I'm doing this in kind of a hurry, so apologies if I miss replying to anyone's points.
One of the primary motivations for this is the (untested) possibility that the CPU adapter will allow the Turbo040 to work properly with a Micron Exceed card. My understanding from Gamba's page is that there are compatibility issues between the Turbo040 and Micron card with a PDS adapter.
Of course, it may turn out that compatibility is no better with a CPU adapter.
The other motivation is simply that it leaves the PDS slot free for normal expansion, without trying to pile too much stuff on top of it.
Bunsen covered flat flex cable pretty well. It is available in precut lengths with the ends properly terminated (fabricated to be connector ready). But 50 conductor seems to be as large as the pre-made jumpers come. Larger numbers of conductors seem to require custom arrangements and this will never be a large enough run to justify that.
I think that Artimix's design could be reproduced for less than he sells it for. However, in ones or twos, it would not be much less.
Earlier postings discussed why Artmix (or anyone) charges as much as he does, when having several hundred made. It boils down to two points. First, the materials cost more than folks probably thought. Two, he's got something like $10,000 sunk in the manufactured boards, and he has no way of knowing how many he'll sell, ultimately. So high price slows his sales, but it also helps to get all of his money back.
I don't think that Artimix's type of adapter could be done for less in materials than this CPU adapter would cost. Mainly because PCBs are charged by the square inch, and there are a lot of square inches in his design.
However, note that his design also extends the CPU lines a long length. He has them on a board instead of on a cable, and so gets some advantage because presumably the board has ground and power planes, but at these (low) speeds, it's not much of a difference. In other words the signal integrity on his board is not going to be substantially better than one with a cabling solution.
A pair of 68 pin SCSI cables might be doable, but I'm not sure it would lower the cost much. I'll look into that some time.
The Ultra-ATA 40/80 conductor cable would be ideal because the extra 40 pins would provide signal shielding. However, three cables would be needed, the density is lower so it takes up more room and more PCB space, and I have not been able to identify manufacturer's numbers for the cable connnectors. Each of the three connectors on an Ultra ATA cable is unique and only one of them is really suitable for this application.
I posted a request for the part numbers, but no one here seems to know them. Perhaps I should try over on Ars Technica?
I understand that there are some differences in SE/30 frames. The one I looked at last night does not have anywhere where one could run a vertical circuit board up to the PowerCache or Turbo040. The only decent spot is the slot for the PDS slot, and I think any such scheme would interfere wtih installing cards in the PDS slot.
I'll take another look at that possibility, but the type of high density connectors that might be needed probably can't be hand soldered. You can't get enough density in the space available for 120 pins using a plain card edge connector. There are some really cool board to board connectors available now, but they tend to be expensive and require professional soldering.
The other advantage of cable is that it's a lot easier to plug in the CPU socket board with the cables attached, install the logic board. Thread the cables up through the holes. and then connect them to the remote board. With board to board connections, you would end up either having to wedge something into place, or pressing a board down into another board after logic board insertion, and possibly dislodging the receiving board in the process.
The .5 mm pitch flat flex cable does not come terminated for soldering, only for use with connectors. The larger pitch FFC is available with solder terminals, but it takes up more space and does not come in as numerous of conductor sizes, so more total cables are required.
This may be a project with too many compromises to be viable. Plus, if it doesn't give some advantage in Turbo040 compatibility, then there may be no point.
I'm at the feasibility study point right now and the geometry was my next project marker. Needing to do more than one board may have killed it right there. I really wish it were possible to do it as one board.
I wonder if there is a board mount mate for that logic board/analog board cable connector? I'm thinking in terms of a CPU socket board that plugs into the CPU socket and into the logic/analog board connector. That way it can cover that connector. Then the analog/logic board cable could plug into the adapter board somewhere else. Hmmm.
That would add the cost of two molex connectors though.
Ooops. I digressed. So the next step is to pull the IIcx out of the attic and test it with the Daystar IIcx adapter and PowerCache and Turbo040. It should work, since that's what it was made for, but it's a good assurance test. After that I'll see if there's any way to fit the IIcx adapter in a dismounted SE/30 board. If there is I'll test the assembly on the bench. If not, then I'm at a stop/proceed point as to whether to spend money on a prototype circuit board, which would cost about $200.
