Cromemco Cyclops 88 CCC board set reproduction

[Hugo Holden]

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?

I'm working on the calculation, I think shave the board at one end and add a shim at the other (about 0.6 to 0.8mm, I'll post the exact amount later) and it will still be retained in the slots in the housing, I think. It is off by about 1/4 the width of the light sensitive array.

 

[tech58761]

So, you corrected an error on one board, but on the second board, you further modify the design to allow people to choose between 'as produced' or 'as designed' behavior?
That could be confusing when you have people developing new code for 'as designed' behavior but you have people with boards set for 'as produced' mode.
I almost wonder if it would be possible to 'hide' the correction, the only clue being an 'apparent' extra trace by circling around pin 9 to reach pin 7.
Or I could just be rambling on about nothing.

 

[hmb]

I'm working on the calculation, I think shave the board at one end and add a shim at the other (about 0.6 to 0.8mm, I'll post the exact amount later) and it will still be retained in the slots in the housing, I think. It is off by about 1/4 the width of the light sensitive array.

For those of us who have not yet drilled and tapped the mounting hole for the lens, would it not be best to place that hole directly over the current location of the sensor array vs cutting and shimming the circuit board?

 

[Hugo Holden]

For those of us who have not yet drilled and tapped the mounting hole for the lens, would it not be best to place that hole directly over the current location of the sensor array vs cutting and shimming the circuit board?

Yes, one could move the lens hole center in the aluminium casing, although having said that, unless the casing is perfectly clamped and milled, a drill hole center can (usually always) drifts somewhat. So what I did was to do it in mostly 0.5mm steps and measure it with a dial micrometer to the 4 hole edges & 4 case edges, making the two measurement pairs on the X & Y axis match, and if the center shifted a tad after a drilling, correct that with a half round file before the next step, so the final hole was very close to bang on center. That could still be done, but it would be a tad more awkward with two different measurements on one axis, rather than just going for a match. Probably it is better to target the lens mount for a perfect center and move the IC/pcb to match that. But whatever works.

 

[Hugo Holden]

So, you corrected an error on one board, but on the second board, you further modify the design to allow people to choose between 'as produced' or 'as designed' behavior?
That could be confusing when you have people developing new code for 'as designed' behavior but you have people with boards set for 'as produced' mode.
I almost wonder if it would be possible to 'hide' the correction, the only clue being an 'apparent' extra trace by circling around pin 9 to reach pin 7.
Or I could just be rambling on about nothing.

The main schematic errors on board 1 was not showing the 560R resistor connected to the correct circuit point and two missing dots on the schematic where wiring lines cross that should be connected but were not. It was obvious because there are a pair of IC's with their inputs connected, but they were both not connected to any IC outputs. (This is the kind of thing the DRC in Kicad knows nothing about and could never detect, it is only detectable by noticing the input-output anomaly or by examination of the original foils)

The thing with this unexpected board 2 "error" where the schematic looks ok but the foils not, I am somewhat conflicted about it myself. I would in fact be happy with a replica of what Cromemco did, because that is "how they did it" at the time and we know it worked (Of course the purpose of analysing the foils was to find and correct errors in the documented schematics, I did not expect to find what appears to be a reverse example, a foil error)

I suspect that the effect of it would be merely to disorganise the grey scale order but only in a least significant bit way, and it must not have been very obvious, or somebody would have spotted it at the time.

But somehow I know that when I get it all running (When that happens!) curiosity will get the better of me and I will want to see what happens with it running the way their schematic was wired vs runnng the way they made their pcb.

So what I decided to do in this case (attached) was add a few extra pads and have a pair of links, one with the original wiring and the other with the fixed wiring to correspond with the schematic, so I can then examine the camera image in both cases and decide what to do about this, if anything. At the moment I am not 100% certain about the effect on the image of this unexpected foil error, but I'm guessing it cannot be enormous.

Aside from some gate and IC pin numbering errors, there was only one schematic error on board 2 where the foils did not match the schematic, involving one gate that drives the output terminal PR OUT, but unless the option terminals "PR IN & OUT" were linked, it would have no effect at all. And all historical board 2 photos I have seen, the PR IN & OUT are not linked, so it is academic probably.

