• Please review our updated Terms and Rules here

CBM PET 4032 - "Garbled screen, OK inside"

I managed to get a PET cassette deck when I bought Rusty. At least I assume it's PET compatible! 😅 Cost me the grand sum of £3 and I haven't even looked at it yet so I suspect it may need a new belt.
 
@daver2 I added the 6522 (replacement) and you were spot on. The distortion has gone.

Then I added the second 6520 (actually a 6821) and it still boots fine. Do you have a link to a 488 basic program? I found a doc on primrose.net that looks useful but not had a chance to dig into it yet.
 
Everything now works, except the iee488 interface. The test program is reporting bits 0 to 3 and bit 7 as faulty. The other functions appear OK. IMG_20250827_185227109_HDR.jpg

If I've read the schematics correctly that means there are two of the three MC3446 chips in trouble in some way. And although they seem to be available they aren't cheap..

And it means that SD2PET and similar obviously wouldn't work. 🙁

On the plus side, the £3 cassette deck works flawlessly now I fitted a new belt into it. 🙂
 
I would double-check the program you typed in first.

We can test things out with the PET and a multimeter to find out what is actually wrong.

Dave
 
Also maybe modify the program to report more clearly on if the detected faulty bits are stuck high or low, and use a volt meter (or oscilloscope if that would be easier for you) to check what actually happens on the port, and on the pins on the I/O chips. I'm not sure how the 6520 chips react to an external short circuit, but the 6522 chips actually show what happens with a signal (shorted to ground or +5V) if read even when set as an output port. Thus if the 6520 acts like a 6522 then software can detect shorts on output pins.

Schematics:
https://www.zimmers.net/anonftp/pub/cbm/schematics/computers/pet/2001N/320349-2.gif
https://www.zimmers.net/anonftp/pub/cbm/schematics/computers/pet/2001N/320349-3.gif
A first simple step would be to just swap the two 6520 chips. If the keyboard starts acting up, then something is broken with a 6520 chip.
It could very well be the case that you have two bad 3446 IEEE driver chips.
With all four signals detected as bad on one of the 3446 IEEE driver chips it kind of seems like a problem with that socket, if it's in a socket. Like bad contact on +5V or GND making all signals fail the test. Or if it's not in a socket check for bad solder joints and/or PCB damage.

Sorry that I'm a bit of a goldfish here, but did you replace the socket for the "non-keyboard" 6520 (C6) or at least clean that socket?

Btw one of the first random google results say that the MC3446 costs £12 each! At that price it kind of seems like it would be worth to cobble together some replacement. A problem is that they are rated at 48mA nominal output current at a logic low level, which is quite a bit more than most 74xx chips. You could probably replace the driver/transmitter part of a MC3446 with one 74xx inverting buffer that has a TTL compatible input, like say a 74ACT04, and then use it's output to drive a ULN2003/2803. For the input the only requirement is that it won't load the bus if power if turned off, I.E. no CMOS chip with built in ESD protection diodes.
https://www.littlediode.com/components/MC3446.html
 
Uh Oh, don't even dream of buying vintage chips where the seller does not post an exact picture of what you are actually buying.

I have spoken about this many times, yet the punters still keep falling for it.

What can I do, leave a trail of bread crumbs to lead people to genuine vintage IC's ?

Well I guess I might have to.

I cannot now because my wife says my Lamb Chops are ready (which every traffic officer knows is the only legitimate excuse for speeding) but later I will send a link.

Get these and you will have no issues, click on where it says pardon our interruption:

www.ebay.com

MC3446J QUADRUPLE BUS TRANSCEIVER IC CERAMIC Texas Instruments | eBay

MC3446J QUADRUPLE BUS TRANSCEIVER IC CERAMIC Texas Instruments
www.ebay.com
 
Last edited:
.....The thing is that when sellers don't show the actual part they often have a different M/O. They don't have the actual part in stock. When you order it, they then seek to acquire it and on sell it. The parts can be re-labelled and or refurbished. If found about about this when I inquired about a late arrival of a part, and they said it was still in the "production department" It was a re-labelled unsuitable device.

I have noticed that the MC3446 IC's are starting to get rarer, especially definite new old stocks. The place I got original NOS parts from in Europe have sold out. This means of course the prices will be going up. If you can get ceramic body ones, like the ones in the link I suggested these are also less likely to be fakes, but they are costly. Still its better to pay more for a perfect part than less for a not so good one.
 
MiaM said:
Also maybe modify the program to report more clearly on if the detected faulty bits are stuck high or low, and use a volt meter (or oscilloscope if that would be easier for you) to check what actually happens on the port, and on the pins on the I/O chips. I'm not sure how the 6520 chips react to an external short circuit, but the 6522 chips actually show what happens with a signal (shorted to ground or +5V) if read even when set as an output port. Thus if the 6520 acts like a 6522 then software can detect shorts on output pins.
Click to expand...
So the PET (Rusty) has two new 6520's (or equivalent), a new 6522 and a new 6502. With the motherboard I have, they are all towards the back and slightly exposed to whatever environment he was stored in. So I'm not surprised there's issues with the 3446's. The three of them were not in sockets either, but they are now. :)

Interestingly, if I swap them around, the faults move. So instead of bits 0 to 3 and bit 7 being bad, I get bits 0 to 3 bad and ATN bad; or DAV, NFRD, NDAC and ATN all bad with bit 3 bad and so on. From that, my brain tells me that there are two dodgy 3446s.. It's payday so I've ordered a couple from Cricklewood. We'll see how they turn out... :unsure::)
 
It is easier to test if they are on the data bus rather than the control lines...

