Flip Chip Boards Replaced

View Entire Thread	Return to Threads
SHARE
	by Bill Degnan - 06/29/2024 23:13
Bit accumulator 7 - R210 was found to be faulty. Note that the R210 in the photo is not part of the memory buffer row, so one should start counting 2nd from the right. When memory buffer 8 was set memory buffer 7 was also being set. Docs indicated that we should test where the switch register is transferred to the accumulator. Dave pulled the R450 clock generator (on memory console side) and put in his 9600b custom card, and his RS232 converter card, to hook up the PDP8 to his laptop for further testing. Dave downloaded DEC-08-LHAA-D and toggled in teh HELP loaded from 027: 6031 5027 6036 7450 5027 7012 7010 3007 2036 5027 Run & Send file. Memory buffer bit 2 fault found, and thus cannot load HELP loader. Next Dave ran the RIM and BIN loader. Sent tape for diagnostic. Put system to 7756 to start the RIM loader. Next Dave loaded the HELP loader and examined from 27 to confirm it was stored correctly. He RAN the program from 27 and the program failed. Next Dave tied running the RIM loader manually from 7756. When one runs this program it will stay running. So when you send tape it will act on it (watch the accumulator) and hit stop when it appears to be done. Next we loaded MAINDEC-801-1-PM loaded at 200. We put all switch registers up (all 7's) and Ran the program. It should stop at 201 with ACC = 0, 206 with ACC 0. Next Dave toggled in the "Group 1 Micro Instruction" test program from the toggle-in programs hand-written sheet. This test passed. Dave mentioned the BSW does not work on original model PDP-8's. Next, toggled-in test "Operate Instructions" - OK. Toggled in Group 2 Micro Instruction - passed. Fetch->Defer->Execute (FDE) testing check to see if the machine is going through all 3 cycles. To do se we set up a probe PB25. After probling Dave found that the program works. Dave said intermittant issues, perhaps a timing issue. Dave found that the R602 was not FDE-ing and we replaced P28 Start PC 0041 TAD Fetch MA 0040 MB 1410 A 0000 CONT PC 0041 TAD DEFER MA 0010 MB 4002 A 0000 CONT PC OO41 TAD EXECUTE MA 4002 MB 5555 A 5555 Still had issues. Memory corrupting randomly, setting to all ones when reading or writing. Could be write back does not occur. Suspicious that read is occuring but the write back is not, setting an address to all 111's. So next we tried the toggle-in echo program. failed. Memory corrupted. Wondering - could certain instructions trigger failure? Final thought JMP to increment accumulator is OK. JMP and IOT seems to be causing memory to go all 111's intermittantly. IOT is "talking to a peripheral" so as long as the program does not attempt an IOT it appears to be less likely to corrupt the memory location being read to be executed. Next - Do we see this corruption in other programs? Run a memory test to confirm that no errors occur, as this program does not attempt an IOT. What Else? Test all cards' forward voltage of all diodes" to 5.5 -> 6V. Transistors - emitter junctions .6, .8 drop when leader are in the right direction. You can use a diode tester for both diodes and transitiors. It's likely there are numerous faulty or marginal diodes and maybe some transitors that need to be replaced. Worst case, we'll improve the electrical state of the computer and perhaps fix the intermittant memory issues with a shotgun approach of going through all diodes on every FlipChip board. Backside of the same flipchip logic card. Reviewed this page https://homepage.cs.uiowa.edu/~jones/pdp8/man/registers.html Reply