The Mystery of the Golden Gnat

A few months ago, intrepid MITERers discovered that a number of Honeywell GG480 “Golden Gnat” rate gyroscopes had been released to the surplus market. Taking the opportunity of a sale at Electronic Goldmine, we snagged a box of 10 just in case we ever decide to make cruise missiles…uhh, animatronic teddy bears.

Yeah.

That’s it.

The box proceeded to sit in the same spot for the entire summer because nobody could figure out what wires to connect to where. Like all the cool electronic components I have had the luck to deal with, datasheets are nowhere to be found, and I don’t possess the course 6 testicular fortitude to find out.

Recently, information has been slowly surfacing regarding how these gyros function. The most active discussion is on the RCGroups DIY Cruise Missile Unmanned Aerial Vehicle forum. Using this as a springboard, and armed with everything that the Intergoogles could supply me, I decided to attempt an extraction of useful data from the units so I can finish my cruise missile DIY balancing vehicle by IAP.

Here’s a picture of the little buggers. Each gyro is a gold-plated cylinder about an inch across and two inches long.  They are single-axis gyros.

I used the free circuit that Electronic Goldmine supplied to drive the gyro rotor. I made one change to the frequency of oscillation, to bring it to about 750Hz, because faster is better…to a degree. Interestingly enough, if the input frequency is too high, the gyroscope takes forever to spool up. It turns out that the 500 to 700Hz range brought the spinup time to around 5 seconds. I’m fairly certain higher rotation speeds improve the sensitivity.

These things are neat. You can actually hear the rotor spinning internally – it sounds like a tiny, gold-plated cylindrical  hard drive.

I had an adventure getting the halfbridge output to work because it turns I was inserting the PNP output transistor backwards. And here I thought that PNP BJTs and P-channel MOSFETs just worked with opposite input signals. Nope, it turns out they conduct backwards too. This is going to fuck with my mind so much, because I’m emotionally traumatized by plugging things in backwards.

The two outputs of the circuit create a square wave at the half bridge output and a what-the-fuck wave behind the capacitor which looks like wants to be a sine wave badly but just can’t manage it.

The nature of a capacitative circuit, in which voltage lags behind current by 90 degrees, says to me that the gyro rotor is like a single phase, capacitor start/run induction motor. It wants a sinusoid voltage input at one terminal, and a delayed voltage input of the same frequency at the other. Alternatively, it resembles a bipolar stepper motor drive.

Either way, it wants two inputs – everything seemed to operate fine with square-what-the-fuck-wave drive.

And here is an output signal.

According to RCGroups sleuths,  the gyro is built around a variable differential transformer. Essentially this means it will return you a scaled and phase-shifted version of your input signal depending on movement of the transformer core, which is connected to (usually) a measuring tool. They are known for being obsessively sensitive. In this case, the variation comes from what I presume is the displacement of the rotor when the case is rotated.

The input (channel 1) sine wave is 10 volts p2p at 5000 Hz, and the output (channel 2, really fuzzy) is roughly 0.4 volts p2p.

So this is how the signal varies.

At maximum positive angular velocity (Ω), the sine wave increases in magnitude to a point of saturation. The ch2 phase, relative to ch1, remains the same.

At  maximum negative Ω, the magnitude decreases to another point of saturation, but the phase flips. Notice that the peak of the output is now on the other side of the peak of the input.

At a certain negative Ω, the output is flat. The output wave does not time-shift at all, it just flips across the X-axis after going to complete zero.

A video is worth (one thousand * frames per second * number of seconds of video length) words, so here’s a video of me wiggling the gyro. Watch the fuzzy output.

Well, this is certainly troublesome. To use the gyro properly, I would have to monitor both the phase AND the magnitude, because certain output magnitudes correspond to the same Ωs in opposite directions. Worse yet, it’s not like the signal is centered around zero or anything – zero is actually some negative Ω.