One of the primary motivations for this is the (untested) possibility that the CPU adapter will allow the Turbo040 to work properly with a Micron Exceed card. My understanding from Gamba's page is that there are compatibility issues between the Turbo040 and Micron card with a PDS adapter.
Of course, it may turn out that compatibility is no better with a CPU adapter.
The other motivation is simply that it leaves the PDS slot free for normal expansion, without trying to pile too much stuff on top of it.
Bunsen covered flat flex cable pretty well. It is available in precut lengths with the ends properly terminated (fabricated to be connector ready). But 50 conductor seems to be as large as the pre-made jumpers come. Larger numbers of conductors seem to require custom arrangements and this will never be a large enough run to justify that.
I think that Artimix's design could be reproduced for less than he sells it for. However, in ones or twos, it would not be much less.
Earlier postings discussed why Artmix (or anyone) charges as much as he does, when having several hundred made. It boils down to two points. First, the materials cost more than folks probably thought. Two, he's got something like $10,000 sunk in the manufactured boards, and he has no way of knowing how many he'll sell, ultimately. So high price slows his sales, but it also helps to get all of his money back.
I don't think that Artimix's type of adapter could be done for less in materials than this CPU adapter would cost. Mainly because PCBs are charged by the square inch, and there are a lot of square inches in his design.
However, note that his design also extends the CPU lines a long length. He has them on a board instead of on a cable, and so gets some advantage because presumably the board has ground and power planes, but at these (low) speeds, it's not much of a difference. In other words the signal integrity on his board is not going to be substantially better than one with a cabling solution.
A pair of 68 pin SCSI cables might be doable, but I'm not sure it would lower the cost much. I'll look into that some time.
The Ultra-ATA 40/80 conductor cable would be ideal because the extra 40 pins would provide signal shielding. However, three cables would be needed, the density is lower so it takes up more room and more PCB space, and I have not been able to identify manufacturer's numbers for the cable connnectors. Each of the three connectors on an Ultra ATA cable is unique and only one of them is really suitable for this application.
I posted a request for the part numbers, but no one here seems to know them. Perhaps I should try over on Ars Technica?
I understand that there are some differences in SE/30 frames. The one I looked at last night does not have anywhere where one could run a vertical circuit board up to the PowerCache or Turbo040. The only decent spot is the slot for the PDS slot, and I think any such scheme would interfere wtih installing cards in the PDS slot.
I'll take another look at that possibility, but the type of high density connectors that might be needed probably can't be hand soldered. You can't get enough density in the space available for 120 pins using a plain card edge connector. There are some really cool board to board connectors available now, but they tend to be expensive and require professional soldering.
The other advantage of cable is that it's a lot easier to plug in the CPU socket board with the cables attached, install the logic board. Thread the cables up through the holes. and then connect them to the remote board. With board to board connections, you would end up either having to wedge something into place, or pressing a board down into another board after logic board insertion, and possibly dislodging the receiving board in the process.
The .5 mm pitch flat flex cable does not come terminated for soldering, only for use with connectors. The larger pitch FFC is available with solder terminals, but it takes up more space and does not come in as numerous of conductor sizes, so more total cables are required.
This may be a project with too many compromises to be viable. Plus, if it doesn't give some advantage in Turbo040 compatibility, then there may be no point.
I'm at the feasibility study point right now and the geometry was my next project marker. Needing to do more than one board may have killed it right there. I really wish it were possible to do it as one board.
I wonder if there is a board mount mate for that logic board/analog board cable connector? I'm thinking in terms of a CPU socket board that plugs into the CPU socket and into the logic/analog board connector. That way it can cover that connector. Then the analog/logic board cable could plug into the adapter board somewhere else. Hmmm.
That would add the cost of two molex connectors though.
Ooops. I digressed. So the next step is to pull the IIcx out of the attic and test it with the Daystar IIcx adapter and PowerCache and Turbo040. It should work, since that's what it was made for, but it's a good assurance test. After that I'll see if there's any way to fit the IIcx adapter in a dismounted SE/30 board. If there is I'll test the assembly on the bench. If not, then I'm at a stop/proceed point as to whether to spend money on a prototype circuit board, which would cost about $200.