Repairing ZX Spectrum issue two

[1024MAK]

Yes, that’s too much ripple on the +12V rail. And yes, replacing C44, C45 and C74 (the extra 4.7µF capacitor that is not part of the PCB layout) should help.

Capacitor C74 should have it’s positive (+) lead to a suitable +5V point (e.g. the positive lead of C34). And it’s negative (-) lead connected to either the left lead of R59, or the left lead of R58.

IMG_6923.jpeg (35.27 KiB) Viewed 3984 times

And yes, it’s definitely worthwhile continuing with the repairs to this board.

DRAM chips are complex little boxes, and have many failure modes. One thing that some DRAM testers may not pick up on is a failure of the control inputs and three (tri) state output drivers to work correctly. Or the chip to operate slower than it’s access time (the 4116 ‘lower’ DRAM should be 150ns or faster parts).

If you have multiple chips, it’s definitely worthwhile swapping them around.

If the Z80 is unable to write to screen memory and the border is already white, it may appear that pressing reset does nothing. But that doesn’t necessarily mean it’s not working. Just that there is no visible change.

Until you get the board working as a 16K RAM model, it’s better off leaving out the ‘upper’ DRAM chips.

Mark

[Pygmalion]

Thank you for the suggestion.

I have unsoldered only one side of the above capacitors and measured them with a multimeter. C44 and C45 are 135μF and 138μF, C74 is 31μF (its nominal value is 22μF instead of 4.7μF). [And yes, I tested the multimeter on a modern 100μF just in case, and it read 100μF.] Now my multimeter can not measure ESR, but is ESR important for this function? Should I still replace the capacitors?

Actually, this proves that these original capacitors are indeed of high quality...

The ripple on the 12V line is exactly 50.0 Hz. That must be coming from the mains. So I also tested the power supply. I have an original ZX Spectrum power supply and when it is not plugged into ZX, it delivers 15 V with no ripple. [Damn, ground is on the pin of the connector, careful!] I have also opened it up. Nothing seems to have been tampered with there.

Now I am completely clueless... Does the coil pick up EM waves from the mains?

[1024MAK]

The ripple on the unstabilised/unregulated Sinclair UK1400 PSU should be 100Hz, as this PSU is made up of a thermal fuse, mains transformer, full bridge rectifier (four individual diodes) and typically one 4700µF electrolytic capacitor (as there was more than one PSU manufacturer, some may have two smaller value capacitors instead).

When a full bridge rectifier is used on a 50Hz AC supply, the ripple frequency becomes 100Hz because of the acton of the bridge rectifier (the negative half cycle is now positive).

Testing the ripple when this type of PSU is not connected to a load does not tell us enough to determine if there is a fault, as the capacitor will charge up to the peak output voltage of the transformer secondary. There is nothing to discharge the capacitor.

So please test the ripple with the PSU powering the computer.

If the ripple is 100Hz, that’s normal. If the ripple is 50Hz, then one or two diodes in the PSU are faulty.

The PSU will produce some ripple, but the lowest voltage (as measured on a ‘scope, not a multimeter) should not be less than 7.5V for a normal load of a ZX Spectrum on it’s own. At the maximum rated current this voltage should be no lower than 7V.

Your question on the capacitors, yes the ESR is a better indicator of the health and quality of an electrolytic capacitor. If the +9V (nominal) supply has excessive ripple (50Hz or 100Hz), changing C44, C45 or C74 will not help. The capacitance of these still being very good means they should still be okay.

The amount of induced mains frequency in the coil should be insignificant.

Mark

[Pygmalion]

This part of the diagnostics was fun, because even I with my limited amateur electronics knowledge immediately understand why a defective diode in the Graetz bridge turns a 100 Hz ripple into a 50 Hz ripple.

Well, I did what you suggested. The output of the power supply is 10.5V with a ripple of 1.1V pp at a dominant frequency of 100Hz. Yes, there is also a 50Hz harmonic which I attribute to an imbalance between the diodes. And also a 38kHz that I just can not explain. At first glance, this does not seem too concerning or require the replacement of all four diodes in the Graetz bridge.

