• Please review our updated Terms and Rules here

Microchannel IDE controller

Chromedome45

Veteran Member
Joined
Jul 6, 2009
Messages
3,993
Location
Central Florida
I was wondering how difficult would it be to make one of these like was done with the XT-IDE?
I know MCA was not widely used but was just wondering. :confused:
 
It could probably be done--I think there are still some Xilinx reference designs around for using a CPLD as an MCA interface.

But how big is the market for MCA boards?
 
I was wondering how difficult would it be to make one of these like was done with the XT-IDE?
I know MCA was not widely used but was just wondering. :confused:

Hi! If you want to see this become real I recommend you take on the project personally and make it happen. It is how XT-IDE or any other home brew type project works; someone took an interest in it and invested the effort to develop and produce it. Waiting around for "someone else" to do it almost guarrantees it will never happen.

Just my $0.02 Thanks and have a nice day!

Andrew Lynch
 
Isn't MCA IBM proprietary so that anyone developing an MCA card needed to pay royalties to IBM? I thought that's why MCA died in the first place.
 
As I understand the arrangement, MCA is like PCI or USB--if you're making a card for commercial exploitation, you need to have an ID registered. I thought the licensing fees only applied to motherboards, not to expansion cards, but I could be mistaken--it's been a very long time since I did any research on the subject. I do recall Xilinx EPLD reference designs on the card interface.

I wonder if the licensing apparatus still exists at IBM in any case.
 
Last edited:
Hi! Given the small number of hobbyists and high technical hurdles of an MCA IDE board I think the project is quite unlikely. Many factors made the XT-IDE project feasible like existing designs or near solutions with minor modification. The existing designs used commonly available parts and were supported in the EDA tools libraries (KiCAD) for the board outline, card edge connector, simple DIP construction, etc. ISA is a simple and open architecture and making home brew components for it is quite easy.

An MCA IDE project would likely require custom CPLD/FPGA work, custom board outline, SMT construction, hard to get parts and technical data, possibly licensing, etc. Compared to an MCA IDE board XT-IDE was an almost trivial exercise. Not impossible but highly unlikely. It would require a ton of effort with little benefit outside of a select few. IMO, the effort is better spent on projects with wider appeal such as other ISA boards, S-100 boards, home brew projects, etc.

IBM's MCA was a closed proprietary dead end from the very beginning and never took off as far as I know. There were relatively few computers built using it, few boards, and almost no hobbyist community. Although MCA did briefly see some limited acceptance with industrial systems IBM did not do themselves any favors with MCA regardless of its technical merit. Closed architectures rarely prosper and MCA follows the common path. The whole PS/2 and related debacle nearly drove IBM out of the PC business entirely.

Thanks and have a nice day!

Andrew Lynch
 
Hi Andrew!

I think calling MCA a "dead end" is overstating things somewhat. While it's true that it rendered the PS/2 series at a competitive disadvantage, it was a fairly successful architecture deployed in IBM's RS?6000 and AS/400 systems. Pricewise, it couldn't compete with cheap ISA-based designs and ultimately, PCI rendered it obsolote. But it was significant enough that Intel produced a chipset for it. Compared to ISA, MCA is lightyears better.

There were a number of other open architectures that didn't fare well at the time. EISA, VLB are two that come to mind. The other force dooming it was the increased integration of peripherals onto the main motherboard. PCI was just the final stroke.
 
Hi! Rather than expending effort on an MCA project, I think a PCI home brew project would make a lot more sense. I've seen PCI board outlines in KiCAD and the parts are still relatively available especially since they are going to tend toward SMT style parts. I don't recall ever seeing a DIP PTH construction PCI board. Get one of those Winbond/VIA/SMC etc SuperIO/MultiIO controllers for PCI and place it on a board with IO connectors. That would be much more useful than an MCA board and helpful in adding things like legacy IO to current computers.

Whether MCA was truly "dead end" or not is probably debatable but it was never common IMO outside of some fairly rare speciality applications. There were probably only a tiny amount of MCA boards actually shipped compared to ISA, S-100, or maybe even MultiBus. They were rare during their heyday and even more so now. I hardly ever come across them and I remember the big hub bub IBM made back in 1987(?) when the first rolled out the PS/2 machines with MCA. I still have one of their cups they were handing out by the score.

I had the same reaction when I saw the first PS/2 MCA machines I had when I saw the Mac 128K for the first time... weird connectors, closed, proprietary, required special tooling, and even opening the case was difficult. Ugh. What a disappointment.