Not very useful, so I started playing around with frequencies and what not. Here’s where I discovered that the gyro won’t do anything with a 24,000Hz input on the rotor. I’d need a VFD or something to get it that fast.

But an interesting phenomenon occured at a VDT input frequency of about10,000Hz (with the rotor back at the usual 700hz frequency). I noticed that the -Ω saturation magnitude began decreasing with increasing input frequency. At around 10Khz, maximum -Ω resulted in a totally flat response. The maximum +Ω voltage increased corresponingly.

Hey, this is something useful. A nominal magnitude at zero Ω, a high magnitude at maximum +Ω, and zero at maximum -Ω. If you rectified the sine wave, you’d get a constant Ω=0, +Ωmax, and -Ωmax voltage with linear variations with Ω in between.

Unfortunately, the intervals seemed to be unequal. That is,  the voltage difference from  Ω=0 to +Ωmax was higher than Ω=0 to -Ωmax. Further observation of the frequency response showed that these intervals changed as the frequency was increased. So I tried to find the frequency where the voltage swing would be approximately equal.

The magical frequency was around 18Khz. At this point, the voltage deviations from Ω=0 were (to my eyes and MITERs’ ancient oscilloscope) essentially equal in both directions. The output sine wave no longer flipped, nor did it reach zero, but it only deviated from the rest amplitude slightly.

Here’s a video where channel 1 is initially suppressed so the full range of the voltage swing is variable.

I threw an op-amp at the roughly .5 volt p2p output signal so it was in a more useful range.

So now I know that at about 18Khz sine wave input, the output of the gyro is a sine wave of a certain magnitude at rest. Moving the gyro about its sensitive axis changes the magnitude positively or negatively to relatively equal voltage limits above and below the rest voltage, but the output is still a sine wave.

This is now alot more useful to me, because the signal can be rectified into a positive DC voltage, then the “zero voltage” subtracted from that. Then, a positive angular velocity will yield a positive DC voltage, and a negative one vice versa. And zero volts should be zero movement. There’s probably a less Byzantine way of reading the signal, but that’s a first-order guess using my limited EE knowledge.

So, the bottom line is, so far…

A +/- 15 volt power supply and something that generates and accepts sine waves is needed.

Black wire: Ground. Legit ground.

Red wire: One rotor input. According to the Interwebs, this should be a 400 to 700hz square(sine?) wave.

Green wire: Other rotor input. One of the two should be capacitatively coupled to the other. Here I assume that depending on which lead gets the capacitor, the rotor spins in opposite directions.

Blue: Signal ground. I tied it to Legit Ground™

Yellow wire: VDT input, referenced to Blue.Takes a sine wave.

White wire: VDT output. A chopped and screwed version of your input appears here. Intepretation is an exercise left to the EE-gifted.

First person to build a cruise missile cool gyroscopically-guided thing wins!

Also, comments, corrections, and additional information is welcome and encouraged. I don’t doubt at all that I misinterpreted or abused something, so please point it out!

Swapfest Revisited

SWAPFEST!! (capitalization and emphasis added) is the monthly flea market that MIT hosts for area radio enthusiasts, electronics tinkerers, cruftseekers, and frightened but daring onlookers. It’s more or less a hybridized radio hamfest. Because it occurs in the equivalent of my back yard, I try to attend every month.

Occasionally I fail because it involves waking up during the daytime, but historically speaking, Swapfest has been a boon for my cruft-hoarding tendencies. It has turned up stuff as weird as the alleged gyroscope gymbal out of a B-52, to mutant semiconductors to literal dozens of old 5.25″ full-height hard drives.

The season runs from April to October, because nobody wants to push their cruft while it’s snowing outside. The first and last Swapfests always seem to produce a more eclectic selection. I figure this is because nobody, in addition, wants to drag all their crap back home and store it all winter, so they dispose of it in October. Likewise, April is the time when everybody brings all the crap they collected all winter.

