Friday, April 30, 2021

Aligner board not quite right, but should get it on the next revision


I received my poly foam board from the laser cutting site - 1/2" thick foam with the holes cut to allow the lamp holders to sit properly aligned for a quick easy insertion into the honeycomb matrix.

Poly foam aligner mask in place

First observation was that the foam is just too soft to push the lamp holders fully. It does help with some of the tilt such that they are very nearly correct even with this material.

Second observation was that with 1/2" foam the bulbs barely stick up out of the foam. I need some of the nylon lamp holder to be proud of the board in order to slide into the honeycomb cells and ensure the light is directly into each hole. 

The brightness was adequate and this could have been made to work, but the PCB could slide up, down, left or right of its intended position because the foam surface doesn't let the holder drop into each hole. 


It is time to take the design file and send it back to the fab, selecting a more rigid and less thick material. I believe 3/8" would be perfect for the height, but I will do more measurements before ordering the new board.

Sunday, April 25, 2021

Created design for the aligner board and sent to fab


My original design for the aligner board was to have truncated pyramid shaped openings that would gently bend the various lamp assemblies into alignment as it was slid down to contact the face of the PCB. However, since the aligner is placed on the PCB while it is laying upright in front of the maintainer, it did not have to accomplish the bending.

Instead, the person putting the boards together can touch and move various lamp assemblies as needed while the aligner is tilted into place, first working on one edge then tilting the aligner board increasingly parallel to the PCB as it engages more lamp assemblies. A pencil eraser or similar tool is all that would be needed to move the errant lamps into position, column by column. 

This greatly simplifies the production of the aligner board, since a simple rectangular opening is all that we need. No bevel cuts or sloped cutting/adding are required. 


Based on the relaxed design approach, I decided that laser cutting was the right choice for me. The material had to be around 1/2 " thick to provide enough alignment for the assemblies, not extremely hard but not so soft that the lamps flop over out of alignment. 


I drew the laser cutting paths using Inkscape, duplicating the board since my template was large enough to produce two copies of the part. I can send the other copy to the museum that is replicating my lamp system for use in the 1130 they are restoring. 

Laser cutting pattern


I chose one day manufacturing and overnight delivery since the total cost was very reasonable. I picked a 1/2" polyurethane foam for the material, which I hope will be stiff enough to maintain the alignment. We will see when it arrives, then at worst case change the material and send the job off again to the fab. 

Thursday, April 22, 2021

Almost there! Fix one position and move two bulbs


With essentially every lamp holder in place, I tried to slide the PCB with lamps right into the honeycomb matrix. I still had to do a bit of wiggling and playing with holders that were skewed a bit, so this still doesn't meet my expectation for a rapid and effortless insertion, but it is much closer than any previous incarnation including the original IBM mechanism. 

Test fit of PCB with new holders behind panel


Using the Lamp Test signal to turn on all the lamps, I had a chance to check that all the bulbs I installed were working properly. As you will see, two of the CE positions are dark because I was short two bulbs from having a fully populated board, but in addition the lamp for Index Register 2 is out. It doesn't light with a direct signal input either, causing me to suspect that it is an issue with the newly installed bulb in the holder assembly.

CE 4, CE 5, and IX2 are dark (for different reasons)


If you look closely at the left side of the control panel, where the six major 16 bit registers are displayed, you will see that the leftmost (bit 0) lamps do not shine through for the top two rows. These are for the Instruction Address Register and the Memory Address Register. That is consistent with a system that can only address a maximum of 32K words, thus needs only bits 1 to 15 to contain all possible addresses. 

Since the machine does not contain signals to drive bit 0 of IAR or MAR, and the panel does not have an opening to allow the light to shine through on lamp test, there is no reason to populate those two positions. 

That gives me back two lamp assemblies, exactly what I need to fill in the missing two positions of the CE signal area of the panel. I don't need to wait for a new supply of 2114 bulbs to finish this up and install it in the IBM 1130.


First, the two lamps from IAR bit 0 and MAR bit 0 will be moved to the CE signal area. Second, I need to check out the IX 2 circuit and lamp, repair it and retest to demonstrate all 162 lights active simultaneously.

Beyond that, I am exploring ideas for an alignment honeycomb that will slide over the lamp holder assemblies on the PCB, holding them upright at the correct position, so that the sandwich of PCB and the alignment part will slide easily into the display panel honeycomb matrix. I envision something produced with a laser cutter, material to be determined but likely some kind of plastic. 