Thanks and have a nice day!

Andrew Lynch
 
You know, if I look at most of my systems with PCI slots, several of them have only the video slot populated (PCIe or AGP); the others have perhaps one other slot populated (wireless network adapter most often; maybe a SCSI controller). Everything else that I need is on the motherboard.

I suppose that what I'm observing is the death of internal expansion slots and the rise of mobile computing.

I submit that PCI is probably on its way to sunsetting and USB is probably the most forward-looking option for a designer looking to hit the broadest base.
 
I wouldn't lump VLB in with MCA and EISA, VLB was pretty much standard on most 486 systems of the time unlike MCA (just IBM and a couple PS/2 clones) and EISA (mostly servers).

PCI development is kind of dead but some interesting cards are still being designed for PCIE (short slot not the 16x ones used for video cards). I have seen some video capture cards, RAID controllers, audio boards, HD TV tuners, etc for PCIE. Granted everything seems to come built in these days so just niche products are being produced outside of video cards.

I can't see spending the money and time to make an IDE card for MCA. There are plenty of SCSI cards for MCA which you can then rig with an IDE to SCSI adapter if you need more space.
 
[snip]

Everything else that I need is on the motherboard.

[snip]

I submit that PCI is probably on its way to sunsetting and USB is probably the most forward-looking option for a designer looking to hit the broadest base.

Hi! Yes, that is exactly why I am asking.

Rather than expending energy on an MCA solution that fixes a problem for a small number of hobbyists, why not go for something a bit larger?

A lot of the "legacy IO" ports are disappearing on new machines. It is becoming more common to see even desktop computers without floppy drives or even the FDC connector on the motherboard. Same for PS/2 keyboard and mouse connectors, parallel, and serial ports. I haven't seen an old style joystick (15 pin D shell connector) on a PC in quite a while.

A PCI SuperIO/MultiIO would be very handy for accessing floppy disks and playing legacy games, using old software etc. However I haven't seen such a beast in a long time if ever. I sort of recall seeing one a long time ago but it was like mid 1990's or so or maybe earlier.

Thanks and have a nice day!

Andrew Lynch
 
It was just a thought really. I think the number of people on this forum using an MCA bus machine could be counted on 1 hand. I am not technically inclined enough to try to tackle this
on my own so consider the matter a dead issue. ;)
 
Unknown K said:
I wouldn't lump VLB in with MCA and EISA, VLB was pretty much standard on most 486 systems of the time unlike MCA (just IBM and a couple PS/2 clones) and EISA (mostly servers).

VLB could be pretty twitchy if you were trying to use more than one vendor's VLB boards in the same box. I have a 486 at the bottom of a pile here with an Adaptec 2740 VLB SCSI card and a Tsenglabs VGA board, but that's about as far as I went. When Pentiums came in, VLB went away. You could just about call VLB a processor-specific bus, judging from its deployment.

Rather than expending energy on an MCA solution that fixes a problem for a small number of hobbyists, why not go for something a bit larger?

A lot of the "legacy IO" ports are disappearing on new machines. It is becoming more common to see even desktop computers without floppy drives or even the FDC connector on the motherboard. Same for PS/2 keyboard and mouse connectors, parallel, and serial ports. I haven't seen an old style joystick (15 pin D shell connector) on a PC in quite a while.

A PCI SuperIO/MultiIO would be very handy for accessing floppy disks and playing legacy games, using old software etc. However I haven't seen such a beast in a long time if ever. I sort of recall seeing one a long time ago but it was like mid 1990's or so or maybe earlier.

Rather than employ a SuperIO board, how about just making a PCI-to-ISA bridge board to allow folks to use whatever ISA boards they happen to have around?

Google 'PCI-to-ISA Bridge" and you'll get a lot of hits. Apparently, it's pretty straightforward to implement in an FPGA.
 
I have a couple VLB network cards, so I have made 3 VLB cards work in a 486 (Video, IDE or SCSI card, network card). Most systems that had VLB just had a generic VLB video card and generic IDE/IO VLB card, the higher end video cards, caching IDE controllers, and SCSI cards were a bit rare (but that has not stopped me from collecting a nice bunch of them).

I guess some point down the road computers with ISA slots are going to be hard to get, better stock up on P2 and earlier motherboards now. How will you mount a ISA card with PCI bridge in a case anyway?
 