In the 'book' with the test program, you can find the POKE to write data to the IEEE488 data bus and the corresponding PEEK to read data from the IEEE488 data bus.

Whatever value you POKE you should also PEEK.

You can follow the data bus path from the IEEE488 6520 to the associated MC3446 input to the IEEE488 bus and the read back from the MC3446 back to the 6520.

Whenever you POKE a login '0' bit (in the byte) you should observe a 0V level at the in, bus, and read back at the associated MC3446.

Conversely, if you poke a logic '1', you should read from 4V to 5V at the corresponding pins of the MC3446.

Is this an adequate explanation or gibberish? If gibberish, I can post a concrete example for one of the data bits.

Dave
 
To avoid the gibberish...

Data bit 0 of the POKE and PEEK takes the following path:

From 6520 UC6 pin 10 (signal DO-1) into MC3446 UA8 pin 3.

The bus pin is UA8 pin 2 (signal DIO-1).

The read back is from UA8 pin 1 (signal DI-1) back to 6520 UC6 pin 2.

If you POKE a byte with bit 0 = '0' then you should read 0V (ish) at all of the points above; and 5V (ish) if you POKE a byte with bit 0 = '1'.

This will test the 6520 and the MC3446 out.

You don't need (or want) anything connected to the PET at the time.

I am sure you can work out the other data bits from the schematic yourself?

Dave
 
Crashedfiesta said:
So the PET (Rusty) has two new 6520's (or equivalent), a new 6522 and a new 6502. With the motherboard I have, they are all towards the back and slightly exposed to whatever environment he was stored in. So I'm not surprised there's issues with the 3446's. The three of them were not in sockets either, but they are now. :)

Interestingly, if I swap them around, the faults move. So instead of bits 0 to 3 and bit 7 being bad, I get bits 0 to 3 bad and ATN bad; or DAV, NFRD, NDAC and ATN all bad with bit 3 bad and so on. From that, my brain tells me that there are two dodgy 3446s.. It's payday so I've ordered a couple from Cricklewood. We'll see how they turn out... :unsure::)
Click to expand...
Well, then it's more or less obvious that you need two chips. For the one with only one bit bad, maybe mark it that the high bit is bad. In the future you or someone else could cut the pins for that bit, and combine with another chip where that bit works but other bits are broken.

For £7.80 it's probably not worth the effort to create some replacement for just two of them. Longer term it seems like a good idea to create small PCBs that fit in the sockets and uses modern chips. I don't know about taxes, tolls and whatnot for imports from China to the UK (for EU IIRC it's just VAT that the exporter has to pay), but I would think that if you need more than just a few chips it would be worth the effort to create and use some modern replacement. (Or at least it would be worth the effort as long as it's a hobby and not a business you intend to make a profit off (as adding the cost of designing a circuit, all handling and such, quickly makes things more expensive than using the somewhat expensive old chips).
 
There are actually a lot of good & genuine chips coming out of China for reasonable prices, some are not genuine, you just have to know how to pick them.

The attached photo is a good case in point. It is actually a photo of a genuine Motorola IC.

But the thing is, if you look more than one seller uses the same photo.They cannot all have the same part unless they are in a parallel universe.

So the thing is, you just have to ask the particular seller to send you a photo of the exact parts they will send you.

Why is the part in the horribly low res photo photo attached a genuine Motorola MC3446 ?... you just have to know what to look for...

The interesting thing about Motorola IC's is that they had fairly unique epoxy packages, with rounded edges that were different to TI, National, RCA, Signetics and other manufacturers. They also had a characteristic relatively large inverted hemispherical shaped package indentation or Dot to identify pin 1 , which had a shiny concave surface, even if the remainder of the package was flat or matte surface.

So it is fairly easy to spot a genuine Motorola package. And they made a lot of parts in Korea in the '80's. So I have no doubt at all, that the IC shown in the attached photo is "The Real McCoy" but you just have to make sure that is what the seller sends you, or you might get something completely different.

But as I mentioned, if you get ceramic package parts, there is a much lower chance of issues.
 

Attachments

  • Rea.jpg
    Rea.jpg
    41.5 KB · Views: 4
fachat said:
If all sources dry out, I've had success in replacing the MC3446 with 74ls641-1 and pull-up/pull-down resistors. I use them in my new PET rebuilds.
Note the -1, this denotes the extra mA they can draw as required for IEEE488

André
Click to expand...
Interesting.... Thank you for that, I shall take a closer look at those. 🫡
 
fachat said:
If all sources dry out, I've had success in replacing the MC3446 with 74ls641-1 and pull-up/pull-down resistors. I use them in my new PET rebuilds.
Note the -1, this denotes the extra mA they can draw as required for IEEE488

André
Click to expand...
Can you detail how you can drop those into a genuine vintage Pet computer as a direct replacement ? I did not know that was possible because the MC3446, it quite unique part to interface with the GPIB. Have you got some sort of adapter pcb., how did you do it ?
 
It looks like the 74LS641-1 is a 20 pin DIP. It also comes in different surface mount packages. So it will need an adapter. It has eight circuits rather than four so that is nice. But there is no buffer between the receiver input and the buss. But with pull up and pull down resistors as André stated, it should work.
 
You must log in or register to reply here.
Back
Top