Homebrew 6502 c1977

[Eudimorphodon]

Your CCS board says 1976, so maybe they existed unofficially for a while? Or perhaps I'm muddying the waters? It's hard to know.

It’s quite the mystery. The oldest ads I could find for CCS in the Byte magazine archive were from 1979, and all the references to the company talk about a 16k S-100 RAM board being their first product, but one of those early Byte ads claim they have a lineup of products for *other computers* like the TRS-80… but I haven’t tracked down what those actually were…

Does anyone know if this card cage/form factor has a common name? It’s probably some kind of very common industrial/ communication thing, but maybe if I could search for *form factor name* + “computer” it might turn up some hits.

 

[ic2024]

When you have a chance, and are ready to apply power to the rack, could you please check the voltage levels on the VFD PCB? I'm reasonably certain the plate rail is at +40v. But no idea what the two filament rails are using, and if one may be at ground or both floating at near 5v?

 

[Eudimorphodon]

When you have a chance, and are ready to apply power to the rack, could you please check the voltage levels on the VFD PCB? I'm reasonably certain the plate rail is at +40v. But no idea what the two filament rails are using, and if one may be at ground or both floating at near 5v?

Sure, I’ll look at it as soon as I can. Didn’t get a chance to mess with it all weekend.

I’ve been thinking of taking that unpopulated board that was in the box that has a mating card edge for the backplane and adding an IDC pin header to it so I can piggyback a stack of the proto cards I have on hand into there to facilitate development of a replacement CPU. Need to get on formally mapping the slot pinouts…

 

[Eudimorphodon]

@ic2024 : On the topic of the VFD display, do you know if there's a maximum switching speed for those lights that might actually damage them? Wondering if there's a cap on how fast I should try driving the CPU bus lines. (I assume it must be "okay" to at least get into the 1Mhz ballpark, given that was pretty typical for early 6502s?)

 

[ic2024]

The data sheet (assuming the Tung-Sol 6977 is the right one) does not mention any frequency limit. The phosphor persistence curve shows light output build-up reaches 75% in 8 microseconds, with decay from full brightness down to 25% faster at 4 microseconds.
Interelectrode capacitance: grid-to-plate 1.8 pf, grid to filament 1.0 pf. Given the 100k grid resister, this puts the RC time constant at 0.1 microseconds. Perhaps this provides a natural dampening for input frequency above 10mhz?

 

[ic2024]

I'm going by this spec sheet: http://www.r-type.org/pdfs/6977.pdf

 

[Eudimorphodon]

Okay. It sounds like *technically speaking* if we were filming with a fast enough camera even the slowest normal speeds for a 6502 you could cycle through memory at speeds that would technically leave the lower address lines in a sort of PWM'ed state of semi-brightness, but it doesn't sound like it should hurt it. (If we interpret 12 microseconds as a "full" on-off that caps it around 100khz, I think a 6502 at 1Mhz can churn through NOPs at around 500khz.)

Just checking, because normally microcomputers would have LED based front panels.

I should probably look at implementing a single-step (or slow-step) circuit, just for fun. Probably wouldn't be super useful without a full input front panel, but it would make a cool demo and let the display really show itself off.

 

[ic2024]

I suspect the VFD display panel from CASINO was actually the DEC PDP-3. Only one was ever built, looks like this was it! See https://gunkies.org/wiki/PDP-3
This PDP-3 went from Scientific Engineering Institute, Waltham Mass, to M.I.T Cambridge Mass, to the Oregon Zoological Research Center (where I worked on it), to my home computer, now to San Jose!
It was 36-bit, the dates and technology match up. It had 1-inch mag tape drives for Radar data. See PDP-3 section on https://en.wikipedia.org/wiki/Programmed_Data_Processor

 

[Al Kossow]

This PDP-3 went from Scientific Engineering Institute, Waltham Mass, to M.I.T Cambridge Mass, to the Oregon Zoological Research Center (where I worked on it), to my home computer, now to San Jose!
It was 36-bit, the dates and technology match up. It had 1-inch mag tape drives for Radar data. See PDP-3 section on https://en.wikipedia.org/wiki/Programmed_Data_Processor

Could you elaborate on what "to my home computer, now to San Jose!" means? (ed. I found your Oct 24 post since writing this)


Do you have any pictures of the OZRC system?

Before Harlan died, I verified from him the two planned PDP's were PDP-2 == 18 and the PDP-3 == 36 bits

Where at MIT was it before going to OZRC?

 

[Eudimorphodon]

Yesterday I was finally able to spend some time on the system, first time in a couple weeks. I found there was some shipping damage to the power supply mounting that was allowing the PSU to shift around, which I’m glad I found before I applied AC to it because it was putting some bare wires to the transformer *very* close to the case metal. Luckily I was able to find the screw/stand-off that had popped out and get it reseated… although I am still kind of thinking I might want to add some electrical insulation to the bottom of the supply to make sure I don’t accidentally sneak a finger in there while it’s hot.

