Friday, March 27, 2020

An attempt to improve the output of the DSKY power supply module, attempt 1

Since the power supply was not delivering the target voltage even with boosted voltages for both the +13 and +24 inputs to the module, I decided to try a change in how I am delivering the 800 Hz reference clock input. If there was not enough drive in the method I first used, it could result in smaller waveforms with a lower peak voltage.


The 800 Hz signal is produced by a counter that ticks to alternate the binary state of an output signal. That essentially produces a square wave at 800 Hz. The type C interface circuit delivering that signal out of the AGC consists of a PNP transistor with its emitter at ground and a 2K resistor in series between the collector and the output pin.

Thus, it acts to pull the external line down to ground or let it float depending on whether the clock signal is on or off. Within the DSKY power supply, +13V passes through half of a center tapped transformer and out to the circuit I just described. Thus, 800 times per second it is alternately pulling 13V down through the transformer towards ground or letting the transformer pull up to 13V.


I set up a function generator to produce an 800 Hz square wave, 0 to 5V, which I fed into an interface circuit I had built for use with the restoration of the Apollo Guidance Computer we undertook last year. The surface mount transistor that I used on those interface boards were designed to drive very small currents for inputs to Arduino or similar devices. It was not engineered for higher current levels such as the use of the clock to drive the DSKY power supply.


I constructed a replacement to the interface circuit since my first approach may not have had enough current capacity to adequately drive the transformer in the power supply module. This was a 74F05 integrated circuit, an open collector chip which takes a TTL input signal (from my function generator) and whose output is hooked through a 2K resistor to the DSKY supply.


Before I put this into the circuit to drive the power supply module, I set up the open collector inverter with the 2K resistor pulling it up to +5V and observed the signal on the oscilloscope. The results show that this chip is inadequate, having too high an internal resistance.

Yellow is input square wave and blue is output pin of the inverter
As you can see the input is a well formed square wave between 0 and 5V while the pulled up output bobs over a very small range near 4.5V. Back to my parts bin to look for a more suitable buffer that will handle 13V with enough drive to kick energy into the power supply transformer.

Wednesday, March 11, 2020

Investigating recreation of DSKY power supply circuit

Since NASA published the schematic and most of the component values for the power supply module that fits in the DSKY (Display/Keyboard) that is the human interface to the Apollo Guidance Computer in the Apollo spacecraft, it seemed possible to build a replacement.

I was loaned a power supply module from the block I DSKY, which is close enough to the version from the block II DSKY used with all the missions flown with people aboard. I did some testing of the pins, but the parts are encapsulated inside a metal rectangular solid which makes it infeasible to probe inside or repair anything that might be malfunctioning.

My first tries to power the EL panel used that original power supply module, driven by an equivalent circuit to the one inside the AGC that generates an 800 Hz square wave which drives the DSKY power supply. The module should produce at least 250V AC at 800 Hz, but at best I was only getting 85% of the target.

That lower level didn't seem to be enough to light the biggest single element in the electroluminescent (EL) panel - the COMP ACTY rectangle - but did give me glowing output on all other elements,. Building a replica power supply circuit would let me produce the full voltage, or debug and correct any issues I might find.
Block I DSKY Power Supply Module
The circuit seemed to be fairly straightforward. The 800 Hz square wave drove one half of a center tapped transformer primary, with the secondary creating a 13V AC signal. The AC signal was fed into a pair of transistors to drive the primary of a second transformer (T2), whose secondary windings boosted the AC signal to 24V but with more current flowing.. The 24V AC signal is fed into a third transformer (T3), whose output would be the full 250+ V that drives the LEDs.

The EL panel works by energizing a capacitor, the top plate being a transparent conductive material and the bottom plate having the shape of the segment to be illuminated. The insulator material luminesces under the high voltage, high frequency field. There are more than 150 elements or segments that can be energized, with the number varying as the panel displays different output values or blanks.

