Validated design for 1130 core loader (DMA version) sent to fab after testing session

BREADBOARD TESTED NEW DESIGN

I changed the breadboard to the new design I created for the DMA loader circuitry and put it through its paces. Initially I had a problem with switch bounce but a debouncing pushbutton on my Cadet did away with that issue so that I could test with confidence. 

I expanded the test setup to show the state of the control outputs I generate as the state machine operates, to be certain they were correct. I drive both -CS Level 0 Request and -File Data Entry Gate low with open collector chips - temporarily added pull up resistors to be able to see the outputs. I also drive the enable line for the Address and Data words; when a bit of an address or data word is 1, it will pull the output low but only when enabled. 

PCB DESIGN SENT TO JLCPCB

I packaged up the Gerber files and drill information in a zip file and uploaded it to the fab. I chose a blue color for the solder mask with white silkscreen images. It will take 3-4 business days to complete manufacturing, then 3-5 days for shipping. The components are all waiting at the shop for the boards. 

Slowly digging through boxes and setting up the shop; testing the DMA loader logic uncovered an issue

SETTING UP THE SHOP

As this is a new space, I am not certain of the best arrangement of equipment, tables and the many boxes of components I have. Further, the boxes were all taped shut and are piled in random order in large heaps. I am still pulling tape off boxes, although probably eighty percent have been done by now. 

With the heaps piled along the walls, each time I decide to place a table I first have to drag boxes and equipment away from that area to make room. I have five of over twenty tables in place, at least to a temporary location. Then when all the boxes and other items are moved and the tables are placed, I have quite a few boxes which are either unsorted or that need further division into smaller boxes to make searching easier. 

BREADBOARD TEST RUN ON THE MAIN SEQUENCING LOGIC OF THE LOADER

I used my Cadet breadboard to set up eight of the chips from the design, ignoring all the output buffer chips for the time being because once the enable signals are emitted at the right times, they are certain to work correctly. 

I found that my scheme to reset at the end of a cycle was timing sensitive and potentially glitchy. I revamped the circuit, which had the side effect of cutting two chips from the design. All the online logic simulations show it to be solid, but I will go back to the breadboard and make sure it is rock solid before committing the updated layout to the PCB plant. 

BUSY DAY AT THE CAPE TOMORROW - NO TIME FOR THE SHOP

I am supporting two launches tomorrow - a SpaceX Falcon 9 launching Starlink satellites, and a ULA Atlas launching the Boeing Starliner capsule with two astronauts. I am helping secure one viewing location. In between, the monthly SF Museum volunteer meeting will take place. Basically I will be onsite the entire day and evening. 

Quick update; regulator is indeed working properly

HAD A SUSPICION THAT MY TEST SETUP WAS AT FAULT

One of the possibilities to explain the voltage dip when I went up from 1A to 4A was that my test setup was not a sufficiently accurate replacement for the attachment to the raw DC supply in the 1130. If so, the regulator might have maintained its voltage while connected in the 1130 although it wasn't doing that on the bench. 

The first thought I had was that the AC bias voltage somehow compensated for load differences, so I varied the bias supply but saw zero change in the delivered voltage. Just before I broke down the setup because I believed the 086 transistor replacement (2N214) was unable to drive enough current due to its lower beta, I had another idea.

Putting the multimeter on the output terminals themselves, rather than depending on my electronic load to display the voltage, I was surprised and pleased to see that the voltage coming from the regulator itself was precisely the same voltage at 1A, 4A and no load at all. Clearly, the voltage drop was occurring on the cables between the regulator and my electronic load, due to the resistance of the cable. 

This is great news, as it means that I have only one test left, the full load test while bolted into the machine. I placed the regulator in the machine again. My resistor network produces a resistance of 0.2 ohms and can handle 200W across all the resistors that comprise it. For the +3V regulator, that means we will draw 15A from the supply. If the voltage stays essentially flat under this load, then the regulator is completely repaired and ready to go. 

0.2 ohm load for testing

Load connected to the supply

I just need to wait until I get my 230V outlet installed to bring up the 1130 and begin some functional testing. 

Progress on voltage regulator for VCF 1130; made plexiglass cover for SSM console printer

QUICK SCHEMATIC DRAWING ATTEMPTED FOR EBAY VOLTAGE REGULATOR CARD

SMS cards are very easy to reverse engineer, because all the components are discrete and visible. More importantly, the PCB has traces only on one side, with wire jumpers soldered when signals have to move across the component side. The parts are all labeled so that I can tell their value with the exception of three diodes - two Zener and one normal diode (that rectifies the AC bias input). Given the similarity to the previous SMS card, I expect the Zener diode values to be identical in the new card. 

