Not quite... The end of the ROM based PLA for the 64

[eslapion]

eslapion do you have any 90b6's left? i would like to purchase one to use for comparison with a PLA the same as the rest.

I have 2 left and I keep them preciously. On the other hand, I have a pile of M27C256B-90B6 and you can combine 2 of them in piggyback along with a 74HCT14 inverter to get the same behavior as a single M27C512-90B6.

The ugly but perfectly working contraption I called 'sPLAtt' is installed in one of my old 250407.

[levoman]

two resistors and a tansistor will make an inverter just need the eproms to go with it , how much for six and uk postage? PM might be better.

[eslapion]

two resistors and a tansistor will make an inverter just need the eproms to go with it , how much for six and uk postage? PM might be better.

You absolutely can't check how fast your inverter will act when you use 2 resistors and a transistor. Both EPROMs may end up signaling at the same time and this can damage/destroy them. Two of the spare inverters in the HCT14 are used to increase the latency on CASRAM giving a response speed that's nearly identical to early 82S100 PLAs.

I'll sell you my sPLAtt, assembled, tested and working for 10 Euros. Shipping to the UK is 8 Euros. I have a single spare and no more.

[levoman]

Don't worry , i think i will leave it . I was just giving you a chance to sell a few.

On the other hand, I have a pile of M27C256B-90B6

They can be used for other things as well .

Thank you for the offer , but i will decline.

[eslapion]

Don't worry , i think i will leave it . I was just giving you a chance to sell a few.

If I sell you the blank EPROMs, I have no idea how you're going to use them and the exercice could turn into hundreds of exchanges about doing this or that to get it to work. I'd rather you have a working solution in your hand from the start.

As for having a chance to sell a few, I sell dozens of PLAnktons every months.

You first asked for an M27C512-90B6 and I won't part my last 2 units. The next best thing is an operational sPLAtt. If you absolutely want the M27C512-90B6 then I think Ray Carlsen has a few left in stock. All those found on eBay are fakes/rebrands.

[levoman]

I will have to get hold of him. Just wanted to include it as it is the one that started the thread.

[eslapion]

I will have to get hold of him. Just wanted to include it as it is the one that started the thread.

The point of the thread is that in 2008, I discovered the M27C512-90B6 as one of the very few (E)PROMs which could safely be used that could be used as PLA and a group of people did everything they could on Lemon64 to discredit this working solution and they even managed to get me banned from that forum.

In 2011, this '1$ PLA solution', which incidentally can also work with the C16 and Plus/4 if programmed accordingly, was discontinued by ST.

Bad mouths deprived the Commodore community of a perfectly reliable and incredibly inexpensive solution to a nagging problem of Commodore 8 bit computers. They even punished me for persisting in telling the truth.

Now, the M27C512-90B6 is certainly not unique and the Super Zaxxon cart as well as the 74LS279 test circuits are methods for finding other memory chips (PROMs, EPROMs, Flash, etc...) which can act as reliable PLAs. Apparently, there is this 'Axel81' guy on Amibay who found a trustworthy alternative but he won't say exactly what the chip is. He grinds the surface before selling.

Presently, if I disregard the zoo of antiquated, fakes and dirt cheap junk EPROMs offered on eBay and limit myself to reputable distributors such as Mouser, Avnet and Digikey, there is a good list of 64k x 8 non-volatile memory ICs that will operate at 5V.

These include a bunch of different speeds and package AT27C512R from Microchip, the AT27LV512A also from Microchip and the AT49F512 Flash ROM. I don't have time to test them but maybe you guys do.

[levoman]

Exactly the point i am getting to. First to test roms against an original PLA and then to test against a relible replacement like the 90B6. Hopefully we will find a few that people can use and weed out those that cannot be used while showing the reasons why!

[eslapion]

First to test roms against an original PLA ...

Would you care to explain what you have in mind ?

When I created PLAnkton, I examined the behavior of early C64 PLAs that were provided with boards 326298 since those truly work with everything. I didn't 'test them against' anything. I adjusted the latency and output voltage to resemble those of this early PLA as closely as the hardware allowed.

Eventually what appeared on the logic analyzer captures was identical at 500MS/s.

With memory chips, you can't adjust much at all. You might be able to add filters if there are glitches. You can add buffers on paired inverters to the CASRAM output if the latency is too small.

... and then to test against a relible replacement like the 90B6.

Why would you want to do that ?

That's absolutely not what you're looking for.

