Ricoh chipset PCI to PCMCIA adapter card - possible?

[Trash80toHP_Mini]

No doubt about that being a good project that's well within K.I.S.S. parameters,

After seeing jamesmilne's fabulous MacToTheFuture project, I can't help but think that a 10/100/1000 solution for the TAM's Comm Slot 2 connector might be more easily accomplished. Unfortunately I've yet to find CS2 developer documentation, but I do have the developer CD collection from about that timeframe. Anybody remember seeing that info anywhere? For straight up PCI development we're covered, It's now a question of how the PCI Bus Communications Slot (CS2) is woven into Gossamer and its adaptability to vanilla PCI.

Just another notion way outside my skillset, but I'm thinking along the lines of bridge board development for an off the shelf WiFi card for the PCI Express Mini-Card Slot. Might that form factor fit on a low profile CS2 card within the standard back's cubic in the TAM? Such a card would be usable within any CS2 PowerMac, not just the TAM.

Apple Developer Connection: PCI Card Services


Designing PCI Cards and Drivers for Power Macintosh Computers 1996



Revised Edition 3/26/99 has corrections, additions and irrelevant New World complications.



Designing PCI Cards and Drivers for Power Macintosh Computers

 

[Trash80toHP_Mini]

Couldn't resist  .  .  .  looks like the bridge card might be 5mm or 12mm lower in profile than CS2 spec. Might either fit inside the TAM's standard back?





Maybe now I can let it go  .  .  . 

edit: Drat! Forgot about the TAM TwinSlot PCI Riser part of the equation. :blink:

What's the PCI PCB to CS2 NIC PCB separation within the Fat Back conglomeration? Wondering how far to cut down an L-shaped USB card so a fast mass storage interface card could occupy the lower slot?

How tall is the CS2 riser? wondering if a snap-off  extension of the proposed low profile CS2 NIC might allow for mounting an SSD within the considerable cubic betwixt the CS2 and PCI risers?

Got pics of TAM PCI riser or is it the same as one in the pics above?

 

[Trash80toHP_Mini]

I've been overthinking this again, just cut the notion back a bit closer to KISS compliance.

A generic CS2 WiFi card would have a huge market as compared to a TAM specific solution, but what the heck? A fat back compatible TwinSlot PCI riser for the TAM would offer plenty of real estate for a "standard" PCI to Mini-Card slot adaptation for WiFi implementation. Mounting points for an SSD of some form factor or other should fit within vacated CS2 cubic. 2.5" is probably too much to hope for, but something smaller? Are any of the TAM compatible USB cards in the "L" config if the top slot requires a low profile PCI card?

edit: Just found a "1/2 height" WiFi Mini-Card that would allow a CS2 solution to be 16mm lower than CS2 spec, Might that be low enough to fit within the standard TAM back?

No time for a pic, but for playtime at work  .  .  .

 

[Trash80toHP_Mini]

Looks like it's about time to start another thread. I've started hacking away at increasing PCI Slot/expansion options for TAM/Gazelle. For now, I've started looking at 6360/Alchemy which is the basis for trag's three PCI slot + Comm Slot 2 Umax C600. No sense spinning wheels on Gazelle's incompatibility smeared 50MHz racetrack until something can be accomplished in the Alchemy architecture.

View attachment 13545

My three slot riser doesn't work, other than the bottom slot tied directly to the Mobo's Slot $A1 Connector. The Riser has two jumper/nubbin cards and wiring to feed eight signals from each of two other slots on a system board to activate the upper two slots on the riser.

The middle slot shouldn't be too hard to figure out from the 6400/6500 riser, it's a question of encoding the level triggered INT bits A-D to assign it to Slot $B1. That's exactly the same methodology and bit count for NuBus Slot assignments.

Thankfully, the CS2 NIC shows up in ASP as a PCI Card in Slot $E1 and all necessary signals appear present on the Comm Slot 2 connector for hijacking to the third Riser Slot or in the case of the Interrupt coding bits that are probably built into the logic board, are easily implemented on the third slot of the riser itself.

Adapting PCI to PCIe for using a Mini-Card WiFi NIC looks to be a less impractical endeavor than I'd figured. Mouser sells The TI XIO2001 PCI Bus Translation Bridge (bi-directional) for $6.41.

trag, do you remember the ID of that third PCI Slot in your C600? That might make four PCI slots!

