The Eudimorphodon Challenge: Builiding a PCI interface card for the Quadra 950.

The purpose of this blog is to document my attempt to answers Eudimorpodon's challenge to my big mouth in respect to a post in which a member asked if it were possible to get USB on a NuBus Mac. My response was that it was possible, but would be challenging. Others added insight on the arguments for and against.

Details on the thread are at http://www.applefritter.com/node/view/5167

In short, I pointed out the existence of a PCI card that provides a NuBus breakout box. The challenge was then laid out to

Design a PDS card using the it (Tundra's QSpan II) to fit my Quadra 950, write the drivers and OS support, and gloat when you're finished.

And so the challenge begins...

Comments

doug-doug the mighty's picture

The first step here is to obtain necssesary permissions from Tundra to view all published doc. Like many manufacturers, they require downloaders of the doc to register themselves. Once done, I will begin downloading select docs and reviewing them.

Also of note, I already have on hand the pinout information for the NuBus slots ('90 standard) (google) and PDS slot (Apple Devnotes) for the Quadra 950. These will be referenced to later.

I originally stated that I would acquire the Refence Design Kit for this chip. Upon contacting Tundra, I learned that the price for this is $995(US). This is not an option at this time and I will proceed without it.

Action item:

  • When home and able to access my own system, setup id with Tundra and download file no. 8091862_MA001_06.pdf. Review other files as questions arise.

doug-doug the mighty's picture

...from back when I was working on the iMac mezz hack. I pulled down the doc, and will start reading.

doug-doug the mighty's picture

Having disposed of my catalogs prior to the last house move (they were out of date anyways), I called Allied and Newark to get updated catalogs. I will need these to order parts from. I will also need to see if I kept my Molex catalog. The trick will be identifying the exact connector for the PDS slot. Catalogs should arrive in 7-10 business days.

doug-doug the mighty's picture

[*]I need to construct a breakout box to accomodate any PCI cards. I need to identify a good connector for the back of the card/box.
[*]Construction will be easiest on prototype board with point-to-point wiring. Given the pin configuration of the QSpan chipset, a socket may not be a bad option and could make wiring simpler, thus preventing heat transfer to the chip when soldering.
[*]Use of point-to-point wiring may be rather cluttered. Coating the board with a potting material when done may contain the wiring and provide a nice look to the finished product.

Tom Owad's picture

This all started in an effort to get USB, right? Building a USB PDS card would be really ambitious but possible project. (Building a PCI interface is more of a joke.)

doug-doug the mighty's picture

I was at first sceptical that this could be done at all. Given the geometrey, I figured I would be better off anyway building a external PCI device housing anyways as this would allow me (at the very least) to plug in any singular card of my choice (100 BaseT for example). So the idea of the breakout box has some merit.

I am having trouble locating the exact section in which I made the interpretation that I could connect up to four PCI devices, but I clearly recall reading something to that effect. I will call Tundra tech support Monday and verify my understanding. I was a little surprised at this as well. From what I understand so far, it has to do with the PCI bus arbitration and running in master v. slave mode against the Motorola processor.

Bottom line:
Can I build a single device PCI interface? I am confident that I can.
Do I think I can build a working four-device PCI breakout box?
I will answer that moday, but am pretty sure that I can.

--DDTM

Eudimorphodon's picture

This all started in an effort to get USB, right? Building a USB PDS card would be really ambitious but possible project. (Building a PCI interface is more of a joke.)

This chip would probably work.

http://www.cypress.com/products/datasheet.cfm?partnum=SL811HS

The hard part would be writing drivers. And don't underestimate how hard that would be.

--Peace

doug-doug the mighty's picture

According to Tundra and the PCI specifications, the QSpan II will support no more than 8 loads as it is running at 33Mhz. Building a PCI breakout box is possible, provided the software is written for adequate system support.

--DDTM

doug-doug the mighty's picture

Have not finished flipping through to check the latest stuff.

