A4000T Zorro slot order

[torsti76]

Hi,

a while ago I built (i.e. soldered) an original A4000T mainboard. While the board works fine now, there's one thing that puzzles me: There seem to be subtle differences in the handling of the different Zorro slots.

All slots are working, BUT the order of the cards seems to be very important:

- If only the lowermost slot (#1) is occupied, cards in the uppermost slots (#4 und #5) aren't detected. They neither show up in early boot nor in AmigaOS.
- If there is no card in the lowest slot, cards in the two topmost slots are detected without problem.
- If there is a card in the middle slot (#3), cards in both the two topmost slots and the lowest slot are being detected ok.
- If all slots are full, all cards are detected, but not in random order. E.g., the BigRAMPlus doesn't work in the topmost slot then.

I'd just like to know if this is normal behaviour or if I did something wrong while assembling the board.

Perhaps someone with a factory built A4000T can check...

Thank you and best,
Torsten

[seb132]

Hi mate. I have never made a thorough testing and appraisal of what you say but clearly noticed on my A4000T some « funny behaviour » and some non working cards depending on their location inside the machine.
I wanted a fully stuffed A4000T with all bells and whistles (PPC, Pocasso4 with all modules, Tocatta, Deneb, Buddha...) and when I saw that I just thought my A4000T had a problem and decided it was a too complicated machine for me, lol.
I still have it but need to put it back together as I took it out for service, capacitors, etc.
I’m interested in that topic as I have another A4000T, NOS this one, that is unfortunately 13,000 km away from me at the moment...

[Fastdruid]

AIUI that _shouldn't_ be the case as the autoconfigure doesn't care about card order.

The Autoconfigures are performed in order _but_ should also ignore an empty slot

http://amigadev.elowar.com/read/ADCD.../node02C7.html

The schematics I have of the A4000 (tragically I don't have a real one) don't show the expansion slots but the A3000 schematics do and they show a 74HCT32 which controls the handover of _CFGIN (pin 12) / _CFGOUT (Pin 11).

So, each card on reset should "negate" _CFGOUT and should configure each in turn when the _CFGIN is "asserted". If a card is not there is means that _CFGOUT is not negated and so the OR gates mean that the next _CFGIN is "asserted".

[Kin Hell]

I've had the Pleasure of an A4KT & several A4KD's over the years....

On both types of System from my experiences, fully loaded ZORRO's can cause issues & order of cards can make a difference. Add in PCI boards & it can become a nightmare.

Board Revisions can also add complications & then there are Library versions as well....

[torsti76]

Originally Posted by Fastdruid

AIUI that _shouldn't_ be the case as the autoconfigure doesn't care about card order.

The Autoconfigures are performed in order _but_ should also ignore an empty slot

http://amigadev.elowar.com/read/ADCD.../node02C7.html

The schematics I have of the A4000 (tragically I don't have a real one) don't show the expansion slots but the A3000 schematics do and they show a 74HCT32 which controls the handover of _CFGIN (pin 12) / _CFGOUT (Pin 11).

So, each card on reset should "negate" _CFGOUT and should configure each in turn when the _CFGIN is "asserted". If a card is not there is means that _CFGOUT is not negated and so the OR gates mean that the next _CFGIN is "asserted".

Yeah, that's what I thought, too. OTOH, having 5 Zorro slots is unique to the tower models. And while the 74HC32 is there in the A4000T, too, there are some literal oddities. The last (topmost) slot, e.g., has got its _SLAVE signal terminated by another resistor network than the other slots. The schematics call the connection between this signal and its terminator "ODDBALL". :-)

And as it seems, real life differs from theory of operation as it does most of the time.

I guess that beside PSU issues in a fully loaded system the bus termination is one of the biggest problems, especially in combination with PCI busboards. The lowest slot in my A4kT is occupied by a Mediator that in itself already is a complex system (as most Zorro boards are, anyway). While in general not important for the kind of frequency an A4000 runs at, I think in corner cases the different trace lengths on the mainboard and several other factors come into play...

What puzzles me nonetheless is the behaviour of the topmost two slots, which depends on the occupation of the lowermost one. While with the Mediator this is somewhat understandable I can 'turn off' the topmost slots by placing a single random card in the lowest slot.