Date   : Sun, 14 Nov 2010 16:33:58 +0000
From   : percy.p.person@... (Ed Spittles)
Subject: Electron Ferranti ULA reverse engineering progress

Hi Philip, Theo,

I've rotated the photo of the ULA cell, and labelled some pins:
http://img337.imageshack.us/img337/2070/ulacellcompare.png

It seems that the two cell designs (as documented, and as
photographed) are logically equivalent.

On 12 November 2010 19:03, Philip Pemberton <philpem@...> wrote:
> On 12/11/10 15:54, Theo Markettos wrote:
>> According to the Spectrum ULA guy who went to visit, there isn't much
>> technical info there - it's all marketing etc stuff.
>
> Ah, shame.
> (and I'm still waiting for my copy of the ULA book, bah.)

I too hope to have a copy of this book soon:
http://www.zxdesign.info/book/ - I got a vacation message when I
ordered.

>> Given the constraints, I wouldn't be surprised if the logic is designed at
>> the transistor level rather than at a gate level mapped onto transistors.
>> So there might not be gate-like standard cells to extract, each gate may be
>> designed specifically for that application. ?Though there will be some
>> regularity.
>
> Another idea successfully scotched :(

I still hold hope that this is a logic-level design.

In my annotated picture, you can see four logic inputs and two logic
outputs, so each cell can
implement combinations like
 - one or two INV
 - one or two NOR2
 - INV and NOR2
 - one NOR3 or NOR4
I'm not sure about more complex logic, nor have I figured out the
idiom for storage (latch or flop)

For a given logic gate, you need to connect a current source to one or
both double transistors and connect a pullup resistor to the output
node.  Also need a pullup for the current sources (a connection which
I didn't label.)  As there are two current sources and three pullup
resistors, it's possible to make two independent logic gates.

>> Not sure there's much I can do to correct for pincushion: that's a
>> microscope+camera optics problem. ?Probably easier to software-postprocess.
>
> Hence why I suggested lensfun -- it's an open-source library to do lens
> distortion correction.
>
> If you could get a slide with a pre-printed grid onto the microscope,
> that would be the ideal reference to use for calibrating things.

But we already have a grid!  I think it's not too hard to make a
reasonable estimate - I did it once before.  I suppose the pipeline of
necessary processes is well understood by those skilled in the art: I
imagine it's something like
 - adjust exposure
 - correct vignette (flat field)
 - correct for lens distortion
 - adjust for rotation
 - adjust for tilt
 - stitch multiple images
If we're lucky, Hugin (http://wiki.panotools.org/Hugin) does everything for us.

Cheers
Ed