Some of the things I think I might need do not jump out at me right away, like the PDS connector or the NuBus connector. Will crocss-check with Molex.

doug-doug the mighty's picture

Generic sub-thread for my posting updates on my progress...

doug-doug the mighty's picture

Have spent last few weeks reviewing parts catalog to see if some of what I am expecting to need is available.
[*]I was hoping to find a socket for the QSpan chipset - I may have to google for the right one.
[*]Did not really see a Nubus style connector, or a PDS type connector. Saw something close, will check specs later. Stil need to find Molex catalog in garage.
[*]Spoke to Tundra about using the QSpan II to perform its own PCI bus arbitration. While there answer did not really make sense to me, I suspect the manner in which I phrased the question introduced confusion as to my intended needs in an answer.

Next step:
Continue reviewing doc and begin conceptual design of card.

doug-doug the mighty's picture

Well the last few weeks have been rather poor for progress. But I have:
[*]nearly finished a paper that may allow me to continue my job in my current role in 2005 (as opposed to being floated some I care not go)
[*]contended with some family issues
[*]repaired two or three eMates
[*]done the Christmas shoppping
[*]made some progress in clearing a much needed space on the workbench

Basically, I have had a million reasons not to make progress on this thing and I have made progress at removing those reasons from my path.

Talks with others (on AF and over in the barracks) may yield two more Mac Quadra 950s, allowing me to continue with my Quadra monster I am planning as well as work on this card.

Oh well, it is Chrismas Eve morning and that's where things are at. I expect to make more of this same progress through the year end. Hopefully I will post something more specific to the card for the next update.

--DDTM

*EDIT* 12/29/2004
Having just finished another eMate, I now have only one more to get through before I can clear the bench and move my Quadra out there! Things will also be easier to work around once the Christams stuff gets back to the attic.

doug-doug the mighty's picture

Well, I finally got the eMates out of the way and got the Quadra out to the workbench. I have determined that the burned trace belongs to pin #12 on the PSU. I will try to locate the pinout info for this and then repair the trace with a fused connector.

doug-doug the mighty's picture

I repaired the trace, but did not add a fuse due to the difficulty in finding one that fit the need. I can always add this later between the PSU and the mobo connector. Machine boots. Also waiting on deal to acquire a second Quadra 950 mobo. One of these will become dedicated to this project.

eMates are going out the door, space is clearing up on workbench.

doug-doug the mighty's picture

Having picked up a fresh (working) machine, I am now ready to put some serious work on to this.

Progress will still be slow as I have many irons in the fire, but I know have less in the way - if that makes any sense...

doug-doug the mighty's picture

The Quadra has been working fine, but I am in the process of moving to a new house (cross-town move). I have gotten all of my notes and research together in the same spot (at last) and will resume this effort by mid next month.

I have already been evaluating some hardware for a protoype. I need to do a lot of work on the components on the bridge's design before I start attempting a build.

doug-doug the mighty's picture

Whilst trying to add a few hard drives and test power up of my test machine, I blew a trace on my mobo (the same spot as the last mobo). This is highly frustrating since I now have to try and fix it. (Too bad I did more harm than good to the last one while trying to fix it.)

I think the problem was on of the NuBus cards I had installed. My speculation is that it was drawing to much current and the trace became a fuse which melted. The good lesson here is that I can be sure to limit the current load I allow on the card knowing that the PSU will happily deliver more than the mobo can handle.

As far as the card goes, I have been doing a little bit of research, but have not made much progress, mostly due to time and higher priorities.

Next steps:

  • Repair trace and add a fuse to the line
  • Update block architecture

Eudimorphodon's picture

I think the problem was on of the NuBus cards I had installed. My speculation is that it was drawing to much current and the trace became a fuse which melted.

You might want to consider throwing out the card in question, before blowing up *another* machine. Just a thought.

doug-doug the mighty's picture

I *think* I know the problem card now too.