• Please review our updated Terms and Rules here

PET 3016 memory upgrade

JonB

JonB

Veteran Member
Joined
Jan 26, 2014
Messages
1,652
Location
South Herefordshire, UK
Hello

I have a 3016 PET with both banks fully populated with 4108 DRAMs. I want to upgrade it to 32k using 4116 DRAMs.

I think it's possible but I'd have to alter the 12 pin "DIP SHUNT" device (which looks like a 6 way dip switch but is in fact a group of jumpers with some connections broken and some made. I have currently got !A B !C D E !F (where ! means boolean NOT or open link) and this is consistent with the schematic here for a 16k PET:

http://www.zimmers.net/anonftp/pub/cbm/schematics/computers/pet/2001N/320351-6.gif

Possibly, then, and according to the schematic, I could swap the 8kbit DRAMS for 16kbit and alter the shunt accordingly, but I would need to find out for sure. It looks like I need !A !B C D !E !F set to do this.

Anyone know?
 
JonB said:
Hello

I have a 3016 PET with both banks fully populated with 4108 DRAMs. I want to upgrade it to 32k using 4116 DRAMs.

http://www.zimmers.net/anonftp/pub/cbm/schematics/computers/pet/2001N/320351-6.gif

Possibly, then, and according to the schematic, I could swap the 8kbit DRAMS for 16kbit and alter the shunt accordingly, but I would need to find out for sure. It looks like I need !A !B C D !E !F set to do this.
Click to expand...

Look at the Top Assembly drawing you posted at the very right side. You will see jumpers needed in both 6 and 10 pin DIP Shunts for 32K. For location of shunts see Detail A and D on top assembly.
 
In other words... Yes, I can!

Thanks. Those pictures are much clearer than the ones on the schematic. Now all I need are 16x4116 DRAMs.... Hmm, anyone got some spare?

:D
 
You may be better off with a PETvet than buying working RAM chips, but weigh this out for yourself. I bought a PETvet thinking I would need it, but by the time it arrived, I found I didn't need it.
 
JonB said:
In other words... Yes, I can!

Thanks. Those pictures are much clearer than the ones on the schematic. Now all I need are 16x4116 DRAMs.... Hmm, anyone got some spare?

:D
Click to expand...

eBay, of course! Look for 4116 ram or 8116 ram (my 3032 is using 8116 ram, they should be 100% compatible), you will spend about 32 euros (1 euro per KB, not a bad price in the late 1970s :) ) plus shipping...
 
JonB said:
Ooh, another question. Is it possible to expand the PET above 32k, and if so what's the easiest way to do it?
Click to expand...

Not really in a useful way. The 6502 has only 64K of address space (16 address lines). In the PET memory map, the lower 32K is reserved for RAM with the upper 32K for video RAM, ROM and I/O.
 
JonB said:
Ooh, another question. Is it possible to expand the PET above 32k, and if so what's the easiest way to do it?
Click to expand...

as far as I know, you can expand it above 32k using a memory expansion (huge) card only, reaching the incredible amount (for 1979!) of 64 KB...

out of curiosity: why do you need *so much* ram??? :D
 
I don't - it's just curiosity. I really need to get to 32K and that's fine for my purposes (completion of a program I started when I was 17, that ran on a 40 column 4032 PET).

Did you get my PM about the PETdisk?
 
Well, that's fine - no apps are going to use more than 32k anyway.

I ordered 18 4116 chips (a lot of 10 and another of 8) and a stack of sockets, total cost < £15. Looking forward to a soldering session!
 
Oh dear...

Oh dear...

Well, I removed all the 4108 chips but had to use a heat gun. Also, used the heat gun and a vacuum cleaner to suck out all the solder from the holes. For some reason, my solder sucker was not pulling the stuff from the ground / Vcc holes (I expect this is due to the iron not melting the solder properly).

Anyway, I soldered two banks of sockets in after cleaning the pads and refitted the 4108 RAM to test it. Nothing, a dead PET. This, you might say, is the price one pays for impatience!

Maybe the 4108 chips got fried. So I reconfigured the board for 16k of 4116 RAM (double checked and tested with a meter) and plugged my 8 chips in. Hmm, all I see is a reverse character (a question mark) on the right hand side of the screen about half way down. It's not booting. If I turn it off then on again, I get a screen of garbage characters before it clears and I'm back to the reverse question mark.