That alignment part will be slid over the lamp assemblies while they are facing up and at the maintenance person, who can easily manipulate them into position to have the alignment part touching the PCB face. I have some measuring and noodling to do before I design and fab this out, but I suspect this will be the way to make this control panel upgrade a complete win. 

Wednesday, April 21, 2021

Missing a handful of bulbs to complete all 164 positions of the 1130 Console Panel


The supplier who told me that my bulbs were on order and would be shipped next week, followed up today with an email informing me that the bulb is discontinued by the maker (which I knew) and therefore they were refunding my purchase. Drat.

I hunted down some other suppliers who claim to have stock on the 2114 bulbs and have ordered another 20 from them, but in the interim I discovered a box of 20 bulbs that I owned. Between the 100% success rate I had reusing the bulbs from the old holders and these new bulbs, I was able to complete all but two lamp assemblies. 


I recently arranged a sale of a Brainiac K-30 as I slash the size of my collection of retro items. This is a derivative of the Geniac, a kit from the mid 1950s that was the introduction to boolean logic and digital computers for children of the 50s and 60s. 

The kit was a large masonite board, predrilled, and six round masonite dials that were also predrilled. These dials formed rotary switches using bolt heads as the contacts on the main board and staple-like jumpers on the dials. You could make various switches from the two pairs of rows and sixteen rotary positions. A battery and flashlight bulbs in sockets completed the active components of the kit.

Quite primitive, but through a series of increasingly complex sets of switches you wired up, solving more and more complex issues, the user could experiment with boolean logic and experience how digital logic could address a wide range of real life problems. 

The weakness of the kit was always the quality of contact between bolt heads and jumpers as you turned the dials to various positions. That weakness remains, but added to it is more than five decades of tarnish on the brass parts.

The buyer of the kit is a museum, to which I impulsively offered to wire up a three bit full binary adder to show the kind of problem this could tackle. Given the six dials that came in the kit, that was about the largest adder it might handle. 

Seemingly an easy task, I discovered anew the difference between digital logic and these switches. No resistors, no diodes, no inverters and in fact a signal did not operate in binary. That is, there was not a 0 and a 1 level on the signal. There was a connection, or there was no connection. No such thing as a 0 level connection. 

The boolean equation for a binary adder stage (except the rightmost) is A xor B xor CarryIn and there came my first design challenge. Implementing an XOR is a matter of wiring a switch on one dial to a switch on the other, but criss-crossing the wires so that a connection is made only if the dials are 0,1 or 1,0 but not if they are both 0 or both 1. 

However, how do you wire another XOR to that when you have one wire when the CarryIn is active but nothing when it is false? Instead, I had to produce two different signals from the stage to the right - CarryIn and InvertedCarryIn - where I had voltage on one of the wires but not the other at any time. 

Then I had to wire up the XOR driven by InvertedCarryIn to produce the sum of the bits for the case where no carry occurred to the right, which indeed sets the output to 1 if the two dials are 0,1 or 1,0. 

When a carry has occurred to the right, the output is an inversion of the XOR such that we get an output only when both dials are 0 or both dials are 1, but not when they differ. This is wired as a switch on each dial with the 1 positions connected together and the 0 positions connected with a different wire. That means four total switches are used to light the output lamp for a 1 value for this adder stage. One pair is powered from CarryIn and the other pair is powered by InvertedCarryIn. 

The boolean equation to calculate CarryOut and InvertedCarryOut were:

CarryOut =                                 (A and B) 


                                       ( (A xor B) and CarryIn )

InvertedCarryOut = (     (notA or notB) and InvertedCarryIn    )


                                     (    (A or B) and CarryIn    )

That required another eight switches, four on each dial, for a grand total of twelve two position switches for each binary adder stage. The rightmost position was easier because it had no carryIn to consider, thus I only wired the switch paths that were fed by InvertedCarryIn, which was itself hooked to the battery. 

The kit had an almost full complement of parts such as bolts, nuts, jumpers, bulbs etc but there were not quite enough of the jumpers to populate the third binary stage. I thus had to settle for a binary adder that took a pair of two bit values and produced a two bit output plus a carry signal. 

Building this and testing all the wiring was a bit more work than I initially had imagined, but once I was done it was time to deal with the poor quality of the connections. I tried to use deoxit spray and ensure the best fit possible, but the result was as disappointing as I belatedly remembered from my youth. 

The bottom line, however, is that I did wire this up to produce a full adder for a pair of two bit binary values. Satisfied, I documented everything and shipped it out today.

