Jump to content

Photo

Where Will the Chips Fall (Come From?)


31 replies to this topic

#26 Level42 OFFLINE  

Level42

    Stargunner

  • 1,903 posts
  • Location:Ridderkerk, The Netherlands

Posted Mon Jan 1, 2018 5:24 AM

I didn't check but if there are small form pure 6502s being made those four external chips required to make it a SALLY could easily be added in the smallest SMD cases they come in today and put on a PCB the size of the old DIL40.

I also haven't checked a lot about the SID replacement but this looks like a drop in replacement with DIL40 PCB size.

The true sad thing is that all of the dedicated Atari VLSI's apparently have a very good design, because the vast majority simply keeps going, way, way, waaaaayyyynbeyond their expected lifetime....I'm actually convinced most VLSIs die due to power supplies going bad and/or Commodore 64 power supplies being tried on A8's.

The Commodore designers weren't as good at designing reliable chips. History tells that they tried to get a video chip running when the 800 was already on the market for two years but they couldn't get it to work.
Their later working chips like the VIC-2 and SID run really hot and thus die a lot sooner than Atari's VLSIs.

There would be two ways to get replacement chips:
1) find the original films at one of the companies , (or deliver new ones that have been reverse engineered )that produced these chips for Atari and have them produced again. This will be very hard to get done for a good price nu it is not impossible. F.I. Certain DAC chips (Digital Analogue Converters) are still so popular that prices for them went up to about 27 dollars. The company who produced them in the 80s apparently noticed this, figured there was good money in this and restarted production. Guess at what price they sell them for (while there is zero development costs). Their bad luck is that the Chinese sell copies which work just as good and cost about 1,50'dollar.

Which would be another idea: if there are complete and 100% correct films in the hands of "us"....we would "only" need to find a Chinese factory........not impossible.....

2) FPGA. Very often I hear and read that a FPGA emulates another chip, or complete hardware. This is not true....it is equal to the original hardware. Emulation is a processor/computer mimicking another (old) computer or chip, but howevermfast, it is still,doing it step by step....an FPGA is direct, electrically exactly like the original. Anyway, it has been possible to create FPGA replacements for the A8 VLSIs for quite a long time already....but until now it was simply a lot cheaper to buy new old stock chips. I think for POKEY we are now getting the situation where the price of an FPGA replacement would be about the same as an original NOS one....so hopefully this will change the situation.

However....there are already FPGAs around (Eclaire)....one chip containing an entire A8. Imhave no experience with them yet, but I have owned two FGPA based arcade (Williams) PCBs. These still used an original separate processor....but everything else on the original PCBs (loads of standard TTL chips) was in FPGA and this worked absolutely 100% like the original boards.

So that's a great development...(and I'd love to see a portable A8 coming from this )....but I prefer to keep the original,hardware running for as long as possible, with as much as original,parts as possible because to me this is the charm of the hobby.....
You don't put a modern day engine in a 60's classic car....but if you need to use a replacement part if there simply are no original parts available anymore to keep that old engine running.....

#27 TroyQ OFFLINE  

TroyQ

    Space Invader

  • 14 posts

Posted Wed Jan 3, 2018 11:47 PM

Just read a post where a dude chased his tail trying to fix his beloved 130XE.  He found a bad Antic Chip. {snip}

 

Off topic: Not sure if I would phrase it quite like that. ;)

 

Back on topic: Agreed, there is a finite number of these chips and some day supplies will dwindle.



#28 slx OFFLINE  

slx

    Stargunner

  • 1,371 posts
  • Location:Vienna, Austria

Posted Thu Jan 4, 2018 10:22 AM

The true sad thing is that all of the dedicated Atari VLSI's apparently have a very good design, because the vast majority simply keeps going, way, way, waaaaayyyynbeyond their expected lifetime....I'm actually convinced most VLSIs die due to power supplies going bad and/or Commodore 64 power supplies being tried on A8's.

 

Doesn't sound sad to me ;) My son has been running a stock 130XE continuously for at least three years, using it as an alarm clock and only switching it off for a couple of weeks vacation ever year. That means it's sometimes running for three to four months straight and it's still working fine. It's on a modern PSU, though. I wonder if he will be able to pass on that machine to his kids one day  :-o



#29 ACML OFFLINE  

ACML

    Dragonstomper

  • 797 posts
  • Location:USA

Posted Sun Apr 8, 2018 7:58 AM

I have secured a lifetime supply of VLSI chips for the 8-bits.  Most come from salvaged 400 and 800's that have for one reason or another become too problematic to fix.  Some are just spare boards and chips off eBay.  The cheapest way I've found to procure these (ANTIC, GTIA, PIA, POKEY, SALLY) is to watch eBay and pick them up from cheap 400's or separate spare parts boards.  I just bought all five VLSI chips for $15 delivered.  In 30+ years of using the Atari 8-bits, on only one occasion did I have to replace one of these chips.  It was the GTIA on a almost new 800 in 1983.  Since them, all mine have been rock solid.  I'm only buying them because "they don't make them anymore" and when they are gone, they're gone.