I also repeated the measurement on the 12V line. It is 12.5V with a ripple of 0.5V pp at a frequency of 50 Hz. This is better than before. Maybe one of the diodes in the Graetz bridge does have some occasional problems after all and replacing them would be a good idea?

Anyway, I suspect that these ripple problems should not affect the function of the ZX Spectrum. After cleaning the contacts of the DRAMs, the situation has improved and I get a completely black box after the red rain disappears (typical for startup) and only later blue lines appear. With two suspicious DRAM chips. I think it's time to wait for "new" chips from China.

https://youtu.be/sI-J3GgxpQA

[redgatemodels]

This part of the diagnostics was fun, because even I with my limited amateur electronics knowledge immediately understand why a defective diode in the Graetz bridge turns a 100 Hz ripple into a 50 Hz ripple.

Well, I did what you suggested. The output of the power supply is 10.5V with a ripple of 1.1V pp at a dominant frequency of 100Hz. Yes, there is also a 50Hz harmonic which I attribute to an imbalance between the diodes. And also a 38kHz that I just can not explain. At first glance, this does not seem too concerning or require the replacement of all four diodes in the Graetz bridge.

power_supply.png

I also repeated the measurement on the 12V line. It is 12.5V with a ripple of 0.5V pp at a frequency of 50 Hz. This is better than before. Maybe one of the diodes in the Graetz bridge does have some occasional problems after all and replacing them would be a good idea?

12V_line.png

Anyway, I suspect that these ripple problems should not affect the function of the ZX Spectrum. After cleaning the contacts of the DRAMs, the situation has improved and I get a completely black box after the red rain disappears (typical for startup) and only later blue lines appear. With two suspicious DRAM chips. I think it's time to wait for "new" chips from China.

https://youtu.be/sI-J3GgxpQA

Any progress on this or are you still waiting for your new DRAM?

[Pygmalion]

Any progress on this or are you still waiting for your new DRAM?

Thanks for showing interest for my problem. I ordered new chips at AliExpress, so this usually takes 2-3 weeks. In the moment 4164s are already in Hungary, so I expect them by the end of next week. 4116s are still in China. But I will probably modify 4164s to use them instead of 4116s to proceed with the troubleshooting ASAP.

[Pygmalion]

I am preparing my own DRAM test sketch that will take into account the 32k DRAM memory used by ZX Spectrum.

As far as I know, 32k DRAM memory is actually 64k DRAM memory, with the lower or upper half of the memory being faulty.

But what does upper/lower half mean? I assume it means A7 is fixed 0 or 1. But is it fixed for CAS or for RAS?

[1024MAK]

I don’t have a datasheet for a Texas Instruments TMS4532 32k bit DRAM, but do have a datasheet for a 8k bit DRAM. 8k bit DRAM is similar in that it’s ‘half working’ 16k bit DRAM chips.

With the Texas Instruments 32k bit chips on an issue two board:

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

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

LK3 or LK4 selects the logic level to an input of IC26 (multiplexer), the output of which then goes to pin 9 on the DRAM. This is an address pin shown as “AR”.

I can’t remember if the fixed logic level from LK3/LK4 is used as a column or row input. But it should be possible to work it out by comparing the issue two schematic with a TMS4164 datasheet.

See also
http://www.breakintoprogram.co.uk/hardw ... rdware/ram

Mark

[1024MAK]

Attached is a datasheet for a OKI MSM3732 32K bit DRAM. Please note that the memory matrix in these are internally laid out differently to the Texas Instruments TMS4532 chips. They are not the same. But it gives you an idea.

Mark

[Pygmalion]

Attached is a datasheet for a OKI MSM3732 32K bit DRAM. Please note that the memory matrix in these are internally laid out differently to the Texas Instruments TMS4532 chips. They are not the same. But it gives you an idea.

Mark

The datasheet suggests that column A7 must be fixed for the OKI chip. So assuming that OKI and TI 32k DRAMs are interchangeable (are they?), the same should be true for the TI chip. I don't understand how the memory matrix comes into play here, my knowledge is insufficient in this regard.