Tuesday, April 20, 2021

More than halfway done building lamps into holders


Since my supply of new 2114 miniature incandescent bulbs is going to be delayed at least another week or so, I am reclaiming bulbs soldered onto my prior approach to a lamp holder - wire leads wrapped around a 2 position header strip then insulated with goop since the bulb can bend and twist bringing its wire leads together. 

Once the bulb is removed from the old header, it is cleaned and ready to solder into a new holder. To construct the holders, I cut two pin segments from a new header strip, insert one end in a breadboard to act as a jig, and snap the nylon holder down over the header segment. This creates a holder which has the two pins separated by a nylon barrier, two side arms that nestle the T 1 1/4 size bulb in place, and allows access to solder.

Fortunately, the bulb sitting in the holder needs only a very short segment of wire lead and it can reach the top of the header pin. Something in the range of 1/4" is adequate. That allows the reuse of the old bulbs since I can cut away twisted ends without loss of any bulb. 

The goop makes this desoldering much harder than if I simply had to heat and remove leads that were wrapped around the header pins. I make use of a large tip on my Weller soldering station to do the removal, transferring heat through the goop but gumming up the tip quickly. Lots of retinning, tip treatment and cleaning is needed as I work my way through more than 150 of the old bulb assemblies. 

Installation of the cleaned bulb in the new holder is quite easy. I use my thinnest tip on the Weller station, push the bulb down within the holder arms, push the lead in contact with the pin, apply the iron and solder the connection. A quick flip of the holder and I repeat for the other bulb lead and header pin. 

I set up a test station on the breadboard connected to my ohmmeter. Popping each new lamp into the station tells me if this is an open circuit, a short, or the bulb is displaying circa 10 ohms of cold resistance. I know that every lamp I build is working properly before I install them on the PCB.


I have cycled between the three process steps - cutting and snapping header strip segments into holders, removing bulbs from the old assemblies, and soldering the bulbs into the new assemblies. This is tedious work due to the need to build 164 lamps for the fully populated IBM 1130 console panel. 

Partially populated board, unfinished holder below

I need to reclaim 66 more bulbs from old assemblies, move them into new holders, and then when my bulb supply arrives I can finish the last 16 lamps which illuminate the unused CE (Customer Engineer) and Synchronous Communications Adapter (SAC) positions. 

I don't have the SCA on my machine and the CE lamps are set up to display diagnostic signals by the use of jumpers when the CE is working on the system. I have better logic analyzer and scope tools, thus am unlikely to make use of the 8 CE indicators. 


From time to time, I check how difficult it will be to slide all 164 bulbs into the honeycomb matrix behind the display panel. So far, it has been pretty easy to wiggle it a bit and have all the lamp holders slide into their cells. 

Monday, April 19, 2021

Received lamp holders, beginning to build lamps for the IBM 1130 control panel

 VERY SATISFIED WITH THE NYLON LAMP HOLDERS is the manufacturer of the nylon lamp holders I designed with Autocad Fusion 360. The holders arrived today and I am very pleased with the quality of the holders that were produced. The prototype I had built by, the fab house associated with Autocad, was rougher and they only offered black. The new white ones very good and totally satisfactory for the job they will do.

Counting the lap holders that arrived today


I had cut apart standard sized header strips, although with round rather than rectangular shape pins, to form the two pin segments that fit inside the holders and mate with the sockets on the PCB. They snap into the nylon holder with a satisfying click, producing very good alignment when plugged in. 


I bought a supply of the 2114 miniature wire lead incandescent bulbs from but will have to wait one to two more weeks to receive them, apparently. I placed the order on April 12 and according to the web site, in stock items ship within 24 hours unless noted at time of order. No notification either at order or the next day when the email status informed me they were 'in fulfillment'. 

As it was a week later with no indication they had been shipped, I called to speak with the customer service group. He told me they had not been in stock and were on order. I asked him to check on an estimate of arrival from their supplier and he indicated it would be a week from today. 

The delay is reasonable and understandable, the only complaint I would have is that nowhere on the order status or in the various emails I received from the company did they inform me these were not stocked and going to take two weeks before shipment. Lack of communications left me uncertain and concerned. 

To move ahead, I decided to begin removing the wire lead bulbs from the current holders, which are just the header strips without any support structure. I did daub some insulating goop on the leads to protect against shorts if the bulb twisted to bring the leads into contact. That makes the removal a bit messier but the process is simple - just multiplied by 150+ to pull them all off the existing headers. 


The process of installing the bulb is slow, requiring care to position the bulb inside the arms of the nylon holder then manipulate the leads one by one into contact with the pin for soldering. It will take a couple of minutes per holder, thus about five hours in total to mount them all. 


