Memotech 32k RAM pack failure

Discussions about Sinclair ZX80 and ZX81 Hardware
Post Reply
ujko
Posts: 9
Joined: Wed Aug 18, 2021 5:07 pm
Location: Croatia

Memotech 32k RAM pack failure

Post by ujko »

Hi,

I have one of this in mint condition, boxed, with manual, no scratches - real beauty :mrgreen:

Of course, something must be wrong - it does not work, produces blank screen when attached to ZX 81. I tried it on three of them so it really does not work.
20211125_222341.jpg
So, I have replaced both 74157's and instead of OKI M3732L-20 I put KM4164-15. It still does not work, likely due to a different matrix but can somebody confirm this ? 3732 is crippled 3764 and it seems it is organized as 256 rows x 128 columns matrix (according to datasheet). But what about 4164 ? It does work as a replacement for TMS4532-NL3 (at least the type which I have), so maybe it is organized as 128 rows x 256 columns ?

In the mean time I'll replace the rest of IC's (except PROM), just to be sure.

:?:
User avatar
1024MAK
Posts: 5101
Joined: Mon Sep 26, 2011 10:56 am
Location: Looking forward to summer in Somerset, UK...

Re: Memotech 32k RAM pack failure

Post by 1024MAK »

I've been inside a 16k Memotech 16k RAM pack and reverse engineered it, but have not done so for their 32k RAM pack.

The KM4164 should work, as they are also 128 cycle refresh DRAM chips. The logic level on pin 9 (A7) should not matter (as long as it is a valid logic level).

This being the only address input that is not required for refresh. But note that there may be a difference between the A and B versions of the DRAM chips. What is the full part number of the chips you are using?

The ZX81 display system means that the Z80 refresh register (and hence the refresh address) does not cycle through all 128 counts as often as it should. So if the DRAM chips are sensitive to not being refreshed as often as they should, this may cause problems.

Have you tested the removed chips?

What test gear do you have?

By the way, Texas Instruments (TMS) 64k bit DRAM chips (such as TMS4164) have a 256x256 matrix and so need a 8 bit, 256 cycle refresh. The 32k bit versions are connected up such that the multiplexers supply the correct logic level to the relevant pin via a fixed track connection or via a link wire.

Mark
ZX81 Variations
ZX81 Chip Pin-outs
ZX81 Video Transistor Buffer Amp

:!: Standby alert :!:
There are four lights!
Step up to red alert. Sir, are you absolutely sure? It does mean changing the bulb :!:
Looking forward to summer later in the year.
ujko
Posts: 9
Joined: Wed Aug 18, 2021 5:07 pm
Location: Croatia

Re: Memotech 32k RAM pack failure

Post by ujko »

Hi,

so I have replaced 74LS14 and 74LS00 IC's but with no success. :( Then I checked are there any shorts between pins and are there proper connections between address and data lines and edge connector. Everything OK.

I have no dedicated test gear for DRAMs, only a few 48k Spectrums and Sinclair's test IF2 cartridge. My, so to speak, test machine is v2 which has no jumpers for OKI DRAM. BTW, in the mean time I have found that, according to a ZX Spectrum service manual, TI DRAM matrix is 128r x 256c but OKI DRAM matrix is 256r x 128c. As you said, it is a refresh problem for a long run, but for a short test it is OK.

Anyways, 2 out of 8 3732's had been defected and after I replaced them with KM4164B it has started to work. :mrgreen:
Why it didn't work with all 3732's replaced ? Because not all of 4164's were in working order. How come ? Don't know, I had checked them all before. :?
One of possible explanations is that I was out of luck with the latest DRAM purchase - 7 out of 20 were defected and it seems it was just the beginning. :cry:

At the end, it is not clear to me how KM4164B can work in a ZX Spectrum upper memory configured for 128x256 matrix and in this 32k memory expansion as a replacement for OKI M3732 which has 256x128 matrix ?

Mark - thank you for your support ! Regarding A and B version of the KM4164 - I was not able to find datasheet for the "A" version. Do you have it ?

Damir
User avatar
1024MAK
Posts: 5101
Joined: Mon Sep 26, 2011 10:56 am
Location: Looking forward to summer in Somerset, UK...

Re: Memotech 32k RAM pack failure

Post by 1024MAK »

Sorry, I was also not able to find a datasheet for the KM4164A either.

One point. All Texas Instruments (TMS) 64k bit DRAM chips (such as TMS4164 and the TMS4532) have a 256x256 matrix. However, when used as the ‘upper’ 32k byte ‘expansion’ RAM in a ZX Spectrum only half of the DRAM chips is used. The result is that the used part of the matrix is 128x256. With RAS supplying the first 7 address bits plus the AR / A7 input on DRAM pin 9 and CAS supplying the remaining 8 bits of the address. On the ZX Spectrum, the DRAM AR / A7 input comes from a multiplexer chip as normal, but the two relevant multiplexer chip inputs are fed via selection links. Depending on which half of the 32k bit DRAM chips are working, determines how the links have to be set.

Here’s the type numbers for the 32k DRAM chips typically used in a ZX Spectrum:
* TMS4532-20NL3
* TMS4532-20NL4
* TMS4532-15NL4
* TMS4532-15NL3
* OKI M3732H-20RS

On Issue 2 two boards, LK3 LK4

DRAM used with LK3 (0V) fitted:
* TMS4532-20NL3

DRAM used with LK4 (+5) fitted:
* TMS4532-20NL4
* TMS4532-15NL4

Later boards can be configured more flexibility, thus allowing more choice of which DRAM chips can be used.

Mark
ZX81 Variations
ZX81 Chip Pin-outs
ZX81 Video Transistor Buffer Amp

:!: Standby alert :!:
There are four lights!
Step up to red alert. Sir, are you absolutely sure? It does mean changing the bulb :!:
Looking forward to summer later in the year.
ujko
Posts: 9
Joined: Wed Aug 18, 2021 5:07 pm
Location: Croatia

Re: Memotech 32k RAM pack failure

Post by ujko »

I investigated a little bit...
Well, some of 4164 chips indeed have 256x256 type of matrix thus need 8-bit refresh. However, as you said, TMS4164 is organized as 128x256 in two banks where A7 (at RAS) controls which bank is used (0/1, lower/higher).
OKI3764 works in a bit different way - it is organized as a 256x128 matrix in two banks. What is interesting is that it is internally organized as 2x128 rows which is important because it allows 7-bit refresh to be used. A7 (at RAS) is connected to A14 to form 8-bit row address. In this case, the bank control occurs at CAS when A7 is set to 0/1 for lower/higher bank respectively, and A0-A6 form 7-bit column address.

So what happens when OKI3732 is replaced with TMS4164 ?
Instead of fixed value at A7, it gets A14 value from bus so some data are stored in the lower bank and some in the higher bank, while the highest bit of the column address remain fixed so only half of the columns are used. 8-) The similar logic may be applied if considering TMS4532 to be replaced by OKI3764.

Damir
Post Reply