Another thing that came loose was the 40v line to the VFD panel. (Stripped end of the wire broke off in the screw terminal.) With that repaired I can verify that all the lights in the panel will illuminate if it’s unplugged from the interface card… which honestly baffles me slightly because the three wires going from the VFD to the PSU are terminating on +5v, +40v, and 1.4v AC. Where is it getting a ground, through the AC line?

(Picking this up some hours later.) I’ve been fooling around some more with things. Putting a VM on the power supply outputs showed the following:

+40v: around 56 volts
+5v: 6.1v (that’s a bit high for comfort…)
+12/-12v: close enough
1.4v AC: around 1.4v AC

Looking at the VFD board it does seem to have ground (and +5v) connected straight across from the bus (although I haven’t yet been able to trace out where they actually terminate), so I reseated the board in the chassis and tested things again. This time +5v is reading closer to 5.6v, everything else is about the same, which seems a little spooky. With the VFD board connected to a powered up IF card but nothing else on the bus most of the lights are on, but a group of seven are off, along with a couple singletons.

I’m a little worried more of the lights aren’t off with the interface powered on, since my gut feeling would be that the inputs of the 7417s would float high? (And thus pull down the line going to the VFD on the output.)

I need to take a break from this for a while, but I took the VFD display out of the enclosure and will beep out all the connections to map out which lights are linked to which bus pin. The display has 32 lights on each row, ten groups of three and a singleton on either end; this means about two thirds of the bottom row are undriven and thus on constantly. (Wondering if I have any open collectors in my chip stash, maybe I can add drivers to them and plug it straight into another board?) Have some pictures on my phone of everything, I need to read that VFD datasheet and make sure 56 volts isn’t going to cook them.

Analog and PSUs are not my strong point, so the 5.6v on the 5v rail has me worried. I’m going to put the scope on it as well, since the hertz counter on my VMM is telling me it can see 60hz hum on it…

 

[Eudimorphodon]

Looking at the VFD board it does seem to have ground (and +5v) connected straight across from the bus (although I haven’t yet been able to trace out where they actually terminate)

I did some more digging, and... I guess I'm still kind of confused how the VFD can run without a ground. (Again, unless the AC line takes care of this.) Digging into what's actually bridged over from the bus interface card it kind of looks like ground *isn't* coming over; +5v is, but, well... take a look at this:



The gray wire with red and white stripes is the +5v line that's directly connected to the PSU, and the red line labeled "To Bus" is the one coming along with the each-controls-a-light-wires; they're just joined together where those power feeds are joined to the "power bus" bars feeding each light. There really doesn't seem to be a dedicated ground.

Anyway, while just poking around now I had a forehead slapper of a realization: I had noticed from the other side (see this picture):

That the bus lines took a weird "swerve" in the middle. I'd assumed that those slots were "wired differently"; apparently the truth is that those two slots are using different pitch connectors:



The board we were calling the "select" card was in one of those slots, and you can see how it has different fingers from all the others cards that were still in the chassis:



I'm wondering if there's some significance to the fact that these different pitch card edges are carried over to the hacked OSI CPU board and the modified TV Typewriter; the pitch of the "Select" card matches the card edge that was added to the OSI board, while the card edge added to the TV Typewriter matches all the other slots.

Currently debating whether I can live with 5.6v on the power rail and proceed with snooping out the probable bus layout and seeing if I can cobble up something to drive it, or if the next thing I have to do before anything else is take the power supply apart and try to see if I can adjust the regulator somehow.

Really a shame the original CPU bits are gone.

 

[ic2024]

Could you elaborate on what "to my home computer, now to San Jose!" means? (ed. I found your Oct 24 post since writing this)


Do you have any pictures of the OZRC system?

Before Harlan died, I verified from him the two planned PDP's were PDP-2 == 18 and the PDP-3 == 36 bits

Where at MIT was it before going to OZRC?

Here's photo evidence of the PDP-3 "Casino" while at OZRC. As a student, one of three working on it, I was able to get some CPU opcodes working, and wrote some basic graphical pattern generation on the large radar-like vector display, with optical light-pen. I can confirm it had the Fabritek core memory (never got as far as getting that part operational), paper tape and flexowriters, and a couple 1-inch mag tape drives.
The germanium transistor cards were temperature sensitive, so debugging involved a can of cold spray in one hand, a hair-dryer in the other. The power rails had variacs to tweek the voltage, the main logic voltage was negative. The large front panel displayed all the registers using the triode VFDs mentioned previously. Sorry I don't know any details of its time at MIT.

