*** M E M O R U M ***

Another repair its a 4000-06 mainboard, it have clearly problems.

Not only is the graphics wrong, but also wrong letters came to the screen. Even I know I type REM it give me a mistake.

So the chip controlling the keyboard might be bad or something else.

The pictures are not so good, but i hope to get some clues from you guys.

I got the same problems with mine which were found to be VRAM issues

http://primrosebank.net/computers/mtx/r ... repair.htm

Hi Claus,

I agree with David, it's likely a VRAM fault, see this page too :-

http://www.primrosebank.net/computers/m ... _fault.htm

I never did figure out how the VRAM could affect the data bus, but it can as my webpage describes

regards
Dave

ok, then i need to identify the bad ic, it seems to be bit1 right ?

You can identify the bad IC by using this method

"The Service Manual suggests using an oscilloscope to check the signal on Pin 14 of any of the VRAMs, but this will not be an option for folks without a 'scope. The Service Manual also describes how to disable each of the VRAMs in turn by earthing Pin 14 (via a 10-15 ohm resistor!) and checking whether the screen corruption reduces, the chip with the least screen corruption when it is being earthed is likely to be faulty and should be replaced."

The resistor "trick" should only be done for upto around 5 seconds at a time...

You could also write a simple test program that writes known data bytes to the screen RAM.
Then by comparing what is displayed to what was written, it should give you a better idea about which bits of the data byte are corrupted.

This idea can be extended further if you have either a parallel printer or a serial interface and suitable terminal (e.g. a PC running a VT52 terminal emulator or similar). As well as writing to the screen RAM, your test program can read the previously written to addresses. Then send the written and read data values to the printer or terminal.

By carefully looking at the differences in the binary digits, it should be possible to work out which data bus bit(s) are wrong and therefore which DRAM chip(s) is/are faulty.

Mark

ok did a bit of research.
There is something with bit4. its always set to 0

means when i press the 1 key it returns the !

ascii value of 1 is 49 decimal or 00110001 binary. but since bit 4 cant be set it will return ! which is the ascii value to 00100001

same is for ascii values for P-Z, P is the first value to use bit 4, and since its always 0 it returns @

when I count bits I always use this:
76543210

so bit7 is 128dec. an bit0 is 1dec.

Does all this mean its also the vram ic controlling bit4 that is broken?

Crazyboss wrote:Does all this mean its also the vram ic controlling bit4 that is broken?

Yes, this is the most likely cause.

Any data that is stored or processed via the video RAM will be affected. However, the CPU RAM should not be affected, hence why the CPU does not crash.

If the CPU RAM was affected, the stack would most likely be corrupted and the CPU would crash.

With 1980's DRAM chips, each chip stores typically just one bit for each address. That's why you see them in groups of 8. In the MTX, the video DRAM chips are 16k X 1 bit. The CPU RAM is either 32k X 1 bit, 64k X 1 bit, or 256 X 1 bit.

So if there are no memory expansions fitted, with most failure modes of a single DRAM chip, it will cause a single bit error throughout the address range of the whole memory. So a video DRAM chip failure will most likely affect ALL the video RAM. A CPU DRAM chip failure will most likely affect ALL the CPU RAM.

Mark

You also have visual confirmation on the screenshots that bit 4 is stuck low.

With text mode using a 6 bit wide character, bits 0 and 1 aren't displayed.

Memotech used bits 3 to 7 for the character, bit 2 being the divider. The on screen display has nothing in any character in the next to last line, confirming bit 4 as suspect.

TI marked the data and address bus of the VDP backwards for some reason, so I _think_ it's the ram chip connected to AD3 and RD3 that needs looking at position 6F.

As far as I can tell from experiments with the MTXplus+ and looking in the ROM source, there's no line input buffer in main ram. Instead the data is pulled direct from the VRAM, so corrupted VRAM data will stop BASIC interpreting what you've typed in, even if is was typed correctly.

The MTXplus+ has the same issue with corrupted VDP data confusing the interpreter when the CPU is running too fast for the VDP.

1024MAK wrote: With 1980's DRAM chips, each chip stores typically just one bit for each address. That's why you see them in groups of 8. In the MTX, the video DRAM chips are 16k X 1 bit. The CPU RAM is either 32k X 1 bit, 64k X 1 bit, or 256 X 1 bit.

Is it possible to get a single 16k x 8bit chip and simple desolder all 8 and fit a single chip to take care of all 16k ?

At the Colelcovision they use the same setup, cause its the same VDP.

People like to change the 4116 to the 5v version, they said they drain less power and let the psu go cooler.

It should be possible to fit them too i guess I think its 4516 or something.