I installed the first ten lamp holders with bulbs into spots spread across the PCB positions fairly evenly, then did a trial fit of the PCB with lamps into the honeycomb matrix. It seemed to slide into position very easily. I am becoming quite confident that this will provide the ease of insertion and removal I want to have a truly maintainable console light assembly.

Thursday, April 15, 2021

Glacial movement of supplies while I bide my time


I had a quote from to produce 165 holders for an attractive price, with a production completion date of April 14 and delivery by the 21st using a medium priced shipping option. 

I was contacted a few days ago by customer service letting me know that the manufacturing partner informed them that the job will take an extra day to complete - April 15. As a consequence, upgraded my shipping at no charge to 1 day which was more than fair.

Today, when the parts should be shipping, the status remains 'in production' and not yet moved to 'quality inspection'. Hopefully they will not be further delayed. On the other hand, having shoddy unusable parts early is not a good alternative; happy to wait a bit to get good parts.


The bulbs were ordered from on April 12, listed as in stock and per their website should have shipped within 24 hours. It is now April 15 and the order is still showing 'in fulfillment' status. I can theoretically remove all the bulbs from my existing header strips and remount them on new strips in the plastic holders, but I won't have much tolerance for any that are damaged or can't be remounted. 

Friday, April 9, 2021

Experimenting with the 3D printed lamp holder prototype


I had whip off a single lamp holder for me, allowing me to test its fit and functionality before I ordered 164 of the final versions. This site offered a relatively quick turnaround, although at a higher per unit price that other fabs would charge. I would tweak the design, if needed at all, then order from a more cost efficient vendor. I also want the final product to be in white nylon, whereas the prototype only had a black nylon option.

Bulb fits between two upper wings

Opening in bottom for two position header

View of top which will fit into honeycomb cell


The process of building a lamp starts with the insertion of a two position header strip through the bottom of the holder. The strip must have round pins in order to fit in the round sockets I put on my PCB. I can use a breadboard as an anchor for the strip (and the holder above it) for the next step.

Header inserted into the holder

I then took a 2114 lamp, which has wire terminals, and placed the leads around the header strip upper pins. When the bulb is situated properly inside the two protective side wings, with the leads wrapped on the pins, I can carefully solder the bulb in place. 

Bulb in place but wires not soldered to pins yet


I then plugged the test lamp into the PCB in various positions, some with absolutely perpendicular sockets and some where the socket is tilting to the side. That would give me a sense for how sensitive the system will be to tilt, since the holder ends have to find the cell and slide into the round hole inside the cell of the honeycomb matrix. 

Final check was sliding the holder into the cell just to see whether it might hang up on an edge, requiring more curvature of the part, or if it aligns itself automatically upon insertion. 

Holder in place and bulb lighting 

Placed behind the honeycomb matrix and the lamp illuminated


While unsoldering a bulb from the header strip prior to re-using it in the holder, the plastic of the strip got soft and the pins bent sideways slightly. This left the pins at an unexpected angle, tilting in the direction of the longer edge of the holder. I was only expecting alignment issues when the sockets on the PCB were not perpendicular, leaning the holder in the direction of the short edge. 

The pins on the header strip only project out of the holder base a relatively small amount, less than I imagined. This reduces the space available to coil the wire leads and to get the soldering tip in place. I believe it is workable as it is. If I were to make a relief cut on the sides, there would be less bracing for the header strip which would aggravate any leaning tendency. 

I decided that I should straighten the sockets on the PCB as much as feasible, just to improve alignment, but don't find them critical. The holder slides easily into the opening of the honeycomb but does NOT slide into the hole. That causes the completed assembly to stand back a bit from the honeycomb, but eases the alignment requirements substantially. I think this will do just fine.

With that determined, it was time to place the volume order through 3D Hubs which knocked the price down to $0.91 per holder delivered. I expect them to arrive no later than April 21 at which point I will install all the headers and bulbs to finish the project.

Wednesday, April 7, 2021

Tested brighter 8610 bulb against 2114 - noticeably brighter but 4X the current


I have supplied the board for the IBM 1130 Console Display with 2114 incandescent bulbs, 6.3V 50ma parts. The total consumption of the board is relatively low as a result and the console lights show through just fine in a normally lighted room. 


In the same form factor and voltage rating, there is a higher power bulb - the 8610 - that consumes 200ma but is brighter as a consequence. If I used these exclusively in the console, the power requirement would zoom up to 232 Watts and draw a bit over 32A whereas the current bulbs would eat under 60W with correspondingly lower current. 