 

[ic2024]

Yesterday I was finally able to spend some time on the system, first time in a couple weeks. I found there was some shipping damage to the power supply mounting that was allowing the PSU to shift around, which I’m glad I found before I applied AC to it because it was putting some bare wires to the transformer *very* close to the case metal. Luckily I was able to find the screw/stand-off that had popped out and get it reseated… although I am still kind of thinking I might want to add some electrical insulation to the bottom of the supply to make sure I don’t accidentally sneak a finger in there while it’s hot.

Another thing that came loose was the 40v line to the VFD panel. (Stripped end of the wire broke off in the screw terminal.) With that repaired I can verify that all the lights in the panel will illuminate if it’s unplugged from the interface card… which honestly baffles me slightly because the three wires going from the VFD to the PSU are terminating on +5v, +40v, and 1.4v AC. Where is it getting a ground, through the AC line?

(Picking this up some hours later.) I’ve been fooling around some more with things. Putting a VM on the power supply outputs showed the following:

+40v: around 56 volts
+5v: 6.1v (that’s a bit high for comfort…)
+12/-12v: close enough
1.4v AC: around 1.4v AC

Looking at the VFD board it does seem to have ground (and +5v) connected straight across from the bus (although I haven’t yet been able to trace out where they actually terminate), so I reseated the board in the chassis and tested things again. This time +5v is reading closer to 5.6v, everything else is about the same, which seems a little spooky. With the VFD board connected to a powered up IF card but nothing else on the bus most of the lights are on, but a group of seven are off, along with a couple singletons.

I’m a little worried more of the lights aren’t off with the interface powered on, since my gut feeling would be that the inputs of the 7417s would float high? (And thus pull down the line going to the VFD on the output.)

I need to take a break from this for a while, but I took the VFD display out of the enclosure and will beep out all the connections to map out which lights are linked to which bus pin. The display has 32 lights on each row, ten groups of three and a singleton on either end; this means about two thirds of the bottom row are undriven and thus on constantly. (Wondering if I have any open collectors in my chip stash, maybe I can add drivers to them and plug it straight into another board?) Have some pictures on my phone of everything, I need to read that VFD datasheet and make sure 56 volts isn’t going to cook them.

Analog and PSUs are not my strong point, so the 5.6v on the 5v rail has me worried. I’m going to put the scope on it as well, since the hertz counter on my VMM is telling me it can see 60hz hum on it…

Thanks very much for your investigation!
I'm concerned about the 1.4v heater level. The VFD specs say should be 1.0v, max 1.1v. I also see from photos the heater is visible when the grid is off. Is this normal? Maybe we should not be able to see the heater. Careful not to cook these rare tubes. The grid cutoff is normally -5v, so I suspect the cathode is "floating" up at +5v rather than ground, so the driver card OCs can pull them down to logic ground and the VFDs will see it as negative relative to the +5v, else the driver board would need additional pnp transistors to pull lower to actual -5v. Just an idea....

 

[Eudimorphodon]

I'm concerned about the 1.4v heater level. The VFD specs say should be 1.0v, max

The output from the PSU is labeled 1.4v, so it’s always been that level? I didn’t see anything that looked like a power resistor on the PCB holding the tubes.

I don’t think I’ll be able fix that without either replacing the transformer in the PSU, adding a resistor big enough to handle the collective heater current, or seeing if I can find a beefy enough linear or switching regulator that can do 1v and changing it to DC.

Edit: running the numbers using the usual resistance dropper formula to get from 1.4v to 1v results in fairly absurdly tiny resistor values. Maybe I need to measure across the actual heater connectors on one of the tubes and see if the resistance of the board is knocking it down to close to 1v…

 

[Eudimorphodon]

I'm concerned about the 1.4v heater level. The VFD specs say should be 1.0v, max 1.1v. I also see from photos the heater is visible when the grid is off. Is this normal?

I dug some more into the situation with the heater voltage. Running the numbers through the usual calculation suggested that when you're talking about the difference between 1.4v and 1.0v the amount of resistance that would make significant changes to the final voltage was *really small*, and I confirmed that by powering the VFD up again and measuring the voltage at the PSU output and where it terminated on the VFD board; I think that's less than three feet of wire, but going through that dropped the voltage from 1.45-ish volts to 1.29-ish.

Based on the datasheet I ballparked that for the number of tubes in the display I'd need about a watt at... basically zero ohms. I had some 1 ohm resistors that are only rated for a quarter watt, so after turning one in to charcoal by trying it by itself (the display didn't light up, either), I went for an aggregate 1/4 ohm x 1 watt resistor:


With the VFD plugged in this drops the voltage at the PSU end to about .95v:



The termination on the VFD panel now reads about .8v, which is less than the datasheet says they want....



but the panel lights up fine; I don't think it's any dimmer than when running at the 1.3v: (Which if I understand how it works, which I'm a little fuzzy on... tubes were a *little* before my time, mostly makes sense.)