 

[trag]

I'm wondering if anything at all needs to be jumpered from CS2 for adding a third PCI slot? I've got a three slot riser that works in my Beige G3 board, looks like it's time to test it in one of the 6500 boards? Dunno, whatcha think, trag?

I think it's been too long since I looked at the PCI  spec.       As long as all the signals needed for a PCI slot are present, it should be trivial.  Might check that REQ, GNT, are unique to each slot on the riser card.   And IDSEL?  I can't remember if that's a per slot or a bus signal.   Also would be interesting to know whehter the INTs are tied together but unique per slot.  That's how they are on X500 and X600 machines.   Each slot has its own unique INT, but all four INT pins in the slot are tied together to that single INT.

There's something different about the 6X00 machines that make them not play nice with PCI-PCI bridges.   Oh, the X500 and X600 machines have the same problem, but one must have two levels of bridge to see it.   On the 6X00 machines, often putting in just a card with a bridge on it is enough to cause issues.

Could also be that the PCI controller in the logic board chipset was built with an internal PCI-PCI bridge.

Apple's first PCI implementation has a serious bug that doesn't handle multiple levels of PCI bridging properly.   They didn't bother to fix it in the X600 machines.   I don't know if they fixed it in the Beige G3 -- of course the PCI controller in that machine was designed by Motorola, so that may have solved it.

 

[Trash80toHP_Mini]

I'll have to buzz the 6500 riser I'm currently using. Then I'll map out the signals on the keyed 9 pin header setup on my complicated (more versatile?) riser. I'm guessing I can hack INTA-D high or low on the headers on the riser for the jumper wires. Mouser and others sell straight up three slot risers, so I'm wondering if the risers with nubbin card jumper setups are meant to allow Slot ID flexibility?

I'm leery of trying to test the INT lines with the system hot. Might it be as simple as sticking four resistor/LED pairs into the female IDC header connections on the nubbin card cable, twist the ends together and insert one lead into the ground connection to make a remote Slot ID readout? At TTL hig/ low levels, do I even need the resistors? That explains the circuitry, jumpers/breadboard implementation would be the way to build it.

I've got a single slot WrongAngle (Apple config left angle) riser that doesn't work, testing that one for Slot ID incompatibility will be interesting.

Picture's worth a thousand words.



These risers are for use in a 3U or 2U rack case, so they're RA to splay the cards out across Mac risers are WrongAngle for compact vertical expansion over the logic board. I've partially rebuilt my 6400/digitalSTARION 910 as a testbed.

Mouser sells a straight up, no jumpers RA three slot riser analogous to Apple's (ostensibly passive) two slot LA riser.



Sintech Electronics has a a fancier (my KISS compliant card seems better suited to the purpose) three slot riser that's interesting.

edit: the speaker gave me startup chimes a few times, then stopped working entirely. Do TAM & company need recapping?

 

[Trash80toHP_Mini]

trag, I think the third slot on your C600 riser may be C1. Took a couple of interesting screen shots:



Full loadout on a Gazelle board shows PseudoSlot Video as F1, PCI cards in slots A1 and B1 with the CS2/NIC at slot E1.



TattleTech has both Alchemy and Gazelle reporting Slot C1 as a possibility. Riser slot and CS2 were empty. That probably won't change with hardware state.



Built my INT signal slot snooper kluge with the resistors  .  .  .  dunno, maybe it'll work? Maybe we'll see tomorrow. :blink:

 

[Trash80toHP_Mini]

Back to TAM networking basics:

First off, did Apple design the TAM so that a Sonnet Crescendo L2 that had been available for 6360/6400 would fit within the confines of the standard back or did they opt for a "profile under function" approach, leaving only enough room for their Cache Card and low profile memory?

Byrd, could you check something out for me? I need to know how much clearance there is between the top of the PCI slot connector and the standard back? Is it as least the height of the Sonnet L2 accelerator cards? If you could make up tall mockups with thin cardboard of a CS2 NIC or Modem "PCB" it would be a big help. Cut it back at the ends so it will fit straight down into the slot and the use a pencil to run the outline of the case curve to model a max height CS2 card. Do the same with the PCI Riser to show the maximum for factor for a standard back RA "PCI riser."