Following this pattern, I abandoned all inhibitions for this past Swapfest and actually… you know, brought some money. Apparently, the incremental gain in awesome parts when you go from “free” to “cheap” is nontrivial. This resulted in the best Swapfest hual in recent memory. And it even includes stuff I can actually use. Imagine that.

Naturally, I sprung for more mutant semiconductors first.

I buy the guy out of these every time I see him. They are the same ginormoFET modules I use on LOLrioKart. I wish they were higher voltage, but regardless, you just can’t beat 200-300 solid amps, 2 milliohms, and 100 volts for $7.

The downside is the incredible 0.1uF gate capacitance, and the lack of a real datasheet. The part number apparently only exists in the archives of shady Far Eastern electronic component dealers.

I decided to invest in some dial meters for Segfault. The variety of things measured ranged from the relatively tame “PERCENT OF RATED CURRENT” to … leak rate. How do you measure leak rate? Dribbles per minute?

Ultimately, no matter how weird the scale, these things tend to be voltmeters in disguise. Finding a straight up ammeter (voltmeter with a scaled reading) was actually pretty difficult considering most of the meters read kilovolts or milliamps (or milliRems per hour). I finally wound up with a 0-50 amp scale meter. There was also an oddball that showed both positive and negative amps. The weird thing, however, was that it required a voltage just to read zero. I thought it was a dial zeroing issue, but nope… the little adjustment screw doesn’t crank that far. Maybe it will be handy some day to read regenerative braking current.

You’d figure an easier way to do this is to center the scale at zero, like the +/- degrees meter. I liked the degree-o-meter  because the scale is very similar to the numbers I want to pursue for Segfault. 15 degrees forward tilt gives a comfortable torque margin for the motors to recover if it becomes necessary (The calculated maximum steady state tilt angle is about 30 degrees, which is substantial). If I ever get to 45 degrees, I’m probably on the way to a face/ass/extremity-plant already.

Speaking of meters, here’s a 750 amps DC meter. With the associated shunt, a very precise 3 and a half pound block of brass. Ammeter for LOLrioKart to go along the amp-volt-microns meter?

While on the topic of mutant components, here’s the functional component inside a variac. To me, this looks like a meaty inductor to use for a high-power boost converter. It weighs about 10 pounds and is 8 inches across.

Using an LCR meter, I found that the inductance of this toroid changes significantly with frequency. At 1kHz, the inductance was around 170 milliHenries. This fell to 60 at 10kHz, but was pretty damn close to 1 full H at 100Hz.

Considering the thing was designed to operate at 60hz, I’m not too surprised. If nothing else, I now have yet another door decoration.

Enough EE shenanigans. This is what happens when you let Charles near a bin of unsorted endmills.

He takes 20 minutes and multiple rusty-tool lacerations to the hand to mine out all the carbide cutters.

You can easily distinguish a carbide endmill from a steel one just by weight. The final tally is one 0.5″, one 5/8″, one 3/4″, and a huge 7/8″… carbide tipped cutter? I didn’t know carbide tipped endmills existed,  but… well, now I know. This looks like it dates from before people figured out how to make whole cutters from carbide.

A carbide spotting drill also falls into the mix. The whole load cost me a few Mountain Dews Equivalent (MDe, a new unit of currency used only in my universe).

The most exciting and highly valued find of the day was this Tapmatic self-reversing tapping attachment for your choice of spindle machinery. They drive a tap through a clutch and automatically pop into reverse, backing the tap out, if you start lifting the spindle up. The reaction arm stays stationary and hopefully braced against something to allow for the reversal.

I investigated these things a long time ago to use the underlying mechanism in a stored-energy flipper ‘bot design, but could not understand what went on in them at all, because smart people designed it.

I still don’t know, but now I have an excuse to powertap things smaller than 1/4″-20.

Now my goal is to find a use for everything before April!