I wired up one of the 8610 bulbs and installed it next to a 2114 bulb, allowing me to visually compare the two in terms of brightness. Indeed, the new bulb is noticeably brighter but it comes at a heft price in terms of power. 

2114 on left, 8610 bulb on right


Because the brightness is sufficient with the bulbs I already used, I decided to keep them in the machine. I will make the additional 14 lights to fully populate the board using the 2114 bulbs. The 8610 remain as an alternative if I ever find that the brightness must be substantially stepped up. 

Minor issue spotted and corrected


Lamp Test failed for the Accumulator Extension Register bit 10 position. It was not the bulb, thus I needed to inspect this further and repair whatever is wrong. It wouldn't light with either the Lamp Test activated nor with a +3V signal on the input pin. 

That suggested that my problem was likely another bad solder joint. Having selected 2 ounce copper pour for my PCB, I had to deal with the higher thermal mass of the pads. It took a longer time applying heat for the pad to flow solder, while the gate lead of the thyristor would flow much earlier. I ended up with a couple of heaps on the component lead that didn't attach to the pad underneath.

Reflowing the joint fixed the problem immediately. I verified that both the signal input pin and the Lamp Test control would cause it to light.

Tuesday, April 6, 2021

Almost fully populated board tested with adequate transformer I received today


I had ordered at 6.3VAC transformer rated at 6A, which is big enough to handle the 7+A of the full board for short periods. I wired up the primary to a plug, popped it into the wall and verified the voltage of the secondary was 7.6 VAC; a bit hot but even better as a test since it draws a bit more current than we will in the running 1130 system.

Transformer rated at 6A

Wiring the transformer to the board was next on the task list - tying one side to the common of my 3V DC supply while I was at it. A quick activation to ensure nothing bad was happening, and then I could turn on the Lamp Test switch.


Due to a shortage of bulbs, until the new supplies arrive, I didn't have bulbs for every position. I got close, illuminating two of the Synchronous Communications Adapter (SCA) positions but leaving the rest of the SCA and all of the Customer Engineer (CE) positions empty. Every other position that is active on an IBM 1130 had a bulb in place.

I flipped the Lamp Test switch, which injected +3V to the control line on the board that caused all of the thyristors to conduct and illuminate their bulbs. Prior to this I had hooked up three random signal inputs to +3 just to have a brightness to compare against. The result was a bit of dimming but the transformer voltage sagged less, only down to 5.76V. I could see the current jump up initially then drop to just under 6A. As I added in the last few rows of bulbs, the final current seemed to have settled to that same level suggesting that the transformer is saturating.

Lamp Test mode

Comparison of three bulbs fully lit

The board remained cool to the touch, as did all the thyristors and the wires bringing the AC current onto the board. I left the Lamp Test switched on for quite a while, with everything remaining as it was. I am now happy to consider this board fully tested and ready for its eventual connection into the IBM 1130 computer. 

Current with three bulbs lit, full voltage delivered

Current with 147 bulbs lit, noticeable sag in voltage

Sunday, April 4, 2021

Measured bulb current draw, pleasant surprise, but purchased junk ammeter


I bought a 20A ammeter through from Uxcell, claimed to have a proper internal shunt and accurate to a reasonable degree. With the ammeter in the circuit, even with 44 bulbs installed it barely registered 1/2 amp draw. However, I was suspicious about the correctness of this given the dimming that happens with that many bulbs plugged in. I am returning it to Amazon since it doesn't work properly.


My VOM has an AC current measurement but a max range of 200ma, not very useful for measuring more than a handful of bulbs, but it did give me a baseline for comparison. With four bulbs inserted, I measured almost 7V from the supply and 172ma of draw. 

The defective ammeter on the other hand is off by a factor of 4 to 5x. As a second check, I grabbed a second VOM, one with a higher range of AC current measurement, and put it into the test circuit with 10 bulbs installed. I should see 400+ ma of current draw if my other measurements are correct. I recorded 440ma with the unloaded voltage of 7.06 dropping to 6.89VAC when the lamps were lit.

Current consumed by 10 bulbs lit by Lamp Test

Populated the board with just 10 bulbs to do the measurement

Voltage while supplying 440ma to light the lamps

Unloaded voltage of the AC supply

Gate draw is going to be pretty negligible by comparison to the lamp current. The resistor voltage divider to each gate consumes about 0.2ma of current, 34ma for all 164 circuits simultaneously conducting. The thyristor adds some microamperes to the draw, which we can ignore. 


That suggests that at full supply voltage in the IBM 1130 these will draw about 40-50ma apiece when lit. All 164 positions, if the same type bulb is used, would only draw around 7 amperes, well within the capability of the new beefier transformer that will arrive Tuesday. 

Some T 1/4 size bulbs have a spec of 200ma, others 50ma, and it is possible that I have a mix of both types in my supply. I have been measuring their cold resistance (10 ohms approximately) to verify lack of shorts, but I decided to set up a jig to test all bulbs for current draw and sort them into groups.

The consumption of the bulbs directly wired to the AC feed is different from their draw in circuit, in part because of the approximately 1V drop across the thyristor while it is conducting. I think that IBM specified 7.25VAC for its lighting circuit with that drop in mind, yielding about the nominal 6.3V at the bulb that matches the bulb specifications. 

The measurements confirmed the expectation that the entire board will consume between 7 and 8 amperes depending on the exact voltage inside the IBM 1130, with all lights illuminated simultaneously. 


If I determine that I have about 150 bulbs that draw 50ma and add 14 that draw 200ma, that would bump the board consumption up over 9 amperes but still within the short burst capability of the transformer that will arrive soon. I should look for a source of bulbs at the proper current, since each burned out bulb will force me to bump up the board load by 150ma, which would scale up the draw quite a bit over time.

As an alternative, if I could find 2114 bulbs, which are the type I already bought, that would be a preferable component to use with my console. In the end, I did manage to buy a couple dozen of the desired bulbs and thus can do a full up test of the board with Lamp Test once everything arrives here. 

Saturday, April 3, 2021

Finished first version of bulb holder design for 3D printing


This holder is designed to take a two position standard (.100" pin distance) header strip in the bottom, with the longer pins sticking up through the holes with a barrier between them. A wire lead incandescent bulb is placed astride the barrier, its wire leads wrapped around the two pins and soldered in place. The side walls keep the bulb protected and aligned. The fillet of the tops of the side walls helps insure that each holder slides into the opening in the 1130 console honeycomb cell. 

View from bottom where header is inserted

Drawing views

View from above where bulb is placed

Friday, April 2, 2021

Difficulties having a full Lamp Test verification of the board


The bulbs I am using draw 200ma at their rated voltage. With 164 eventually installed, that would total about 33 amps. Since the IBM 1130 lighting circuit is fed with 7.25VAC, the actual current will be higher, up to almost 38 amperes.


I used a bench power supply to provide 6.3VAC for the initial testing. That will light up individual lamps or small numbers of lamps to a satisfying brightness. However, it is rated at only 1 amp thus woefully inadequate to test the full board.

When I populated the board with 158 lamps and tried the Lamp Test, they all glowed, but very dimly. The power supply made a groaning sort of sound which was the transformer overloading, and when I checked the AC voltage it was down substantially due to the (over)load. 

By removing bulbs, I was able to confirm that Lamp Test will cause bulbs to glow as brightly as will logic signals, but only within the capacity of the power supply. That is, I could run up to about 10 lamps, which will draw 2A, and the brightness is nearly identical to a single lamp. Any more and the dimming becomes quite noticeable. 


I have bought a beefier transformer, one that is rated at 6A for 6.3VAC. Based on my experiences, I should be able to fully light 30 positions all day and could push up to around 60 without too much dimming. That is of course as long as I limit the time I am overloading and heating the transformer. 

I also bought an AC ammeter to insert into the line. I want to confirm the current being consumed by bulbs at full brightness. I should be able to measure the current demands of 30 bulbs, then scale it up by 5.47 times to estimate the full load on the board at 6.3VAC. 


As a third and final bench test, I will try to make use of a variac to produce 7.25VAC and satisfy the full current demands of the board. I am short about 13 bulbs, but I have a new supply of 50 lamps coming later next week. In that way, I will have the means to accomplish a full Lamp Test of the board, checking for issues and hot spots.

Since I don't have an isolation transformer, one side of the variac output, which has to be hooked to my common ground on the board, will also be tied to the wall supply neutral conductor. I shall need to test my bench supply that delivers the +3V for Lamp Test control to be sure that it won't be harmed by having its ground terminal hooked to wall neutral. 

I guess at worst case I could make use of two batteries to produce the +3V to feed the gate current of all those triacs. I project they will draw in the range of 60ma from the +3V control line to have them all firing, which is within the capabilities of a pair of D cells. 

The Variac that I own is only rated at 5A. I would need one rated at quite a bit higher level. It doesn't have to be rated at a full 40A, but can't be all that much less or the heating effect could damage the windings and the slider that sets the output voltage. I could easily live with a 30A unit, but it isn't worth the $75 to $100 it would take to buy one. I think that after the second test, I will have to hook this into the IBM 1130 and validate the thermal and full current behavior of the board while installed in the console and active. 

Thinking about improvements for lamp holder for the incandescent control panel upgrade


I use mini incandescent bulbs with wire leads, which I wrap around a two position header strip and solder in place. On many of the lamps I built, I applied some kind of insulating goop to protect against shorts if the leads touch but it makes the lamp a bit ugly. These plug into two position sockets on the PCB, enabling easy replacement of any dead bulbs. 


Because the bulb is simply suspended on its leads above the header strip, it is not held in exact alignment with the opening in the honeycomb matrix into which it has to fit. Bulbs can be cocked because one lead is a bit shorter than the other. They can flex over to the side as well. 

Some of the sockets are tilted over, rather than at right angles to the PCB since I didn't have a jig that assured proper alignment. This tilts the lamp plugged into the socket. 

This degree of variability makes the insertion of the PCB with all 164 lamps installed a bit fussy; the bend in various lamps may cause them to not slide into the honeycomb cell without some encouragement with a thin screwdriver. 

The leads can short together, particularly if the lamp is bent over because it fails to enter a honeycomb cell during installation of the PCB into the matrix. The goop is not a perfect solution to this problem.


I will go back over the board to manually realign any sockets that aren't perpendicular, reflowing the solder while I position the socket with a suitable tool. That part is easy to address.

It will be necessary to build a holder for the lamps, something that keeps the wire leads separated, holds the bulb upright and perpendicular, and ensures that each lamp is well aligned with the honeycomb matrix so they slide right in. 

I am exploring having them 3D printed, after I build a design that I feel will properly maintain alignment for all 164 bulbs. Over the coming day or two I will noodle around with various designs until I find one that is promising enough to have some samples made. 

Thursday, April 1, 2021

Powered tests of the functionality of the board - everything working very well


The board requires a common ground between the SLT logic level DC (0 or 3V) and the 7.25VAC lighting power. The other leg of the AC supply is hooked to the top terminal block. Finally, the Lamp Test block on the bottom is hooked to a switch that ties it to ground in the normally closed position and flips the Lamp Test line to 3V DC when moved to the normally open position. 


With the AC turned on but no lamps or inputs connected, the board should not consume any appreciable current on the AC level - and it did not as it was less than one ma. I also checked that AC was not presented on any of the 164 logic input pins. 


I inserted one bulb to verify operation of one of the 164 lamp circuits. With no connection to the input pin, the lamp did not light. With a simulated logic level applied the bulb lit up. This was done with +3V applied through a diode and a 6.2K resistor, matching the output driver inside the 1130 that provides the signals to the control panel. 

I flipped the lamp test switch, putting +3V on that control line, and my bulb lit with either logic 1 or logic 0 applied to the input pin. This represents a full test of the operation of one lamp position. I know that the basic circuit is working properly


I installed five bulbs onto the board and switched on again. Nothing should light with Lamp Test off and no logic 1 applied to the input pins. None did. I hooked up three logic 1 levels to three of the circuits and left the other two at logic 0. Those three bulbs lit and the others did not. Finally, switching on Lamp Test caused all five bulbs to illuminate. Again, all good. 

Three signal lines switched to high, Lamp Test switch in off position


I built a quick and dirty tester for lamps to make sure that the wire leads were not shorting before I plugged them into the board. It was a bit of header strip in a breadboard with my ohmmeter in line to check the cold resistance of the bulb. I could distinguish between dead (open) lamp, shorted wires and a workable lamp using this setup. 

I inserted the board with all the lamps into the honeycomb matrix of the front panel. This would let me access the signal pins on the rear, protect the lamps inside and let me watch the illuminated positions from the front. With all lamps installed, I repeated my check to verify that none of the 164 logic input lines were seeing AC. This was key to protect the 1130 circuitry from damage when I install this upgrade in its control panel. 

I powered up and verified that no lamps were lit in the absence of an SLT logic 1 level on its input. I tested all the circuits to verify that they operated correctly with my logic 1 inputs on their pins. Finally it was time to flip the Lamp Test switch and verify that all the lamps lit up properly.

The current measured during steady state of the lamp test was hard to measure since my ammeter didn't support high enough current and none of my AC supplies have current measuring capabilities. 

Even worse, the puny AC power supply couldn't put out the current demanded by the lamps when all were on, thus the voltage dropped to about 5VAC and the lamps were all quite dim. I checked for signs of hot spots and was satisfied with the thermal performance of triacs and the board since everything was cool as can be, loafing alone under the dim version of the Lamp Test. 