Because each illuminated element is another capacitor added in parallel across the power supply, it increases the load on the output. With more segments lit, the output voltage might droop, causing the EL panel to vary its brightness based on what is displayed. To adjust for this, the third transfer (T3) is a reactor whose core can be saturated or controlled by a DC current in a control winding. That change in core saturation varies the inductance thus changing the step up ratio of the AC windings of the transformer.

The clever circuit puts a load resistor in series with the high voltage, high frequency going to the EL panel. As load goes up, the current changes and the voltage drop across that load resistor will increase. A full wave bridge rectifier converts the voltage drop of the load resistor to a DC value which is fed into the control winding. The load resistor value was tweaked with each instance of the power supply to cause the output voltage to change the least as the number of energized segments varied.

Finding (or winding) a transformer with a core which can be saturated with the current produced from the load resistor is going to be essentially impossible. We don't have winding counts, stats on the core, manufacturer part number or even manufacturer name. One document hints that the part was bought from "Bush", but google is not my friend when I search for any sign of such a company or product.

I suspect that I would have to design a different circuit to produce an invariant high voltage across a varying capacitive load. It would have been much more straightforward and period authentic to build an exact replacement power supply.

Tuesday, March 10, 2020

The EL panel has a variant color, different from later production DSKY panels

The color of the panel is more blue with less yellow than we had expected. I dug through the NASA drawings that comprise the specification control documents used to procure the part. I did find that the first version of the drawing, 1006315, listed 5100 angstrom as the color. This is consistent with what I see on my panel.

EL panel showing the drawing number 1006315, without a revision
The back of the part has to include the NASA drawing number along with the revision letter. As you can see above, it is 1006315 with no revision. I judge this to mean that the part was a very early procurement.

The revisions of the drawing for the panel ranging from A to G all require that the light be emitted at 5300 A, thus being closer to the color we all expected.

Monday, March 9, 2020

Testing additional segments for operation after tweaking power supply


The voltage output was a bit too low. The panel expects over 250V at 800 Hz to drive to full brightness, but I was about 20% low even with the dimmer potentiometer completely out of the circuit. I tried increasing the +14V supply a bit, which gave me a very small but real increase in output voltage. When I varied the +24V supply, however, it seemed to make very little difference. I will look further into this.


I tried connecting the power to a number of the other segments on the panel, such as the signs for R1 and R3, some segments on digits in VERB and R2, with every segment I checked lighting just fine. My problems appear to be solely with the COMP ACTY feature, which only illuminated a small dot inside the COMP ACTY rectangle.

I was able to get the static elements to light up after I tweaked the voltage from the supply up a bit. Very pleased. Even if the computer activity segment is damaged somehow, I seem to have a decent display in all other particulars.
Static features lighting up!


I expect to wire up all seven segments of a digit and the two portions of a sign, hooking them to the relay module which in turn will be powered by the power supply module. I anticipate driving this an Arduino based setup allowing me to cycle through digits and the sign values. This will take a few days to wire up.

Improving connections and first testing of the EL panel


I crimped the sides of the connectors together to tighten their grip on the DSKY pins, which gave me enough security to complete some experiments. These connectors form a rectangular box, with one of the short ends missing, which allow me to bend those ends closer together forming a sort of triangular shape.


The Electroluminescent (EL) panel I am working with has two known problems. The front glass has a crack in it, whose effect on the internal structure and function is unknown. Also, the two ground pins 155 and 156 should be a virtual short to each other, as they collectively provide a ground connection across the entire face of the panel; these have a resistance of more than a kilohm between them.

I hooked my power up to pin 156 for ground and to pin 157, which should illuminate the steady features of the panel. These are three horizontal lines that separate the three five-digit register values, plus a rectangular lit box to highlight the text labels NOUN, VERB and PROG. No sign of a glow.

I hooked the power up to pin 154 which should light the larger rectangle under the COMP ACTY label. I saw a tiny green dot, probably where the wire is bonded to the segment, but no glow across the rest of the feature.