When running on the dropped voltage the heater is *barely* warm; I had to darken the room to see any glow at all (you can just see a little bit of ruddy red in the middle):


(To make that tube go out I dropped it to ground through a 500 ohm resistor; chosen because the control board has 430 ohm resistors on the open collector outputs of the 7417. But the tubes are *very* sensitive; they'll go out if I touch the terminal with my finger as long as my other hand is touching ground, and even being close to the terminal can make them flicker.)

To compare to before I reconnected the panel directly to the 1.4v feed and took another picture of the same tube; the heaters are definitely visible on the higher voltage; in a dark room you can see the redness pretty clearly on even the lit ones.



So... I guess at this point even though we're running a little below the datasheet's heater voltage now it looks like they still function fine. Does that help allay your concerns about the tube life? I suppose I could keep fiddling and see just how low it can go and still light up; I think based on the fact I didn't see *anything* before the resistor cooked with 1 ohm I'm going to guess I probably can't go too much lower.

 

[ic2024]

Great investigation! re: "The output from the PSU is labelled 1.4v, so it’s always been that level? " - yes, most likely my amateurish design hack at the time, I'm fairly certain I did not have the tube specs as reference.
I'm happy with your lower 0.8v - should avoid the risk of early burnout.
So the grid swing is between +5v and ground?

 

[daver2]

According to the datasheet, the grid voltage should be 0V for maximum brightness and a negative voltage to switch off the illumination.

The grid voltage being relative to the filament/cathode.

According to the data sheet, the grid voltage maximum limit is 0V.

Therefore, there must be some sort of negative voltage present somewhere - or some means of offsetting the filament voltage so as to generate a relative negative voltage on the grid to turn it off.

EDIT: See: https://www.radiomuseum.org/tubes/tube_6977.html. The supply for the grid is definitely shown as negative, and the text clearly indicates that a negative voltage on the grid will switch OFF the tube brightness.

The filament voltage can be ac or dc. If dc is used, the brightness should change across the length of the filament. The reason being that the brightness is a function of the voltage between the grid and filament (given a fixed anode voltage).

This is one reason why a centre-tapped dc or ac supply is sometimes used for the filament supply. This limits the maximum voltage difference between the grid and filament.

Dave

 

[Eudimorphodon]

According to the datasheet, the grid voltage should be 0V for maximum brightness and a negative voltage to switch off the illumination.

And that is where I’m still puzzled with this. The *only* voltages supplied to the board are the 50v-ish anode, the now sub-1v AC heater, and a +5v line. There isn’t even a wire for ground to the PSU or chassis. My best guess what’s going on is this thing is wired so it’s biased so 5v is “0” and thus pulling the control terminal to ground is effectively giving “negative” voltage by comparison to its “local “ground state?

(I took some readings with the VM and I saw I could see the heater’s AC voltage if I put a probe on that and used *either* the chassis ground *or* the 5v line for the “ground” probe, so… again, it feels like it might be “riding” the heater voltage on top of 5v?)

 

[daver2]

That makes sense.

If the filament has been lifted up to (say) +5V, then a logic HIGH (+5V) would mean that the effective grid to heater/filament/cathode voltage is 0V - and the tube will be illuminated.

Conversely, a logic LOW (0V) would mean that the grid is now at -5V relative to the heater/filament/cathode - and the tube should be extinguished.

Dave

 

[Eudimorphodon]

If the filament has been lifted up to (say) +5V, then a logic HIGH (+5V) would mean that the effective grid to heater/filament/cathode voltage is 0V - and the tube will be illuminated.

Thinking about this a while I kind of got a little spooked, because an implication of this would appear to be that if the AC line were accidentally grounded to the chassis while power was applied you could end up ramming +5vdc back through the heaters. Checking the resistance between the AC and 5V line appears to confirm the rationality of that worry, so I went ahead and shrink-wrapped my resistor bodge to make that a little less likely:



I think for now I'm happy to move on to other hacking, although some more poking did raise another question:

I'm happy with your lower 0.8v - should avoid the risk of early burnout.

Would insufficient heater voltage affect the frequency response of the tubes? I started working my way through the bank to see which tube corresponds to which finger on the edge connector, and I've noticed that *some* of the tubes take a significant amount of time to glow at full brightness again after they've been pulled down. (By "significant" I mean "noticeable", like, I dunno, a tenth of a second?)

It's possible I might need to gang up a couple more resistors to get the voltage back up another tenth or two if doing so fixes this slow response, although I guess it's not really critical.