Sunday, May 31, 2020

Further testing of power supply shows it healthier than originally thought


When I looked at the effect of ramping up the two voltages provided to the power supply module, +14 and +28, I discovered that there was no point looking for a higher level supply as I originally thought. Changing the +28V supply had essentially no effect on the output AC voltage of the power supply module. It was, in fact, the +14V supply whose boost would visibly brighten the electroluminescent panel.


I then hooked up some digit segments and noticed that the lines and text legends I had lit up were intentionally dimmer than the numeric digit and sign segments. In fact, the brightness was quite acceptable with input voltages of 16 to 17V. 


Each digit on the panel has seven segments with a pin on the rear for each segment. With 21 digits, it takes 147 pins to drive the segments. Each of the three signs requires two pins each, then the line/text features take a pin, the COMP ACTY area takes a pin, and a few pins are assigned to provide ground. The panel as 160 total pins on the back, constructed with the Malco mini-Wasp connectors.

The power supply module needs 7 pins between +14, +28, 800 Hz input, ground and the AC output pair. One side of the AC output is routed to the panel ground pin, while the other AC output line is hooked to between 1 and 155 pins to light up the desired output.

In between these two modules, power supply and electroluminescent panel, the DSKY uses four relay modules. These modules will connect the AC output to the individual segments on the panel, latching in the connections so that the digits and signs stay lit using many tiny latching relays. The guidance computer sends commands to select a particular row of relays and feeds a word to them that set or reset the various relays in that row. 

Each relay module has 137 pins on its connector. There are other circuits besides those that light segments on the electroluminescent panel (obviously since they collectively have 548 pins while the panel has only 160). Adding to the complexity, the seven segments for each digit on the panel are controlled by only five bits coming from the guidance computer. 

I currently have one relay module which I have fully tested and found completely functional. It can drive only a portion of the panel, for example, six numeric digits out of 21 plus one or more sign segments. To do even this subset, I need more than 45 connections from relay module to panel, seven pins for the power supply, and quite a few control and input pins to cause the relay module to latch various values into the digits. 

There are no sources for the mini-WASP connectors, other than a modest number that were built at great expense by Samtec to support our restoration of the Apollo Guidance Computer. I am loathe to deplete the supply building permanent connectors to produce a demonstration with six digits, signs with my one relay module.

I have tried to use some connector pins I found that form a U or open box shape slightly larger than the mini-WASP size. The can make a loose connection but easily pop off. Trying to wire up more than a hundred to link the three modules is very difficult.

Our custom built mini-WASP pins have a wire-wrap tail, which potentially could be used to temporarily wire these together. The connectors could be used for another project afterwords. The way that a 137 or 160 pin connector is built using the mini-WASP is unusual.

Each pin has a nylon holder that is pressed into a hole drilled into an aluminum plate. The holes provide the spacing and arrangement of pins into the desired grid. Pressing a pin in the plate commits the pin to that use, but I want these to be fully available for future projects. Therefore, I need to find a way to arrange the pins and hold them in place without using a press fit. If I can solve that, I can hook up the demonstration. 

Tuesday, May 26, 2020

Attempt 2 at an alternative driver for 800 Hz to the power supply module of the DSKY


If I had a decent NPN transistor that would be the most straightforward way to drive the circuit. With the virus lockdown, I can't go out and buy any parts but I may have something in the garage that would work. With a bit of hunting I located a 2N5551 and wired it up.

This is equivalent to the interface circuit in the Apollo Guidance Computer that provides the 800 Hz drive for the power supply. It is an NPN transistor through a 2K load resistor This restricts the current to a maximum of 2.5ma when tested with 5V but using the spacecraft power level of 14V it can sink up to 7ma, a tenth of a watt dissipation.

Up first was a direct test of the driver transistor delivering the appropriate 800 Hz signal at a few milliamps current, which is well below the max I want to drive the single transistor. It was breadboarded and hooked to the scope, with exactly the results I wanted.

Then I wired it to the power supply module through this driver circuit. The function generator gave me the 800 Hz square wave, then the driver transistor will sink the 14V to ground, after it passes through the primary of the first coil inside the power supply module.


For this test, I hooked the power supply output to a segment on the electroluminescent (EL) panel to provide a proper load. I wanted to monitor the output waveform and measure the voltage using the scope. I started with the potentiometer wires open to reflect the cabin lighting pot set to max brightness. I ccould then insert a fixed resistor or short across the wires to simulate various dimmed settings.

The segments were no brighter than before - pretty dim actually. Their wasn't enough oomph to light the fixed lines and legends with the standard input voltages of +14 and +28. I had to see what the output voltage looked like on the scope. It should be 800 Hz AC with an RMS value of about 275V.

The high voltage is isolated from the input side and spacecraft ground, thus the scope has to be grounded only on one side of the output. I turned on the test bench and observed the output waveforms. It appears to produce a reasonable waveform but a bit low, just as the previous circuit did.
800 Hz with 250V peaks, therefore below desired level
The math in the scope isn't doing well interpreting the RMS AC value nor the frequency since it isn't producing clean sine waves. Looking at the time between peaks of about 1.25 ms confirms that this is operating at 800 Hz. The peak voltages are a bit above 200V each way which, with sine waves, would be an RMS reading of perhaps 145V. With no segments active it jumped to 240V each way and an RMS equivalent of 170VAC, consistent with the dim illumination.

Since there seems to be excess resistance in the ground side of the EL panel, that voltage drop might be confusing the current compensating circuit in the output, where the supply could be working just fine. To test this, I left it disconnected from the EL panel and measured the output voltage. Since the peak to peak approached 480V, it shows there is some effect but not enough to account for the low output.

At worst case, I will have to work with the El panel in a darkened room to get acceptable brightness. There are two other possibilities to brighten this up. First, I could recreate the entire circuit in discrete components, ensuring I can reach the voltage targets that I want. Second, I could find a way to 'amp up' the power supply output. The discrete component alternative is quite hard because of the specially wound transformer that is used to ensure constant brightness with a variable number of EL segments illuminated.

I can pump a bit more out of the circuit by boosting the supply voltage to the power supply module, although it introduces the risk that I will exceed the max voltages of the transistors inside causing a failure. Because of this, I can't ratchet the voltages up too much, but I should be able to go up about 25% with some safety. This was designed to operate in a spacecraft which did have some transients and high side excursions in power, thus the transistors would not have been chosen right at the margins.

Bumping up the +14V supply to 17.5 volts and the +28V supply to 35 volts would implement my 25% boost.  I gave it another try, with 17.5V and 32V since my supply can't produce the full 35V.  The segment was noticeably brighter even without the supply generating the full 275VAC that it should.

Partially boosted voltages improved brightness
I am not sure of the cause of the anomaly on the positive peaks. I decided to reverse the connection just to see if the problem shifts or is a scope/measurement artifact. Indeed, the strange waveform flipped. This is definitely a characteristic of the module under test.

Scope leads reversed, symptoms also reversed

I need to pull out another power supply to inject the full 35V for my original brightening plan. This is not readily at hand but the current illumination is adequate for viewing indoors in normal room conditions.

Decent illumination level