Does this mean anything, or is it just telling me the RAM chips are borked? These chips are second hand, cheap, but look like they came off a board they were soldered into. there's some solder left on the legs. Maybe they are toast. I've checked the board and my soldering several times and I cannot see anything wrong. It is possible the 74LS04 inverter to the right of Bank 1 got toasted. I did check the crystal to the right of Bank 2 and it is oscillating at 16mhz per spec. I also tried piggy backing a 74LS04 onto the one I just mentioned - it made no difference. It was a guess anyway.

I still have a batch of 10 chips to come. I can test with those too. Failing that, I can scavenge some off a ZX Spectrum I have here, and I can fit sockets to the Spectrum and use it to test the RAM chips I received.

But for the time being, I'm asking for some advice. Does the single inverted character on the screen mean anything? What should I check next?
 
Last edited:
JonB said:
If I turn it off then on again, I get a screen of garbage characters before it clears and I'm back to the reverse question mark.
Click to expand...

If the PET was working with 8K originally, then I assume there may be a short (solder splash) in the data or control lines (read/write) with the new sockets. Or perhaps the heat gun melted one of the traces and there is an open circuit.

The clearing of the garbage characters indicates the processor is running long enough to clear video RAM before it goes south due to perhaps bad zero page RAM cause by a circuit trace problem. A heat gun used in excess can easily overheat a PCB and a vacuum cleaner can create thousands of volts of static electricity which can lead to ESD. This is probably just a temporary setback but now you need to some continuity testing unless you have a scope handy.
 
Ok, out with the multimeter again... I have checked the board under a magnifying inspection lamp very carefully for solder splashes and removed all I could see. I also performed continuity checks for all RAM chip sockets to make sure there are no bridges between the pins. The PCB looks ok. There's no visible sign of singes, burns or other discolouration.

Is it likely the 74LS04 I mentioned before is toast? Are they particularly sensitive to heat?
 
JonB said:
Ok, out with the multimeter again... I have checked the board under a magnifying inspection lamp very carefully for solder splashes and removed all I could see. I also performed continuity checks for all RAM chip sockets to make sure there are no bridges between the pins. The PCB looks ok. There's no visible sign of singes, burns or other discolouration.

Is it likely the 74LS04 I mentioned before is toast? Are they particularly sensitive to heat?
Click to expand...

Are you sure you didn't break a trace under a ram chip? Because a previous owner did it in my 3032 and it gave me a lot of troubles....
Did you removed and checked one chip at time? I mean, remove the ram chip, put a socket, put again the same ram chip in place and switch on the computer... If you didn't it, it will take a lot of time to find the problem, admitting it's some kind of broken trace....

aaah, a suggest about 30xx pcb: I've noticed the traces that run between two IC pins aren't masked. So you can easily get a short between a pin and its opposite.
In other words, if we take as example a 14 pins IC, you should check with your multimeter not only continuity between pin 1 and 2, 2 and 3, etc. but also between pin 1 and pin 14, 1 and 13, 2 and 14, 2 and 13, 2 and 12, and so on (admitting of course -in example- the trace from pin 1 passes between pin 14 and pin 13).
You can hardly see the short because it will be masked by the socket.
I hope my explanation was good enough to be understandable!!
 
Tonight I tested with a meter and there are no shorts in the places you suggested.

I performed a RAM rest, using a Spectrum. I removed one of the Spectrum's 4116 chips and fitted a socket. Then I swapped each chip into the Spectrum and tested boot up (the Spectrum uses the 4116s for screen RAM amongst other things - the bank is at the bottom of memory). What I found was that NONE of the chips I had bought were faulty, apart from the one I mentioned before which got very hot. So I have a single bank of known working RAM in the PET, and still this single inverse question mark.

I also double checked the RAM configuration jumpers . They are correct and tested for continuity.

I have gone over the board with a magnifying inspection lamp again. Cannot see any solder splashes, broken traces or other anomalies. I have tested that each RAM chip is getting the correct voltage (they are). I was able to test most of the traces under Bank 1 because they end in vias; they are OK, no breaks. The only thing I see is a missing decoupling capacitor which must have fallen out when I desoldered the chips, but I do not think that is critical. It is labelled C37 and it decouples voltage to U16 (one of the 4116 chips in the populated Bank 1).

So... I've run out of ideas.... is the inverse question mark itself significant? Probably not, but it always comes up (without the RAM installed the screen is blank).

Hmmm, frustrating....
 
Last edited:
You must log in or register to reply here.
Back
Top