If anyone who could do this for me it would be a big help. Scanning the two cutouts along with a metric ruler would be the point, that way I've got something to mess around with.

Another way of looking at it:

EvilCapitalist, I wasn't sure about the labeling on your comparison pic. It seems like the front/back pics you posted are of the TAM riser?

View attachment 24003

Can the board be extended to left or right? Will your 5500 riser fit within the Standard TAM Back? 

We've been ignoring the vacant cubic between PCI and CS2 connectors. My single slot 1U RA Riser's only about 19mm higher than the top of the connector and could be lowered to 5500 height if absolutely necessary. The PCI connector could be desoldered and replaced with Rows of headers for plugging in a "PCI daughtercard" to fit within the TAM's available PCI to CS2 cubic.

Dunno about the TI PCI to PCIe bridge chip, but there's another that's compatible/transparent to the original PCI spec. intended for legacy equipment. If we can find a Mini-PCIe WiFi card with a chipset that's compatible with OS9 drivers we could be in business. It's nothing at all like something I could do, but it's looking more and more like the puzzle parts  seem to be coming together.



The really nice thing about using that cubic would be the possibility of setting the Mini-PCIe NIC up to fit in that cubic allows for the headers to fit on the available area between top of the mobo slot and the thruholes of the Fat Back PCI riser's connector. If the PCIe NIC/bridge is set up at the CS2 NIC Slot E1 location or the phantom Slot C1 location we'd be in TAM PCI Slot heaven. A customized Fat Back riser would have the usual Slot A1 connector and a connector for Slot B1 putting a PCI card in the cubic intended for the lowly 10bT CS2 NIC.

Enough for the musings and back to the tangent at hand:

Likely no time to test the LED kluge today, so info would be appreciated: oes anyone know if a logical high signal is adequate to power an LED so it's enough to be visible in a darkened room? If not I guess I can meter the resistor legs for high/low or do I need to find and re-learn how to use my ancient logic probe?  Board's set up for doing that or adding a couple of header rows on the other end of the board for such testing or amplification of the logical high to LED driving level.

edit: might as well ask up front in case I need one, anybody have a suggestion for a DIP IC amplifier that will drive four logical high inputs to LED power level outputs? Resistor and capacitor specs to drive such a circuit from an FDD Molex takeoff would be much appreciated. I have plenty of room on my snooper board for such shenanigans and almost no design experience for this. HELP!!!!!!!!!!

 

[Trash80toHP_Mini]

One more then off to work:

With a customized twin slot PCI Riser in place of the CS2 NIC/Riser, about how "tall" could a low profile PCI card be while still fitting within the sloped Fat Back's cubic? KISS compliant approach would be a modded 6500 type riser. Gigabit NIC and Vonets WiFi. With low component height on the lower card in Slot B1, slot to slot distance might be reduced considerably.

Dunno, just wondering about it.

 

[Trash80toHP_Mini]

Curioser and curiouser. Slot A1 (lower connector) of the 6500 riser works in slots A1, B1 And C1 in my BG3DT testbed, It takes on the Slot ID of whichever host slot in which it's installed. Video card in upper Slot B1 of the 6500 riser boots with proper display up to the point of freezup after the first extension loads. With Extensions Manager activated with the spacebar the control panel loads, but ADB ceases to function. Removed the EM control panel and the system freezes up at about the same point. Holding the spacebar dow fails to load the System with extensions off, somewhere right before that ADB gets poleaxed and the first extion loads and freezup occurs..

Again, it's amazing to me that the VidCard's working fine off drivers in ROM When installed in 6500 riser Slot B1 in whatever host slot I try, but having it chime in at some nebulous Slot ID (WAG) borks the boot process.

I'm using a 1U single slot RA riser to tip it over to the right so the cards are upright. With the right side panel removed, the kluge sticks out the side of the plexi case.

I'll  keep hacking away at the problem from the standpoint of the nubbin card's jumpered signals for the heck of it, but since the physical slots don't exist in the TAM/Gazelle/Alchemy machines, I think I'm at pretty much of a dead end with this riser. I was hoping that the 6500 riser inserted in slot A1 of the G3 board it's B1 implementation in the top slot might work. Not the case.

