Thursday, March 11, 2021

Back home and starting on projects

 SUMMARY OF OLD INCANDESCENT LAMP REPLACEMENT FOR 1130 CONSOLE

The IBM 1130 computer has a display panel that sits above the console printer and shows the status of the major registers and important internal state of the machine, through a matrix of approximately 150 incandescent lamps that are individually driven by that many triacs. 


The earlier machines, such as mine, built small PCBs with 8 or 16 triacs that mounted 8 or 16 lamps. These lamps had to be pushed into a plastic matrix, but if any of the lamps was cocked at all then the assembly wouldn't seat. 

Original IBM design for lamp mounting and replacement

With six rows that are about a half inch apart, that meant six PCBs that had to be maneuvered to seat their lamps with little room to see or manipulate them. Thus, changing the burned out lamps became a tedious and time consuming activity. There was one column of six rows of the 16 lamp PCBs and two columns of 8 lamp PCBs, with a rats nest of wires carrying the roughly 150 signals plus power and lamp test wires to each PCB. 

Later models relocated the triacs to separate PCBs mounted on the rear wall, but there was still a maze of wires and lamps to be individually guided into the holes in the matrix. The bulbs were individual lead bulbs which were mounted in small nylon sockets that had holes for larger pins to press in. One pin was on the end of the triac and the other was wired to a common line on the PCB. Even the task of pushing a socketed bulb onto the pins was fussy work. 

I decided that I needed to build a more workable alternative for the display. Since each circuit had just a bulb, triac, resistors and wiring, with additional resistors, signal drivers, test switches and other components elsewhere in the machine, it would be easy to duplicate the functionality but using a larger PCB to sit across the face of the honeycomb matrix with lamp sockets at all the proper positions. 

Honeycomb matrix into which the lamps are inserted

Using smaller bulbs with a less unwieldy socket for replacement, I could simply guide a PCB onto the matrix to have all 16 columns by 6 rows of lights (or 8 columns by 6 rows) slide into place in one operation. I found a suitable triac, by which I mean one that will trigger reliably but could handle much higher power and voltage, that I could surface mount on the PCBs I designed. 

The AC power, common ground and lamp test leads are connected once to each of the big PCBs, while the circa 150 logic signals that are displayed push onto individual header pins on the rear of the PCB. The circuit would work exactly the same as the IBM original methods, but with a mechanically better design that is far easier to maintain. 

Test fit of one PCB onto the honeycomb

I soldered smaller individual lead incandescent bulbs to two pin headers that will plug into sockets on the face of my PCB at each light position. The one area that I remain unsatisfied with is the protection from the two leads twisting into contact with each other and blowing out a triac. On the original IBM design, this could easily happen and I lost a few triacs as a consequence, while wrestling lamps and the boards into place on the matrix. I tried some insulating goop but it looked ugly and was cumbersome to apply. 

original IBM bulb on right but went with left bulbs instead

STATUS OF THE PROJECT

Because of the remaining issue with bulb leads potentially shorting, I put the boards aside after building them. I didn't complete testing them because I then conceived of a alternative using full color LEDs. That would have the advantage of long life without bulbs burning out periodically requiring replacement. It had the technical challenge that reproducing incandescent bulb dynamics with LEDs is hard yet I wanted the console display to look just like the original while in operation. 

Recently I was contacted by a museum that is restoring an 1130 system. The console lamps where a mess due to the poor design and they were seeking an alternative. I decided to finish testing my design, so that if it worked satisfactorily I could send them the PCB design files and triac part number. 

TESTING SETUP

I want to test the brightness of the lamps, the ability to independently light them and the functionality of the Lamp Test switch that would cause all lamps in the display to light simultaneously. I chose one of the smaller PCBs to minimize the scope of the test wiring. The particular board I chose has 33 triac controlled light sockets, although 8 of them are for CE display lights I won't use initially so no lamp was installed in those sockets. 

