B-52 Bomber Astro Tracker - Part 3: Figuring out the optical system
Hello and welcome back. If you follow the channel, you know that we opened up a Kollsman Astrotracker, with the welcome help from our friend Michel, from the "Labo de Michel” YouTube channel. This astrotracker is but only one part of the complex Kollsman MD-1 astrocompass, a once secret military astro-navigation system that was flown on B-52 bombers. When we cracked it open in the first episode, we were greeted by an amazingly complex arrangement of electro-mechanical components. Designed in the 1950's, it used analog computing and synchro servo-motors techniques.
Our star tracker is in essence an automated sextant. It is able to sight, center, and maintain lock on a star, or the sun, while the bomber is flying. It then provides a continuous true heading information to a classic inertial navigation system, as well as a running position error estimate. This dramatically improved the navigation accuracy over long flights. We have only scant information about the system, mainly in the form of a partially helpful patent. In the previous episodes, Michel was able to reverse engineer enough of the beast, that we were soon able to power it up. [Marc] So, are you ready? [Michel] I'm ready! [Marc] Enable... Oh, it works!
We then found a way to connect the tracker to an external vertical gyroscope and some home-made electronics, as we were missing both from the original system. And lo and behold, we had a gyro stabilized tracker platform, which would stay pointed straight up while the plane is flying. The next step is to reverse engineer the optical system and figure out how to get a star signal. From the patent, we expect that the signal should be modulated in a somewhat unclear and complicated way. If we manage to understand it, we'll hope to recreate the demodulation and servo electronics that we are also missing. Only then will we be able to lock and track a star as originally designed. One more note before we start, given the difficulty of the restoration, Michel and I had abandoned English and reverted to our native French, so I'll do my best voice acting to dub my own video.
[Marc] So, for the next step, we flipped the contraption. We have a light that shoots into the prism. It shoots back up, we removed the photo multiplier tube, and we look at what comes out. And what you see is a combo of a rotating aperture and a disc, a chopper.
How do you say chopper in French? Un hacheur. They are both coupled together, so I can't rotate one without the other. So the chopper modulates the light at 48 kHz, and the disc does it at 400 Hz. And since this is intelligent dubbing, I can add that we got the frequencies completely wrong, as we'll see later. So, this chops the star signal, it modulates it.
And that allows to find the position of the star, to center it apparently. So I hope you all appreciate the scrappiness of the setup, I reused my sponsored Andonstar digital microscope to provide us a magnified view of the optical output. Last time I made a microfiche reader out of it. I keep finding unexpected uses for this inexpensive microscope aid. So Michel hooked up the chopper AC motor to the 400 Hz supply, and it started to spin, but not making the most reassuring of sounds. At this point, I should probably interrupt for an elevator music section, to explain what we think is going on with this steampunk gear and aperture system.
As we saw in the previous episode, the optical system consists of a fixed telescope with a rotating prism on top, a star periscope of sorts. The light detector is a Photomultiplier Tube, a vacuum tube which is still one of the most sensitive light detectors out there, able to count single photons. However, it is not an imaging sensor. It just tells us how much light we are receiving, but nothing about the exact position of the star within the field of view. So, how do extract the star from the background and get the position information from a single pixel detector? Well, you don't have many choices, you have to somehow encode the position in a temporal modulation of the signal.
That's what our two rotating apertures do. The first one, with the fine teeth, is the high frequency chopper, or raster as they like to call it. It's a background light removal system. When the star image is focused on the disk, the image is small enough to fall in-between the lines, or be obscured by a line. As the disk rotates, the light from the star is blocked by the lines at a steady frequency. In this case, as we found out later, producing a tone at 190 Hz.