It's time to look into snagging a three slot riser without the jumper boards. Guessing they're intended to install in the most inboard, Slot A1 to save space and will be set up to automagically splay Slot A1 out to Slots B1 and C1. With some luck it will "just work" to implement the architecture's ephemeral Slot C1, may be worth trying?

BTW, the slot snooper's LEDs work so I got polarity of the basic circuit right, but I'll need to try it with a single jumper/LED readout at a time. Getting STRANGE results that make me think the circuit needs diodes for isolation downstream from the LEDs. Getting all VERY bright in A1 of the 6500 riser, but all VERY dim in any other slot, including Slot A1 of the Beige. Most strange. :blink:

 

[Trash80toHP_Mini]

I think it's been too long since I looked at the PCI  spec.  

LOL! Never thought I'd be looking at DCaDftPMF or the PCI spec, but there it is!

As long as all the signals needed for a PCI slot are present, it should be trivial.  Might check that REQ, GNT, are unique to each slot on the riser card.   And IDSEL?  I can't remember if that's a per slot or a bus signal.

They all must be unique to each slot. If I can get my slot snooper to tell me how to set up INTA-D for CS2's Slot E1 we might try something very interesting.

Static signals not found on CS2 pinout (they'd be implemented on the logic board as CS2 is a subset of the PCI spec?)

INTA = header pin 1

INTC = header pin 2

INTB = header pin 3

INTD = header pin 4

Active signals on CS2 that match the other four lines required to be jumpered from the nubbin cards to my 3 slot riser:

CS2 pin 101 - CLK = header pin 6

CS2 pin   92 - GNT = header pin 8

CS2 pin   91 - REQ = header pin 9

CS2 pin   78 - IDSEL = header pin 10

I'm guessing the CLK signal needs to be on the same length jumper as REQ, GNT and IDSEL for them to latch to the leading edge properly?

Also would be interesting to know whehter the INTs are tied together but unique per slot.  That's how they are on X500 and X600 machines.   Each slot has its own unique INT, but all four INT pins in the slot are tied together to that single INT.

INTs are what it's all about ATM, that and cutting down one of my single slot 1U RA adapters to fit into the Comm Slot. If I desolder the connector and install four headers that's half of the Slot $E1 equation.

There's nowhere to hijack signals from the phantom Slot C1, that'll have to be checked by using the only riser I've found that has no jumper cards. The only thing that makes sense for that one is that it goes into A1 and rolls its own B1 and C1 from there. That's a tech support call for the morning, but just maybe I can get the riser I have on hand to speak E1 at some point. [}]

 

[Trash80toHP_Mini]

So far so good. Advantech support told me there are no jumpers on this card's schematic, but I'll have to wait until Wednesday to talk to the tech who directly supports it.



It's the only three slot riser I've found that might work in Alchemy/Gossamer for implementing Slot C1.

 

[Trash80toHP_Mini]

Sounds like the card ought to do the job, so one's on the way.

Had another space saving, PCI slot producing notion for the TAM and my beloved 6360. From this Evil's pic:

View attachment 23999

it looks like a Mini PCI connector + WiFi card might possibility crawl up the ass end of a custom TAM riser? That could leave room for a second PCI slot and card to fill the vacated ComSlot2 cubic? That would depend on the (fingers crossed) utility of the three slot riser that's on the way for providing Slot C1 in these machines. If not we could dial it bask to using of Slot B1 on the same custom riser.



http://www.oldschooldaw.com/forums/index.php?topic=3795.0

The shot above shows the NIC notion, but the adapter isn't the right kind, The connector needs to be installed as low as possible on the riser:



Might need to find a low profile Mini PCI NIC to clear available overhead within the Fat Back? Dunno, I've got another blurry notion for fitment if that doesn't pan out. OS9 drivers for the chipset on a NIC in this form factor would the limiting factor in either scenario.

It's not the GigaBit chipset riser hack of my caffeine deprived, over the top dreaming, but Gossamer couldn't possibly make full use of such a crazy thing. I've seen Mini PCI WiFi cards rated at 300Mbps with 200Mbps sustained throughput, so FASTx2 on an available WiFi card would work for me.

edit: coffe's kicking in, AirPort wouldn't be a straight up PCI derivative would it?

