6502 built from discrete components!

[SNGLinks]

http://www.evilmadscientist.com/2016/6502/

[AntiProtonBoy]

I'd love to have one of those.

[XOIIO]

Next: core i7 quad

[blacksheeplogic]

http://www.evilmadscientist.com/2016/6502/

That's quite impressive, I've done much simpler CPU's but nothing like that or anywhere near as clean either.

[mikeselectricstuff]

More info :
http://hackaday.com/2016/05/24/how-the-dis-integrated-6502-came-to-be/

The transistor arrays on the back are from http://www.aldinc.com/

[bitseeker]

Wow, that's pretty amazing. I can't imagine going through the process of designing let alone assembling it.

[Melt-O-Tronic]

That is way cool!  I'd like to see that mounted vertically between glass panels with a ribbon cable connecting it to the CPU socket of an Apple ][.

[kc8apf]

Eric's talked about that a bit.  MONster 6502 is very unlikely to ever run at >100kHz or so.  The Apple II and similar designs relied on specific clock frequencies for correct operation.  It'd be freaking awesome but just isn't going to happen.

[bitseeker]

Ah, I wondered about that too. Is it the physical distance of the components that's the limiting factor? It is HUGE in comparison to the area of the actual IC.

[JoeN]

I wonder why they used a number of quad transistor arrays instead of discretes when most of the thing is discrete.  That takes the whole "disintegrated" bit away from the project.  Other than that, I quite like this. 

I finally found the justification for the transistor arrays, it is because they have body contacts and they can be used to build transmission gates:   http://tubetime.us/?p=346  .  Not sure if I like that, he could just redesign those subcircuits in CMOS (or maybe even NMOS?  not an expert here) like the rest of the chip and bask in the warm glow of victory, significantly cheaper parts, and even more little discretes all over the place.  Win-win-win.

I love their 741 and 555 projects and I cloned them into smd variants myself just for fun a while back:



Then I moved on to CMOS circuits, following the 74xx patterns.  Have several working now, and not completely trivial ones.  A 74151 in discrete CMOS (BSS84 and 2N7002), two boards stacked up, 122 transistors total:

[T3sl4co1l]

Ah, I wondered about that too. Is it the physical distance of the components that's the limiting factor? It is HUGE in comparison to the area of the actual IC.

Computers faster than this have been implemented in discrete logic many times; the Cray-1 ran at 80MHz, and was a modest fraction of a wavelength across; that's probably about at the practical limit for designing a speed-of-light-constrained CPU.

The limit here is largely power dissipation (tolerable supply consumption, and dissipation capacity of all the SMT transistors) versus the rather large gate capacitances, which I'm still guessing but still don't know if they are 2N7002s or not, but if so, there's your problem.

If transistors were used that were comparable to the original (i.e., with ratings like 10Vds(max), 5mA Idss and 100 ohms Rds(on) -- you'd only find discretes for RF with ratings like this), it would be able to run full speed, or nearly there.

Tim

[mikeselectricstuff]

The limiting factor on frequency is capacitance

[Psi]

Advanced Computer Architecture was the most interesting paper i took at Uni.
Ya come out of the course knowing you could build your own cpu from logic gates if you had to.

[tszaboo]

The limiting factor on frequency is capacitance

So crank up the current and it will run faster.

[botcrusher]

It could be intersting watching it MELT while trying to crunch Oregon Trail..

[Red Squirrel]

That's pretty awesome!  Now redo it with valves.