Ideas on non-working Mega 4?

[troed]

Acquired a (known) non-working Mega 4 today. Won't have time to open it up and look at for two weeks, but I thought I should post what it looks like when booting for some ideas on the most likely cause :) It's somewhat interesting.

non_working_mega4.JPG

So, GLUE produces a stable image and the Shifter at least believes it's doing something of value. Not really sure what data it's getting fed with though (but I have a nagging feeling I've seen exactly this issue in a post somewhere, at some time).

The same pattern is repeated, each line, four times. The pattern is 40 bytes in size.

edit: Oh, it's not just a visual problem - it hangs like this. However, it does so at 50Hz so TOS has gotten a few things done first.

[PhilC]

One of mine did something like this with a bad blitter.

[frank.lukas]

test cartridge + null modem cable (9600Baud8N1)

[exxos]

Isn't that the pattern used by TOS when trying to size RAM ?

[troed]

Isn't that the pattern used by TOS when trying to size RAM ?

I was actually reading through the TOS code before to see if it was :) The below code could possibly do it, although it's not an exact 40 byte pattern. However, at this point screen memory is at $10000 and there's no way I can see for this code to overwrite that address.

Code: Select all

/*
 * We now need to size the actual memory present.  We test memory,
 * starting at 128K and increasing by 128K each time.  For each area of
 * memory, we start at the current address, and work downwards, storing
 * a series of values; we then verify that the values have been stored
 * correctly.  An invalid value (or a bus error when we try to access a
 * memory location) causes us to stop and size the memory accordingly.
 */

        lea.l     $00008000,a7
        movea.l   (_buserror).w,a4
        lea.l     st_ramsize_done(pc),a0 /* bus error means end of ram */
        move.l    a0,(_buserror).w
        move.w    #$FB55,d3           /* d3.w = difference between adjacent stored values */
        move.l    #$00020000,d7       /* d7 = address of first memory test zone & gap between memory test zones */

        movea.l   d7,a0               /* a0 = address of current test zone */
next_zone:
        movea.l   a0,a1               /* a1 -> memory being tested */
        move.w    d0,d2               /* d2.w = test value */
        moveq.l   #42,d1              /* d1.w = loop count-1 */
store_mem:
        move.w    d2,-(a1)            /* store a value */
        add.w     d3,d2               /* update the value for next time */
        dbf       d1,store_mem        /* rinse & repeat */
/* now check the values we stored */
        movea.l   a0,a1               /* a1 -> test area */
        moveq.l   #42,d1              /* loop count */
check_mem:
        cmp.w     -(a1),d0
        bne.s     st_ramsize_done   /* miscompare, ram invalid, exit */
        clr.w     (a1)                /* else zero it out */
        add.w     d3,d0               /* update for next compare */
        dbf       d1,check_mem        /* until we're done */
        adda.l    d7,a0               /* point to next test area */
        bra.s     next_zone
st_ramsize_done:                    /* bus error or bad ram  */
        suba.l    d7,a0               /* (a0 -> first zone that failed) */
        move.l    a0,d5               /* d5 = st ram size */
        move.l    a4,(_buserror).w

test cartridge + null modem cable (9600Baud8N1)

Yeah I must confess I don't think I have a serial cable ... :/

[troed]

test cartridge

Since TOS does manage to set 50Hz I quickly checked whether there was any difference booting a test cart instead, but no. This ... confuses me a bit.

[exxos]

This is where diagrom comes in useful, as it can test out RAM without TOS needed...

[troed]

Solved: Needs a new GLUE.

[troed]

trivia: Hadn't seen this hack before.

megast_psu.JPG

[PhilC]

Bridge rectifier the old fashioned way