Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[smilesdavis]

Apple officially sold the G4 mini as having “no PCI slots,” yet internally it has a 33 MHz PCI host and Apple routed that PCI bus right to a removable expansion module.

https://leopard-adc.pepas.com/documentation/Hardware/Developer_Notes/Macintosh_CPUs-G4/MacMiniG4/2Architecture/q88_archi.html

- the mini already contains a real conventional PCI bus
- The Intrepid I/O controller has a 32-bit PCI bus with a bus clock speed of 33 MHz.
- The connector for the AirPort Extreme Card uses the PCI bus.
- The mini's host bus is exactly the sort of bus your older Digi cards expect: 32-bit PCI at 33 MHz.
- Apple also used the same generation AirPort Extreme card across many G4/G5 Macs. Available Apple schematic material for that interface shows essentially the conventional PCI ingredients: AD address/data lines, C/BE, FRAME, IRDY, TRDY, DEVSEL, STOP, clock, reset, interrupt, REQ/GNT, IDSEL, etc. (https://www.informaticanapoli.it/download/SCHEMIELETTRICI/APPLE/Q45-G5-IMG5-051-6482-080304-RC.pdf)
-


Mac mini G4
    │
AirPort PCI
    │
custom PCBway interposer* for CardBus
  │
Magma Chassis
    │
 ┌─┼───────────────┐
PCI cards.....

a theoretical adapter should have its own regulated PCI power supply

Mac mini AirPort
  PCI SIGNALS ──────────────────┐
  GND ──────────────────────────┤
                               ▼
                         PCI connector
                               ▲
external PSU:
 +3.3 V ────────────────────────┤
 +5 V ──────────────────────────┤
 +12 V ─────────────────────────┤
 -12 V if required ─────────────┘

Grounds obviously have to be common.

G4 minis don't put the AirPort card directly onto the main logic board. They have a removable mezzanine board (820-1660-06) which can simply be removed from the motherboard, and replacement boards exist as Apple parts 922-6671 / 922-6893. The late 2005 1.33/1.5 GHz revision is less attractive, because Apple revised the mezzanine and integrated the wireless hardware more tightly. Earlier 1.25/1.42-era hardware gives us the nicest physical starting point.

That means I would not touch the Mac mini motherboard initially.

Mac mini motherboard
       │
 original mezzanine connector
       │
 donor mezzanine PCB
       │
 [remove AirPort connector /
  attach interposer here]
       │
 normal PCI

┌────────────────────┐
│    Mac mini G4     │
│                    │
│ Intrepid PCI       │
│      │             │
│ AirPort connector  │
└──────┼─────────────┘
       │
       │ custom PCB
       ▼
┌────────────────────┐
│  32-bit PCI slot   │
│  external power    │
└──────┬─────────────┘
       │
    Magma
   host card
       │
       ║
       ║ Magma cable
       ║
┌──────▼─────────────┐
│ MAGMA 4-slot PCI   │
│                    │
│ MIX Core           │
│ MIX Farm           │
│ DSP Farm II        │
└────────────────────┘

[laulandn]

This sounds plausible and possible, and something I have wondered about also.

The trick will be if the firmware will query roms on the PCI bus that the airport card uses.  If it does, you're gold.  If it doesn't some work will be needed. 

On the other hand, if you don't need cards on the new PCI bus at boot time, the drivers can just be loaded as needed.

[IIO]

@lauland: what is the main difference between the PCI card expansion slots and all other PCI buses on a desktop G4 mobo (or possibly any other personal computer?)

[smilesdavis]

This sounds plausible and possible, and something I have wondered about also.

The trick will be if the firmware will query roms on the PCI bus that the airport card uses.  If it does, you're gold.  If it doesn't some work will be needed. 

On the other hand, if you don't need cards on the new PCI bus at boot time, the drivers can just be loaded as needed.

well in times of UnexpectedBomb i think that shall be the least of our problems

The magma 4slot would then be a selfcontained mac os 9 mix cube system with usb2.0

i will scan the 1.25ghz mac minimezzanine

[laulandn]

@smilesdavis wishing you the best of luck.

@IIO: PCI is limited to the number of devices, before you need a bridge.  If a new bridge is needed, the firmware may or may not scan it.  (So this slot may only support a single device, etc).

For something like a video card that was needed at boot time this would be unfortunate.

