• Please review our updated Terms and Rules here

Anyone else building an Apple-1 ?

I considered doing a clone with current-production CMOS logic, but aren't sure what to replace those huge shift registers in the video generator with. I assume when Woz designed this thing those registers were cost-effective alternatives to then available parallel SRAM. There can't be any other valid reason for hobbling the capabilities of the video display by choosing to design with what is effectively serial memory.
I could serial-ise modern parallel SRAM with address counters and I/O control logic to emulate the function of those original shift registers, but it would all look rather inelegant in the end, with a significantly higher chip count than an otherwise straight-forward and much superior memory-mapped design.

One could just emulate the whole video generator with a suitably fast modern micro-controller, but then it becomes just another embedded project lacking much of the retro appeal.
 
GK2001 said:
One could just emulate the whole video generator with a suitably fast modern micro-controller, but then it becomes just another embedded project lacking much of the retro appeal.
Click to expand...

That's the route taken by most not-trying-to-be-original Apple I clones, including the Briel Computers Replica-1 circa 2003-ish, which makes it maybe the first "modern" Apple I clone?

Honestly, considering the way the Apple-1 treats its video circuitry I'd consider it a reasonable route to take because, as you note, unwinding the original design to use SRAM instead of those shift registers results in a pretty silly mess. An Apple I with memory mapped video is pretty close to a Commodore PET, so you might as well just make a PET replica.

(Actually this has me thinking, are the schematics for the Sphere-1 out there in the wild? It was a 6800-based machine that predated the Apple-1 that had memory-mapped video, it might be an interesting replication target.)
 
Got a package today, two transformers, capacitors and all the DIL sockets. Going to use an Apple ][ keyboard (not a datanetics, that was just too pricey) but will need to wire up a key to do the clear screen.

Not decided on the case yet. Koa is out as its just to hard to get in the UK so its probably for Mk1, pine boards.
 
GK2001 said:
That's an interesting solution to the single 2519, but you'd need ninety six of those HEF4557s to replace the six 2504s!
Click to expand...

To sub with SRAM, ignorantly speaking, might you not need much more than a single 1Kx8 RAM chip and a binary counter that goes to 1024 to increment the address lines? IE, increment the counter based on the O3/O4 clock phases so it circulates like the original, and then follow the WRITE signal that's fed to the 74157 multiplexers that control whether the bit that's fed into the register is the recirculated one or one from the bus to simply switch the date lines from READ (normal state) to WRITE as appropriate? Obviously we don't need to actually feed read bits back in, so if we're optimizing the replica design maybe we replace the multiplexers with a simple '244 buffer between the data bus pins feeding the 2519 (replacement) and the RD lines coming from the PIA.

(FWIW, seems like the same technique would work with the 2519, IE, a RAM chip and a count-to-40 circuit.)
 
Last edited:
Eudimorphodon said:
To sub with SRAM, ignorantly speaking, might you not need much more than a single 1Kx8 RAM chip and a binary counter that goes to 1024 to increment the address lines? IE, increment the counter based on the O3/O4 clock phases so it circulates like the original, and then follow the WRITE signal that's fed to the 74157 multiplexers that control whether the bit that's fed into the register is the recirculated one or one from the bus to simply switch the date lines from READ (normal state) to WRITE as appropriate? Obviously we don't need to actually feed read bits back in, so if we're optimizing the replica design maybe we replace the multiplexers with a simple '244 buffer between the data bus pins feeding the 2519 (replacement) and the RD lines coming from the PIA.
Click to expand...

Yes, that is basically what I suggested in my first post, but with much less detail. But no matter how efficiently you replace both the 2504s and the 2519 in totality with currently available discrete logic it still looks at bit like poo in the end. I bought all of the parts required a while ago to assemble a prototype of a tweaked Jupiter Ace clone on breadboard. I'll be getting back to that in the not too distant future. Now there is a stark contrast to the Apple-1 when it comes to innovation in video generator circuit design !
 
GK2001 said:
Yes, that is basically what I suggested in my first post, but with much less detail. But no matter how efficiently you replace both the 2504s and the 2519 in totality with currently available discrete logic it still looks at bit like poo in the end.
Click to expand...

Yeah, I figured you'd probably had it worked out, this was mostly just me saying out loud "hey, wait, maybe this could be done without making the chip count much higher than the original, at least...". But yeah, architecture-wise it'd be pretty ridiculous.

