Signetics 2518 replacement with FPGA

11 posts / 0 new
Last post
Offline
Last seen: 1 year 8 months ago
Joined: Jul 15 2022 - 09:34
Posts: 2
Signetics 2518 replacement with FPGA

I had an idea to replace the hard-to-find expensive chips with a new design.  Like the 2518 and MK4096.  If we use a small Lattice FPGA, we could do the same things as ROMs where we put the surface mount chip between the pins on a tiny circuit board so the form factor is the same.  It's definitely more complicated, but with the scarcity and price of these old chips, it's probably worthwhile.

 

I was hoping to have one layout for all the hard-to-find chips for each package, but even among the DIP-16 chips, there's different locations for power and ground.  There might be enough room to have jumpers to select the pinout to emulate different chips.  And there's the problem of level-shifting since the FPGAs are 3.3V.  So there's some problems to solve, but I think it's technically possible for a reasonable cost. 

 

I'm thinking like $10 in parts for a chip.  The FPGA is like $2-3.  There's a dual voltager regulator for another $2.  There's level shifting which may be as simple as a resistor, or maybe another chip or two.  With the tiny size of the board, it might be a possible cost target.  If the device was generic enough, it could be use to emulate many chips which could also help to bring the cost down.  Ideally it could replicate any 7400 chip in a 16-pin package.

 

What do you guys think?

Offline
Last seen: 7 months 3 weeks ago
Joined: Mar 1 2019 - 04:38
Posts: 115
I think people have had the

I think people have had the idea but never quite got into mass production with the A1 shift register at least.

 

A device that could emulate almost any 74 or similar logic would be a fantastic thing. Especially if it was user selectable and looked similar enough not to spoil the look of a vintage PCB. It would not only keep some machines running, it would be a great set of devices for testing too.

 

Lot of work though, I imagine.

Online
Last seen: 26 min 17 sec ago
Joined: May 4 2021 - 06:35
Posts: 153
Replacements

Hi,

There are already some "modern replacements" available for:

2513:

https://p-l4b.github.io/2513/

 

2519:

https://p-l4b.github.io/2519/

 

RAM can be entirely replaced (and expanded) with a modern 62256 and a few other components, as I usually do on my projects, for example:

https://p-l4b.github.io/sdcard/

 

Anyway, it would be interesting to create replacements for the 2504 (and its equivalents/compatible), but could be not so easy.

First, those SRs have current-sink power supply, a consequence of the interfacing techniques to/from TTL logic in use in those years.

This means that you cannot have a plug-in replacement... 'cause GND connection will be missing. (Same happens for 2519 replacement, this is why PCB needs also GND from 7404 socket).

In addition to this they also have a non-overlapping dual-phase clock, also with non-TTL voltages.

 

Probably just the level-shifting circuitry would make the project more complicated (and probably more expensive) than you maybe anticipated.

 

But it might be worth a try.... good luck!

 

Enjoy :-)

Claudio - P-LAB

 

 

 

Offline
Last seen: 20 hours 34 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
About the perceived "scarcity" of Apple-1 parts ...

... which does not really exist.

 

Toksin wrote:

 

"I had an idea to replace the hard-to-find expensive chips with a new design.  Like the 2518 and MK4096.  If we use a small Lattice FPGA ..."

 

Uncle Bernie comments:

 

First, you should do your homework. There is no Signetics 2518 in the Apple-1. What you probably mean is the 2519N or 2519B 40x6 PMOS dynamic shift register in a DIL-16. They are not rare. Manoshewitz in Jerusalem sits on ~5000 of the 2519N with date codes from 1977 (and maybe other datecodes, too, but AFAIK he has no or only a few 1976 date codes as in the original Apple-1, because he would sell them for $200 each, as far as I know Mr. Avi). I found that 20-25% of them which I bought from him through other brokers are bad, possibly a reject lot, but that was OK for these ~45 year old parts as long as he sold them for below $17 a piece. Nobody can expect ICs to last forever, especially in plastic encapsulation. But now he wants $65  a piece on Ebay, plus shipping, and he does not test them, people have complained. But as you can see, they are not rare. We could increase the number of Apple-1 clones in the world by a factor of 8 at least before the 2519N run out, which would take about 70 years at the current pace of builds. I've already factored in the 20-25% bad ones.

