PDP-8efm console switch logic failure -- eg, LoadAddress does not work

[AK6DN]

View attachment 999690

For those of you that have not looked at this circuit, here it is. Basically the 74175 is used in RS latch mode. The clock and data inputs are grounded. The CLR~ input has a pullup to +5V, and a daisy chain connection to ground thru all the momentary console switches. When a switch is depressed (like EXAM) the chain to ground on the CLR~ input is broken, so the CLR~ input is pulled high. Once the momentary switch makes, it forces the Q~ output of the 74175 to ground. This is expected to set the corresponding Q output high, which then is inverted into the priority encoder and decoded.

Problem is that this circuit does not work with all 74175 chip designs as was previously noted in an earlier post. It requires an unbuffered output (so that it is also an input). It also then requires an explicit circuit behavior within the 74175 internal logic, acting as an RS flip flop.

My analysis is that DEC could have replaced the 7404 hex buffer and two 74175 devices with three 7400 quad nands using them as crosscoupled RS flipflops. And then the circuit would have worked as expected using only documented behavior.

Of course the only worse design would have been what was done in earlier versions of the console schematic. There they did not do any debouncing of these switches at all, but just routed the switch output with a pullup into the priority encoder.

So it makes me think this console design was given to some noob engineer just out of tech school ...

Don

The above response was posted back in 2014 but it was based on something that happened to my PDP-8m back in 2006.
Basically the LoadAddress and LoadExtendedAddress switches stopped working.
All the others (Exam, Deposit, etc) worked fine, just could not load the console address.

Long story short, this week I had the exact same failure as I did back in 2006 on this exact same console board.
Ultimately resolved to a bad 74175 on the console board. The above post quote details the problem.
Back in 2006 I replaced the two 74175 parts on that board with TI 74S175 345B parts using sockets.
Luckily I still have the remainder of the tube of that specific part/date code that I bought, so pulled the
console, popped the two S175 parts out, popped in two new ones, and now the console is back to working again.
All the front panel switches (esp LA and LAX) work exactly as expected.

Here is a post from an online discussion I saved at the time back in 2006:

Subject: Re: PDP-8m Console Switch Problems - fixed!
From: Don North <ak6dn@mindspring.com>
Date: Sat, 09 Sep 2006 00:08:30 -0700
To: General Discussion: On-Topic and Off-Topic Posts <cctalk@classiccmp.org>
CC: General@priv-edtnaa06.telusplanet.net

Brent Hilpert wrote:
> I was curious ... pulled out a bunch of NOS 74x175s
> and did some tests. I noticed that the behaviour was
> sensitive to Vcc and ended up with the following table.
>
> The values 0<=n<=4 in the matrix indicate the number of
> flip-flops in the IC which worked as desired, so 4 is 'good'.
>
> <--> repeat across
> --> progression up
> <-- progression down
                                    Vcc
Unit MFG  DEVICE  DATE   4.1 4.3 4.5 4.7 4.9 5.1 5.3
---- ---  ------  ----  |---|---|---|---|---|---|---|

1.   TI      175  7340    4   4   4   3   3   3   3
2.                 "      4          <-->         4
3.                 "      4   4   4   4   3   3   3
4.   TI      175  7624    0          <-->         0
5.   Hit     175  6G46    4   3      <--          0

6.   TI     S175  7340    4          <-->         4
7.                 "      1      -->      3   4   4
8.                 "      0           -->         3
9.                 "      2   3   4      <-->     4
10.  TI     S175  7936    4          <-->         4
11.                "      4          <-->         4
12.                "      4          <-->         4
13.                "      4          <-->         4
14.  NS     S175  8742    4          <-->         4

15.  NS    LS175  8332    0          <-->         0

                                    Vcc
Unit MFG  DEVICE  DATE   4.1 4.3 4.5 4.7 4.9 5.1 5.3 5.5   Notes
---- ---  ------  ----  |---|---|---|---|---|---|---|---|  -------------
16.  NS      175  7923    4   4   4   4   4   4   3   3
17.  NS      175  8948    4   4   4   4   4   2   0   0
18.  FAIR    175  8001    4   4   4   4   4   3   1   0
19.  SIG     175  8202    4   4   4   4   4   4   4   4    Works in PDP-8m console
20.  TI      175  8705    0   0   0   0   0   0   0   0
21.  SGS   LS175  8241    0   0   0   0   0   0   0   0
22.  TI     S175  345B    4   4   4   4   4   4   4   4    Works in PDP-8m console
23.  SIG    S175  8421    0   0   0   0   0   0   0   0    Fails w/ no pullups

> Observations:
> - behaviour may differ between IC units even within valid Vcc range,
> - a given IC unit may change its behaviour within or near valid Vcc range,
> - 175 class may pass at lower Vcc and fail at higher Vcc,
> - S175 class may pass at higher Vcc and fail at lower Vcc.
>
> Interesting the way 175 and S175 devices differ in their response to Vcc change.

I didn't see any of the fail at lower VCC on any S175 parts, but I only
tested one sample from each date code. I have more parts, but I got bored :-)

> The group of TI S175s from 1979 did seem to show reliable behaviour over the
> entire range.

I found this to be true of the TI S175 date code 345B (which I don't know how
to decode) and the Signetics 175 from '82. Both were rock solid in the test
circuit and work reliably in the PDP-8m console socket with no add'l rework
(ie, no pullups required as with the previous S175 SIG part).

> One way or the other, the DEC front panel is relying on unspecified behaviour
> of the device. I wonder if the designers were just relying on old habits of
> setting flip-flops via collector triggering from the discrete days, and just
> got lucky that it worked.

I agree, this circuit is more complex than it really needs to be. The two 175s
and the 04 hex inverter packs can be replaced by three 7400 quad nand packs (plus
six pullup resistors) that implement simple cross-coupled latch debouncers.
Sometimes a designer gets a bug in their head to implement a 'cute' circuit
when really doing something more simple and straightforward is the better solution.
Of course I would never do that :-) , I'm only reporting that I've seen it done.

Anyway, for now it appears replacing the existing flaky S175 dc '76 parts in my
PDP-8m console with TI S175 dc 345B parts has fixed the problems, no add'l rework
or circuit hacks required.

Don