The boards are powered by a 7.5VAC source with one of the leads tied to the logic ground of the system. Each logic signal that is connected to a triac is isolated with a 6200 ohm resistor, that resistor is mounted elsewhere in the machine. The triac gate is also connected through its own 6800 ohm resistor to the common Lamp Test wire. That wire is switched to either ground or +3V. 

The logic levels in an 1130 are 0 and 3V for binary 0 and 1 respectively. Thus, the driving gate in the machine delivers 3V or 0V through the resistor divider of 6200 and 6800 ohms to ground, making the triac gate either slightly over 1.5V or at ground. When the Lamp Test wire is switched to +3V, it makes every triac gate sit at a bit under 1.5 or up to 3V depending on the state of the driving gate. 

To properly test the board, I have to provide the Lamp Test wire at ground while connecting all the triac gates to what appears to be independent driving gates in series with a 6200 ohm resistor. I want many off and a few on, so that I can switch those wires around to prove that I can light specific lamps. 

I did this by using diodes in series with 6200 ohm resistors, three of them hooked to +3V and the remainder hooked to ground (with the diode direction set appropriately). This will appear to be 33 discrete driving gates delivering 3 or 0V. 

Once the individual lighting control is proven and the bulb intensity is judged sufficient, I will connect the Lamp Test line to +3V to verify that all the bulbs light up simultaneously. I soldered 33 resistors to 33 diodes, hooking one end of each of those pairs to a wire with a socket to fit my signal pins on the board. The other end of the pairs were wired together, 30 to a wire that will hook to ground and three together to a wire that hooks to +3V.

6.2K and diode to simulate each logic driving gate

I then needed to hook up an AC source. I chose my bench supply with a 6.3VAC, since in real operation that bulbs will be even brighter. If it works okay at 6.3 I should be good in the real world. Finally, I wired the wire for Lamp Test which I can move between ground and +3 power supply connections for that part of the test. The wire from the three leads that will light chosen lamps are connected to the +3VDC supply and the wire from the 30 dark lamp leads is hooked to ground. 

Power supply for 6.3VAC and 3VDC

With all that wired in place, I connected the leads with three lamps chosen to be lit, but obviously I can swap these around to check other lamps. The Lamp Test was set to off (ground). The +3V supply was turned on first and then the 6.3VAC supply was energized. The first tests judge lamp brightness and individual control by moving the leads around. The second test will move the Lamp Test wire to +3V and verify that all 27 lamps are illuminated. 

TEST RESULTS

Individual control - the lamps I hooked to +3V lit, the others did not. Moving the hot leads moved the lamps that would illuminate. Test passed.

Brightness - The bulbs seemed bright enough, but I inserted the test board into the matrix to see what the front panel brightness would be. I was pleasantly surprised, with the panel indoors with ordinary room illumination I could read them just fine. 

Test of illumination of T7, X7 and W(ait) status lamps

Lamp Test mode activated - when I first switched the Lamp Test wire from ground to +3V, the lights lit for a brief moment then the whole board went dark. I disconnected everything to do some diagnosing, hoping that traces didn't burn or something else fail from the current of all bulbs lighting.

I discovered that the pin for ground to the PCB wasn't well soldered. After touching it up, the board passed with both ordinary and lamp testing mode. I checked the brightness with all active to see whether voltage drop would make them unacceptably dim, but again this passed just fine. In the image, two of my bulbs were folded under and not inserted properly into the matrix, thus they are dimmer than the rest, but that will be solved by proper insertion. 

P2 and ADD lamps didn't insert into matrix properly, but all light up

REMAINING TASKS TO USE THE BOARDS IN MY 1130

  1. I have to wire these boards into the 7.5VAC, Lamp Test and Common Ground lines. 
  2. I must set up holders to hold the board up in the right place and against the rear of the matrix of holes. 
  3. The logic lines have to be pushed onto the pins on the rear of the PCB in the proper places. 
  4. I will refine the PCB design slightly to improve the power connections to each board as that was the weak link. The new PCBs will be shared with the other museum and I will probably rework the components from the old PCBs to my new design before I button up the 1130 display panel.

I am quite pleased with the results of this project. 

2 comments: