IISI nubus adapter in an SE/30

[IamSpartacus]

Does not work. The SE/30 zebra striped and wouldn't boot. A definitive bust. There's no mention of this in any of the forum searches, so I thought I'd throw it into the public knowledge.

My only guess is that it has something to do with multiple 68882s in the system at once. But the only way to test this theory is to desolder the FPU from the IISI adapter, and I'm not going to do that.

 

[Trash80toHP_Mini]

Actually, there are several mentions of the NuBus SE/30 notion. Bunsen was looking into doing what you just tried.

I said it'd be problematic due to the /NuBus signal being N.C. on the SE/30, but I was mistaken. It's the IIfx where this pin/signal is N.C. which makes perfect sense in light of the following.

The SE/30 PDS has an interesting spin on the signal at C2 where the IIsi has the same /NuBus signal on the connector.

Pin C2 on the IIsi 68030 PDS /NuBus NuBus space address

Pin C2 on the SE/30 68030 PDS /NuBus

Signal descriptions for the In the PDS: machine specific signals
In the Macintosh SE/30, indicates address in the range $6000 0000 to $FFFF FFFF. This signal is active when the CPU addresses the built-in video display. (Expansion cards can use this signal and further decode the slot ranges to avoid conflict with the video logic)

The IIsi NuBus Adapter has no provision made to "further decode" this signal, because its NuBus PseudoSlot interrupt/memory address range/whatever is fixed at $9 while its Vampire Video PseudoSlot ID is fixed at $0.

So it's no surprise that your Video display wigged out and apparent memory conflicts caused boot failure . . .

. . . but this need not be the case! }

It makes sense that this signal is N.C. on the IIfx PDS, as all six NuBus Slots are already implemented on the MoBo and the Address Mapping handled in ROM.

 

[IamSpartacus]

Bunsen mentioned in the other thread that the IISI nubus adapter might be hard wired to work with a 20mhz system (like the IISI). Over clocking the SE/30 to 20MHz isn't an option, so I under-clocked the nubus adapter to match the SE/30's 31.3344MHz oscillator. No effect. Still zebra'd.

Now I'm thinking it's either a difference in the IISI's ROM or the previously mentioned FPU thing. It would be nice if I had a IISI ROM to try out. It would also be nice if either of the FPUs were socketed.

 

[Trash80toHP_Mini]

The /NuBus signal addressing, or the lack of decoding thereof, would be your current problem. The NuBus Adapter conflicts with the Video SubSystem of the SE/30.

The "custom NuChip 30 and associated transceivers" could be the problem, as they're used on the IIci and the IIsi Adapter. Both are faster, and more capable I/O buses than the SE/30's, but I really doubt that the speed aspect is your problem.

The IIsi, and very likely the IIci, has the ability to do burst transfers across the PDS when operating as bus master. The SE/30 lacks this feature and I'm guessing the Mac II, IIx and IIcx also lack this feature. These earlier/slower members of the Mac II family used the original NuChip ASIC and transceivers, probably without the ability to do burst transfers over their '030 I/O System (essentially one big "PDS") as well.

What may become necessary to do NuBus on the SE/30, would be to design a NuBus Adapter based on the old NuChip Chipset.

Dunno, more fine reading of the Docs kinda research necessary to come up with a real solution.

Look at it this way NuChip < NuChip 30 < The IIfx NuBus Chipset < The NuBus 90 ChipSet for the Quadra and X100 PPC series.

Backwards compatibility of the ChipSets ain't very likely in my book . . . :-/

My NuBus Adapter is AWOL for the moment, but it only has a socket for the CoPro, IIRC.

What was the rate of the oscillator you under-clocked for the SE/30 trials? 10MHz?

What other thread? :?:

 

[Trash80toHP_Mini]

IDidn'tRC, my CoPro is soldered down as well.

Instead of removing the whole thing, try lifting just the VCC leg and installing a jumper header.

This works fine for disabling the DuoDock's DeclROM. It's not the prettiest of hacks, but hey, it may work!

At the very least, it'll allow for that particular test to be done AFAIK.

 

[Trash80toHP_Mini]

I don't have a II,IIx or IIcx in my collection, so I need someone to do me a favor for the info in this thread.

Can someone take some really high resolution pics of the NuChip and Transceivers that make up the NuBus Bridge Chipset that would be appropriate to the SE/30's IIcx Architecture roots.

Just the PinConnt of the NuChip ASIC can likely answer some questions, but pics are WAAAAY better!

TIA,

jt

 

[IamSpartacus]

The thread previously mentioned is here http://68kmla.org/forums/viewtopic.php?f=7&t=14479&p=141865

I replaced the 40mhz oscillator on the nubus adapter (same as a IISI's main oscillator) with a 31.3344 oscillator (same as the SE/30). It was a long shot, so I kept my expectations low.

At this point I've used up all the time I can spare for this, and don't want this to turn into a major endeavor.

 

[Trash80toHP_Mini]

First, thanks for that link.

Second, try replacing the Oscillator that'll give you exactly 10 MHz from the 16MHz SE/30 clock rate . . . yours seems off a bit to these coffee starved eyes, byt MMV . . . |)

Third, I've already turned this into a major endeavor:

SE/30 NuBus/PDS Expansion Hack Thread . . .

THX, for the kick-start, comrade! [] ]'>

 

[Bunsen]

Do you have a Nubus video card and external monitor? Could be worth a shot.

I'm a trifle surprised the IIsi Nubus adapter has an onboard osc - I would have thought that would be derived from the IIsi's own osc via the PDS. Maybe they were hoping to use it in both machines until the roadblock appeared.

 

[zuiko21]

Can someone take some really high resolution pics of the NuChip and Transceivers that make up the NuBus Bridge Chipset that would be appropriate to the SE/30's IIcx Architecture roots.

Will these suffice? (click on the images for a large version, around 3.5 MB each)







I thought the main culprit of the SE/30 not working with the IIsi NuBus adapter was the extra FPU... but did some experimenting and got interesting results:

NuBus adapter with soldered FPU: chimes, but zebra pattern and no boot. Pressing reset repeats everything, including the chime.

DaynaPORT E/si30 NIC socketed: does not chime at all. Shows more or less distorted version of the last displayed image, if not disconnected for long. Pressing reset doesn't change anything (no chimes either). Further testing did chime sometimes, or made strange sounds, or death chime in a standard, slow-tempo, distorted or low-pitched way...

SuperMac PDS pass-thru: does chime the first time, and zebra pattern appears. But it does not chime after pressing reset, only after a hard power-off.

Radius Pivot: zebra strips, but no chime ever... although it started making awful sounds after pressing reset, but just once.

It goes without saying, all of these socketed cards do boot fine on the SE/30 once the FPU is removed. SlotInfo recognizes them properly. And the IIsi works fine with the FPU on them, and it's recognized too.

 

[Trash80toHP_Mini]

Yup! No surprise there, too many FPUs is bad news, but lifting the VCC pin on the NuBus Adapter's FPU ought fix that problem, while helping to isolate the real culprit.

Or it just might work. )

I'm snagging you pics for a look-see in GraphicConverter as my next step. [] ]'>

_______________________________________________________________

edit: that was particularly unhelpful, #^**&^*%&@ Flik-whatever is limiting the file size to a useless 200k or so each. Check PM for my RealMail addy so you can send them to me as attachments. :approve:

 

[IamSpartacus]

zuiko21 - interesting. Given the damning evidence of multiple FPUs I'll find some time to do surgery on the IISI card.

Bunsen - agreed about the oscillator. Looks like the nubus hz is independent of the PDS hz. Testing a nubus card in the SE/30 is problematic though, the monitor tube is *directly* in the way and I don't have any 96 pin right angle sockets. FYI I'm using one of artmix's pricey PDS riser things.

Trash80 - are you sure the FPU only has one vcc pin? The CPU has a dozen or so

 

[zuiko21]

According to the datasheet, the FPU in 68-pin PLCC has not one, but eight VCC leads :'( And twelve Ground leads.

However, I think a less intrusive way to disable the FPU would be applying VCC to the /CS input (Chip Select, pin 29 -- third from the right of the top row if the index dot is at the bottom). Doing that thru a low value (say, 330* ohm?) resistor probably won't harm much the associated logic, at least for a brief testing. Desoldering a single J-shaped lead of a PLCC chip doesn't seem a trivial task to me...

*) I'd put first a reasonably high value (≈ 1 Kohm) and then go down gradually until the FPU stays safely disabled. Haven't actually tried this, though.

 

[Trash80toHP_Mini]

BBraun has identified traces connection the CoPro to the NuBus 30 Bridge, these connections are not present on any other PDS card. For them it's just a flat out second "PDS" extension for use by the MoBo Proc.

bbraun has posed a question about the CoPro being an integral part of the NuChip 30 implementation on the IIsi adapter. I responded by saying it might give access to the CoPro from the NuBus, which would be present, but but not apparent on NuChip 30 based MoBos.

Dunno, interesting questions! [] ]'>

 

[trag]

According to the datasheet, the FPU in 68-pin PLCC has not one, but eight VCC leads :'( And twelve Ground leads.
However, I think a less intrusive way to disable the FPU would be applying VCC to the /CS input (Chip Select, pin 29 -- third from the right of the top row if the index dot is at the bottom).

I don't think that would be a good idea, without confirming where that wire comes from on the adapter board. It might lead all the way back to the /CS pin on the FPU on the logic board, in which case you'd be disabling both FPUs...

 

[Trash80toHP_Mini]

I'm still convinced that a harvested NuChip Bridge ChipSet transplanted to a new PCB layout is the only way to go . . .

1) near certain compatibility with SE/30 do to architectural considerations

2) no extra FPU to get in the way/muddy the waters

3) placement of TWO NuBus slots in the space of the relocated HDD's cubic is possible

4) reverse engineering the SE/30/IIsi PDS <-> IIci cache adapter slot for a new PAL layout seems eminently feasible for PowerCache jockeys

5) Using an Asante Nic with with the removable Thicknet passthru card allows for breakout of a HOT 7"+ NuBus VidCard's DA-19 lead or doing an internal cable conversion to HD-15 VGA sticking out the modified backplane I/0 panel

6) a 10/100 Nic might be feasible in the second slot, freeing the third address for a card installed in a straight PDS passthru somewhere else on the card

. . . and best of all, the adapter can be made to fit the inside of the SE/30 like a glove! :approve:

edit: bbraun told me he ripped the CoPro off a NuBus adapter, but needs to see if the traces damaged in the process are a doable fix.

 

[zuiko21]

Good point, trag.

However, just checked the SE/30 schematics and its FPU's /CS signal comes directly from the GLUE chip, and nothing else is connected to it, so that one won't be disabled.

Also, from the IIsi PDS/030 slot specification, I can't see anything directly related to the FPU, so it should be decoded somewhere into the adapter's logic. Still, there's the problem of choosing the appropiate value for this pull-up resistor...

 

[IamSpartacus]

I was able to cleanly remove the FPU from my IISI nubus adapter. I verified it still worked with my IISI. It still zebras my SE/30, with no boot.

 

[Trash80toHP_Mini]

Have you checked if the problem is the documented conflict with the SE/30's Video Addressing being active on the /NuBus signal line whenever Video is active, as I mentioned above?

That REALLY needs to be nailed down before trying hardware hackery as a solution. Address conflicts are bugaboos that don't give up lightly.