Repairing ZX Spectrum issue two

[1024MAK]

They are not interchangeable without modification to the issue two board.

The matrix comes into it, because the Texas Instruments 64K chips require a 256 cycle refresh, while the 64K OKI chip only needs a 128 cycle refresh. Hence the address bits for the column and row are allocated differently. Because of this, the particular input to select between the working and ‘non-working’ sections may be different.

If you look at the schematic for a later issue ZX Spectrum board (issue 3 onwards) you will see a lot more links that are used to configure it for the different DRAM chips.

Mark

[redgatemodels]

My issue 2 has this mod pcb so that OKI upper DRAM can be fitted

[1024MAK]

Yes, there are at least three variations on the modification of an issue two board to enable OKI DRAM to be used. Ultimately they all so the same thing electronically.

Mark

[Pygmalion]

Well, it seems that it is a fixed row A7 after all. I have two 32k chips that are OK if you fix row A7 to 0 but not OK, if you fix column A7 to zero. These are all the lower half 32k DRAMs.

But if someone knows and can confirm that, that would be nice.

More worrying is that there are four 32k chips OK in the lower and upper half, which probably means that the Arduino DRAM tester is not good enough to detect errors...

[1024MAK]

What date codes are on the chips that test okay for both halves? Are they different to the chips that only have one half working?

And keep in mind that the chip manufacturers tests would have been very extensive. They would have rested at the upper and lower limits of the specifications. The last thing they would have wanted would have been to have supplied faulty chips and wreak their reputation.

Mark

[Pygmalion]

What date codes are on the chips that test okay for both halves? Are they different to the chips that only have one half working?

And keep in mind that the chip manufacturers tests would have been very extensive. They would have rested at the upper and lower limits of the specifications. The last thing they would have wanted would have been to have supplied faulty chips and wreak their reputation.

Mark

They are all TMS4532-__NL3.

[1024MAK]

The date code is a four digit code (it may have a letter as a prefix), the format is YYWW
where YY is the last two digits of the year of manufacture,
and WW is the last week number within the year of manufacture.

So for example here

IMG_7091.jpeg (23.47 KiB) Viewed 3706 times

the date code is P8423 which is week 23 of the year 1984.

Mark

[Pygmalion]

The date code is a four digit code (it may have a letter as a prefix), the format is YYWW
where YY is the last two digits of the year of manufacture,
and WW is the last week number within the year of manufacture.

So for example here
IMG_7091.jpeg
the date code is P8423 which is week 23 of the year 1984.

Mark

The chips that have both halves fine are 8223, 8327, 8332 and 8336, the ones that have lower rows fine are interestingly both 8314 the one that has both lower rows and lower columns fine is P8332. The one that is completely faulty is 8336. So there is simply no system. Also, both 8332 are 150ns, and the rest are 200ns. Complete mess.

[1024MAK]

Yeah, but now we know that you have a mix. Machines fitted from the factory normally either have all the same size DRAM chips with the same date code, or a mix of just two date codes.

Mark

[Pygmalion]

I received 10 HYB4116 chips for a total of $7.38 and 10 KM4164B-15 chips for $4.37 from AliExpress.

After replacing the 4116 chip that proved faulty in the Arduino tester and two 4116 chips that I suspected were faulty because changing their position changed the picture, the ZX Spectrum came to life!!!!

I also replaced a 32k chip that proved faulty in the Arduino tester and now I can play 48k games.

All the AliExpress chips are obviously harvested. Out of 10 HYB4116, 7 are OK, 2 were proved faulty in the Arduino tester and 1 is faulty, despite being tested OK in the Arduino tester.

It is interesting to note that

1. Of the 6 faulty chips, only 3 were detected by the Arduino tester, the remaining 3 only proved faulty when used in the ZX Spectrum.
2. I was able to correctly determine the additional faulty chips by changing the position and seeing if the image changed.