The maximum credible accident in October 1997:

A TTL-IC SN7400 burned off with a red glowing fire and a strong smell.
After that, I've found at least three more malfunctioning ICs. From 
this syndromes you can gather that there is an external error for 
example at the power supply, that caused the destruction of 
multipe ICs.

I keep on searching for destroyed ICs and I hope that none of the ICs, 
which is inserted to core memory, is boken.

Unfortunately, I don't own a circuit diagram of the machine, only a 
maintenance manual of the PDP8-I. If someone owns a circuit plan
or just parts and would like to send it or a copy to me(of course, i 
would pay for the costs), would help me a lot.


========================================================================


News from the PDP8/L

Many colleagues and friends were interested in this report and 
wanted to know the condition of the patient. I was very pleased and 
that's why I make a progress report here:

Something about the phyiscal architecture of the machine:
The whole machine consists of a frame containing 4*36 plug-sockets.
Into this plug sockets, so called Flip-Chip-modules are plugged in. 
These are circuit boards with the dimensions 60*125mm(single height) or
128*125(double height). There are three different colored modules:


 White : Cable connections. (f. e. connection to the power supply and
	 to the core)
 Green : Analog modules. The read-/write-amplifiers as well as the 
         matrix decoder of the core.
 Mauve : Digital modules, containing 3 to 6 SSI-TTL-ICs.
	 The double-height modules contain up to 18 ICs.

Alltogether, the CPU contains 54 single-height and 9 double-height
M/Modules.
 
 
The Repair

I started with testing the most simple modules, where the ICs are not   
linked with each others.


21.10.97:

6 modules M111 (each 16 inverters), 6 modules M113 (each 10*2 NAND),
6 module M115 (each 8*3 NAND) tested.
There I've found 7 broken SN7400. It was the first time that it was
possible to save a value on the memory adress 0200(f.e. 7200 = CLA), 
that still was this adress after switch on/off-cycle, after changing
the ICs. Pressing the DEPosit- or the EXAMine-Key, the program-count is
not incremented but approximately doubled.  

At least the whole thing is an evidence, that the memory control did 
not suffer from the catastrophe.



8.11.97:

2 modules M117 (each 10*2 NAND), 5 modules M617 (each 10*2 Power-NAND)
tested. Found 2 broken SN 7440 and chenged them. No appriecable
improvement of the status.



11.11.97

5 modules M160 (each 3 AND-NOR Gates with 4,2,2,2,3 with 4,2,2,4 and
with 2,2 inputs. At this, found 2 mbroken SN 7450 and one broken 
SN 7453.
These types of ICs were not available at my own pool.
But: The Dräger company had them still stocked(Date of production:
1973 and 1974(!)). Thus almost in proper style.
After changing the ICs, there were real improvements:

The program
  ADRS	OPC
  0200  7200    / CLA  (Clear Accumulator)
  0201  7001    / IAC  (Inc Accumulator)
  0202  5201    / JMP 0201

can be executed with the restriction, that the accumulator-bit 09 has
the value 1, in the Single-Step Mode. A change of the rings of the 6 
registerslices doesn't cause a translocation of the error, what from 
you can see that the error does not come from the registerslices but 
probably from the frontpanel.

By the way, it is remarkable that always the inputs and the outputs
are broken at the damaged ICs.


15.11.97

The error did not come from the frontpanel but from a gate at the
TTY-Terminal (module M706). Besides the bit set to 1, there were
two more bits hanging on 0 after some more tests. Of course, this was
only recognizable with this program:

  0200 7200	/ CLA
  0201 7040	/ CMA Complement Accu 

This error came form a I/O-bus-driver M623, where again the two inputs
of two SN7400 were broken. Both the TTY-transmitter and the 
I/O-bus-driver are directly linked with the Q-outputs of the SN7474 
double D-flipflops, serving as accumulator.
Thus the outputs of these flipflops are not buffered, but from them the
feedback line is lead back to the corresponding complementary gate in 
the inner of the ICs, you can force the ICs to change its state by 
impressing a logical level, which has low enough impedance 
This happens asynchronous and has a higher priority than SET/RESET 
or CLOCK.


Now the machine is ready for executing the following program

  0200 7201	/ CLA
  0201 1100	/ TAD 0100  add constant from Adrs 100
  0202 4300	/ JMS 0300  jump to a subroutine
  0203 7001	/ IAC	    increment Accu
  0204 7440	/ SZA	    Skip if Zero Accu
  0205 5202	/ JMP 0202  jump there
  0206 5200	/ JMP 0200     and there

  0300 0000	/ return adress of the subprogram
  0301 5700	/ JMP I 0300
		/ A stack is something only for softies, a 
		/ real computer does not need a stack. 
		/ Here, the return adress is saved at the first 
		/ word of the subprogram.

What doesn't function is the serial interface, which consists of 
modules of double height (M706 (TTY-Transmitter and M707 (TTY-
Receiver)). These circuits boards are also only transistorized with
16-pole ICs((NANDs und doppel-D-FF).


22.11.97

The TTY-transmitter was really maggot-ridden: 7 of the 18 ICs were
measurably broken, two of them I changed by suspicion. The error 
detection is now eased by the fact, that the CPU works again and that
short 'Scope Loops' can be programmed and executed. These are short
program loops consisting of 2 to 3 commands, whose effects can be 
whatched at the oscilloscope.

Since today, the TTY-transmitter works again.


26.11.97

At about 20.15 o'clock, the TTY-reciever works again, too. The module
M706 with double height is transistoriszed with 17TTL-ICs. An input 
of a NAND gate was broken. Search- and change-time per IC: about 
2 hours, a time that also marks the average for the other damaged 
components. 
The second malfunction wasn't that obvious: The Preset-inputs of two
SN7474 are just blank at the M706 module. This is simply a raw 
design error, that by the way also occurs on the M220 register module.  
Every practitioner knew in former times, that a standard input of a TTL
is logically set to 1, that though the voltage level is located in the illegal
range and that open inputs are very sensitive for interfering signals.
In a short term: The TTY-reciever is working again from exactly this 
moment, when I impressed a 5V voltage on the inputs via a 1k resistor.
Beacause the machine worked for over 28 years without these resistors,
I suppose that the ICs were damaged either by deterioration or the 
incident, that destroyed 25 ICs, so that the blank inpurs were not set
to 1 any more.


27.11.97

Result: After changing 13 SN 7400, 2 SN 7402, 3 SN 7410, 1 SN 7440,
2 SN 7450, 1 SN 7453, 3 SN 7440 FOCAL-69 interpreter worked today!.

By the way: The letter combination SN is an acronym. What does it mean?


back to the PDP8/L-page
back to the computermuseum