I then wired up two hot lines to a pair of segments on a digit of the PROG section and saw them light up. Not super bright but clearly working. This tells me that I need to do more investigation of both the delivered power from the AC supply and to test many other segments to see what lights up properly.
Lighting PROG digit 1, segments F and K

Sunday, March 8, 2020

Results of DSKY power supply testing and power-up


The equipment I set up on the bench for the test of the power supply was:

  • Signal generator set to produce 5V 800 Hz square waves
  • Oscilloscope to observe and measure the outputs of power supply and signal generator
  • My AGC interface board
  • Dual bench power supply delivering +14 and +28 to the DSKY power supply
  • VOM as a voltage level test

The interface board I build contains converter circuits between the Apollo Guidance Computer I/O levels and TTL/LVCMOS for use with modern technology. In particular, I had a circuit designed to hook to the AGC's D type inputs which was an open collector inverter gate with a pull-up to spacecraft signal levels (+28V). By leaving the 28V disconnected, it was a simple open collector that could be hooked to the DSKY power supply and accept a TTL square wave input.


In order to fit onto the Malco Mini-Wasp blade connectors on the power supply, I needed to create some wires with push-on connectors that almost fit. These don't grip really firmly but wedge on enough to make electrical connectivity. I had to crimp the connectors to many individual wires, then put heat shrink over the end to prevent shorting as the connectors can tilt and wobble.

Male Malco Mini-Wasp connectors, rectangular ended
Pin with a suitable rectangular shape to press over the Malco blade
Building jumpers between Malco pins, prior to heat shrink application


I hooked the bench supply of +14 and +28 to the pins of the DSKY power supply. The signal generator output was wired to my interface board, whose output was hooked to the DSKY power supply input for 800Hz pulses. A put a 10K resistor across the DSKY power supply contacts for the dimmer to represent a pot set to max brightness. The DSKY supply output pins were wired to my oscilloscope where I could measure the output being produced.

As a safety precaution, I made sure that my oscilloscope ground lead wouldn't damage either the DSKY power supply or my scope. This is a quick test with a VOM from the ground lead of the scope to the planned ground attachment point in the circuit under test.


First I looked at the waveform coming from the signal generator. All was good.

Output of my signal generator
Driving this 5V signal through my interface circuit with a temporary pull up to 14V gave me a clean output signal, ensuring that I had a proper input for the DSKY power supply.

Amplified to 14V swing by my AGC interface circuit
Next I checked the output with it hooked to the DSKY power supply with +14V applied to it.  This too looked good.


I then switched on the +28V and +14V levels, watching the output signal on the VOM and oscilloscope. My scope maxes out at 50V per division, which isn't enough to show a full 275V AC signal. Still, I could set it to measure the peak to peak and other values.  I was happy to see it developing a peak to peak well over 450 volts.

Output of the power supply at full brightness on dimmer control

I wired up the two ground pins of the Electroluminescent Panel to one side of the AC from the DSKY power supply and then applied the other side to the fixed lines and the Comp Acty pins. Nothing showed. Since I had an unexpected non-zero resistance between the two ground pins 155 and 156, I may have enough problems with the panel that it will be difficult to light it up.

Worse, the funky pins I am using are very difficult to keep connected to just the nine pins on the power supply, such that I kept losing the high voltage because one of the supply voltages, ground or the input pulses went bad.

It is time to rethink how I make the connections before I continue testing.

Tuesday, March 3, 2020

Complications during DSKY module testing and power up


I had been working from a manual produced by AC Electronics, the prime contractor who managed Raytheon in producing the Apollo Guidance Computer and its Display/Keyboard (DSKY). It was ND-1021041, Project Apollo Command Module Guidance and Navigation System Manual, Volume II. On page 4-508, figure 4-233, it shows the connections for relays K24 and K30 which drive the four segments that light up the Plus and Minus sign in front of a five digit register.

