Cromemco Cyclops 88 CCC board set reproduction

[hmb]

Excellent news that you entered the schematics into KiCAD and ran the DRCs.

That probably means that the errors are with the original schematics. I have just printed out the original schematics and will give them a check.

We already know there is at least one error in the original schematic (discussed at post #104) which was replicated in the KiCad schematic and shows up in the production boards.

Another set of eyeballs should also verify the KiCad schematics to make sure (a) they match the original schematics and (b) check for drawing errors. Here's why:

Just now I looked into the reason for the board error that Hugo identified in post #123 noting "on board 1 there is missing copper trackwork pin 4 of the 74193 (U38) and pin 4&5 of the U27, the 7420". The original schematic correctly shows this connection. I reviewed the KiCad schematic for the board and discovered why this happened - if you look closely at the pin 4/5 connection of U27, you can see the connection drawn from pin 4 U38 terminates just short of being connected to pin 4/5 of U27. This explains why KiCad did not route the connection, and why it is missing on the board.

In case anyone is looking for them, the KiCad files are here:

GitHub - nullvalue0/Cromemco-Cyclops-88-CCC-Reproduction: Kicad reproduction of the 2-board set "88 CCC", which are S-100 boards meant for use with the Cromemco Cyclops digital camera and Dazzler video board

Kicad reproduction of the 2-board set "88 CCC", which are S-100 boards meant for use with the Cromemco Cyclops digital camera and Dazzler video board - nullvalue0/Cromemco-Cyclops-88-CCC-...

github.com

 

[nullvalue]

Just now I looked into the reason for the board error that Hugo identified in post #123 noting "on board 1 there is missing copper trackwork pin 4 of the 74193 (U38) and pin 4&5 of the U27, the 7420". The original schematic correctly shows this connection. I reviewed the KiCad schematic for the board and discovered why this happened - if you look closely at the pin 4/5 connection of U27, you can see the connection drawn from pin 4 U38 terminates just short of being connected to pin 4/5 of U27. This explains why KiCad did not route the connection, and why it is missing on the board.

Wow I have no excuse for that! it was so close too.. weird

 

[daver2]

Agreed.

Checking the original schematics against the KiCad schematics is something everyone can do. It requires a methodical approach and a highlighter pen.

You can also extract a netlist from KiCad and compare the nets to the original schematics. If there is a missing PCB track, not all of the IC pins on the original schematic will be connected to the same KiCad net.

We are (in this case) looking for IC inputs and outputs that are inter-connected - but they form two (2) separate nets that do not cause a DRC conflict.

How should we procede with debugging?

Hugo has built his boards, and they present a certain fault scenario. Has nullvalue built his boards, and do they present the same fault scenario as Hugo's or not?

I have a few basic signals that we can check.

Dave

 

[daver2]

>>> Wow I have no excuse for that! it was so close too.. weird

Because the track is not present in the schematic capture part of KiCad, the original track forms two (2) separate networks within KiCad - each with a different 'name'.

The KiCad PCB layout tool will actively prevent you from interconnecting these two IC pins together, because they do not form part of the same net...

Dave

 

[nullvalue]

Hugo has built his boards, and they present a certain fault scenario. Has nullvalue built his boards, and do they present the same fault scenario as Hugo's or not?

I have not yet. I do finally have all my parts in and am going to continue building. In the meantime though I can go through and confirm GND/+5V for each IC. I'll work on that later this evening.

 

[daver2]

So if I work with @Hugo Holden for the time being?

Dave

 

[Hugo Holden]

So if I work with @Hugo Holden for the time being?

Dave

I will do the tests but what I want to find out first is if the original pcb's match the schematic or not. We know the original pcb's worked, regardless of any defects in Cromemco's documented schematic (or the two defects in the documented firmware for that matter) if I can regenerate the schematic from the pcb's alone I can use it to screen out any defects in both Cromemco's and @nullvalue's schematic. I'm more interested to do this initially because, there could still possibly be a remaining fault where the boards may seem to "work" but not work exactly the way the designer intended, like the pin mix up error in Pong which created the Ghost in the Machine Bug.

 

[daver2]

That's the big problem, I don't know of any scans of the rear of the original boards.

I agree that would be the 'gold standard' - but it may not be attainable.

Dave

 

[Hugo Holden]

That's the big problem, I don't know of any scans of the rear of the original boards.

I agree that would be the 'gold standard' - but it may not be attainable.

Dave

All off the scans of the rear of the boards were on the github link previously posted. Those and others are on the Deramp (Mike Douglas's website). I am working on these images to reverse engineer the schematic from those. It just takes a while. The challenge is actually the pcb top, because of course the IC's & sockets conceal the tracks under them. But there were a couple of unpopulated board top photos, even though one was in a plastic bag, of the orginal versions on Github. I have done this sort of thing before, I can usually regenerate the schematic from board photos with reasonable accuracy, in conjunction with the original schematic too. There is always the potential for an ambiguity, but if I avoid most of those, it has to help.

