DEC VT100 - characters misplaced in line

[RickNel]

I've recently got a couple of logic boards to replace an original that seems to have lost not only chargen but also can't read Non-volatile Ram (setup info). I suspect the program PROMs may have decayed, they are the same ROM chip as the chargen.

Anyway, VT100 now fires up nicely, responds to keyboard and comms, accepts and saves setup options, and forms characters correctly including AVO attributes.

But it messes up the placement of characters in a consistent pattern on every line.

The test command (ESC # 8 ) is supposed to fill the screen with 80x24 matrix of "E".

It actually displays 24 lines of only 56 characters.

I typed in sequential digits (1,2,3,...)to keep track of what was happening in the line formation.

Character pos Result

1-9 initially blank
10-35 appear normal
35-43 appear normal PLUS fill positions 1-9 with same characters
47-71 initially blank
72-89 fill positions 47-89
90+ no line wrap, each char replaces previous char at position 90.

So 40 chars of an 80 char line are not displayed - chars 1-9 and 47-71

Also, the cursor can be moved into the "blank" positions but displays there as default, without attributes.

In the "live" positions, the cursor shows attributes (block, blink, reverse).

Get the same result with two different logic boards swapped in and out, and also may have been present with the original logic board though disguised by other problems.

Does this sound like a problem in the analog display section? Or something buried in the settings? Hard to believe it would be an identical logic fault in two/three different logic boards.

Could it be a timing issue related to caps?

If logic, maybe an address issue in the display RAM? Something is not counting right.

Benefit of experience sought...

thanks
Rick

 

[Lou - N2MIY]

Rick,

It is indeed odd that the same problem occurs with both basic video boards (the main board). However, I am very sure that it has nothing at all to do with the analog video board.

I think that one of the line buffer address lines from the DC011 is broken. Signals LBA0H through LBA7H go directly from the DC011 video timing chip to the two 2111 line buffer rams. On the prints, these signals from the DC011 are in zone A3 of sheet 4. The line buffer rams are zones A-B/7 on the next sheet.

You might think that the buffer addresses would count sequentially from zero to the screen character width, but they do not. Cf. figure 4-6-11 in the VT100 technical manual. I think this is why nine character locations get screwed up instead of eight (you would expect some power of two of positions to be screwed up if the counting was properly binary sequential.) The rationale for this is in paragraph 4.6.2.7 .

How two DC011s would have the same address bit to fail is quite weird. The DC011 is socketed - would inserting it backwards kill the same bit?

At any rate, use your scope to look at LBA0H through LBA7H. They should all toggle. If any seem stuck, you have a bad DC011. Although they are unobtanium, all is not lost, for the missing signal may be able to be fabricated from the remaing ones with some gates provided by you. I have had this problem myself before and done just that.

Hope this helps,

Lou

PS. Do you have the same AVO connected to both boards in the trials you mentioned? What behavior do you see with the AVO disconnected?

 

[RickNel]

You were right as usual, Lou.

I confirmed the analog video was innocent by connecting an external monitor to the VT100's composite output. Same stable but messed up lines.

Scoping out the Line Buffer Address (LBA) lines on the DC011 timing controller chip, I found that LBA6 (pin 5) showed no signal on EITHER of the boards.

If the chips had been reversed in their sockets, that pin would have gone to socket 19 - just a TTL line taking the Vid Write signal. Could that have fried a gate? Hard to believe that the same thing would have been done to two chips.

I'm still wondering whether LBA6 may be being held low by some other combination of signals. The documentation says that the LBA6 signal can be used as a clock for vertical synch. I've got no vertical synch problem.

The Vid Write signal is used to write display parameters, including 60/50Hz refresh, into the DC011, but it also resets the LBA counters (Service Manual 4-55). I'm in a 50Hz power zone but powering this VT100 via a standard ATX PSU with an extra -23v source, so I don't see how the 50/60Hz would be visible to the logic board. The original PSU is pure DC output so how would it be transmitting 50/60Hz to the logic board? Could that be the problem?

In any case, it beats me why two chips would show the same address line failure.

LBA6 pulse is high for 9 times LBA0 and for 1/2 of the total line cycle. That might explain the 9-character blank at line buffer beginning, and the 50% character loss - LBA6 is correct half the time.

If the only solution is to remake a correct LBA6, how would I synthesize that?

Rick

 

[Lou - N2MIY]

Rick,

This will work to make LBA6 from LBA 4 and 5. Two chips.

So, remove the DC011 from the socket. Fold out the bad LBA6 pin. Wire wrap onto the Vcc, Gnd, LBA4 and LBA5 legs to bring out to your little daughterboard where the two chips go. Put the DC011 back in the socket with LBA6 leg hanging out. Bring the new LBA6 back to the socket, poking in where the old pin used to go.

Lou

 

[RickNel]

I took this offline to exchange a few PMs with Lou as we chased the problem down.

I can now report an interesting diagnosis and a simple cure.

We discovered that line 6 of the line buffer address (LBA6) was always low. This was the same even when the timer chip that generates the LBA6 signal was swapped out for a supposed good one, or was put in a different board.

What could be holding that one address line low?

Turns out the VT100 Basic Video Board (main logic board) has a trace that runs direct from pin6 of the Line Buffer SRAM 2111 chips (2 of them) to finger J of the power supply connector.

That finger supposedly carries a unique signal +12 GND. 5v and -12v and -23v each have separate GNDs, though they each terminate eventually to chassis.

The LBA6 signal used to have a function connected to power supply synchronisation, but not implemented on VT100. Maybe on VT52?

I have swapped my faulty original DEC PSU for a standard ATX PSU with added -23v rail.

Problem is, ATX has only one common GND, so the connection to finger J goes straight to TTL GND.

Rather than cutting any traces or synthesizing any signals, all that was needed was to disconnect finger J from my improvised power supply connector.

Instant perfect VT100 display!

Next I will cram the new PSU components into the space of the original, preserving the external faceplate with power input, fuse and switch.

Then I will have sitting on the bench one working VT100, one remnant original PSU and one messed up basic video board.

Not sure when I will find the mojo to sort out those relics.

Thanks, Lou!

Rick