Lamp Test switched to on position, power supply struggling to keep up


Accumulator Register Bit 6 did not light with the Lamp Test activated. The lamp lit properly for a logic 1 applied to its input pin, but did not react to Lamp Test. I traced this down to lack of connectivity from the 6.8K resistor to the gate of the triac. After a quick repair, I reinstalled a lamp to verify that ACC 6 does light under Lamp Test. 

Operation Bit 2 did not light with Lamp Test nor with direct logic 1 to the signal pin. This was a poor solder joint on the gate pad of the triac. Resoldered and tested.

Tag Bit 7 did not light with Lamp Test but did work with logic 1 on the signal pin. The 6.8K resistor was not connected properly. Resoldered and tested.

REC signal of the Synchronous Communications Adapter did not light with Lamp Test nor with logic 1, but did when I wiggled the socket. Poor solder on one of the pins of the socket. Resoldered and successfully tested.

Odd Status lamp was always illuminated even though logic 0 on the input pin and Lamp Test not activated. Found a bridge on the triac between the output 'anode' and the gate pad. Removed and replaced triac, then tested successfully. 

My original designs did not attempt to light the 8 Customer Engineer (CE) bulbs nor light the signals from the Synchronous Communications Adapter (SCA) since that feature is not installed on my machine. However, this new board is universal and includes those positions. However, when I built my lights - wire lead light bulbs soldered onto two position header strips - I didn't build the extra 16 needed to fully populate my board. 


I have to build another 16 bulbs and install them into their positions for the CE and SCA displays. 

I want to work out a good fastening or clamping method to hold my board onto the honeycomb. 

It would be reassuring to find a high capacity AC power supply with current measuring capability and hook that to my board so that I can verify how it operates with the 148 currently installed bulbs. That should be close enough to the fully populated load of 164 to feel comfortable installing this into the IBM 1130.

The console should be put back together, my board wired into the 1130 with all the signal lines attached to the signal pins on the back. That will allow for a full power on test with the actual computer, the final certification of this design to replace the kludgy IBM incandescent lamp mechanism that came with the system. 

Anomaly tracked down to failed triac


When measuring the resistance from the Lamp Test control line on the terminal block to the common ground, I found a surprise. It was 6.8K, the value of one of my resistors whose other end was somehow shorted to ground. All of the resistors should run from Lamp Test to the gate of its triac and there is no ground trace near these that could account for a solder bridge causing a short.

I checked the resistance from each triac gate to ground, which should have been infinite but was 13.6K for almost every triac - the resistance if its local resistor plus the mystery circuit whose resistor was sunk to ground. Eventually I found that the gate of the triac for Interrupt Level 1 lamp was at ground. No place where a solder bridge could have joined it to ground, leading me to conclude that the triac itself was internally shorted, gate to ground. 

I had reclaimed most of the triacs from my earlier boards to use here, they were not new other than 30 I still had in the box unused. Somehow this one was damaged and needed to be replaced. 

I whipped out the hot air rework tool and the soldering supplies, removed the triac and cleaned up the pads for installation of another of my reclaimed triacs (after testing it for a similar short of course). I set up a number of air dams to avoid loosening any adjacent parts when I unsoldered the bad triac. 


 Very puzzling indeed - with the triac removed, the short to ground was gone on the board but the triac itself didn't show a short. Nothing on or around the pad could have bridged to ground, so it must have been some failure inside the triac, perhaps some package stress that was relieved when I removed it. 


I used a different triac, installed it into the position and checked over everything again.  All appears well with the board so it is on to the powered testing phase.

Finished build of the incandescent light upgrade to the IBM 1130 control panel


There are only two remaining steps to finish installing all the components on this board. The first is another large tedious operation - placing 164 single header pins on the board for all the logic signal inputs. The second is quite small by comparison, installing the three wire terminal blocks that provide 7.5VAC, Lamp Test signal and common ground connections. 

By mid-day I wrapped up the signal pins and moved on to the terminal blocks. Since this board used 2 ounce copper pours for each layer, it has a very noticeably higher thermal mass than the usual PCB. It takes quite a bit of heat, with a wide tip, over a longer than usual time to flow solder properly through the pins and holes for the terminal blocks. 


I carefully examined the soldering of all the sockets and pins by eye before starting tests with an ohmmeter. I then checked for zero ohms across the three power/control input terminal blocks. Finally I tested every socket and every signal input pin for signs of a short. Everything was good. It was time to apply some power and check things out.

Finished board ready for testing