The sky background, even if not uniform, will not produce any modulation as long as the non uniformities are larger than the chopper line spacing. So, if you filter out the 190 Hz frequency, you have isolated the signal from a star. Meanwhile, the second rotating aperture, the circular one, encodes the star's position. It wasn't clear to us how that worked, until we realized that this rotating circular aperture is not in the focal plane. Instead, the image of the star is de-focused when it passes through this aperture, forming a circular blob. If the blob is centered,
the aperture always samples an equal amount of light and there is no modulation. A steady, continuous, 190 Hz tone is generated by the subsequent chopper. But if the star is off-center, the de-focused blob also moves off-center. An amplitude modulation of the tone appears. The further off-center, the stronger the modulation. So you end up with a signal like that of figure 7B of the patent, with a 190 Hz chopped signal modulated by a 30 Hz wave. The stronger the modulation, the further off center you are. If the star is dead center, there is no modulation, just a steady carrier, and you are golden.
And now that you know how far off center you are, how would you know in which direction to go? Easy! You simply look at the timing of the AM modulation. The peak of the modulation wave will happen when the circular aperture is in the direction of the star. In other words, the phase difference between the disk position sensor and the peak of the modulation tells is your offset angle. Which explains why we found a disk position sensor monitor in the first place, to allow for that phase angle measurement. Now it's up to the follow-on electronics to demodulate the amplitude and phase information, and derive an AC error signals for controlling the synchro servo motors. Which is far from being obvious if you think of it.
But for now, Michel and I are interested to see if we can get a star signal out of the photomultiplier at all. And for that, we need a star simulator. Modern lasers to the rescue. [Marc] We got our first star sighting! We made a little star simulator here, with a laser that bounces around. And then this is my star. And after a little bit of difficult alignment, we got it showing up over here. So, here is my incredible star simulator. It's a helium-neon laser, nice bright beam. This is just a glass plate. This is way too powerful,
so I just want to pick one reflection off. This is a polarizer, which I use to control the intensity. Then it goes on this mirror, then it hits this mirror. Then it goes on the star tracker. And I can see where the reflection from the start tracker is.
So, it's a little bit on the left here. Right there. So I'm approximately aligned. And once we have it good, and we can see it, we'll put the PMT back in place. We'll put that in front, that will be our star. And then using the polarizer, I can reduce the intensity, so it's very low. And that will be our little star.
So it's much easier to align with a bright laser light. And then we do the experiment with the photo-multiplier tube, and that little source of diffuse light. Meanwhile, Michel had reverse engineered the photomultiplier biasing circuit. On the left, you see the resistor ladder biasing the multiplier, which we have to power with minus 900 Volts. On the right, there are two tuned pickup transformers for getting the signal out.
We then re-installed the photomultiplier, after we aligned our star and attenuated the laser beam greatly. [Marc] We're locked, okay. So, we have the high voltage supply we repaired in the previous episode. Oh, 1,000 volts, that's no good! 100 volts, okay. Conveniently, we had restored a 3 kV HP power supply in a previous episode, when we played with CRT tubes. We were not brave enough to give it all the high
voltage beans right away, and started with a conservative 100V, to see if we got anything at all. In PMT land, lowering the voltage of the multiplier just lowers its gain. [Marc] Yeah, yeah, there it is! But it's not the biggest signal... The green is the 1 MHz, and the yellow is 48 kHz? At least we get some signal already.
OK, at 100V bias it's not super convincing, but we are getting something. So let's try full chooch with the 900V. [Marc] Go for it Michel, turn on the 900 volts. [Michel] 900 volts activated! [Marc] Voila, so we have the chopped signal So the modulating frequency should be at 400 Hz, and the carrier should at 48 kHz? [Michel] Yeah, normally that should be it. [Marc] So, if I get in front of the laser simulator, I put my hand in front of it, and it disappears! So, this is really our little star, which we had to brighten up quite a bit. So,
I'm not sure we have the sensitivity we should have. [Michel] So now the drill is to synchronize the scope on the 400 Hz that's used as a reference, and see if we indeed have a phase difference between what we see here and the reference signal, if we change the position of the simulated star or the astrocompass. [Marc] Which should allow us to make the servo motor work and do the tracking. Okay, let's try! (Moments later...) [Marc] Well, we have a slight problem here. Michel, what what's happening? [Michel] Well, you can see, on the upper green trace, the 400 Hz reference signal.