 

[Hugo Holden]

I worked on the centering of the image IC issue. There was slop in the fitting of the original board to the slots in the case and it could move about 0.35 mm each way, on both axes. Close examination of the chip also showed a small vertical offset too, with the sensor array sitting a little lower than the midline of the IC body (when pin 1 is in the upper R with the IC's long axis on the horizontal meridian and the IC viewed from the top) by about 0.3mm. Simply by putting a strip of scotch 27 tape along the long axis of the pcb and using up the mounting slop on that axis, corrects for that when the sides are on the housing.

The board width is about 60.9mm, I shaved 0.5mm off one end and that amount, plus the existing about 0.35mm of slop was enough when spaced, to put the center of the light sensive array very close to the centerline of the lens & housing. The required spacer is about 1.2mm, I simply used a piece of 18 SWG wire which is about 1.2mm and a piece of polyimide tape to hold it to the board's end, but it doesn't really need that as the spacer cannot escape when the camera sides are on. Still its better not loose when the board is lifted out of the slots.

Obviously with variations in constuction, lens mounting hole, even the fit of the IC socket to the pcb, these details may need to be customised to some extent. But at least now the sensor array is in basically the correct position.

I also checked the distance of the sensor array to the lens mount in the 5th slot from the front, where that board normally sits and interestingly it is very close to correct (using a standard machine pin socket) that was some coincidence which Cromemco must have appreciated. But it pays to keep in mind there are variations in the profiles of some IC sockets. I thought it was off at first because my initial measurements (before I took the tops of some IC's) were to the ceramic surface of the sensor IC body, but with the IC top removed the IC die sits the correct amount below that surface making the total distance correct.

 

[tech58761]

Hugo: You have given it some thought, I'll grant you that, and I guess your approach is the best way to handle it!

 

[Hugo Holden]

Hugo: You have given it some thought, I'll grant you that, and I guess your approach is the best way to handle it!

Time will tell. A new set of pcb's are in the post, so within a week I should know........unless I have missed something.

I checked these pcb's twice against the corrected schematics and they passed. But, as Bob Dylan once said, "there are molotov cocktails and rocks behind every curtain"

 

[Hugo Holden]

The boards arrived today from my pcb maker in China (Storm Circuit, from the very helpful Mr. Kim Chan) much quicker than I thought.

The tracks on these basically match Cromemco's boards and the corrected schematics, except for the fact on their boards many tracks followed slightly curved pathways, but these had to be a little more geometric and angular, so the vias ( I use large masked ones as they make good test points) needed moving a tad this way and that to allow for it.

I will test it all tomorrow with the Camera and see what happens. There may still be some diagnostic work to do.

 

[daver2]

May ?!

Dave

 

[Hugo Holden]

May ?!

Dave

yes may-be

 

[Hugo Holden]

As usual, the plot thickens.

The new boards work fine, (there was only one tiny slip up in a single connection, despite my best efforts, that was easily remedied, those best laid plans of Mice an Men, it is a good thing I like Mice) The data from the sensor IC is now passed to the Dazzler' memory space and the software appears to be working. It takes 2 to 3 seconds to refesh each frame.

There was an issue with the clock pulse sent to the camera via the 560R resistor. Due to that and the camera cable's capacitance (higher in my case no doubt due to the ribbon cable and the extra earths in the cable) the clock signal became so heavily integrated, that by the time it arrived at the LM311 in the camera head, the level was so low that noise became a problem. This resulted in the bias lights randomly switching on and off. I simply changed it to 100R and it fixed that.

In the dark (no light on the sensor array) all the pixels go black, in room light on the sensor, they all go white (after a frame refresh) and with various shades of grey inbetween depending on the light level falling on the sensor array. The bias LED lights also act to lift the back level to grey. But as yet I cannot get the lens to focus an image on the sensor array to give say an image of a face or any definite geometrical object. It could be that the sensors I am using are the wrong kind, the one I was intending to use was damaged as it turned out, possibly my fault. So I am still working on this problem and I will report when it finally works.

 

[daver2]

Not a bad 'first outing'!

Dave

 

[Hugo Holden]

One issue I have found is that the lens I'm trying was a 36mm type rather than the specified 24mm type, which likely means the optical image falling on the sensor is larger and the image over-magnified and not helping, so I have ordered the exact sepecified lens. I'm also working on a plan to check if the sensor chip I'm trying, which is a 1972 vintage 4008-P is a type with a scrambled output or not. As noted the 4008-9 is not working, what happens with that one is that even in no light, the pixels nearly all migrate to white even with the chip in the dark without light excitation, probably I damaged it, not 100% sure.

