboard is a Rev B and had all its capacitors replaced before I got it.

- I measured the voltage on the PSU after cooling the entire system down and after letting it run for a while. It gives stable 5.05V and 11.99V.

- The Fastlane has 16MB onboard - 16 x 1MB SIMMs.

- I run an A3640. It is a Rev 0, but I have upgraded it to a Rev 3.2: I applied the patch and jumper mod on U200 for /STERM, I changed the RC circuit R400 / C400 to a real reset generator and I changed U209 to 391472-03 and U204 to 381470-02. I also changed all capacitors to ceramic ones. Besides this issue with the Fastlane, it runs super stable.

- The Fastlane manual says a Rev 3.0 A3640 requires a 74FCT240 in U103 on the Amiga mainboard, but that one has to check the revision of U209, too. Although my A3640 is a Rev 3.0, it has both the patch and jumper mod and also U209 in Rev -03, so I assume changing U103 to74FCT240 is not required. I have the original 74FCT244 in U103. I also think that the Fastlane probably wouldn't work at all if I had the wrong 74' IC in?

Okay, so here are the things that I tried thanks to your suggestions in addition to what I've tried already:

- I completely took the A4000 apart and removed the mainboard. I pulled Buster out of his socket with a PLCC extractor tool and cleaned its legs with with very fine sandpaper. Then I cleaned the legs again with IPA. Unfortunately, the behavior is the same: Around 30 minutes warm-up and the OS does not want to talk correctly to the Fastlane and / or the SCSI drive anymore.

- When I was upgrading the A3640, I also noticed the "gap-age" on the KEL connectors, therefore I already knew what to do With the mainboard still out of the desktop case, I pushed the A3640 all the way down before putting the mainboard back into the case. By doing so, the two KEL connectors from the A3640 and the mainboard are as connected as possible.

- I disabled the Fastlane's memory by setting the jumpers correspondingly. The memory doesn't show up in Workbench anymore. But still, the strange behavior is the same: Around 30 minutes warm-up and the OS does not want to talk correctly to the Fastlane and / or the SCSI drive anymore. In addition, I ran memcheck in this state and it did not report any errors.

- When the system is in this strange "I don't want to talk correctly to Fastlane / SCSI" state, everything else works just fine. I can still boot from floppy and do things as always.

I am starting to run out of ideas now Maybe re-flow the GALs on the Fastlane and all connections with a bit of fresh solder?

[Paul_s]

The top of the case isn't shorting any components by chance?

[amigasith]

Hi Paul_s, welcome to the bug hunting club Yeah, I had that thought, too, but why would it take almost half an hour for the short to surface? That doesn't make sense somehow...

There is another detail that I forgot to mention: When the Amiga is in this weird state, sometimes the loading of programs fails for obvious reasons. I mean if the OS cannot access the drive reliably anymore, it's obvious that the transferred data might be garbage. Anyway, this loading of garbled data sometimes leads to guru meditations after clicking the respective error requesters away. The interesting detail is that it's always 8000 0004.

Remember, I can still successfully boot from floppy without any gurus when the Amiga is in this state. So it must be related to the Fastlane, SCSI, or DMA somehow. At least this is my conclusion...

[roomeo]

What about pulling out the memory completely off the fastlane, not just disabling it. I think the fastlane draws about 2.5 amps with full ram setup. could drive temperatures up?

just my two cents.

[amigasith]

Hi roomeo, also a big "welcome" to you for joining the bug hunting club and thanks for your post.

Although your tip is definitely worth trying, I am a bit reluctant to actually implement it Even if it was working, I wouldn't want to leave the Fastlane like this, since I'd really like to keep the extra 16MB. There must be some "proper" solution to this...

So here is my next plan: I ordered a 74FCT240 that should arrive in the next couple of days and I will give this a try. Although my Rev 3.0 A3640 has both the cut and jumper patch for /STERM and also a -03 revision for U209, maybe it's not enough for the Fastlane, who knows. Dave Haynie described the A3640's /STERM sampling as "right on the hairy edge", so there is a possibility that the issues that I am seeing is a timing thing. And that it only surfaces when the components heat up. I did some research, and I read that CMOS circuits can change their propagation delays with temperature. If it's really "right on the hairy edge", a few degrees Celsius could cause a large enough shift in the timings to make the A3640 miss some signals.

My hope is that changing U103 to a 74FCT240 _might_ improve the situation. As far as I understand, the worst thing that can happen is that the Fastlane doesn't work at all with the 74FCT240 in place. At least I will be in a position to rule out this possibility if the experiment fails If the 74FCT240 doesn't work, I can still try a faster 74FCT244. The one that is originally on the board is of a slower type.

I'll keep you guys posted about my experiments. The bug hunting stays exciting - at least for me

[roomeo]

thanks for the welcome..

Also... not only could all that ram create heat.. but they are responsible for ALOT of connections that could be bad..
I would pull out everything that has nothing to do with the scsi operation.. then you know for sure where the problem is NOT hiding. every zorro card. bare minimum of memory on the mb. etc

i doubt the fastlane would generate that much heat. My fastlane is fully populated with 4mb chips, and its running rock stable with the case closed.
btw..Do you have active cooling on your cpu?

Im running the a3640 3.0 with the cut'n'jumper applied. I dont think that chip on the motherboard has been changed. I'll have to check.

[amigasith]

The Fastlane is currently the only Zorro card in my A4000D, so there is hardly anything more to remove And yes, there is active cooling on the CPU. I have a heatsink plus a fan. The 040 does not get really warm - not even hand-warm.

