Spectrum 48K Upper Ram Fault

[1024MAK]

It could be that one or more DRAM chips have a fault with their internal address decoding/matrix circuitry. Hence the memory cell for one location may end up being used for multiple memory addresses. This fools simple memory testing tools, as they can write a value and read the correct value back.

Mark

[Tiger]

Thank you Mark - that's what I meant

[hatman72]

That didn’t go well. Having replaced IC15 and IC16 I’m not able to boot any more. The speccy makes more noise than normal and fills the screen with multi coloured blocks. Not dissimilar to some of the memory tests I’ve seen.

I’m not the world’s best solderer but I’ve got a connection on all the pins and none seem to be shorting. It not obvious to me whether I’ve messed up the soldering (most likely I guess) or whether my replacement ram is faulty, or whether something else is going wrong. If I could figure out how to attach a photo I would, but does anyone know what the most likely cause of flashing blocks is?

[1024MAK]

To attach a photo, either drag and drop it from a file window to the edit box. Or use the attachments tab below the edit box.

Flashing blocks are due to the attributes screen memory (in the 'lower' 16k bytes of RAM) having random data in them. Most often because the ROM program has crashed. This can be caused by the 'upper' DRAM being faulty as well as the 'lower' RAM being faulty, or due to shorts on the data bus lines.

Mark

[hatman72]

Thanks Mark. This is what it looks like. The border is yellow.

[1024MAK]

If the border is very rarely white after power on, then the Z80 is unable to run the ROM code.

Check for shorts on the data lines of the chips that you have worked on. Test to nearby circuits not just to the +5V and 0V rails.

Mark

[hatman72]

Ok, so I discovered that a line running under IC15 was just making contact with the pin above it. I’ve no idea where the line goes - it seems to go under IC16. You can see the line underneath the capacitor.

So I now get a white border but the overall picture is blue and there are odd pink blocks smearing to the right. The picture below shows it but the colours are off making it appear more blue than it is.

I guess I am concerned that I may have caused damage somewhere else.

[1024MAK]

The issue two board requires manual calibration of the colour encoder circuitry.

Have a look at this picture:

Issue two VR1 and VR2

The items highlighted in the photo are adjustable/variable capacitors and adjustable/variable resistors. Because the variable resistors don’t have provision for a control knob, they are normally known as presets or trim pots.

VR1 and VR2 are used to calibrate the colour encoder settings. On issue three and later boards, additional circuitry does this ‘automatically’.

There is a guide on how to adjust these variable resistors here.

If you have problems with trying to adjust VR1 and VR2, you could try a VERY SMALL amount of WD40 on the black/brown carbon track where the metal wiper makes contact, then adjust them backwards and forwards several times. Then set them up again. This often reduces any electrical noise from them due to dust.

Normally the adjustable/variable capacitors should not be adjusted unless you have suitable test gear to set them up correctly. But if the colours are unstable, you have nothing to loose if you adjust them just a little bit. Take a photo that you can refer back to if you need to, to see where they were before adjustment.

If you have not already done so, all the electrolytic capacitors should be replaced with brand new ones. If possible try to buy axial types and buy from a reputable supplier. Be careful buying from trading sites like eBay. There are poor quality parts being sold on the various trading sites.

During fault finding and servicing, I recommend the following tests.

Test the voltages on the 7805 voltage regulator and report the results.



Be very careful that a multimeter probe does not slip and short anything out. You can use use the heatsink tab for 0V/GND instead of the middle pin (electrically inside the chip they are connected together). Use this point for the black/negative meter lead unless I specify a different location.

When using a Sinclair 9V UK1400 PSU, the input should be between 10V and 13V.
The output should be between 4.75V and 5.25V.

Then can you please check the +12V and -5V power rails. These are all available on the edge-connector. They are also all available on the pins of one of the 4116 (or equivalent) DRAM chips, for example IC13.

Note that some diagrams in some Sinclair documentation (including some of the manuals) show a "-12V" supply pin as well. Ignore this, the name is supposed to be "~12V" (where the ~ is a wavy line) as this is a AC waveform, not a DC supply. The supply rails that you should test are however, labelled correctly.



A copy of the full data sheet is here

Pin 1 is VBB, which is -5V (-5.5V to -4V).
Pin 8 is VDD, which is +12V (+10.8V to +13.2V) (although lower than 11.5V normally indicates a problem).
Pin 9 is VCC, which is +5V (+4.75V to +5.25V) (the output from the 7805 regulator is normally very close to +5V).
Pin 16 is VSS, which is 0V (ground/GND) Test this by measuring between this pin (black test lead) and pin 1 (red test lead).
(Voltage tolerances taken from here)

Also, let’s look at the +5V and +12V supplies to the video section.
Test pin 14 on the ULA (IC1) for the +5V supply.
Test pins 14, 15 and 16 on IC14 (LM1889) for +12V.

If all the voltage tests are okay, then all the power supply circuitry is okay.

Otherwise please report back your results.

If you are up to doing modifications, there is a Sinclair recommended upgrade to the DC-DC power supply circuitry on the issue two board. I’ve posted details here on my blog.

Note that even after doing all the above, it looks like you still have a problem. So please report back what colour the border is each time after at least five start-ups.

Mark

[hatman72]

Thanks for the info Mark.

I’m not going to try and speculate what’s going on here, but I turned the speccy on a few times to check the voltages as per your instructions. All voltages looked good, so I plugged the screen in to check the border colour and..... I was greeted with 1992 Sinclair Research Ltd on a white background!!

I quickly plugged the keyboard in and checked the RAM - it’s reporting 65535 for the first time ever!!

I’m too scared to unplug it again so I’m going to have a quick game of Manic Miner!!

[Spectrum65535]

1992 Sinclair Research Ltd? I suspect a hacked rom there