ALCA PLANET - graphics failure

darthbob1425

Newbie
Credits
27CR
This machine is giving me a heaache now. With the help of DaveSpicer I have managed to get it running test mode. This involved replacing counters at row 5 (unlike the atari board these were LS191)

Got this
New_Picture.png


Then when you switch it into game mode I get this:

New_Picture_1.png


New_Picture_2.png


New_Picture_3.png


This line will then repeatedly flash but I shut it off to prevent damage to the monitor. Before I filmed it I though only the top portion of the screen was being drawn. NOw it looks like I'm getting through the first frame.

Base on what you were saying Dave do you think this is the HALT failing to reset and anyone with any ideas where to go next??

Thanks

Rob
 

DaveAS

User
Credits
336CR
Could be the top half of the vector RAM. Test mode only uses the bottom for vectors, whereas game mode double buffers with one frame in each half. First frame good, second frame bad would sort of point that way. The RAM test can still pass with marginal timing.

The R4+M4 pair are the bottom half and P4+N4 the top. Try swapping them over and see if you get different behaviour (e.g. swap R4 with P4 and M4 with N4).

There are a million and one things to try to be honest. It could still be a problem with the vector instruction counters or the muxes that follow them.

--
Dave
 

DaveAS

User
Credits
336CR
Lol.

On reflection, the RAM theory probably isn't right. You'd expect it to either crash or eventually flip back to the lower RAM when drawing the bright line. Either ought to cause another flash of sensible-looking vectors.

Something could be hitting the vector reset. As I discovered when experimenting, the board design is wrong and the reset actually ends up drawing the diagonal line of death (tm). Rev 1 ROMs might try to use it when a fault is detected.

--
Dave

davespicer2011-02-03 22:42:24
 

trm

Who loves you, and who do you love?
Feedback
2 (100%)
Credits
2,877CR
Hi Rob,

If you don't get much luck soon then I can probably swing an evening or weekend session over at your place and bring a crapload of test gear along. Might be easier fluking the board or at least using the logic analyser to ensure we're getting the right combination of signals at the right times. I even remember state machines from my Operating Systems 101 classes so maybe we can sniff, record and decode the VSM activity!

Plus it's all good practice for when I finally quit this contract and get a year out to work on my Star Wars project (shh!)
smiley2.gif
 

darthbob1425

Newbie
Credits
27CR
Tim thanks for the offer, I think it would be a great help. Part of the problem for me is that even if I have the kit I would be unsure of what signals/results to expect. I don't know how familiar you are with Asteroids but maybe Dave can throw some suggestions our way of the best places to look.

Ta

Rob
 

P-Man

International man of mystery
Feedback
4 (100%)
Credits
1,273CR
I'm with Dave's initial assessment. Atari's built in ram-test isn't perfect and can pass bad rams (on rare occasions), and speed is definately an issue with vector ram, you need to aim for -10 or -15 or possibly -20 types. You get this bright flashing line when the vector ram is not right. Swap it round if you can, or replace it with fast, known good ram. not all 2114's are the same, a lot are -40 or -30 and are not fast enough for asteroids vector ram, i have had mixed results with -25.
 

darthbob1425

Newbie
Credits
27CR
Following on from Dave and Andys comments over the ram I have tried swapping them about but with no change. I have looked up the test sheet for chips labelled D2114 which appear to be either -20 or -30, its not clear. I think all the other ram I have is of the slower varieties .

Anybody have 4 x 2114-10's that I can buy to try out, if it makes no difference at least I'll have spares.
smiley1.gif


Rob
 

trm

Who loves you, and who do you love?
Feedback
2 (100%)
Credits
2,877CR
Rob - my Dad pimped my last lot of 2114s but I've asked him to check the speed rating. If I can find some then I could probably drop them over this w/e as I now can't make it to the Slam (am scared that g00n3r will jump me!). Will PM you later when I know what's what.

cheers

tim
 

trm

Who loves you, and who do you love?
Feedback
2 (100%)
Credits
2,877CR
I tell a lie - he pimped my fast 2114s :)

I've got here:

6 x GTEu 2114-2CB (which I assume are 20ns)

6 x NEC uPD2114LC-2 (again, assume 20ns)

I'll see if Dad has any faster ones but I could always drop some of these off and give you a chance to try a set of matched -20s. Maybe some of them were ok at a higher speed but the entire batch got marked down as -20s for guaranteed performance.
 

DaveAS

User
Credits
336CR
I've seen several people say Asteroids needs fast RAM and can't really work out why. Cycle time for both the vector processor and 6502 is 666 ns and you ought to still have around 500 ns of that left by the time you factor in the 6502 timings. CS isn't gated with 02 which would otherwise kill half the access time. My board had 450 ns RAM from the factory by the looks of things - no sign it had been changed.

Not saying anyone's wrong, just that I don't understand it.

--
Dave
 

trm

Who loves you, and who do you love?
Feedback
2 (100%)
Credits
2,877CR
Aha. In which case my "more current RAM speed" has made me assume that the -2 is a 20ns which it's actually going to be a 200ns then Dave?
 

Equites

Chief Sheesher®
vacBacker
Feedback
35 (100%)
Credits
3,329CR
davespicer said:
My board had 450 ns RAM from the factory by the looks of things - no sign it had been changed.

Agreed, mine had 450ns RAM fitted at factory too, although since replaced with same manufacturer but rated at 200ns.

trmatthe said:
Aha. In which case my "more current RAM speed" has made me assume that the -2 is a 20ns which it's actually going to be a 200ns then Dave?

Tim that sounds right -2 is 200ns, fast enough!

trmatthe said:
Plus it's all good practice for when I finally quit this contract and get a year out to work on my Star Wars project

Tim you on drugs? Contractors don't take holidays or pull sickies for that matter
smiley36.gif


Equites2011-02-04 23:33:17
 

darthbob1425

Newbie
Credits
27CR
Some new LS191s arrived in the post today - so I have fitted them in place of the existing counters to be sure that they are 100% working - no change.

Never mind
 

DaveAS

User
Credits
336CR
Try giving this test ROM a spin. Burn in place of 035143.

http://www.sparcade.adsl24.co.uk/asteroids/035143.bin

You should see 6 short lines drawn on the left side of the screen. I've arranged the the test so that as long as the bottom 5 bits of the address are working, the vector processor shouldn't lock up due to an address error. Things that will still kill it are: a) failure of the carry from the bottom stage of the vector program counter, b) both banks of vector RAM enabled simultaneously or c) neither vector RAM bank enabled.

Each line can be white or grey and reading left to right, they represent the following:

1st - fault type (white = stuck low, grey = stuck high)
2nd - A6
3rd - A7
4th - A8
5th - A9
6th - A10

All lines are white - test passed
All lines are grey - can't detect fault
First line is white - any grey lines represent a stuck low fault (can detect multiple faults)
First line is grey - any white lines represent a stuck high fault (can only detect one fault)

A6-A10 are the address bits going to the RAMs. Even though it has one less lsb, the program counter is documented with the same numbering in the schematics. Note that A10 actually selects between the 2 pairs of RAM chips.

I've tested the code in Mame where I can frig the driver to create the different failures. It seems to work as advertised. No warranties, etc.
smiley1.gif


If you get the all grey result, probe A6-A9 at the RAMs and look for activity on all bits. Also check that the chip select is going active for both banks.

--
Dave
 
Top