I began beeping out the connections from the SMS card fingers to determine where the circuit would diverge from the previous card to this one. During this process I got a beep where I shouldn't. The card has a large 108 PNP transistor on a heatsink that is the final stage which drives current through the output transistors. The emitter and collector had almost zero resistance - shorted out. This is not true with the other card (or any properly functioning 108 transistor). 

Therefore this is not functional and can't be inserted into the power supply as a substitute. It reverts to its original status as the parts donor. 

DEALING WITH THE 026 TRANSISTORS ON THE ORIGINAL CARD

I desoldered the two 026 transistors that form the comparator on the regulator card, so that I could test them on the curve tracer. I found that there was a 3X difference in beta (gain) between the two transistors. Without a spec sheet or a known industry alternative part number, I couldn't determine which was bad at this point.

I then removed one of the 026 transistors from the ebay card and put it on the curve tracer. I now had two 026 with identical gains and curves, plus a third that was much higher in gain. It seems quite reasonable that the two transistors in a comparator should be as equal as possible, so I put the similar pair back on the card. 

INSERTED CARD IN REGULATOR AND TESTED AT LOW LOAD

With the card now possibly repaired, I plugged it into the regulator and set up the power supplies and electronic load. The regulator is fed 8-9 volts DC from one supply, which in the machine comes from the unregulated raw DC supply. The machine also provides 8-9V AC to the regulator which is turned into a DC bias voltage by a half wave rectification on the SMS card. I used another power supply to provide a DC voltage in lieu of the 1130's AC input. 

My electronic load was set to 1A, enough to be able to assess whether the machine is holding a fixed voltage. I then adjusted the potentiometer on the card to set the output voltage at just over 3V DC. 

This did the trick. The regulator kept the same voltage for long periods of time, unlike its prior behavior where it began to run away after a few dozen seconds. I bumped up the load to 1.5A, 2A and 4A to see how steady the output voltage remained.

I was not completely happy with the outcome. The regulator sat at 3.04V while at 1A, but declined a bit with each bump in current until it was at 2.96V at 4A. This could be one of two things. First, it could be an artifact of how I am powering the supply on the workbench. Second, it could be due to the lower beta of the transistor I used to replace the 086 on the card. 

In the real machine, the bias AC supplied to the regulator comes from the winding of the raw DC supply that feeds the regulated supply. This might change under load and somehow compensate while my bench setup feeds a fixed DC voltage to the regulator. If this is the case, then the regulator card is really working properly and I can consider this fully repaired.

On the other hand, the lower beta of the replacement transistor might not produce as much drive current, therefore the regulator might not adjust as much as it should as the load increases. The solution to this would be to find a better replacement or even try the original 086 back in the card. 

REGULATOR INSTALLED IN VCF MACHINE

I bolted the regulator back into the machine and installed some of the wires to it. I can do load testing first with my electronic load up to 5A and then with my resistor networks for a much larger load. That will determine whether the regulator is actually working good enough or if I have to keep working on a substitute for the 086. 

HIGH LOAD TEST WILL CONFIRM PROPER OPERATION OF REGULATOR

I used my network of power resistors to load up the regulator to 16A, which is higher than the draw from the 1130 it will be used in. The aim is to see if the voltage remains nearly constant when under load, which proves that all four of the power transistors in the final stage are working properly.

The next stage of testing, when all the power supplies are working correctly, is to adjust the voltages to the target levels for the machine by reference to the terminal blocks attached the logic gates where the DC is delivered. I need the 230V outlet in my workshop to power up the machine, but the owner of the unit is waiting on the electrician to arrive to install the outlet. 

DESIGNED AND HAD PLEXI COVER BUILT FOR SSM 1053 PRINTER

The 1053 console printer for the System Source Museum 1130 system did not have the plastic cover that fits in the lid. This protects the printer from objects falling in as well as reducing the sound produced as it types. The IBM part is a complex shape which was unrealistic to copy but I had a flat piece laser cut from plexiglass as a substitute.

The plastic IBM part has notches on the underside where springy metal wires hold it in place and against the front of the lid. I cut some small bits of acrylic to provide the same notch and glued them to the bottom of the cover I just had made. It fits well and looks appropriate. 