After recently researching MCA quite a bit, learning a lot, and righting some misconceptions... I still have a lot of unresolved thoughts floating around. Basic MCA is shocking similar to ISA. I never realized that before. In fact, other than qualifying bus cycles with two extra strobes (CMDn & MAKE24), all 16 address lines for I/O, and adding some simple logic for returning the POS board ID, PIO+IRQ designs are nearly identical to ISA. I've also seen (and purchased) several boards that are PTH DIP designs to probe as references. And you have to implement the jumperless piece of the spec if you are willing to support the board at a more intense level - eg. tailoring an ADF file with fixed reservations based on jumper settings.

A home-brew IDE interface with DIP parts is certainly within the realm of possibilities. A MCA to ISA bridge - even with one that translates DMA request/acks into MCA arbitration requests is also straight forward, but a bit impractical mechanically as ISA cards are typically bigger than MCA. I've also built a test board with an FPGA to work on a bus mastering bridge design in Verilog. I don't have any current practical objective other than learning, and I have too much on my plate to start anything new, but if there is anyone intimately familiar with MCA, I'd like to discuss some questions I have. This app note from National has been the most helpful resource I've found:

http://www.tij.co.jp/jp/lit/an/snla024/snla024.pdf

But things like the fairness design described in both the app note and the MCA architecture book written by the spec authors doesn't make any sense from a functional or practical perspective. In fact, if you take the NS app note design, drop the fairness section, only have to support one DMA request, and update the SPLD deisgns, you could make a DMA aware MCA board with an astonishingly small amount of components. A PIO only design without PLDs would be similar component counts.

The biggest problem with a homebrew MCA design is sourcing appropriate back plate connectors - which is a hard enough task for ISA.
 
Last edited:
A MCA to ISA bridge - even with one that translates MDA request/acks into DMA arbitration requests is also straight forward, but a bit impractical mechanically as ISA cards are typically bigger than MCA.

How about MCA to one or more PC/104 modules? I've actually been considering doing a PC/104 XT-IDE and making a full-length ISA PC/104 carrier -- that way, you could mount up an XT-IDE, a PC/104 serial board, and a NE2000 compatible NIC (for example) and only use one slot. The base board would be huge and therefore somewhat expensive, but the modules would be cheap. PC/104 stuff is available as surplus, and designing a board would probably be easier for most hobbyists than doing a proper ISA board, since there's no edge connector.

The biggest problem with a homebrew MCA design is sourcing appropriate back plate connectors - which is a hard enough task for ISA.

The ones used on the XT-IDE boards are a standard Keystone line item -- Mouser carries them. While I'm certain finding a stock of MCA brackets would be more difficult, it probably wouldn't be too horrible to just have some stamped out in a small production run. One of the other members here had a bunch of ISA brackets laser cut at a local machine shop for his XT-IDE builds, something similar to that could work.
 
How about MCA to one or more PC/104 modules? ...you could mount up an XT-IDE, a PC/104 serial board, and a NE2000 compatible NIC (for example) and only use one slot.

I didn't think of PC/104. But you are supposed to be discouraging me from more rabbit holes and testing XT-IDE instead! Fail. :) But NICs and even serial cards are still relatively available for MCA on eBay. I did the FPGA board so I could eventually play around with a proper SD-card solution with either DMA or proper bus-mastering. Practical storage solutions don't really exist. I have a SCSI card and SCSI2SD in my Tandy 5K. But my PS/2 m70 is more limited due to only 3 slots - currently occupied by RAM, NIC, and XGA2. I'm also really wanting a DOS compatible MCA sound card and they are so rare, I'd have a quicker route building one despite my work load. I was musing today about a Gravis GF1 chip on a MCA card. But it will be just a dream for quite a while. Plus DMA incompatibilities with the MCA implementation will likely prevent any software designed for ISA/8237A operation from realistically working on MCA. But, from what I remember from the GUS SDK and the s3m player I wrote long ago, I believe it might work as DMA is only used for uploading voice samples.

...MCA brackets would be more difficult, it probably wouldn't be too horrible to just have some stamped out in a small production run.
It's a simple stamping with 3 bends - 1 of them in an opposing direction. Anyone have an idea on where to go to get quotes for such a thing? On-line preferably due to economies of scale and long term sourcing. The blue plastic tabs and card extensions can just be 3D printed.
 
Last edited:
Seeing how I'm optimistic since eeguru's posts, if there is a narrow future for a possible "PS-IDE/2" prototype design... make it happen for god's sake! :nervous:
 
Back
Top