Memopak : bad RAM chips ?

[amauget]

Hi to all,

During my tests with different RAM packs, I've got strange display with one Memopak 16 Ko.
The switches are in the master position (Off, On, On, Off).
Here is the boot screen :

Memopak 16 Ko.jpg (69.52 KiB) Viewed 5438 times

And with some text :

Memopak 16 Ko (2).jpg (90.88 KiB) Viewed 5438 times

I've just made a simple loop program and everything seems to work fine.
Could the display comes from bad RAM chips ?
Is there any short program to test RAM ?

[siggi]

Could the display comes from bad RAM chips ?
Is there any short program to test RAM ?

The ram test ist already done by the ROM during startup. If PEEK 16389 is 128, then the ROM "thinks" the 16KB are ok.

The strange display could be a result of something affecting the access to the character generator in ROM during the refresh phase (when the character patterns are read from ROM). So maybe something is wrong with the address decoder in the MEMOPAK, which might disable the ROM too long during refresh????

Siggi

[gozzo]

As I understand it, the 'boot-up' memory test works by attempting to load all possible RAM locations with 02 and then reading them to see when it gets to something that isn't 02 and then setting RAMTOP to that address. Maybe a fault could be possible which doesn't show up with an 02 bit pattern but may with certain other combinations.?????? I'm assuming by memopack you mean a Memotech memopack. You could try checking the rampacks internal -5v and +12v lines to see if they are ok...??

[1024MAK]

Are you using a Sinclair ZX81 power supply unit?
Do you have a multimeter?

The Memotech 16k byte memopack RAM pack further smooths the "9V" from the Sinclair ZX81 PSU and then uses it as the +12V to the RAM chips. This is possible as normally the output from the Sinclair ZX81 PSU is about +11V, and this is within 10% of the specified +12V needed for the RAM chips.
The Memotech RAM pack generates the -5V supply from one of the positive supplies via a switching circuit. This is again needs to be within 10% of -5V.

Mark

[siggi]

As I understand it, the 'boot-up' memory test works by attempting to load all possible RAM locations with 02 and then reading them to see when it gets to something that isn't 02 and then setting RAMTOP to that address. Maybe a fault could be possible which doesn't show up with an 02 bit pattern but may with certain other combinations.?????? I'm assuming by memopack you mean a Memotech memopack. You could try checking the rampacks internal -5v and +12v lines to see if they are ok...??

The ram works good enough that the Zeddy does not crash and shows a normal behaviour. During that the Z80 uses the ram to store the stack, to update the display file and the BASIC system variables. And everything works normally, so it's not a pure ram problem.

The error occurs only, then the character (stored in D-File) are "translated" into bit patterns to be displayed on screen. So something goes wrong during this "translation", which takes place in the refresh phase of the ram, where the character generator in the ROM is addressed to read the bit patterns from it.

Siggi

[sirmorris]

While I was developing ZXpand I saw exactly these patterns and the solution was to include REFSH in the chip selection logic.

I think Siggi is on to something.

C

[PokeMon]

I think +12V or -5V is used for refresh, so could be a problem with these voltages.

[amauget]

Thanks for all your suggestions.

Yes, it's a Memotech Ram Pack and I'm using a Sinclair PSU (1.2 A).
The 16389 location does hold 128.

I have verified the voltages (yes, I do have a multimeter and an oscilloscope too) on the pins of a RAM chip : -4.96 V, 4.84 V, -8.82V.
I'va also checked the voltages on the zeddy bus : 4.84v and 9.50v (with the Memopak plugged in) / 4.86v and 10.10v w/o the Memopak.

I was concerned by the -12v value so I've tested another PSU (0.7A). At first, the screen was clean but the problem appears a couple of seconds later, slowly, point by point, line by line (something I don't mention in my first message : the pixels on the display are moving. It's not a fixed pattern).

My last test was to shut off the PSU for a couple of minutes. On power up, again, the screen was clean. Then the "dancing pixels" came in...

[gozzo]

"-8.82v" !!! doesn't sound right!! ??? Should be about +11v to +12v, if it uses 4116 RAM chips - positive, not negative!!?? Strange!! Maybe I should get out my Memopack rampacks and see what readings they give!! (have 2 which definitely work and 3 others as yet untested)

[1024MAK]

Hi amauget

The "+12V" supply should be positive with respect to the 0V (ground/common).

The output from the Sinclair ZX power supply unit (PSU) does appear lower than normal.

What is the nominal AC mains voltage in your area? Is it much lower than the 240V that the Sinclair ZX PSU is rated for?

The RAM chips used need a minimum of +10.8V on the +12V supply pins according to the manufacturers data sheet (12V - 10% = 10.8V).

I'm sure that someone can point you to a good memory test program (that is a more comprehensive one than the Sinclair ROM memory test).

Mark