On the topic of @nullvalue's boards, I have been unable to fault find those further to find the issues causing the computer lockup, I had to salvage what I could, the capacitors, vintage voltage regulators and all the IC's as I only had enough IC's to make one 88-CCC board set and they were the last of my mil spec AVX 0.1uF bypass caps and wanted to save them along with the expensive 47uF axial tant caps. Unfortunately I could not save the about 58 machine pin IC sockets and reprocess those, it would have used up too much more time. I spent a lot of time trying to solve the issues on those boards, including fixing the missing polygons, but I did not get very far and it was a costly exercise. However, if those boards were modified to definitely conform with the corrected schematics, in theory at least they should work, unless there are some other yet unidentified problems causing them not to conform.

A problem briefly cropped up in my SOL-20 , unrelated to this project. It has performed perfectly since I restored it some years ago with a massive IC pin cleaning and IC socket cleaning exercise (aside from one inverter chip dying), but immediately after power up, some flickering characters appeared to the right of the cursor and then vanished. So far not seen again. Something like that has a habit of returning with increasing frequency. Likely this will be IC socket-pin connection issues returning. I think I know deep down, one day, I will have to replace all of those horrible TI sockets in the computer to get full reliability, but that would be a very big project and on the back burner for now.

 

[hmb]

On the topic of @nullvalue's boards, I have been unable to fault find those further to find the issues causing the computer lockup, I had to salvage what I could, the capacitors, vintage voltage regulators and all the IC's as I only had enough IC's to make one 88-CCC board set and they were the last of my mil spec AVX 0.1uF bypass caps and wanted to save them along with the expensive 47uF axial tant caps. Unfortunately I could not save the about 58 machine pin IC sockets and reprocess those, it would have used up too much more time. I spent a lot of time trying to solve the issues on those boards, including fixing the missing polygons, but I did not get very far and it was a costly exercise. However, if those boards were modified to definitely conform with the corrected schematics, in theory at least they should work, unless there are some other yet unidentified problems causing them not to conform.

Could you post the corrected schematic of the two boards you created after verifying everything from the original board traces? I'd like to figure out whether the boards @nullvalue had manufactured are salvageable, or whether a new set of boards should be made based on your corrected schematic just as you have done.

 

[Hugo Holden]

Could you post the corrected schematic of the two boards you created after verifying everything from the original board traces? I'd like to figure out whether the boards @nullvalue had manufactured are salvageable, or whether a new set of boards should be made based on your corrected schematic just as you have done.

Yes I will post them as soon as I am 100% sure the boards are working as they should. At the moment, although everything appears to be "working" I cannot get anything that resembles an image out of the camera, it responds globally to light. It could be that the data is somehow scrambled, which could be a hardware issue in the boards still, or it it could be my working sensor IC is the wrong type or something with the pixel addressing in the computer memory or maybe that one pixel say is being represented in more than one place place in the image, with some sort of address issue. I have devised a way to check that today. As soon as I am 100% confident we are working with a correct schematic I will post it.

I just cooked up and tried an experiment that I am fairly certain confirms the image is not actually scrambled. It must be something to do with my lens not focusing the correct sized image of enough brightness on the sensor. What I did was to create a fiberoptic test tool, by getting six inches of 0.5mm diameter fishing line and feeding one end of it with a torch and masking over the torch face with carboard and tape. This created a miniature light pipe that I used as a probe to use in a darkened room to project a 0.5mm disc of light that I could move over the surface of the glass in front of the sensor to test it. The sides/corners and center all appear to be responding normally, so I think all the sensor elements and the electronics are likely working as they should and are very likely all in the correct order.

One thing I have observed, though I'm not sure if it is normal, when the frame is updating from top to botton, a few grey flickering pixels are seen horizontally migrating downwards with the leading edge of the new frame (a very coarse analogy of snow if you like) they are more visible when the update causes an increase in brightness, not visible if a darker shade of grey or black. Does anybody know if this effect is normal for this system ?

 

[Hugo Holden]

I have attached the schematic corrections. The effect of the Cromemco foil mixup remains unknown. I may be able to find out later when I get the Camera working better. I don't a have a way of copying the original higher resolution pdf's to modify them directly, but you sould be able to see more clearly with those.

