A 6502 constructed from discrete transistors

[amyk]

http://monster6502.com/

It's actually not that big.

[hamster_nz]

I'ld love to see the other side of the board... (a tiny bit) more info are on the person's blog at http://tubetime.us/?p=346

[arobincaron]

I spec'd doing this myself using a similar discrete component approach.  Some back of the envelope estimation put the effort at ~2 years of hobby time and about $2K in hard cost. That was the end of that :-)

Supercool to see someone followed through and made it a reality.

[westfw]

It's actually not that big.

At about 12x15 inches, it's the largest "hobbyist" PCB I've seen in a long time.
The original 6502 was 168 x 183 mils, so this is about 6000 times larger.

[mikeselectricstuff]

I've just seen this in person & it's awesome!
On the back there are some SO14 or 16 transistor arrays -  I don't recall exactly what Eric said they were for but I think it  to do with needing oddball transistors  with multiple emitters or something similar. 

[T3sl4co1l]

Are they literally all 2N7000s (or close enough, i.e. SMTs, arrays), or did they use something of actually modest size?

I can't imagine an NMOS design, using huge hulking 2N7000s, going much more than 10s of kHz.  Which is about what they'd claimed, really... so... 

Tim

[DJohn]

On the back there are some SO14 or 16 transistor arrays -  I don't recall exactly what Eric said they were for but I think it  to do with needing oddball transistors  with multiple emitters or something similar.

The "tiny bit more" information at http://tubetime.us/?p=346 covers that.  The 6502 uses dynamic logic, which means a lot of its transistors are used as transmission gates.  For "various technical reasons", you can't do this with ordinary 3-pin MOSFETs: a separate substrate connection is required.  The chips are arrays of 4 transistors with that separate substrate pin.

[edavid]

On the back there are some SO14 or 16 transistor arrays -  I don't recall exactly what Eric said they were for but I think it  to do with needing oddball transistors  with multiple emitters or something similar.

The "tiny bit more" information at http://tubetime.us/?p=346 covers that.  The 6502 uses dynamic logic, which means a lot of its transistors are used as transmission gates.  For "various technical reasons", you can't do this with ordinary 3-pin MOSFETs: a separate substrate connection is required.  The chips are arrays of 4 transistors with that separate substrate pin.

Probably SD5001s: http://www.linearsystems.com/assets/media/file/datasheets/SD5000_5001_5400_5401.pdf

[djacobow]

Wow, that is a really ambitious project and the result is quite handsome, too.

When I was a young engineer I often daydreamed of having the superpower to be able to see the state of a latch on a chip just by looking at it. This person has realized that power by putting LEDs on all the register bits!

[JoeN]

It's actually not that big.

At about 12x15 inches, it's the largest "hobbyist" PCB I've seen in a long time.
The original 6502 was 168 x 183 mils, so this is about 6000 times larger.

It looks like they did it at OSH Park too.  12x15x5 = $900  for three boards.  I love OSH Park, but I would go with the cheap guys for the first go-round on this board. 

I spec'd doing this myself using a similar discrete component approach.  Some back of the envelope estimation put the effort at ~2 years of hobby time and about $2K in hard cost. That was the end of that :-)

Supercool to see someone followed through and made it a reality.

I had given some thought to doing this myself also.  But my first thought was change the design from NMOS to CMOS.  Yes, you double the number of transistors.  But at least it will run faster and you won't have to bother with those arrays which I think takes away from the project, frankly, having any non-discrete components on the board.  But this guy did it.  I didn't.  That is a very expensive board.  I think the amount of cost is rather less than what you spec, through.  Even at say 9,000 transistors, that is reel and a half of 2N7002 and BSS84.  Those parts, roughly, are $100 a reel.  So $300 for parts total, and get the boards from China on the cheap.

[westfw]

It looks like they did it at OSH Park too.

Nah, that looks like black soldermask, not purple!

12x15x5 = $900  for three boards.

Except that it's a four layer board, so that'd be $10/in².  But they could use the "medium run" (150 in² minimum) which is cheaper ($2/in² for 4 layers) and would let them do fewer than three boards (if I read that right.)  It's still going to be an expensive PCB!  (which should answer the whole "if you can do this why did we need integrated circuits?" question!)