I managed it with this project and regenerated the entire schematic. But I did have one advantage, I had the original pcb on my desk and I could sound out connections with the meter. It is more difficult without that:

https://worldphaco.com/uploads/THE%20TAYLOR-WILSON%20PET%20PRINTER%20INTERFACE.pdf

Still I managed the RM-65 video card, a 4 layer board without actually having it on my desk, but some X-Rays helped for that one.

So I cannot say I can do it for sure, but I am going to try.

 

[Hugo Holden]

I realised it would be good to have some kind of a Tripod to put the Cyclops camera on.

I looked around for a small tripod and came across an insanely miniature one, the photo angle doesn't show it , its about 6 + 1/4 inches tall and about twice that with its legs extended. The seller didn't mention that one of the threads on one of the multi-axis hand friction nuts was stripped, but with vintage stuff I expect this, So I repaired the thread using a 6-32 Heli-Coil kit, that cost about the same amount as the Tripod itself. Still it is a quality plated all Brass item that will last forever, really cute, most items like this now are made of plastic.

 

[nullvalue]

Man, @Hugo Holden you certainly don't cut any corners! Looks amazingly era appropriate.

 

[Hugo Holden]

I crafted two 1 inch wide lengths of 2mm thick Satin Anodosed aluminium to make the side label plates. As far as I could tell, Cromemco stuck these on with glue or double sided sticky tape. I think they had them made by a professional label/plaque maker store. This sort of thing in not my M/O, especially in Australia where the heat melts tape glues and anything stuck on, has a habit of falling off, for example stick on rubber feet on instrument cases. And it is hard to get a glue to adhere well to a satin anodised surface too, because the surface has a glassy quality to it, typical of aluminium oxide.

So I screwed the plate down, just with two miniature 1mm diameter x 4mm screws for each label plate, with the thread cut into the label plate. The sticker largely obscures this. The sticker was made cear where the writing is, so that the satin anodizing shows through to give a metallic look to the lettering. I positioned the tiny screw heads so that they don't interfere with any of the structures inside the camera.

 

[Hugo Holden]

I'm still working on board 1, it is a very long job. There are a number of issues with Cromemco's schematic not exactly matching what was on their actual pcb. It is almost analogous to the defective software they published, maybe they didn't want people copying it, it is all a little odd. As it turns out replicating the schematic they published, would never have worked, and that is just for board 1. God knows what I will find on board 2. Just one of the few issues I have found on board 1 is that the missing link to the 560R resistor did not connect to pin 8 of the 7420 as we had initially thought, it just looked that way. The track actually passed by between pin 8 & Pin 9 and connects to pin 6 under the IC body.

In about a week or two I will be able to start on board 2 and after that I'll document the issues.

 

[Hugo Holden]

I have been conducting a detailed examination of the cyclops pcb's.
Still working on it !

 

[Hugo Holden]

While working on board 2:

I have come across something intersting that I don't like, but it will not stop the original board from "working".......Cromemco obviously did not pick it up.

Because it would be changing history to correct it, which has an odd feel to it. So I am conflicted whether to correct this error, or not, on a replica pcb.

On board 2 Cromemco fouled up the design of their own pcb foils, and it did not match their schematic. Though in other places the pcb the foils were good, but there were some schematic drawing errors and pin numbering errors, just as there were on board 1, but only one actual schematic error on board 2, that I don't think would count for normal operation. However the board 1 schematic drawing errors were catastrophic and would, if anyone copied the schematic alone, would stop the board from working.

In any case what Cromemco did, on their board 2 foils, they accidentally swapped the LSB's of two output data lines, from U43 to D0 & D1 pin 36 and pin 35 on the S-100 edge connector.

Pin 7 of IC U43, as the schematic shows, should pass to pin 36 (D0) on the S-100 edge connector and pin 9 of U43 should pass to pin 35 (D1) on the edge connector. But the connections are swapped on their acual pcb foils. The photographic evidence (unequivocal) from the available pcb images clearly shows these two connections are swapped over, obviously by accident.

It is interesting the implications of swapping the two LSB's on data lines (it is similar to the pcb accident in the Original Atari Arcade pong game in that case two LSB's got swapped on the foils) The effect of this sort of mix-up is immediately obvious in computational scenarios where digital numbers are presented to an observer beause they are messed up. A lot less obvious if the nibble of data controls the luminance of a pixel of a CRT image. Probably though, if the Cyclops Camera was pointed at a Grey scale image/test card, the irregularity in the image would have just been detected due to this mix-up, but I won't know until I have the system working. But as noted, it would work without this particular issue fixed, and we know the original boards "worked" but perhaps not quite as well as the designer intended.