When using memory chips to replace a slow logic chip, the vast majority of them will cause undesirable transient output transitions and they look like that (see outputs from AT27C512R): http://www.melon64.com/forum/viewtopic.php?f=12&t=8691

If there are glitches and they are brief enough, you could develop some sort of filter. If the latency is too small, you can add buffers/inverters; if it is too long then there's nothing to be done.

You can just as well test against a PLAnkton or any properly working Commodore genuine 906114-01 PLA from whatever generation.

[levoman]

comparing the outupts to a genuine pla , timings , glitches , everything i can to see the difference between the good and the bad.

[eslapion]

comparing the outupts to a genuine pla , timings , glitches , everything i can to see the difference between the good and the bad.

The M27C512-90B6 is 'good enough' but it shouldn't be considered a reference.

It works well on boards 250425, 250466 and later revisions of the 250407 because they have the R42/C204 CASRAM line filter but their response latency is somewhat too short so they won't directly work on the early 326298, ealier revisions of the 250407 and the European KU-14194HB.

The correct reference remains genuine Commodore PLAs. It's the only 'good' you should ever consider. Perhaps PLAnktons too since they are incredibly close.

[banman]

Hi eslapion and levoman,


I like the gold fingers on your ZaxxDual cartridge levoman.


I found this interesting article that GyroGearloose was discussing on another forum....

https://www.computerhistory.org/silicon ... ntroduced/

I found it interesting....



Would you be kind and explain to me what response latency is in regards to a PLA? Am I close when I say that it's the time taken for an signal on an input line to reflect itself on an output line?

[levoman]

Got some 90b6's coming from Ray Carlsen.

This is mainly to check what it does right that others do wrong.

[eslapion]

Got some 90b6's coming from Ray Carlsen.

This is mainly to check what it does right that others do wrong.

That's something I can easily explain to you.

All memory chips will generate glitches when used as a logic IC. The PLA of the C64 is a 16 inputs, 8 outputs combinational logic device whose function is to generate exceptions in the memory access structure - the CPU can access 64k but the RAM, ROM and IO addresses total more than 64k so where do you want the ROMs and IO to be 'visible' ? The PLA decides that.

See: https://en.wikipedia.org/wiki/Combinational_logic

The problem when using memory chips as logic chips is all logic ICs are designed to have their outputs signal continuously when changes occur on the inputs. Not so with memory ICs which normally require the /CS or /OE signal to be disabled when changes occur. A memory chip will normally have a short duration undesirable output if changes occur on the inputs without stopping the /CS or /OE. This undesirable output is called a static hazard. https://en.wikipedia.org/wiki/Hazard_(logic)#Static

The M27C512-90B6 drives very little current (sink/source) on its 8 outputs when signaling so when a change occurs on the inputs, the brief period when the IC effectively outputs garbage is filtered out by the impedance load of the type of lines used in the C64. The combination of low drive current and relatively high impedance load acts as an RC filter. Back in 2008, I referred to this effect as 'lazy output' and since this was not an exact official term, Groepaz jumped on the opportunity to say I needed psychiatric assistance.

Now if you look at the scope captures I posted on Melon64, you'll see the AT27C512R-45PU does generate these static hazards but they are incredibly short in duration. In theory they can be filtered out using an RC filter on all 8 outputs and another logic IC with hysteresis such as a Schmitt trigger. In practice, this makes the whole circuit much more complex than a real logic IC such as the one PLAnkton has.

[banman]

The PLA of the C64 is a 16 inputs, 8 outputs combinational logic device whose function is to generate exceptions in the memory access structure - the CPU can access 64k but the RAM, ROM and IO addresses total more than 64k so where do you want the ROMs and IO to be 'visible' ? The PLA decides that.

I very much like this explanation of the PLA's role in the C64..

when a change occurs on the inputs, the brief period when the IC effectively outputs garbage is filtered out by the impedance load of the type of lines used in the C64.

If I am reading this correctly, would it be fair to say that different revisions of C64 mainboard PCB's have different physical layouts which may in turn have different impedance loads?

I noticed that the dodgy Winbond W27C512 EEPROM PLA substitute in one of the three 250407 variants I tested actually seemed to work ok. However all the other boards didn't go so well.

I noticed that on the 2 Ku style mainboards I have they have added in extra components at the factory that are different from each other when viewed side by side.....
Were they trying to custom match the impedace load to PLA chip used?
Maybe the very early PLA chips exhibited some undesirable behaviour.
The PLA's currently in these KU boards are not the original ones but later 1986 dated PLA's.