And below it, we have the output of the PMT after a filter, and we are not at the right frequency. We are not synchronized. We're supposed to have the same frequencies on the output burst, same as a reference. That means that the synchronous motor is not spinning at the right speed. [Marc] And it's way too slow apparently.
[Michel] And it's not synced at all, you can see it sliding away. Not synchronized at all on the 400 Hz. We need to figure this out.
[Marc] Yep, we can't do a control loop with this. The following day... So, Michel went on to reverse engineer some more and solved the riddle a day later. It turns out that the early patent we were following did not represent the device exactly. In our version of the tracker, the synchro motor rotates at a third of
the speed of the 400 Hz. Moreover, there is a gear reduction train, and in the end the rotation speed of the circular aperture was measured at 48 Hz. We'd find out later that this is not quite right either, but that was our best guess at the time. So, for the phase reference, you have to use the generator that we had found earlier, the position sensor for the aperture. But the designers outsmarted us even more. The signal from the generator is
wired into another device, a resolver. The resolver is mechanically tied to the azimuth of the telescope, and is used to do some trigonometric math. It's an analog computing device. It shifts the phases of the generator, so the reference phases rotate along with the telescope. Clever, this astrotracker, so clever. It's all spelled out in Michel's Indiana Jones notebook, but to simplify, if we take the signal output from the resolver, we should now be able to observe our phase relationship. In Michel we trust.
[Marc] Okay, so, we have the star simulator, shining its little star - over here, where are you?... over here - on the tracker. And it's receiving it. And now we are taking the reference from the generator that Michel found, and it works according to what it should do! [Michel] Now we can see, we are synchronous. [Marc] This time we're good, we have a reference! And if you misalign the star, we should be able to change the phase. It's super sensitive, but you can see that we can move the burst around. How do you say burst in French? [Michel] Une salve.
[Marc] The burst moves compared to the reference. So the electronic circuit is going to do what I do by hand, trying to align the burst around here. That's a significant find, Michel, because it wasn't described like this at all in the patent. Soon after... [Marc] So, Michel found a magic button. It's a solenoid, right? [Michel] Yes, a solenoid, on the side somewhere.
[Marc] We can't see a thing, because we are in the dark. And we are in the dark, because the magic button gains us... So, that was the mode we were in earlier. And we were in the sun mode!
And when you switch the magic button, now you are in the night mode! And my star simulator is now invisible to the naked eye, you can't see a thing. But the detector sees it! So, it seems to be working fine now! So, with zero time to spare before Michel had to fly back to France, he seems to have figured out our tracker. Much later, as in a year later, I found an IEEE paper that described the astrocompass system in more details, and seems to match our production unit. In it, you can see that the shutter motor is not driven at 400 Hz, but at 250 Hertz, by a tuning fork oscillator of all things. So our modulation rate should be at 30 Hz. And the block schematics of the electronics reveals even more analog trickery. Michel has been busy figuring out detailed schematics for the control electronics, so we can activate tracking.
But the story gets even better, as word has it that AvE, our favorite Canadian engineer with a special knack for invented words, is sending us the analog computer we were missing! [AvE] Gentlemen, welcome back to the shop! I was back in the deep storage getting my blender bus, what for taking care of coyotes, they're rather thick this time of year. Here is a couple million dollar worth of ITAR technologuee from the USanA, sent to me by my internet buddy Chris Boden. And he sent one of these also to Dave Jones. Dave Jones had it apart. And it's got all kinds of electromechanical contrivances in there. CuriousMarc down in California, hopefully not burnt right to hell, wanted this for one of his projects - he's got the other bits. So, in the spirit of generosity with which it was given to me by Herr Chris Boden, we're going to send it off to a better home, we're going to re-home this! [Marc] And yep, it made it all the way to California, escaped the Trumpian tariffs.
And it has the AvE sticker (laughter). So, we'll try not to dumb with it, and man, hopefully we'll get to reunite it with the other end of the system. So, this is the pointing computer. Should be able to point the star tracker to the star, from its coordinates. And then, the star tracker should start tracking the star.
Well, thanks a bunch Chris and Ave, we'll try not to dumb with it too much. See you then!
2025-03-08 04:17