Erroneous drawing from the AC manual

When I did some continuity testing with the relays both set and unset, I had several puzzling failures that seemed to require multiple failures inside the module, where the other 40 relays were working flawlessly. Fortunately, I had Mike Stewart to thank for steering me towards more accurate schematics of the relay module, drawing 1006161 which gave a more concise (and accurate) view of the module.
From proper NASA drawing
To interpret the diagram above, you need to match the legend on the right to the blocks on the left for the two relays and draw out the contacts. I did this by modifying the incorrect AC drawing to illustrate the difference.

Modified AC drawing to reflect reality
I realized that the two relays have their contact 7 tied together, thus when both are unset a path exists between 6 and 96, although not shown on the NASA document. When unset, contacts 4 and 7 of each relay are closed, thus connecting 6 and 96 together.  Otherwise, continuity was verified as shown on the NASA drawing.

In practice in a DSKY, only pins 56, 13 and 91 are used to control the sign digit. The other pins exist for testing purposes only, I believe.


The power supply is hooked to +28V and +14V supplies and is furnished an 800 Hz clock signal from which it generates the 275V 800 Hz AC output. This is not a sine wave, instead a digital square wave input. The two DC levels are easy to produce, but I needed to understand the specs for the clock particularly voltage or current levels.

Fortunately I know the Apollo Guidance Computer side, which generates this signal. It uses a open collector circuit that pulls down the 14V from the input transformer to ground as a 800 Hz square wave.  I had built a substitute for that AGC circuit as part of our interface to the computer, thus I can conveniently use a 5V TTL square wave source at 800 Hz into my interface and wire the output directly to the DSKY power supply.

One other aspect of the circuit needed analysis. The spacecraft has a potentiometer that acts as a dimmer for the DSKY illumination. It is wired across the input transformer of the power supply, shunting some of the power from the AGC clock pulse through the pot to lessen the flow inside the transformer. This serves to make the output change.

I don't know what value the potentiometer has, but looked it up in the Interface Control Documents that were defined for how the Command Module and AGC/DSKY were connected. This was a 10K linear taper potentiometer.

I did realize that at one end of the travel, the pot will totally bypass the input transformer causing the segments to be extinguished and at the other end it will set the maximum brightness based on the 10K resistance of the device.

Sunday, March 1, 2020

Preparation to light up an Apollo DSKY electroluminescent panel


The Apollo Display-Keyboard (DSKY) is the human interface to the guidance computer within the Command Module and Lunar Module. It consists of an array of pushbuttons that form the keyboard and two side by side panels of lights. The left panel is a matrix of incandescent lamps which indicate various operational, caution and warning conditions for the computer and spacecraft. The right panel is a single electroluminescent panel that displays 21 decimal digits, three signs and a few lighted text blocks.
Apollo block II DSKY electroluminescent panel
The EL panel forms the digits with seven segments but these are not LEDs like modern displays. The segments are formed with phosphor coated metal traces, as are the segments of the + and - signs or the rectangles that light up with text. Over the top of all of this is a transparent conductive coating.

When an appropriate AC field is applied to the segment traces and the top coating, it causes the phospors to emit green light. Typically the AC is up to 275V at 800Hz with the voltage varied a bit by the cabin dimmer potentiometer. The AC is switched to each segment by contacts in tiny latching relays. Signals from the guidance computer cause the relays to be set or unset, thus illuminating or extinguishing each segment.

The DSKY is supplied with only a few relatively low DC voltages, mainly 5, 14 and 28V. It is also fed with an 800Hz clock signal. One module inside the DSKY produces the high AC voltage which is then routed to the relay modules to drive segments.

Decoder modules look at the 15 bit word coming from the guidance computer, evaluating the top four bits as a row number and the remaining 11 bits as the set or unset value for eleven columns of relay. One column is generally used to hold the sign value, which is +, - or blank. Generally the remaining 10 bits are used to represent two digits. A five bit value is used to control the seven segments of a digit using a bit of clever relay logic.

Each relay module contains 42 relays, six columns by seven rows, most of which are latching types with DPDT contacts but a few are non-latching. The DSKY uses four such modules, thus has the capacity to handle 168 relays. They can be thought of as a 2x2 matrix of modules, so that each of the possible fourteen rows controls 12 total columns.

