Looking for info on Trifecta XL IDE/SCSI adapter (manual, jumper settings)

[berke]

Hi all,

Found myself an A2000 which includes a Trifecta XL controller. This controller has an IDE port (44-pin) and a SCSI port. It also has some RAM.

http://amiga.resource.cx/photos/trifecta,1

It has a number of jumpers. The RAM configuration jumpers are self-explanatory, but the other jumpers are labeled ID0, ID1, ID2, A, B, C, D and E and I don't know what they do. They affect SCSI operation, and maybe IDE operation but I was not able to figure it all out.

The machine came with a Quantum SCSI drive (was 80 megs I think) which worked for a few days and then died. I was able to boot from it.

I was not able to boot from the IDE drive. I had to run ICDMount from a floppy, but then I was able to access (the first ~500 megs) of a "recent" (250 gigs) drive. Never managed to boot from IDE even though I marked the drive as bootable. Transfer speed is 2.5 MB/s according to SysInfo.

Then I orderd an SD-to-SCSI adapter. It works and I'm able to boot from it, although transfer speed is 1.15 MB/s according to SysInfo.

However I still have a problem because the machine does not cold boot.

When I turn it on, it just hangs. I have to Ctrl-Amiga-Amiga and then it boots fine. (Note: I Ctrl-Amiga-Amiga quickly, without switching to Kick 1.3). It wasn't doing that when the Quantum drive was still working - it was booting fine when powered up.

It hangs even if I connect the IDE drive, so I don't think one of those "IDE terminators" would solve the problem.

The machine is pretty loaded, it's modded with a Fat Agnus and 2 MB of chip, has a kick switcher with 1.3 and 2.04, a GVP accelerator card with 4 MB, 68030 and 68882 slightly overclocked at ~27 MHz and a deinterlacer.

- Trifecta card jumper settings
- Configuration of the SD-to-SCSI emulator

If anyone has information on this card or has tips, it would be much appreciated!

[berke]

Follow-up: After some experimentation, it appears that ID0,ID1 and ID2 are used to set the SCSI ID of the IDE drive, as in ID = ID0 + 2*ID1 + 4*ID2. (The IDE appears as a SCSI device.)

[Firthy2002]

How soon after cold boot do you reset? Is it affected by which version of KS you boot with?

[berke]

Originally Posted by Firthy2002

How soon after cold boot do you reset? Is it affected by which version of KS you boot with?

Pretty much immediately.

I don't know how to cold boot into KS 1.3 as the switcher powers up on KS 2.04, and to switch to KS 1.3 I have to reset while holding Ctrl-Amiga-Amiga for a few seconds.


With a soft boot, KS 1.3 can access the drive (but of course it reports "Not a DOS disk" since it's FFS.



Also, I've noticed two strange things:

- The machine hangs if I run Enforcer and then try to access the SD2SCSI. The SD2SCI seems to hang as well (activity LED stays lit). Only power cycling gets it out of that state.
- No issue when accessing the IDE drive.


This makes me wonder if it is some kind of CPU cache / DMA issue. I guess I'll have to try with DMA off.

[Toni Wilen]

What does "version full trifecta.device" return? Only known ROM dump contains v67.1 trifecta.device.
EDIT: I think this was dumped from ROM chip that was labeled as "v4.31".

[Firthy2002]

Can you see the drive if you access Early Startup Control on KS 2?

[berke]

Originally Posted by Firthy2002

Can you see the drive if you access Early Startup Control on KS 2?

I can see the SCSI drives from ES. Never saw the IDE drive, though.


Toni: I'll check the trifecta.device version tomorrow, but from memory it was sixty something (as listed in SysInfo).

[berke]

Originally Posted by Toni Wilen

What does "version full trifecta.device" return? Only known ROM dump contains v67.1 trifecta.device.
EDIT: I think this was dumped from ROM chip that was labeled as "v4.31".

trifecta.device version 64.1. ROM chip is labeled "Trifecta 4.01, © 1992 ICD, Inc.".

[berke]

Update: When I run enforcer (with output to stdio), if I access the SCSI device, I get a ton of hits saying WORD-WRITE to $00B70000, name : "Processor Interrupt Level 2". Chip goes from $00000000 to $001FFFFF, fast from $00200000 to $009FFFFF inclusive. $00B70000 is in the $00A0-$00BE reserved/do not use area (http://amigadev.elowar.com/read/ADCD.../node00D3.html).

WORD-WRITE to  00B70000        data=0000       PC: 0020B6F8
USP:  0028EAAC SR: 2204 SW: 072D  (S2)(-)(-)  TCB: 002AD188
Data: 00200154 00006610 00000018 00000000 00221BF0 00002000 00000000 002A0009
Addr: 00219850 00EA0000 00219850 00EA0000 00F81526 0020B2A4 002007A0 --------
Stck: 00F82520 00F90CA0 0009C1AC 00200936 002AD1E4 002007A0 00F9032C 002124E4
Stck: 002124E0 05F90028 EADC4B34 002A083C 0028EAC8 002AD1E4 00000018 FFFFFFFF
Name: "Processor Interrupt Level 2"

The SR shows we are in an interrupt handler. The PC is in trifecta.device, which has a string that says it's v4.0r1. There is a

MOVE.W #0,$B7000000

instruction there. I replaced the address in that instruction with $DFF180 and I don't get any more Enforcer hits.