If you check the IC in U103 - which would be awesome! - can you please post it's full spec? I am particularly interested in the suffixes (P, TP, ATP, etc.).

[roomeo]

no problem.. give me some minutes and i'll be back with that info.

[roomeo]

74FCT
244TP
9234P

Thats the original one, isnt it?

[amigasith]

Awesome, roomeo Thanks a lot And yes, that's the original part. Nothing modded / changed on your board.

Hmm... My experiment will probably fail with this additional insight. I'll give it a try anyway - the part is already on it's way... Again, thanks for checking - really cool!

[roomeo]

np

[roomeo]

i would put my money on
1. Your harddrive / scsi cable
2. Fastlane mem

Hope you get it sorted out..

[Kin Hell]

Just to endorse one or two feelings here, definitely try with the ram off the Z3 & not just jumped out. GURU Meditation software error 8000 0004 usually means Memory problems. However, it might not just be a physical memory issue, but the lack of U103 having a 74FCT240 fitted to the mainboard. Incidently, your A3640 wouldn't work in your A4000 without a Buster 11 fitted or U209 on the A3640 reading -03; But only with either a Z3 Fastlane or C=/DKB A4091 SCSI card in it. There are some serious DMA conflicts otherwise & the Amiga will mis-behave. This is what you may be witnessing.

Try Ram off Z3 first.

[amigasith]

Thanks for the continued support, guys I really appreciate this! I'll try with removing the RAM once I get to it.

There is another thing that I fear I have to start taking into account seriously: another broken trace on my A3640

Why "another" you may ask yourself. Well, here's the story: When I got the card, it was sold to me as "fully working". Oh well... After I recapped it, I noticed that the Amiga would freeze after running for something between 15 minutes and 1 hour. The exact interval was more or less random. I could, however, reproduce it without closing the case. After many hours of checking the traces and vias, I finally found a via that was destroyed by capacitor / electrolyte spill. I bridged the via and the A3640 was working perfectly ever since. Until I put the Fastlane and the SCSI drive in...

When the Fastlane / SCSI problems popped up, I did not think of the possibility of another broken trace at first. The symptoms are different this time: (1) I can only reproduce the issue when the case is closed and (2) the Amiga does not freeze or crash or whatever. As I said, I can still boot from floppy without any troubles when the Amiga does not want to talk to the Fastlane / SCSI drive correctly anymore.

But the longer I think about it, the more it seems likely to me that there must be another broken trace or broken via somewhere. My theory now is that a broken trace / via does not necessarily lead to a freeze or crash. Maybe it's "just" a line that is related to DMA somehow and that line was not actively used or very rarely used until I put the Fastlane in. I don't know if such a line exists on the 68040, but it seems possible to me. Maybe there are some gurus who know...

Anyway, I will try to find some more time to check traces and vias once again... Sigh

[amigasith]

I have a working system - HOORAY

For the impatient: At the end of the day, I'm not 100% sure what exactly fixed it, it probably was a combination of things. All I can say is that it's running stable now - happily copying files for hours

If you've got a minute, here's the full story: I spend an entire evening doing nothing else but continuity testing. And guess what? I actually found another non-working via! Just as I suspected. The entire area looked a bit suspicious and so I desoldered all the resistor networks next to the CPU. Here is how it looked like right after desoldering:



You can see that the capacitor spill really did quite some damage. After cleaning everything up and also stabilizing a few traces with fresh solder it looked like this:



Much better now. The next evening was dedicated to re-soldering and continuity-testing all the parts again. Here is a picture of half way through:



It worked better than I expected... Here is the final result from above:



Although it was a hell of a lot of work to do, I really like the new traces There are even some extra wires, in order to guarantee better continuity - see the new wire at the right bottom of the largest resistor network, for example. Here is the back of the PCB, also with an extra wire - to the left of the freshly soldered area:



And last but not least, here is my A3640:



A bit franken with all the extra wires, but hey it works great now. I also very much like the home-made 10K "SMD" resistor at R400

But now back to the story: When I put the freshly repaired A3640 back onto the mainboard, I thought I was done. How wrong I was! Although it did work much better than before - it now took around 45 minutes - I finally got the dreaded "Wrong dirblock id" PFS3 requester again...

So my next step was to test the newly arrived 74FCT240. Just as I expected, the Amiga did not even want to boot with it. This is because I already upgraded my Rev 3.0 A3640 to a full Rev 3.2. Nevertheless, I wanted to exclude this possibility.

After putting the original 74FCT244 back in, I removed all the RAM from the Fastlane, just as you guys suggested. And Bingo! For the very first time, I was able to run for 2 hours without any file system errors, etc.

My next test was to clean the memory sticks and put them back in. The guy from which I got this card had mixed memory sticks in all the banks - all of them were 1MB, but they were from different vendors. I managed to get 2 sets of matching 4 x 1MB SIMMs out of this, so 8MB in total. That's what I put back in. And now it's running stable for more than 2 hours, happily copying files with the case closed - awesome

The only thing I'm not sure about is what actually fixed it. It could be that the 74FCT244 simply needed a reseat. I forgot to run a test after putting back in the original 74FCT244, but before removing the RAM. The other possibility is that the RAM needed a reseat. Either way, the fixing of the bad vias on the A3640 was also a good thing to do - makes me sleep much better now.

And thanks for taking the time to read through all this Hope you liked it

[roomeo]