 

[Trash80toHP_Mini]

JB, looks like I've taken this tangent full circle!  :lol:   But I'm burying your chipset alive in a second PCI slot so you might have your NIC and USB too?

http://lowendmac.com/2009/hacking-a-wifi-pc-card-to-work-in-apples-airport-card-slot/

http://lowendmac.com/2009/wifi-pc-cards-compatible-with-mac-os-x/

edit: oopsie, forgot to mention the OS9 compatibility info in that OSX list.

 

[Trash80toHP_Mini]

BINGO?





Looks to me like a custom riser for the TAM with the Type II upright Mini PCI connector and WiFi Card ought to fit within the confines of the Fat Back?

The WiFi cards appear to be lower profile than the 55.8mm full card spec.

 

[Trash80toHP_Mini]

The WiFi cards appear to be significantly lower in profile than the 55.8mm full card height spec. The Mini PCI slot connector could be much closer to the USB/FW PCI card than standard PCI spacing. That's especially true if it is offset as far to the left as possible to clear the Tango's backplane connectors.

How much air space is there between PCI and CS2 PCBs in a tricked out TAM?



The component height of a Mini PCI WiFi card is insignificant as compared to the CS2 NIC. There's even a convenient RJ-45 sized cutout in the standard cover for a WiFi whippet. Methinks I'm within the curvature of the Fat Back. Did Apple put any junk on the inside of the Fat Back in the space I'm exploring? mcd didn't include a shot of the inside of his Fat Back in his great teardown. Anybody got a pic link handy or a chance to take and post a pic for me?

I'll be playing with this in AI later. Would someone volunteer to print a .PDF at full size to cut out a pop-up card to check the cubic for me? you wouldn't need to pull the PCI card, just hold the T-shaped foldup against the standard riser and snap  a pic or two for me?

A lower riser card might fit in the unmodified drawer/chassis of an unmodified 6360/5x00? That I can check myself.

edit: could someone take and post a dead-on profile shot of the two cards/risers with the back open and matching pic with the back closed, please?

 

[Trash80toHP_Mini]

 

[Trash80toHP_Mini]

JB, sorry about hijacking your thread and taking it so far afield. Have you gotten anywhere in searching for compatible PCMCIA chipsets?

A lot of legwork for finding Mac compatible NIC/WiFi cards and drivers has already been done and documented out there, so that should be a lot easier to nail down than the PCMCIA equivalent. Doubling (tripling?) the speed of a 10/100 NIC over WiFi in lieu of the CS2 10bT card/riser combo appeals to me. But from your comments in the other thread I wonder if it would appeal as much to you or to the TAM crowd in general?

I finally got things outlined for an approach to your CS2 riser predicament that's really not a whole lot more complex that recreating just the CS2 riser. I hope the illustration I did over Apple's isometric is finally clear? Card size likely a pretty good approximation. I was up quite a bit later puttering around with that. I cannot imagine fitment being much of an issue even if a Mini PCI turns out to be quite a bit larger than I've shown.

At any rate, for a couple of bucks the BroadCom card pictured in the "AirPort card" thread is on its way from China. I'm mulling over the many simple PCI adapter choices for the 6500 testbed. So it goes  .  .  .

 

[Trash80toHP_Mini]

One last visual applique to Apple's isometrics of the TAM's bits:



Illustration distortions and errors aside in Apple's docs and my doodles aside, I think things are looking very promising.

The first illustration was closer to actual size, but not the 3D didn't pop. This time around I've oversized the Mini PCI slot and WiFi card very significantly. So much so I'm thinking it's possible to move a second, full size PCI connector rearward enough to clear any card connectors and a bit closer than PCI spec spacing on a two slot riser for the TAM. A low profile. no external connector card like my Tempo ATA-133 would tuck very nicely in the top slot, but a much taller card could fit if it's not quite full height, components heights and overhead obstructions allowing.

Breaking any connectors for a second card out wherever empty area on the lower card's backplane might work for a card that needs such, depending on combination. An entirely new side plane cover/support plate for the Fat Back could be modeled and printed with any sort of modifications required as well. Heck in Apples isometric, a low profile cover plate/opening looks like it might be be trimmed to fit.

I'll drop this here and start a hacks thread if any of this crazy stuff is of interest to you TAM lovers. I'm up to the same shenanigans for the drawer equipped Alchemy/Gossamer machines., but that's even farther off topic here. :-/

Thanks for the inspiration, JB, it's been one hella ride. :approve: