ZX81 Homebrew Builders ..

[gozzo]

if they are fake, is there any easyish way to 'speedtest' them? hope they are at least 4mhz and not 2.5s !!

[1024MAK]

Yep, it's possible to test them. In a Spectrum, or maybe a ZX81 (but not tried this), the NMOS and CMOS produce different results when executing OUT (C),0 or OUT (C),255 instruction - see here.

But for the speed test, you really need to make up a very simple single board computer using fast SRAM, and fast FLASH / EEPROM. With a simple decoder and a simple I/O interface, so that you can run them at various clock speeds...

Mark

[PokeMon]

The current consumption of NMOS and CMOS is remarkable different, too.
The NMOS is stated with 200mA while the CMOS takes about 20mA at 4 MHz.
I don't believe that NMOS chips are produced or relabeled.
It is hard to find NMOS Z80 at all, I think.

[gozzo]

so you think they would at least be cmos? and hopefully 4mhz??! if so thats ok, but would be nice to know their real speed capabilities! i might buy a few of the other cheapie ones, and see what they're like....

[gozzo]

hmm, so no quick and easy way to speed test other than using in an actual computer circuit, and see if it crashes or not?

[Andy Rea]

And even if you did build something you would need to exercise every instruction 1000s of tumes to be sure. Also any instruction that effect registers or flags should be verified for correct operation, you may get erroneous results but no actual crash.

[stevendt]

"er, how can you tell a 'fake' one"

With difficulty !

you might find that page useful

http://primrosebank.net/computers/mtx/p ... _fakes.htm

regards
Dave

[stevendt]

The tests that I have carried out were done on MTXPlus+, I have the ability to set clock speeds between 4 and 16MHz to run RAM tests and report the CPU type (NMOS or CMOS). I have had a few that totally failedl to run, this could be because they were U/S, or because they were 2.5MHz parts. As I could not determine which, those ones were binned

regards
Dave

[1024MAK]

hmm, so no quick and easy way to speed test other than using in an actual computer circuit, and see if it crashes or not?

Well, it's very far from a full proper test, but for a quick and dirty test, you could make a simple NOP tester, but use a more useful instruction instead...

This is where you wire the databus up with a fixed instruction, and ignore the address bus. So the CPU only reads a single instruction repeatability. Then monitor the control lines to see if the CPU is behaving.

Mark

[PokeMon]

This is not a reliable test. There are situations, where timing can got quite strong.

Candidates in my eyes are prefixed instructions (with FD/DD/ED/CB), IO instructions, restart vector instructions (with a 5 clock cycle t-state), interrupts and nmi and wait. Even busreq is a candidate to fail but not often used.

A good idea is to put this into a Zeddy to see if run with 3.5 MHz. If you can see the video picture on the oscilloscope while monitoring A15 (I know when and how display is running even if not shown) you can increase clock frequency as well. But I doubt that the ULA will do it. Maybe a good idea to take Andy's ZX ULA (2 ?) to test at least 6.5 MHz.