Cover without notched pieces added

DMA Core Loader PCB has issues - gambled and lost; intrigued by 360 voltage regulator card I bought

DISCOVERED A FEW ISSUES WITH THE PCB DESIGN ALREADY

I have not completed all the testing yet with breadboards and an 1130 but I have already discovered a couple of flaws. One of the issues would be a huge pain to resolve by cutting traces and adding bodge wires, so I won't bother. I knew the risk when I submitted the first design; just a few dollars wasted. 

I did take the time to optimize placement of the decoupling capacitors, which generally involved relocating quite a few unrelated traces to make room for the capacitor as close as possible to the chips they support. 

I have a new PCB design set up in KiCAD but will hold off sending it to be fabricated until all my testing is done. This adds some schedule delay waiting for the boards after the testing wraps up, but it minimizes the chance that the next set of boards requires any rework. 

EBAY PURCHASE OF 360 REGULATOR CARD MAY BE WORKABLE REPLACEMENT

I purchased a card on Ebay that had the 026 transistors I lacked to repair the VCF 1130, thinking that it was incompatible but a good parts source. The documentation I have for the 1130 shows two versions of the voltage regulator, a 'previous' and a 'new' version. The VCF machine has the previous style. The new style is different enough that the regulator card and the supply can't be mixed with pieces from the previous type.

However, examining the new card shows that it is neither the previous nor the new type. It is however pin compatible with the previous style and may work just fine with the supply. This appears to be a newer design leveraging a later IBM transistor than was available when the previous card was designed.

The previous type used an SMS card that had twin 026 transistors as a comparator, an 086 driving an 028 which then drives the final 108 transistor on the card. The 086 is NPN while all the other transistors are PNP. 

The new type SMS card has twin 026, but they are amplified by a chain of two 183 transistors before driving the 108. One of the 183 has a heat sink pushed on it, thus it is the second stage of amplification. I don't have schematics for the card but will produce one by some reverse engineering. The design appears similar enough to the previous SMS card, other than the change to deploy twin 183, that I suspect I will find the circuitry pretty similar outside of the areas directly driving the 183 parts. 

Previous on left, mystery board on right

The workshop is moved! Faster than I thought and a bit more casual

SCRAMBLE TO MOVE TODAY INSTEAD OF TOMORROW

The Chiropractor who is expanding into the storefront that was my current workshop had promised to move everything and had detailed specs from me on the dimensions and weights of all the larger equipment. We had arranged for his movers to come tomorrow to move the heavy items, but yesterday afternoon he called to try to do the move today.

He said the movers only had today available out of the entire month so it had to be today. I on the other hand had two important appointments that could not be moved and I was not going to have people move the items without being there to supervise. Since I could be back by 1PM after the last of my appointments, we settled on that time for the move. 

I envisioned a larger truck, lift gates or fork lifts, thus was surprised to see a trailer on the Chiropractor's pickup truck suitable for hauling a riding lawn mower. He had several friends, his wife and his new partner in the practice to accomplish the move. 

What made this feasible was that my new workshop was only a block or so away thus a very slow careful drive was possible and short travel time supported multiple round trips. We began at 1PM and by 4:30PM every last item from my shop was in the new space!

Everything is there, but I have to figure out how to organize the space and locate everything. That may take some time, but I can fairly rapidly return to working on the VCF 1130's voltage regulator, the SSM 1130's typewriter and the VCF 1130's disk drive. Too, I can test my design for the DMA based core memory loader. 

Ordered components for DMA based core loader; arrived home today

COMPONENT ORDER PLACED

I sent the final order for the ICs, voltage regulators and terminal block to Digikey, so that I should have both PCBs and parts on hand by the time I am set up to do the assembly. The design has eighteen chips, all surface mount parts, which fit on the PCB. 

HOME FROM RHODE ISLAND TRIP TODAY

I landed this morning and am headed over to complete the final packing, labeling and arranging of everything in advance of the move which will begin on May 3rd. With the move itself seeming to be well in hand, I am starting to look ahead to the challenge of working out a layout for all the tables, boxes and machinery in the new location. Once that is done I will need to sort and place the containers then deal with all the partially or fully unsorted last minute containers I had to fill. 

I will be able to set up the bench fairly quickly for work on the 1130 voltage regulator as well as the assembly table to build the DMA loader. Thus I will be able to accomplish productive work interspersed with all the housekeeping activities.