mezzanine slot info

Incidentally, I did get some part numbers on PC Card slots and supported chipsets as well. What stopped me from pursuing this project was time and money, but what research I have is yours if you would like.

--dgdg

I don't know where you got the idea that slot has 64 bit PCI available on it. It doesn't. The iMac in question uses the MPC106 PCI controller that only outputs 32/33 or 32/66. The Blue and White G3's had a PCI-PCI bridge that took 32/66 from the MPC106 and turned into 64/33 slots.

Dr. Bob

I got the idea from information I read about the port at the time. I never did go back and verify the functionality of the on board PCI controller to establish what it supported in terms of 32 v 64 bit. But I can vouch for the following:
- 64 bit PCI has 198 pins (per standard), but neglecting the reserved bits and eliminating the redundant bits (i.e. appx 40 pins as ground, including the 'keys'), there are 112 unique pins. A similar situation is true with the 32 bit space, having about 75 unique pins. Your point about the functionality of the MPC aside, this is interesting to know.
- out of 160 pins, I know 94 of the pins from verifying continuity between chips and the connector on an iProRaid TV card.
- I am missing 16 pins to account for the full set within the 32 bit space (but this includes things like 'Test Logic Reset', 'Test Clock', Interupts B, C, and D, and a few others). Unfortunatly, 'System Error', 'Parity Error', and Initialization Device Select' are included in this.
- If one wanted to place the chasis for a cardbus inside of an iMac, the native hole on the 233 is a near perfect match to accomodate a type 3 card.
- the bus architecture of the early iMacs and 199 G3 Series Powerbooks is virtually identical with the differences being the models on certain chipsets (i.e. USB, video, and Ethernet). But how things are arranged around the Padington is the same.

It stands to reason, and this is where Dan K's idea was sound and where I was trying to go until I gave up, that by utilizing the mezz slot and planting a PCI1210 or PCI1220 (and a few other minor parts required for use of these chips) onto a custom mezz card with a cardbus chasis, one could provide cardbus functionality on an early iMac. Alternaltively, one could hang a DVD decoder or secondary video card or firewire port off of a mezz card (practicality of each choice aside).

Now that I have my notes dug back out, I may try to find time to reverse engineer these other pins off of the motherboard. I am more than happy to share what pins I do know if anyone is interested. Because all of what I did was from the card-side, I left it open that the connector could be full 64 bit. With your info about the MPC, it means that the info I have is much closer to complete than I thought. Perhaps those other bits include some audio pins and perhaps some of the ancillary pins from the Rage IIc (which could be augmented with an optional TV tuner (per chip specs) when an assisting chip was added. I would love to get some insider info on what Apple intended with this port!

--Doug-Doug

The original purpose of that port was to connect a logic analyzer and quickly diagnose board problems. It was intended to make servicing the boards easier. It was never at any time intended for an upgrade card, and thus, the PCI loads were calculated for that board with the intention of no real PCI loads being present.

The rest of the pins are used for various signals on the board. Just because there are more pins, or enough pins for 64 bit PCI, does not make it so. That's only wishful thinking. Motorola no longer supports the MPC106, and has sold the line to Tundra. Here is the info on the part:

http://www.tundra.com/Products/PowerPC/Tsi106/index.cfm

As you will see, it only supports 32 bit PCI.

The Wallstreet, PDQ, 101, Beige G3, B&W G3, and 1st gen iMacs all used the MPC106 architecture. Of which Wallstreet and Beige G3 used Heathrow, and the rest used Paddington. Both are PCI devices. Such commonality of design is quite common within generations of products.

Not trying to argue, I just was answering the question as to where I got the idea from. Thanks for the link for the MPC, I can't find that I ever had this info in my notes, but they are going on three years old and have been through 3 residential moves during that time (still looking for some other things too ;-)

You mention the loads on the PCI bus - I am not sure what the spec tolerances are, but I know that the iProRaidTV actually had two PCI devices on it (the tuner and the SCSI controller). I heard mixed results on how successful these cards were. I am sure the system could successfully tolerate a single PCI device (stablely) long term. Two is a bit much if it could be possible to construct a PC Card adapter for the slot.

My plan at this point is to try to figure out the missing pins for the 32 bit set. I got the pin outs for the MPC, but it is so tight I do not think I will have much luck with this tactic. I am trying to get ATI to give me the pinouts for the Rage IIc. It is a little bit bigger and should be easier to hit with a probe. If anyone has the pinouts for this, let me know.

--Doug

Thank you Doug for reviving this topic, I've read this discussion with great interest! I had previously come across your mez-slot info, but hadn't had the time to do much followup.

Mac-supported cardbus PCI cards are available, I'm wondering about the possibility of having their makers produce an identical board with a mez connector instead of PCI? Hmmmm . . .

Dr. Bob wrote:

These details are news to me! Please tell us Dr. Bob of how you came to hold this interesting and useful knowledge, and if I may be so bold, what the heck else do you know about the mez slot?

I've not seen the above info before, so naturally I'm excited to know of someone (that being Dr. Bob) who knows more.

Dan K

This friday is my last day of just shy of 6 years at apple. Most of our boards have diagnostic ports, but after people started using the mez slot in the iMac, they deleted the connectors and opened the circuit lines going to them. As boards got faster, signal integrity later demanded it anyway.

Most of what I know is not sharable though.

The term "loads" refers to how much current the receiving data devices can draw from the source transmitter before voltage drops to unacceptable levels. Usually, once a board is done in design, (unless it has slots) resistors are added to the data lines to make the signal integrity higher and performance better, at the expense of being able to change the loads better. So in the case of a diagnostic port, the logic analyzer draws an incredibly small amount of power, far less than a normal PCI device.

To add to the bus, upgrade card companies are using buffers to amplify the PCI signals without drawing too much current. However, this approach does weaken signal integrity, and also reduces performance.

You really need the schematics, but those aren't available. I'm surprised ATI doesn't have the pinout on their web site. The rage Pro is pin compatible so check it too.