exxos blog - random goings on

[exxos]

This should make diagnostics a bit easier than with the first version, the idea seems to work though - seems to agree with the other test failing.

Maybe display the address range on screen as well ? Then got a better idea where in RAM the faults are.

But saying that. Maybe a different program which tests the ram and prints on screen then address and amount of RAM per "block" which fails. Like you said earlier, if its failing 1MB evey 4MB then just print it on screen.

[exxos]

In a bunch of those error messages it shows bit 8 (of 16) being wrong. e.g. ($0800 != $0000) and ($f7ff != $ffff)

So at the very least a bad data bit or line?

Yeah but the diagnostic cart only ever showed bit 11. So something is very wrong with something somewhere..

I'm not sure how the RAM is orientated in RAM :lol: on the STFM you have 16 chips 1 bit each. It makes up 512k. So when the other 16 chips are fitted, it uses a high bank CAS1H CAS1L. Then to add more RAM you have MAD9 in play for 4MB. At which point I assume its 1MB banks which are selected with RAS/CAS.

The falcon loses me :lol: I assume having more addresses and 32bit that its in blocks of 4MB then bank switches. So for there to be a 1MB chunk missing every 4MB.. I'd assume MAD10 is broken or a bank select RAS/CAS ... I dunno.. @Badwolf translate :lol:

EDIT:

Might be wrong.. CAS I think selects high/low byte and RAS selects the bank.. Need to get on the PC to look at the schematics :lol:

EDIT2:

So on the STE RAS0 is a pair of simms. Normally the first 2 you put in. Then CAS0L,CAS0H selects the simm. High or low byte.

Similar with 2 more Simms, CAS1L,CAS1H with RAS1.

Each simm can be 1MB. Forgetting CAS then as we not interested in 8 bit. So we would have 2 Simms, 1MB each, so 2MB block based on RAS0. The next 2MB in the memory map would be RAS1 block.

So back to the falcon... :lol:

I forget but think only RAS0,RAS1 is there. So assume it would be 8MB in RAS0 and 8MB in RAS1 kind of thing..

At that point there's no way it could be a RAS/CAS fault as its alternating good and bad blocks of RAM. So its like something ,I'd MAD8 maybe is broken.. Makes sense at 4am.. So I'll go back to sleep now..:lol:

[dml]

Here's an updated 'rampan'.

Actually it's two different versions, because it has no configurable settings.

rampan0 = clears memory, shows black pixels for any set bits on any bitplane
rampan1 = sets memory, shows black pixels for any cleared bits on any bitplane

Aside from this, it implements everything on last night's todo list including the keys. The nonresponsive + flashy colours at the start indicates it is busy either setting or clearing the ram. After that you can move around.


I didn't put any text on the screen because its a big job involving a raster screen split, a font, a number print function and Hatari can't emulate screensplits it so it also means remote debugging. So its not worth the bother - will need to be without for now.

The fact you can now hop to a specific MB and +/- by 256k at a time with left/right arrows should be close enough for horseshoes and the other stuff.

Anyway to refresh - the point of this is quite limited - to get a Videl's-eye view of RAM contents, independent of the CPU. It takes the CPU out of the equation. It's not meant to be precise, just another datapoint.

Getting exact locations of actual faults is best done a different way, with a program that scans and prints. I'll look at a better way of doing that because it's not difficult but I don't think its worth trying to integrate it into something like rampan.

[exxos]

Thanks @dml I'll give it a try when I get back home :thumbup:

[dml]

And heres a new Falcon ram fault scanner too...

It will scan STRam and FastRAM if available, for data bit faults and for address bit faults (address aliasing) starting with a bank size of 4MB, going down to 4kb in successive passes.

I had to test it with fake faults but I think its correct.

Needs to be run from STRam, but it's already set up that way. Don't change the PRGFLAGS. It will not respond to input until its finished.

Let me know how you get on.

[exxos]

Needs to be run from STRam, but it's already set up that way. Don't change the PRGFLAGS. It will not respond to input until its finished.

Let me know how you get on.

Just got home , so here is the first test....

IMG_0228.JPG

So back to bit11 faulty.... But not convinced as such as there was bad blocks before, not just a outright bit which had died :shrug:

[exxos]

Here's an updated 'rampan'.

Actually it's two different versions, because it has no configurable settings.

rampan0 = clears memory, shows black pixels for any set bits on any bitplane
rampan1 = sets memory, shows black pixels for any cleared bits on any bitplane

Using rampan0 is interesting..

Is it interlaced as the screen is sort of juddering.. Also it "sub-judders" every few seconds..

Anyway, using right arrow key..about 4 pages of back and white "garbage" then onto white... then about 50 pages in...

IMG_0229.JPG

IMG_0230.JPG

The "blocks" are flicking like a mono techno demo :lol: Then the same for a few screens.. then think it loops back to the start.

[exxos]

Using rampan1... first 3 or 4 pages mostly black white white garbage.. then leads onto solid blocks of white. Basically same as above but the inverse.

[dml]

Yeah if you use the keys 0-9/a-f it will hop to each MB (16 keys, 16MB).

The first one (0) will take you to the initial TOS state which looks like garbage. The last two (e,f) will take you to the HW regs and cartridge area.

Everything in between should be clear, solid fill, except for a few bytes maybe on page (d).

You can use left/right arrow keys to hop +/- 256k at a time, from each MB. Or use up/down to smooth scroll lines (but easy to get lost)

[dml]

Actually ramscan is reporting bit #12 as faulty 0x1000 on 16bit transfers, at least in that area of RAM. It doesn't complete all other tests if it runs into lots of data faults so there could be more after that. But anyway its bit 12, not 11 (bit 12 counting from 0, so the 13th actual bit). In case you get your logic probes out and attach on the wrong pin :)