Hi all,

I'm posting this in a seperate thread as it might be of wider interest.

As part of Beeb816, we use a 55ns 5V 512Kx8 SRAM in in DIP 32 package. We wanted to try out a faster 10ns part, but these are only available in surface mount (SOJ and TSOPP) packages. Before respinning the whole board, we though we would try an adapter shim.

The design files are in github:
https://github.com/hoglet67/SRAMAdapter ... ster/kicad

There is a two layer version:

And a four layer version:

The SRAM is in an SOJ-36 package, which just fits between the DIP pins:

The decoupling capacitors are hidden underneath:

We ended up going with the 4-layer verstion, which cost just $5.00 for 10 boards from JLCPCB.

This is the first board off the production line, fitted with an a 10ns ISSI IS61C5128AL-10KLI part:

And fitted in Beeb816:

With this, Beeb816 runs stably at 20MHz (previously the limit was about 14MHz):

I've built three of these, for the three prototype Beeb816 boards.


Feel free to make use of this work if you have need for such a thing.

Dave

Well done Dave - very nice repackaging and a fine result!

I think it turns out that the usual 55ns parts are barely able to support 14MHz, so moving to 10ns is a great leap forward! And we keep the DIP format, the green PCB with lots of drilled holes, if we choose to.

Only just, yes, depending on their internal organisation and address decoding (since access time basically starts at address stable, not the strobe).

ReCo6502 ran at 14.7 MHz with 55 ns DIP SRAM. With the same construction you have here (SMD on DIP header) a 25 ns SRAM brought that to 22 MHz. I think others have reported even higher numbers than that. The headroom of 14 MHz WDC chips is astonishing, almost up to the point of having to conclude gross underspecification ...

Interesting datapoint, thanks!

That's because production testing is done at 20 MHz.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

If only Intel did that. I could safely run my i9-7920x at 6.14 GHz ...