PDA

View Full Version : AT&T 7300 RAM repair



AmigaJules
December 30th, 2015, 06:51 AM
Inspired by this post from a few months ago:
http://www.vintage-computer.com/vcforum/showthread.php?49052-AT-amp-T-7300-Unix-PC

I pulled my 7300 out of storage to have another crack at fixing it.

It displays this image 28609 and has motherboard LED's 1(red) & 3(yellow) lit. According to the documentation that's a DRAM failure, I think?

I'm comfortable with a soldering iron and software, can follow directions, don't have any component level troubleshooting experience but very willing to learn.

Is there any way to narrow down which chip(s) may be bad? I'd hate to have to clip out and replace 72 of them. In the other post someone mentioned a logic analyzer?

Would these be suitable replacement chips?
http://www.jameco.com/1/1/1851-4164-150-64kx1-dip-16-dram-150ns-memory.html

Thanks for any insight!

AmigaJules
January 31st, 2016, 09:33 AM
Well, after snipping & desoldering 1152 IC legs, then installing 72 sockets & new RAM chips, I have the same set of diagnostic LEDs lit.

Any suggestions for what to check next? Map RAM?

cruff
January 31st, 2016, 03:32 PM
Any suggestions for what to check next? Map RAM?

Theoretically, it could be anything related to any address and data bus buffers, the DRAM multiplexing/referesh or address range selection circuits. Also power supply related, if the power supply/supplies are out of tolerance, you could see issues. Any obviously bad capacitors?

AmigaJules
February 1st, 2016, 06:11 AM
All voltages measure correct on the power supply. No obviously bad capacitors.

Chromedome45
February 1st, 2016, 09:24 AM
Well in my case it was the video ram. 6 chips if I remember correctly. But I only had led 2 on test 2 fail. Map ram would be fail on test 3. Led 2 & 1 on.

Ram type? In my case my 7300 had 41256 ram in it. If I remember correctly. Been a while! what was in yours originally?

AmigaJules
February 1st, 2016, 10:09 AM
It had 4164 chips. I found a datasheet for them online and replaced with same spec RAM.

I'm grasping at straws with the map RAM since bear said in that other thread that he was mislead by the LEDs since he had both main and video RAM failures.

When I power it on, I hear a click from what I assume are modem relays, then red (1) , then green (2), then red & yellow (3) stay on - I don't see it cycle through tests 3 & 4 unless they happen really really fast.

If anyone has a 7300 and can verify the LED startup sequence and timing, I'd appreciate it.

cruff
February 1st, 2016, 04:40 PM
I don't see it cycle through tests 3 & 4 unless they happen really really fast.

Use your smartphone and take a video of the startup. I had to do something similar years ago with an SGI workstation that wouldn't boot even into single user mode. The error message disappeared off the monitor. The solution was to haul out a video camera and tape the monitor during the startup sequence.

AmigaJules
February 11th, 2016, 11:31 AM
Use your smartphone and take a video of the startup. I had to do something similar years ago with an SGI workstation that wouldn't boot even into single user mode. The error message disappeared off the monitor. The solution was to haul out a video camera and tape the monitor during the startup sequence.

That was a great idea! By taking a slo-mo video of the LEDs on powerup I could see that it does indeed cycle through all of the tests and hangs at the RAM. Oh well... back to the drawing board on a fix.

bear
February 11th, 2016, 02:13 PM
That was how I had to do it.

I didn't leave good notes to my future self on the process of fixing my 7300, but I can definitely say that I had bad main RAM and bad video RAM both, and one of them caused the machine to stop with the grid on the display, and the other caused the machine to stop with the binary count (the fractal fern looking thing) on the display. On the grid, you can see pixels that are lit which shouldn't be (and possibly also pixels which aren't lit though they should be); on the binary count it stops filling when it hits an error. From these, it is technically possible to identify exactly which bit(s) cause the problem, but in practice I found it tedious and error prone. In the end I only got "close" and then had to replace a few extra parts than were probably strictly necessary.

You've got the grid there, and IIRC it shows first. I'm tempted to say, that's the main memory test.

AmigaJules
February 11th, 2016, 02:44 PM
From these, it is technically possible to identify exactly which bit(s) cause the problem, but in practice I found it tedious and error prone. In the end I only got "close" and then had to replace a few extra parts than were probably strictly necessary.

You've got the grid there, and IIRC it shows first. I'm tempted to say, that's the main memory test.

Yes all documentation points to it being a main memory problem. I replaced ALL of the ram chips (shotgun approach) but still have the grid so as someone said above it's likely the fault is in some other component.

You don't happen to have any documentation or hints on reading that grid pattern, do you? I haven't found anything yet.