 

[Hugo Holden]

I have now regenerated both boards by hand in Kicad from photographic images of the original boards and cross referenced the trackwork on both boards so that every track is accounted for. I decided with the Cromemco based foil error to simply make the board with a pair of links so it can be wired as they had it with the two LSB data bits swapped, or un-swap them.

After I have these new boards made I will test them, prove that they work and then publish all of the corrections and an accurate error free schematic of both boards. It should be possible with that to fix up @nullvalue's boards too. So far I have discovered that they have two types of errors, one type caused by Cromemco's schematic being in error and others because of errors entering Cromemco's schematic into Kicad.

There are some things:

Clearly the later revisions of the pcb's were designed to have a more complex interface with the S-100 bus, with connections that I don't think the original board versions had. An additional 7474 flip flop IC's was added to board 2 and the trackwork significantly altered compared to the first version of that board.

On board 1 there are a pair of change-over switches (configured as links) controlled by the PHLDA and PSYNC lines on the bus and also the SINTA is an option and they send a open collector signal out to the BUS on the /PINT bus terminal 73 which is the primary interrupt request signal. Does anybody have an idea of why these option links were provided? There are no "option links" on board 2 to alter the interface with the bus.

On board 2 there are also the priority in & out pads and a 1k pullup resistor was added in the second generation board 2. One photo of a board with a "tested sticker" on it shows these pads to be un-linked.

 

[Hugo Holden]

I have cut glass before and I know its foibles. When I cut the microscope cover slides I planed down their perimeter with fine abrasive paper, I know how microscopic cracks in glass can propagate, But, one issue was, when the tops are taken off the IC's, there is a fairly thick layer of solder and metlazation on the top surface of the IC. So when I glued on the cut down cover slides, they were at the perimeter on the long axis of the IC, proud from the surface of the ceramic IC body for perhaps about 0.2mm. When I installed the IC and pushed it into the socket, it caused the perimeter of the cover slide to crack. So I decided to re-think it.

What I did was to plane the surface of the IC top down using 600 grade paper on a flat surface, until the metalization on the top of the IC was almost removed. I then searched for suitable glass material to cover it and I found a perfect glass section 17 x 7 x 0.5mm option, a little thicker than the cover slide too, but its perimeter is pre-ground and not prone to cracking and it is stronger than the cover slide.

I was not entirely sure of the optical properties of it, and how well it would transmit longer wavelengths, it looks optically crystal clear, but when I got it, I tested it with an infra-red LED and it does not significantly attenuate the light. So now I have gone over to this part to make the clear top for the image sensor IC:

https://www.ebay.com.au/itm/188010777414?var=695686618070

 

[nullvalue]

Incredible work, Hugo! I can't believe you discovered those mistakes all this time later! Is there any indication that they worked around this in software somehow or did they just leave it? Too bad there's virtually no information that I can find about the later revision boards.. hopefully someone has some original documentation (perhaps with an errata?) laying around and comes forward? but I won't hold my breath.

Fingers crossed on your boards though! If all works out please let me know what the modifications were so I can update the github-hosted version.

Thanks also for sharing your update with the glass window. Were you not concerned about getting dust inside the IC while sanding the top?

 

[Hugo Holden]

Thanks also for sharing your update with the glass window. Were you not concerned about getting dust inside the IC while sanding the top?

Yes I was actually, so how I did it was to move the IC body over one section of the abrasive paper, in one direction only, at a time on a fresh section of paper. That minimized any particulate matter dispersed to the chip. Then, I used some low velocity compressed air to blow any residual particilate matter off the die area before fitting the glass seal.

I have found though, there is another pcb "replica" error we are facing, it is purely a mechainal one. The replica for the three pcb's in the camera, were again not exactly copied from the original Cromemco pcb's.

On one of the replica pcb's the socket for the image sensor IC, is placed on its long axis in the center of the board, so that it is symmetrically spaced. However on Cromemco's board, image sensor socket which is available for inspection on the DeRamp site, shows that the socket was not placed exactly there.

This is because, in the image sensor IC, the 2.54mm x 2.54mm die section, where the memory elements are, is not actually centered in the middle of the part IC body, even though the outer die perimeter was close to center. So when the Lens is mounted centrally on the housing, the sensor's middle is not on the optical axis with our current board. Cromemco knew this so they appropriately moved the socket on that board to allow for it.

 

[nullvalue]

I have found though, there is another pcb "replica" error we are facing, it is purely a mechainal one. The replica for the three pcb's in the camera, were again not exactly copied from the original Cromemco pcb's.

hmm do you think that ACC board will need to be redone or could one end of the board be "shaved down" so it fits more to one side? how much is it off?