 

People who want to be cheap and thrifty (Apple-1 builders are cheapskates, me included, otherwise we would buy an original) always have had the idea to replace elusive or expensive ICs in the Apple-1 BOM with other ICs and so there is a 6xCD4557 solution to replace the 2519. They even published a PCB layout.

 

For all other "elusive or expensive" ICs in the Apple-1 BOM there are plenty of drop-in substitutes and last time I checked, the worldwide stock of the 1404/2504, 2513, 4027 DRAM, or their replacements, is in the tens of thousands so there is no way the Apple-1 crowd could ever deplete this stock.

 

To see a typical substitute IC set, look at Ebay for my IC kits. Absolutely everything in them are drop-in substitutes (except for the TTLs and the Signetics 2519). Despite I know stocks of original Signetics 2513 and 2504/1404 in the grey packages numbering over 5000 pcs each, I do not use them because the IC brokers who actually stock them want moon prices, such like $200 for the Signetics 2513 character generator. Or $15 for one Signetics 2504/1404, seven of which are needed, maybe 10 because not all are good. And the buyers of my kits don't want to pay twice the price for the kit. They want the cheapest possible IC kit, period. So I have to use drop-in substitutes.

 

Only the Signetics 2519B in my kits is the real deal. And it even comes with the desirable 7627 date code and the horizontal print ... fits perfectly into the date code range for original Apple-1. Oh, and I don't charge $200 for it. See it as another bonus in my kits. Some owners of original Apple-1 have bought kits from me just to have spare parts.

 

And here we went full circle, back to the Signetics 2519. If you want to make yet another substitute PCB with yet another type of IC - this time a FPGA - go for it but watch the logic levels. You will learn a lot. In the end, with the FPGA tools and the PCB tools and the JLCPCB run and the DIL-16 adapter with the right diameter of the pins to not ruin the IC socket on the motherboard, you will have spent enough money and time to be able to buy a few of my kits. Which I also sell directly, outside of Ebay, just use the "send PM" button. But if you want the FPGA experience, go for it, and tell us about the progress. This is not snarky. I'm really interested in these little FPGAs, $2 to $3 you say ? These could help me with some other projects of mine, if they are easy to use. I do have some Xilinx Spartan and Altera DeO-Nano kits but found them and their software to be too complex and tedious for my purposes. Heck, you could put a complete Apple-1 into one of those if you want to waste several months of writing Verilog or VHDL. And then what ? An Apple-1 in a matchbox ? Vince Briel's Replica-1 is a fine alternative for those who don't want to spend the $666.66 or so (pun intended, but it's close) it takes to build a complete Apple-1 clone nowadays, with motherboard, ACI card, power supply, keyboard. If you want a vintage looking B&W monitor you will find it sets you back by another $100-$150. So in the larger scheme of things where do these little FPGA fit into this peculiar microcosmos ?

 

- Uncle Bernie

Offline
Last seen: 1 year 8 months ago
Joined: Jul 15 2022 - 09:34
Posts: 2
Thanks for all of your

Thanks for all of your insights.  I appreciate the feedback.  That's why I came to you guys before even starting the project - to assess how difficult it would be and how much it is really needed.  I've read all the work that's been put into getting these old chips, and it seems like quite a struggle.  I was hoping to combat the high prices the oligarchs you mention are wanting for the old stock.  Competition is good for lowering prices.  And I'm afraid the angels like you Uncle Bernie won't be around forever, and we'll be left with paying high prices for broken chips.  Plus, I'm an FPGA hammer, so everything looks like an FPGA nail to me.  I was looking for a solution to my idea of a single replacement device which can be reprogrammed to fit any of the discontinued 74xx or similar devices.  The Apple 1 sounded like a good place to start with a cursory look at the chip prices.  

 

Looking into the design further, the small ICE5LP1K-SWG36ITR50 goes for $3.23 on Digikey for single units.  It's 2.1x2.1 mm footprint which will fit between the pins of a normal DIP-16 package.  Doing a little more looking into the layout of the device though, they require via-in-pad microvias with non-conductive epoxy filling (which is expensive).  So even though it's a cheap device, making a PCB for it will not be cheap.  I haven't priced it yet, but it makes me worried about meeting a good price target.  Also, it makes it very not-hobbyist friendly.  Of course, it's just one design challenge on top of the logic levels and power requirements and fitting all of these in a tiny place.  It all seems possible, but I'm not sure if it's worth the time.