I've seen it quoted that the reason the original used the shift register memory in the first place was that those parts had reached the "available as cheap surplus" stage of their lifecycle, but I am curious what the price delta between them and SRAM actually was when they sat down to churn out the schematics for the PCB, I can't imagine it was much. My suspicion as to why they were really used was simply that's what Woz was familiar with at the time; if you read his autobiography the thing that'll hit you is how... un-curious, he can come across regarding what other people in the field were up to. (He in fact comes across as actively hostile, calling everyone "bad engineers", in a lot of his old interviews.) If you took his book as gospel you'd come away with the impression that MOS started churning out 6502s without knowing you could even build a computer out of them, they were just an artifact that fell from the sky in a meteorite that they copied and it was up to him to figure it out.
 
I haven't worked it out to the last detail, just enough to groan about it. Given a single SRAM could eliminate those mentioned six 2504, I guess that is a chip-count advantage to start with that somewhat ameliorates the additional counter and control logic that would be required. The Apple-1's video generator chip-count, though, couldn't really be considered low in the first place, especially considering its limited capabilities. By a rough count it's on par with the PET-2001, which in contrast was memory mapped, could do reverse characters and generated 8x8 pixel character blocks instead of 6x8.

Correcting my previous comment on the number of 2504s in the Apple-1, there's actually seven of the damn things, not six. Designated C11B, looks to keep memory of the cursor position.
 
Last edited:
Al Kossow said:
yes
Click to expand...

A Sphere-1 replica should definitely go on someone's bucket list. It's such an enigmatic piece of computer history that needs more exposure, and it also might actually be a more fun computer than an Apple-1 replica anyway. (From what I know of it it'd roughly resemble a 6800-powered PET in capabilities.)
 
Video appears to be more like a TRS-80 in low-res mode minus the graphics.


You would need to source a NOS 6800 and 6820 PIA unless going the FPGA or other embedded route.
 
Last edited:
GK2001 said:
Video appears to be more like a TRS-80 in low-res mode minus the graphics.
Click to expand...

Yeah, I meant it in the broad sense of it having memory-mapped video like a PET. Maybe a monochromatic TRS-80 MC-10 would be a better analogy. The replica would definitely need some kind of graphics capability as an enhancement.

GK2001 said:
You would need to source a NOS 6800 unless going the FPGA or other embedded route.
Click to expand...

Ugh, I guess that's an issue. Looks like it's still not too hard to lay hands on a 6802 or 6808, which should work with some tweaking? But it's definitely not going to be as amenable to modern builds as a Z80 or 6502 based machine.
 
One attractive thing about building the A1 is the lack of a specific case or keyboard. It means I can build it without having to get into metal bending, riveting and welding.

Still not sure on the keyboard

I have one of these (look the other way Daver2 ;) )

1651219121086.png

and one of these

1651219159756.png

The apple keyboard is probably the easiest to interface but the SD-16534 is older but has extra keys and does include a key labelled 'clear'
 
Here is my Apple I setup. I chose acrylic because I wanted to display the motherboard - what is the point of having a motherboard (signed by Woz, to boot) that you can't easily view? I also went with the Apple II keyboard, from a beat-up B&H Apple II, in black to match the case. It was easy to hook it up and I added a clear button to the upper left of the keyboard. One thing about those big keyboards is most of the keys are useless with the Apple I's primitive line editing, i.e. none, and take a lot of space.
IMG_1555_2.jpeg
 
So bits have started arriving

Uncle Bernies IC kit, Tranformers, IC sockets, PCB, kettle plug socket and switch

Problem is, I ordered a HASL PCB and have got an ENIG PCB which doesnt look the part but the reproduction manual is fantastic as well as the plasticised circuit diagrams.

However, I might be getting a reproduction Newton style board with proper hard gold edge connector.

:)
 
Just started the process! I found an admittedly cheap replica board from Poland and bought it on an impulse. So far I've found all but a handful of the TTLs from a local electronics store. (All dated 1973 to 1975. Which was kinda amazing to find!) Sadly, they didn't have the old shift registers or DRAM but I'll keep looking!

I expect that it will take months to find everything I need. But then, that's the fun! (Also, I can't really afford the preassembled kits right now, so scavenger hunt it is!)
 
Think Unicorn electronics are your best (but expensive) bet for the shift register.

Ebay has them for $60-$100 !
 
You must log in or register to reply here.
Back
Top