But for arbitrary devices that can have their drivers loaded when needed it is no problem at all...provided the firmware sets the device tree up properly...or the driver can scan if it doesn't.

[ssp3]

@smilesdavis, have you, by any chance, been talking to AI lately?  Those graphs look very familiar.. 

[smilesdavis]

J12 looks like the mezzanine connector

Ssp3 try it the new studio m5 ultra is awsome for AI

In addition to the card itself, a Mac Mini needs an additional board connecting it to the motherboard. This is called a "mezzanine board." There are two versions of the mezzanine board – one (older) compatible with the A1027 card (found in "early" Minis) and the other compatible with the A1127 card (used in 1.33 and 1.5 GHz models). I'm not sure if they're compatible/interchangeable (Powerbookmedic says they are). Of course, if you have an A1127, it's a "combo" (WiFi and Bluetooth), while the Bluetooth module is separate in the case of the A1027. Additionally, you'll need antennas (note that the antenna connector for AirPort is different between the A1027 and the A1127).

https://www.ppa.pl/forum/morphos/38908/mac-mini-g4

https://www.macdat.net/files/pdf/apple/servicesource/mac_mini.pdf

J12 mezzanine connector: 80 pins
Standard 32-bit PCI: 124 contact positions
PCI positions excluding power and ground: approximately 78

This makes an 80-pin connector theoretically interesting for a custom PCI interposer.

who wants to identify which PCI signals----if any---are actually routed to J12 on the Mac mini G4 motherboard

[IIO]

(So this slot may only support a single device, etc).

For something like a video card that was needed at boot time this would be unfortunate.

ok thanks for clearing up, that is about what i suspected.

but a card with bus multiplier would be no problem?

[joevt]

but a card with bus multiplier would be no problem?

You need a PCI bridge to be connected to the slot if you want to support more than one PCI card.

Usually a card is connected to the slot. The card has a cable connected to an external PCI chassis. The PCI chassis has a PCI bridge to provide for multiple PCI slots.

Open Firmware supports PCI bridges having PCI device numbers 0 to 15. A patch is required to support PCI device numbers 16 to 31 behind a PCI bridge. For example, the slots in my Netstor NA255A use device numbers 8,9,16,17 so Apple Open Firmware only checks the first two.

Do you have dump-device-tree output from Open Firmware for a Mac mini G4? It should show where the PCI stuff is and your Open Firmware version. Then we can check if Open Firmware will scan arbitrary PCI devices from the slot.

[IIO]

but a card with bus multiplier would be no problem?

You need a PCI bridge to be connected to the slot if you want to support more than one PCI card.

no i mean cards with more than one controller, such as firewire&usb cards have. in regular slots they show up as 2 devices and i wonder that would work when you misuse another PCI port to attach a card.

[smilesdavis]

The next decisive step is a verified J12-to-PCI signal map

[joevt]

no i mean cards with more than one controller, such as firewire&usb cards have. in regular slots they show up as 2 devices and i wonder that would work when you misuse another PCI port to attach a card.

Is there a firewire & usb card that doesn't include a PCI-PCI bridge?

Each individual controller needs separate PCI grant and request signals to do bus mastering.

Now, if the controllers are related and know about each other, then they may exist as separate functions of the same PCI device number. Such as in this example:

Code: [Select]

01:0b.0 USB controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 61)
01:0b.1 USB controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 61)
01:0b.2 USB controller: VIA Technologies, Inc. USB 2.0 EHCI Controller (rev 63)
01:0b.3 FireWire (IEEE 1394): VIA Technologies, Inc. VT6306/7/8 [FireFly] IEEE 1394 OHCI Controller (rev 46)

Or this example:

Code: [Select]

05:00.0 USB controller: ALi Corporation USB 1.1 Controller (rev 03)
05:00.1 USB controller: ALi Corporation USB 1.1 Controller (rev 03)
05:00.3 USB controller: ALi Corporation USB 2.0 Controller (rev 01)
05:00.4 FireWire (IEEE 1394): ALi Corporation M5253 P1394 OHCI 1.1 Controller [10b9:5253] (prog-if 10)

The controllers must have internal arbitration between each other to handle the grant and request signals for bus mastering.

Without the extra logic for arbitration (such as provided by related controllers or by a PCI bridge), you cannot connect multiple controllers to a single slot.