 

 

 

 

 

Offline
Last seen: 20 hours 34 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
In post #5, Toksin wrote: "..

In post #5, Toksin wrote:

 

"...they require via-in-pad microvias with non-conductive epoxy filling (which is expensive).  So even though it's a cheap device, making a PCB for it will not be cheap."

 

Uncle Bernie comments:

 

Herein lies a problem for us hobbyists. You would need to make a panel full of these PCBs and then have it professionally assembled. These type of packages are super difficult and the industry hates them (despite they must use them) because of the assembly and test costs. They must X-ray each assembled board for integrity of the solder joints on each ball. Now, because these board can have 30+ layers and are thick, they crank up the X-ray dose which more often than not shifts precision analog ICs whose functions were painstakingly trimmed to target at their manufacturer. And then the customer complains about these precision ICs being out of spec and of lousy quality. This is where the fun ends. Our "hi tech" has gotten to a point where the electronic components look like bird feed, or worse, dust, can't be handled or installed by humans anymore, and are so finicky that they need inspection methods using radiation doses which destroys the parts. So you have verified all the solder joints but your PCB is useless because it's out of spec since the precision ICs have shifted.

 

There is a name for that. Idiocracy. Like the movie of the same name.

 

Snarky minds have forseen this. I remember a cartoon movie I saw in TV as a kid where the animated figures painstakingly assembled a product on an assembly like (I think it was an alarm clock), every step was shown, in cartoon style of course, and at the end of the assembly line was the "quality test department" which made the alarm clock ring and then smashed it with a hammer to silence it again, the springs came out, and the ruined clocks were packed into the shipping boxes. May have been one of the Looney Tunes cartoons. I don't remember everything after more than half a century.

 

My dad - the proverbial rocket scientist - once designed a ramjet powered rocket to demonstrate the feasibility and performance of the novel propulsion technology he had developed with his team. The guidance system for this missile came from a subcontractor ready to install. The incoming quality inspection department of the company my dad was working for "inspected" the guidance system for quality by applying the line voltage, instantly killing all the delicate electronics inside. This set back the project by months. Eventually the rocket was test fired and the propulsion system was successfully demonstrated.

 

So the cartoon world of lunacy and madness was true even 40 years ago. And it's getting more and more insane as we speak.

 

Some religious people say this planet is purgatory from scripture.

I disagree. IMHO, this planet is a madhouse the Gods built for their amusement. They are observing us and laughing their heads off. This also is the reason why they don't intervene as we self-destruct our civilisation.

Macintosh_nik's picture
Offline
Last seen: 14 hours 15 min ago
Joined: Jan 8 2021 - 05:18
Posts: 418
Hi Tocksin!

Let me give you my opinion too, I don't see much point in replacing anything other than the 2504. For 2513 and 2519 P-Lab has already made replacements, I put them all together and it works fine. The whole 74xx series are not particularly rare components, there are plenty of them in China and they are quite inexpensive. Besides most of them can be easily replaced by the 74LSxx series, I've already tried many of them. But the situation with 2504 is worse, although the respected Uncle Bernie says that they are plenty, but the prices are biting. I know you can buy them at Unicorn for $17.99, but they don't test them there, so you have to buy 10 units, that's $180. And there are suggestions for a 2504 1404 analog in China, but you have to be careful: the ones at $1 apiece are over-labeled 555s, and at $20-30 seems like a crazy deal to me. If you undertake to design a replacement 2504 I'd be happy to help you with that, I could order all the components and assemble a test specimen. But I can't do it myself, I don't know anything about it...

Offline
Last seen: 20 hours 34 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
About the Signetics 2504/1404 from Unicorn:

In post #7, macintosh_nik wrote:

 

"I know you can buy them at Unicorn for $17.99, but they don't test them there, so you have to buy 10 units, that's $180."

 

Uncle Bernie answers:

 

So far Unicorn Electronics have exchanged bad Signetics 2504/1404 they sold free of charge. At least they did it for me. But it is troublesome: you need to tell them to give you a return authorisation number, then you need to send the bad ones back on your own dime using that number, and then you have to hope the replacements they send  do work. Interestingly, I always got 100% functional ones in the replacements. So I suspect they may have an Apple-1 clone to test them. It may have been luck, though, because my statistical basis for them is slim. It's not often that I need "real" Signetics 2504/1404. I used them in my custom builds only if the buyer insisted on them and paid me a king's ransom for all the hassle and waste of time with these duds.