#30 Spaced Cowboy OFFLINE  

Spaced Cowboy

    Chopper Commander

  • 143 posts
  • Location:San Jose, CA

Posted Sun Apr 8, 2018 9:12 AM

"2) FPGA. Very often I hear and read that a FPGA emulates another chip, or complete hardware. This is not true....it is equal to the original hardware. Emulation is a processor/computer mimicking another (old) computer or chip, "

That's not quite how I'd put it. No, it's not software emulation (which is what you're describing as "emulation"), but it is likely to be black-box emulation. Let me explain.

FPGAs are conposed of two main components (ignoring any hard macros like memory-controllers or built-in CPUs). They have lots and lots of lookup-tables, surrounded by a sea of interconnect logic. A lookup table is (these days, it used to be 4-bits wide) normally 6-bits wide so it can take any 6-bits of information and map it to logic 1 or 0.

A simple example would be an OR gate, two inputs, A and B, producing an output Q. Q is 1 if either A or B is 1, else it is 0. Now extend that to 6-bits wide, not 2 bits and you can start to see how logic can be built up. There are literally tens (hundreds, thousands) of thousands of these lookup tables in an FPGA which combine together to make the overall logic of the design.

The interconnect is often overlooked, just taken for granted, but FPGA interconnect is itself something of a miracle of engineering, it typically supports an operating frequency of 100-500 MHz, depending on the chip, has sufficient capacity that you don't normally have to consider routing issues between your lookup table elements, connects each lookup table output to a huge number of other possible inputs and is also responsible for connecting to the FPGA pins via i/o buffers.

Ok, so now we know what they are. How do you get something useful out of them ? The old way was to literally draw a circuit in terms of gates (AND, OR, XOR, NOT,...) and let the CAD software figure out how to map that onto the hardware. These days, FPGAs are too large for that to be useful, so we use HDL (hardware definition languages). I personally prefer Verilog, but in Europe VHDL is more commonly used.

An HDL is trivially comparable to a software programming language, though the mindset when writing in the language is different. In an HDL you can describe the gates in a similar way to the old schematic interface by designating inputs and outputs, and the operation that takes place. You can also write much higher-level code that describes more complex logic behaviour.

And so we get to the point of this soliloquy... using an HDL, you can take a decapped depiction of a circuit and implement every gate, and yes, this will be a faithful reimplementation of the original circuit, no emulation here.

An alternative, when you don't have access to the circuit layout, is to use the published spec of the design and re-implement that. The internals might bear no resemblance to the original chip design, but if for any given input, the output is the same as the original design, that doesn't matter.

I would hazard a guess that's overwhelming majority of FPGA designs that re-implement some other chip take this second approach, and that is generally called black-box emulation. The internals will almost certainly be different than the original, but if (in the case of the 6502 for example) every opcode has the same results as the original chip, it doesn't matter. It's still a drop-in replacement.

I would also offer a third option, (3) a new hardware run. It's almost getting viable now (I'd estimate about 2x the cost of buying old chips from Best). Once you have a verified HDL design, it's actually possible to get chips fabricated on what's called a shuttle-run, where you share the wafer with a lot of other people to spread the NRE costs. Typically it's between $1000 and $2000 for (say) 40 chips. That's because although the 6502, Antic, etc were cutting edge when they came out, they're very much old hat now, and therefore the chip runs are cheap.

I have a long-term plan to re-implement the custom chips in the XL, I have a design for the 6502 done, but it's not going to happen this year, and probably not next, either. Time is the enemy, and I already have projects stacked up :)

FWIW, my 6502 is a black-box emulation, that runs at ~100MHz on a modern FPGA :)

#31 DrVenkman OFFLINE  

DrVenkman

    River Patroller

  • 2,930 posts
  • Back off, man! I'm a scientist.
  • Location:KMBT

Posted Sun Apr 8, 2018 11:20 AM

I have a long-term plan to re-implement the custom chips in the XL, I have a design for the 6502 done, but it's not going to happen this year, and probably not next, either. Time is the enemy, and I already have projects stacked up :)
 

 

Those are ambitious plans. I'd respectfully suggest, however, that the retro community needs new POKEY clones much more than it needs a new 6502. 



#32 Spaced Cowboy OFFLINE  

Spaced Cowboy

    Chopper Commander

  • 143 posts
  • Location:San Jose, CA

Posted Sun Apr 8, 2018 12:20 PM

The 6502 is a far easier "first step" to get the project started with; it's entirely digital, whereas Pokey needs on-chip analogue components (or a DIP-style PCB with a DAC). Analogue VLSI is *not* described using HDL, it's still literally "drawn" onto the chip surface to get the desired analogue effect. The 6502 is also the central control of the system, so once it's up and running, debugging the others is a lot easier.

I'm reasonably sure the analogue output of Pokey is just a R/2R resistance ladder (the simplest form of DAC), but that means literally drawing polysilicon areas onto the chip substrate in Magic  to form resistors on-chip. There's no HDL way to describe that.

So Pokey/GTIA are fact the harder things to get done. Antic/Sally are easier since they're entirely digital. The other possible way to do it I alluded to above - get a small fast DAC, or implement the resistor ladders externally, and make the new chip + DAC/ladder be on a PCB that fits into a DIL socket with the same pinouts as Pokey/GTIA. Video DACs might be pushing it though, in terms of board space.

 

I'd be doing all of this in an FPGA to start with, which would make debugging a possibility before committing to Silicon. It's a big project, and I've got a fair amount on my plate already :)






0 user(s) are browsing this forum

0 members, 0 guests, 0 anonymous users