I also discovered that the 2.7k pullup resistor on the output of the LM311 comparator in the camera head is inadequate/borderline and it should ideally be changed to 1k. Also if there is a long camera cable I would recommend changing the 560R resistor on pin 2 of the camera connector to 100R and also its not a bad idea to use a 470R pullup resistor on pin 6 of U27 too. TTL outputs are pretty hopeless at driving a cable's capacitance without a pullup resistor. Also the 560R resistor on board 2 on the camera data signal should be fitted in my opinion, as per the schematic.

 

[Hugo Holden]

I was working on the camera today and the 9 pin connector fell out. I plugged it back in and the camera started again, the equivalent of hot plugging, at least for a second or two. Then there was a significant power glitch, the SOL rebooted to the cursor, the camera program disappeared from memory and the device I use to power the SOL (A type of constant line voltage machine) made a correction as if it was trying to compensate for a voltage glitch (it uses a motor driving a Variac to run a stable low line voltage to the SOL in the interests of running it cool). So I reloaded the program, but then nothing on the screen. Had the Dazzler failed, one of the 88-CCC boards or the camera head ? I quickly confirmed the Dazzler boards where working by running K.scope and that likely the two 88-CCC boards were ok too on some quick tests with the scope.

Then I went to the camera head, the first thing I noticed was there was no video out of the sensor IC, but the 4040 address counter was running, but the reset pulse to the 4040 chip was AWOL. But the reset pulse was present in the composite clock pulse from the board set arriving at the camera. On scoping the circuit that processes it in the camera head (a pulse width detector circuit) I found the cause, the MC14572 IC had failed. Wouldn't you know it, I don't have another on hand, it was the only one I had in my box of CMOS chips! so I will have to wait a week to get one. These are not common at suppliers in AU other than Digikey or Mouser, but they still have to come via Fedex or similar from the USA at a cost of $24 shipping. There a are a lot on ebay but mainly overseas.

Why did it fail ? The only thing I can think of is that it got irregularly power cycled by the separate + and minus -8V supplies it runs from, perhaps one came on before the other when the plug re-connected and the chip went into a latch up state (when this happens two output devices in series across the power supply get switched on together) or maybe it sourced current via an input substrate diode. Interestingly the nature of the fault in one inverter in it that has failed, the input is dawing current when it should not and the gate output is stuck, but that is probably due to melted junctions. I think the device transiently shorted out the power supply when it failed producing a large voltage glitch. It is a interesting lesson, perhaps the camera head should never be hot plugged or if the plug comes out, shut it down before refitting the plug.

The circuit there is interesting, how they did it was to charge a capacitor at the gate input with positive going clock pulses from the LM311 comparator. If the pulses are only 1uS wide the gate voltage only climbs to a value below the gate's threshold because when the pulse goes low a diode discharges the capacitor to close to -8v, but if a wider pulse is injected into the clock pulse stream the voltage can then climb across zero and the gate's input threshold and into the positive area and that initiates a reset and the acquisition of a new frame starts. And even longer pulse allows switching the LED's on and off. The thing is a tricky and cheap way to gift a clock pulse line two additional functions.

 

[Hugo Holden]

Further to this post above, now I have time on my hands waiting for parts, I have been pondering the notion of how to protect IC's which in this case are running off two independent + and - DC supplies when it is possible that one supply rail comes up before the other. I think the logical conclusion is that there needs to be a high capacity filter capacitor, say at least 1000 uF connected between the +8V and - 8V voltage terminals, so that in the event that either the + or the - voltage comes up sooner than the other the charge is shared on the capacitor and the voltage comes up more symmetrically about zero, and no harm can come from it.

 

[daver2]

The Nascom 1 PSU ( https://80bus.co.uk.mirror.jloh.de/pages/nascom/documentation/Nascom2_3a_psu.pdf) had diodes on the outputs of the +12, +5, -5, and -12 Volt rails to prevent a lower voltage rail from exceeding a higher voltage rail.

I suspect the smoothing capacitors also played a part in the design.

This prevented failures from occurring in devices such as the DRAM should a power supply rail fail (for instance).

This, obviously, doesn't have any immediate relevance to your immediate problem, but it does show that some designers did appear to consider power up/down sequencing.

Dave