 

The upside is that the 2504/1404 I got from Unicorn had really interesting date codes and together with some other rare ICs from Uncle Bernie's treasure vault something like this is possible:

 

 

I still could do such a near perfect fake but I don't do it anymore, it just costs too much of my time, so don't ask, I don't do these custom builds anymore. Oh, and they had to pay with real money - US gold bullion.

 

Since I'm an IC designer I could design a 1404 drop in in about a weekend, layout included. These are really super primitive shift registers. The only problem is that the old PMOS process technology does not exist anymore, and the 1404 works with +5V/-5V supply voltages and +5V/-12V clocks. The clock swing could be reduced on chip to prevent blowing up the gate oxides, but making it run on +5V and -5V supplies would require careful selection of some trailing edge CMOS process around 0.8um (or so) and some simple circuit tricks to make it work. So it's not as straightforward as it seems at the first glance.

 

Last time I looked you could still buy AM2804 for below $2 each, from regular US IC brokers, but you have to buy 1000 or more, and so I have decided to stop selling my kits after my stash of 2804 runs out. You see, the problem is to use up all ICs in stock so that after the last kit is completed, nothing all too valuable is left. I just can't buy 1000 of those AM2804 because this would mean 142 more kits. And I don't have the other parts in this quantity anymore (and can't get them anymore at a reasonable price, the 2519B I have right now were the last in the world available at brokers, now the only known remaining stock of 2519 is at Manoshevitz, and he is too greedy for me. I don't pay usurious prices. And neither should you.)

 

Comments invited !

- Uncle Bernie

 

 

 

 

 

 

 

Macintosh_nik's picture
Offline
Last seen: 14 hours 15 min ago
Joined: Jan 8 2021 - 05:18
Posts: 418
Hi guys!

I don't know if this helps even a little, I found it on a Russian forum, I think it was written 8 years ago.

"I took another look at the 2504 datasheet and the computer schematic.These 7ms form a 7 bit memory of 1024 bytes with sequential read-write and are clocked by the same signals. In idea they can all be replaced by one 8 bit 2K SRAM (like 537РУ10) and one 10 output binary counter (like 561ИЕ16)... True, the 2504 has a separate input-output, but do we need it? If anything you can add an 8 bit register ..."

Offline
Last seen: 3 weeks 2 days ago
Joined: Dec 13 2021 - 12:20
Posts: 34
Replacing 2513, 2519, 2504 & Proms
Tocksin wrote:

Thanks for all of your insights.  I appreciate the feedback.  That's why I came to you guys before even starting the project - to assess how difficult it would be and how much it is really needed.  I've read all the work that's been put into getting these old chips, and it seems like quite a struggle.  I was hoping to combat the high prices the oligarchs you mention are want

I think there would be a demand for a replacement chip the same footprint size as the original even if it didn't look identical.  The S2519 is a good example of a DIP16 chip that sells for $50+ in small quantity and is hard to find that could be replicated.  The 2504/1404 DIP8 footprint would be harder to replicate due to size constraints.  You could replicate the proms in the same DIP16 footprint since they are going for $25 per chip. Maybe the S2513/CM2140 as well.  Keep us updated on your progress should you proceed further.

 

Justin.

Offline
Last seen: 3 months 3 weeks ago
Joined: Apr 9 2021 - 04:31
Posts: 90
I think with 2 of these:

I think with 2 of these: https://www.alibaba.com/product-detail/Best-Sellers-LF9502JC-25-PLCC-Brand_1600164689502.html you can replace all Shiftregisters on the Apple I mainboard.

One replaces the 2519 and the other all 2504.This IC is outdated aswell but still quite easy to get. There is a version with only half of the Bits that still would be enough but is harder to get.As it is impossible to plug 1 IC in serveral sockets same time without creating a very starnge Adpater Interfacing IC C4 and IC C14 might work and use 3 instead of 2 chips.The cursor IC C11B would need some extra wires.Regarding the 2519 one idea might be replace it with a dummy and replace the 2513 with a hybrid function device. But it would still be quite tricky to feed in the two extra signals needed.Or just create a more compact new PCB that also replaces the DRAM with one SRAM and substitutes the PROMs with an EEPROM

 

Log in or register to post comments