48k 800xl

[Jeffrey Worley]

It seems I have an xl that doesn't properly select the ram under the OS?  It acts as though all is well and hangs when programs get to a certain point.  An example is the SIDE2 cart's Spartados X gives an error 179 (memory contention) when trying to load the Side.sys driver at boot.  This is exactly the error the 48k 800 gives ( not xl ).  So I have an XL behaving the same way.  I swapped PIA's and swapped cpu's with no change.  I re-soldered the socket the PIA is in as it showed signs of having been worked on.  There was once something soldered to pin six of the CPU, and there is a 104 capacitor soldered to one side of R97, with a wire leading to pin 5 (Luma) of the video port.

 

Do I have a bad MMU or is one of the 74 series glue logic chips bad?  Any leads would be appreciated.  I'd hate to have to desolder the whole board chip by chip to find the fault.  I'm thinking it isn't ram, most likely, probably something to do with bank selecting ram under the os?

 

Thanks,

 

jeff

[+Nezgar]

See if you can get a disk-based spartaDOS 3.2 to boot - it also requires RAM under the OS.

 

Has the original OS ROM potentially been replaced with an 800 compatible OS?

 

I guess there's other fiddling that could be done - like POKEing 54017 ($D301) bit 0 to 0 which disables the OS ROM, and would normally crash the computer without special measures taken. If it doesn't crash, that would confirm that the OS ROM is not turning off...

 

Edit: I have seen some posts showing 600XL memory upgrades that truly were only 48K, which are rare... What do you have inside the machine?

[+tf_hh]

5 hours ago, Technoid Mutant said:

Do I have a bad MMU or is one of the 74 series glue logic chips bad?  Any leads would be appreciated.  I'd hate to have to desolder the whole board chip by chip to find the fault.  I'm thinking it isn't ram, most likely, probably something to do with bank selecting ram under the os?

 

As Nezgar already wrote, a simple BASIC poke can check that. Boot the computer w/o peripherals attached to BASIC and enter POKE 54017,252. This keep BASIC enabled, but disabled the O.S. and enable (normally) the RAM under the O.S.´ space.

 

If BASIC remains usable (computer reacts to pressed key) and you´ve changed the PIA already, then it could be only the MMU. This is the one and only 20 pin chip found below the cartridge port. Replace it (it can be substituted by a GAL16V8) and this issue should be gone.

 

 

[Jeffrey Worley]

I swapped the MMU for a known-good MMU and get the same behavior.  I'm gonna try the ls138 next I guess.  I have to socketized a mess of chips.  Not a real problem as my desoldering iron is quick and I've done that like a million times.  I'll let you know what the problem turns out to be.

 

Another 800xl has a wierd problem.  It works great but for the Side cart's hard disks.  It displays some junk instead of the CF card information and won't actually display any directory entries.  The side loader gives a 'no media' on refresh, which is actually pretty normal, but several consecutive refreshes does not clear the problem, which is NOT normal.  In fact no number of refreshes will result in a mounted CF card and visible drives.  This one is socketed so I gotta socketize some more chips.  I don't have a stock of 375's ls14's ls08's, gotta get some I 'spose.

 

**TNM **

 

[Rybags]

Maybe the PIA is bad and the relevant OS bit is stuck "on".

 

Try POKE 54017,254

That should disable the OS - expected behaviour is that the computer should hang and the text screen should go blank or have random pixels.  If things continue as normal then the PIA is bad.

[flashjazzcat]

The only time you should see 'No media' in the SIDE loader is when there's... no media. It's not normal to see it otherwise. Severe Phi2 timing issues can cause this, among other things.

[_The Doctor__]

FJC what's my favorite chip to use for the PHI fixes? 

Edited August 12, 2019 by _The Doctor__
used Jon instead of FJC (Garfield was gonna eat my lasagne for that)

[Jeffrey Worley]

The xl with the wierd 48k problem is no chip on the motherboard.  I swapped every part on it, after socketizing them, with known-good parts.  The only things I have not yet tested are:  Basic Rom (:-)) and the ram itself.  Maybe I have bad ram.  Whodah thunk it could be that intuitive?  I'm going to socketize the ram and put some known-good stuff in and see if that fixes the problem.  Boy this machine is consuming sockets like crazy.  This will make like 18 for this board.  I had to modify some 16 pin sockets to make the 20 pinner and to make the 14 pinners.  all machine sockets.  They cut nicely with a pair of scissors if you are careful.

 

The other Xl, the one which would not recognize the CF slot, is fixed.  It had a bad(ish) 74ls08.  I swapped it for the one from the "48k 800xl" and it is working nicely.  Then I put that surplus genuine ICD Rambo 256k in with the gold-top ceramic Mitsubishi 256kx1 ram chips.  Nice.

 

Doing this gave me a spare 74ls158, which I needed to replace one I stole from an 800 ram card.  So everyone is happy.

 

** TNM **

/s

Edited August 13, 2019 by Technoid Mutant

[_The Doctor__]

double check those sockets and the traces/pads for colds/cracks/shorts/and breaks...

if that's all good and there is no OS chip mix up, then the support chips are the only place left to go, unless somehow you have a stuck bit in memory chip that isn't detected.....

 

after re reading, does this mean you've got them all working now?

if so scratch the whole paragraph :)

[Jeffrey Worley]

Well, I solved it.  After socketing every glue chip on the board, plus the mmu, I started on the ram.  It was either u9 or u10, 4164.  One of the chips is bad, and it occupied one of these two positions.  So those two chips are socketed now, with the bad chip replaced from my stock.  The machine works perfectly as a 64k 800xl again.  What an interesting fault that was.  Whoever mentioned a stuck bit is probably right.

 

Anyone know why this machine has, from the factory, the trace cut on pin 2 on each dram and a 100ohm resistor substituted for it?

 

** TNM **

Edited August 18, 2019 by Technoid Mutant

[_The Doctor__]

and just like that, RAM is still the number one failure point... using a third party RAM test is still the best way to go... they seems to ferret out stuck bits and slow ram much better than the cursory self test in the Atari

[JR>]

One of the Sys-Check boards from tf_hh would be a good investment. Very handy and versatile little device.  I love mine.

[toddtmw]

15 hours ago, JR> said:

One of the Sys-Check boards from tf_hh would be a good investment. Very handy and versatile little device.  I love mine.

Plus one on this. That board is the sonic screwdriver for Atari. 

[+DrVenkman]

1 hour ago, toddtmw said:

Plus one on this. That board is the sonic screwdriver for Atari. 

I could have used something like that when I was fixing my Ugly Duckling 1200XL, except for the pesky fact that it has no PBI or ECI connector. ? But the experience was a nice reminder of what can be accomplished with a multimeter, a logic probe, and lots of patience. 

[+Nezgar]

5 hours ago, DrVenkman said:

I could have used something like that when I was fixing my Ugly Duckling 1200XL, except for the pesky fact that it has no PBI or ECI connector. ? But the experience was a nice reminder of what can be accomplished with a multimeter, a logic probe, and lots of patience. 

I guess another option now in a pinch without a SysCheck is Shoestring's RAM & BASIC ROM tester. But less convenient as it would require swapping out the OS ROM chip(s). It also pinpoints memory failures to specific chips.