Maybe the IDE device does not use interrupts, or maybe it raises a different interrupt?

[alexh]

Interesting device. I always wanted one. I always wondered how they were able to make claims of bandwidth exceeding 2x the theoretical Zorro II bandwidth. Marketed with phrases like "hidden DMA cycles"

Here are some interesting trifecta threads if you've not already found them

http://eab.abime.net/showthread.php?t=94980
http://eab.abime.net/showthread.php?t=98091

There are some scans of the manual in this post

http://eab.abime.net/showpost.php?p=...4&postcount=26

[berke]

Originally Posted by alexh

There are some scans of the manual in this post

http://eab.abime.net/showpost.php?p=...4&postcount=26

Thanks so much! That was exactly the information I was looking for.


The boot delay setting might be it.



Now I need to go buy some jumpers to try those settings. I keep losing them, and one of them just broke.

[Firthy2002]

I was just about to suggest that there could be some kind of delay setting which needs to be altered.

[berke]

Originally Posted by Firthy2002

I was just about to suggest that there could be some kind of delay setting which needs to be altered.

Thanks for the suggestion. I tried various combinations of the jumpers and of the scsi2sd-util startup delay to no avail. The boot delay options do slow down the boot process but do not hang it.

The mystery of the writes to $B7000000 is solved, however. They occur only when the AdSpeed option is activated via the jumper. I don't have an AdSpeed board. I suppose writing to that address is some kind of hack to signal the AdSpeed hardware to flush caches or something similar.

Also, I noticed that if I power the machine up while holding both mouse buttons, the early startup screen appears and then the machine freezes.

The scsi2sd adapter I have has a tendency to lock up: the activity LED stays lit, and a reset doesn't do anything - I have to power cycle the machine.

[alexh]

Which version of scsi2sd do you have? There were lots of firmware updates for v6 recently and a new PCB.

[berke]

Originally Posted by alexh

Which version of scsi2sd do you have? There were lots of firmware updates for v6 recently and a new PCB.

It's not a V6, it's a "SCSI2D 4.2c by Michael McMaster" sold by amigastore.eu with a Cypress chip. I've put the latest firmware in.

UPDATE: Downgrading to an older firmware version, using only one SCSI target and making sure IBURST stays off seems to fix runtime problems (so far.) It still hangs at boot.

Turning IBURST on and starting ICDPrepHD reliably causes a SCSI bus freeze. When this happens, the SCSI LED (from the controller card) and the activity LED (on the adapter card) stay on, and SCSI I/O freezes until I power cycle the machine.

[Firthy2002]

Just a wild shot, have you tried it without the GVP board?

[berke]

Originally Posted by Firthy2002

Just a wild shot, have you tried it without the GVP board?

I did. Also with the GVP in 68000 mode, with RAM disabled on the Trifecta and various combinations. Basically, if the scsi2sd is connected with an SD card, it hangs at power up.

[berke]

Further update: I'm giving up on this adapter, it just isn't stable. I can't 100% say it's the adapter, but at least with the IDE drive it doesn't hang at boot.

Might be because it's an old 4.2c version.

It won't boot after a power cycle, but after resetting it via the keyboard I thought it was working more or less alright, so I proceeded to install some software.

I transferred ADFs for the Imagine 3 ray-tracer using ZModem and Term. The idea was to let the machine do some heavy computation to ferret out any stability issues.

I got checksum errors on the transferred files after mounting them with xfs under file.device. Note that these arent serial CRC errors (which would have been caught by ZModem), but checksum errors on the image files on the system disk. I re-transferred them, saving the downloaded files to RAM: instead. Then I managed to install them, but later I got checksum errors on the drive again. I even did all of this with the data cache disabled; the burst mode was off too (it's the default.)

I'm still wondering if it could be :

- The kick switcher?
- The missing battery clock causing initialization issues?
- A termination issue?

Any tips would be appreciated. In the meantime, I'll transfer everything to the IDE drive, boot from a floppy and see if I get any problems.

[berke]

Originally Posted by alexh

The Amiga Gayle IDE driver in scsi.device needs newer drives to be partitioned with the max transfer setting set to 0x1fe00. Otherwise large writes cause data corruption.

http://eab.abime.net/showpost.php?p=759918&postcount=25

I wouldn't have thought the IDE driver for the Trifecta would need this work-around but worth a go?

The thing is that the IDE drive works perfectly fine. The problem is with the SCSI. Just having SCSI causes problems, even if IDE is disconnected. Also the SCSI was working fine with an old Quantum drive.


Also the trifecta.device driver somehow makes the IDE drive appears as a SCSI target. Maybe it uses an IDE-to-SCSI chip or something...

[Jope]

Most likely the scsi target thing is totally in software. :-)

These older controllers really don't seem to like the scsi2sd target for some reason, I've had those LED on + scsi2sd hung experiences with my scsi2sd v6 and various Amiga scsi controllers.