A row most commonly controls two digits, thus 11 rows are needed to control the 21 decimal digits on a DSKY. Controlling the signs for the three registers takes the first column of two rows each. Other relays in that first column are used for functions such as lighting the Uplink incandescent light or flashing the verb and noun fields.

Not every contact of every relay is brought out to an external pin; some are interconnected inside the module and some are not used. There are a number of select pins that pick the row to control, set and unset lines that affect an entire column but only act on the relay whose row is selected.

Other relays have individual or small group select signals as well as set and unset. Finally, the non-latching relays simply expose the coil through two pins. All told, there are 137 contacts on the back of a relay module to handle 42 DPDT relays.

Two main versions of the DSKY (and guidance computer) existed, block I and block II. The later block II versions were the ones used on all the manned missions. I have been loaned a block II EL panel along with one power supply and two relay modules from a block I DSKY.

Apollo block I DSKY power supply
Apollo block I DSKY relay module
Although the two versions are overall incompatible, they are similar enough in using 275V 800Hz AC and relay modules that I can combine those to control about half of the EL panel at any one time. I would need more relay modules to simultaneously power all 21 digits and the other lighted features.


I set up a lengthy test procedure to verify the operation of the relay modules. I began by verifying that all the coils had continuity, all the steering diodes for set and unset lines were working, and then that every single relay latched and unlatched (or switched under power if it was a non-latching type).

The relays operate with about 13VDC applied to the set or unset line, grounding the select line simultaneously. My bench power supply provided that power and let me watch the current levels as they relays operated.

For every set of relay contacts, I verified that they were closed or open under the appropriate condition, set or unset, of the relay. I entered the codes for displaying a blank and some digit values to check that the segments were properly switched. For the special cases where a segment can be lit with two different patterns of relays, I tested both.


I checked continuity, freedom from shorts and for appropriate values on the few pins on the outside of the power supply, ensuring that they corresponded to what I expected from the schematics. I then prepared to power up the module by feeding it +14V, +28V and a reference 800Hz signal. I used a signal generator and a bench power supply.

I did have to wire up a potentiometer to simulate the cabin dimmer knob and bridge a few contacts that needed external wiring. With everything ready, I applied power while watching for signs of trouble such as excessive current or magic smoke escaping.


I created an elaborate testing plan for the EL panel as well. I verified that no segments were shorted to the top plate or two each other, but that any top plate contacts were connected to all the other 'cathodes'.

My initial power-on will illuminate only the three glowing lines that separate the three register values, and the three text blocks highlighting VERB, NOUN and PROG.

The output was hooked to a scope where I could validate both frequency and voltage level. As soon as that was good, I delivered the AC to the initial segments (lines and text blocks) to see whether the panel would glow.

A second phase of testing would statically route AC to a chosen set of segments for digits and signs and other features, to be sure that the crack in the top glass layer has not compromised the function of the panel.


The modules and the EL panel all use a special connector that was used in the Apollo program and a few other aerospace applications for a short period of time in the 1960s to early 1970s. The company who made them is Malco and the connector is a Mini-Wasp. They are essentially unobtainium, although Marc's company built a run of these for our use while restoring the AGC.

Power supply module with 12 MW male connectors

Relay module with 137 male MW connectors

Apollo EL panel with 168 male MW connectors on rear
Given how precious they are, I won't use them until a final 'production quality' connector is ready for long term use. In the interim I have found some flimsy connectors that I can press into service for testing and temporary connections.

makeshift connector for testing and prototyping

Rectangular cross section similar to shape of male MW pin

Relay contacts to control the sign segments

The relay contact DS1B1 in the top picture is hooked to the B segment of the middle picture. The relay contact BS1A1 in the top picture is hooked to both A and C segments of the middle picture. These relays are controlled by selecting rows 1 or 2, then powering the SET line to latch the relay to power the segment(s) or the UNSET line to switch them off.