The Golden Gnat

As my first blog entry, I have decided to write about the Honeywell GG440A18 “Golden Gnat” mechanical rate gyroscope and how to get it working.

According to unofficial information, this cold-war era device has been used in various missile and radar systems since the 1960s. US Dynamics Corporation (USD) now make “Form, Fit and Function” replacements [1] for the Honeywell GG445 series which I can only assume are very similar to the GG440.

These gyros have been circulating internet surplus stores for some time. I purchased mine from Goldmine Electronics [2] late last year. They (I bought three) are dated June 1981, so this surplus stock has been around longer than me. Nevertheless, they appear to be in excellent condition, and indeed golden.

The device uses a polyphase hysteresis type AC motor to spin a mass (the rotor itself) within a gimbal structure, whose position controls the coupling between the windings of a differential transformer, known as the pickoff. By exciting the primary pickoff winding with an AC signal, the angular rate and direction are deducible from the magnitude and phase of the current induced in the secondary side windings [3][4]. I figure this is typical of a mechanical rate gyroscope, but is certainly very different to the MEMS gyros I’m used to 🙂

Anyhow, a couple weeks back I finally decided to bring one of these little golden critters to life. Here is how I went about it…

Step 1: Getting the motor to spin.

The gyro’s motor requires two AC bipolar drive signals separated by 90 degrees at presumably 400 Hz. As I happened to have a couple of STMicro’s L165 power op-amps at hand, I went ahead and built a simple direct-drive sinusoidal quadrature oscillator for this purpose:

You might be wondering why the six zener diodes? The purpose of these is to prevent saturation of the waveform. I also happened to have exactly six 3 V zeners in my drawer at the time. Also note that resistor R8 was added to initiate oscillation on power-up, and that the op-amp output snubber values were arbitrarily chosen and probably have no real significance.

Not the most stable design in terms of frequency drift, but for a test circuit I was satisfied. I think the following screenshot speaks for itself:

Step 2: Exciting the pickoff.

The primary winding of the pickoff is driven by a single phase AC signal. I couldn’t find any specifications for the Honeywell GG440’, so it was up to the generic guidance of USD documentation and a bit of experimentation to work out.

Typically one would want to oscillate the pickoff at a frequency much higher than the response bandwidth of the gyro to avoid aliasing. I figured this would be maybe 50Hz at best, though highly unlikely. Additionally, the inherent cog-free (cogless, uncogful, without cog?) motion of the hysteresis motor practically eliminates the impact of motor drive frequency on sampling error. With this in mind, and in my initial haste, I simply tapped off one of the motor drive phases to provide an excitation signal for the pickoff winding. This was simple and worked OK.

By this stage I had already advanced to Step 3 by building a demodulator, so it was time to decide on an optimal frequency to drive the pickoff. It was an iterative process, and as you’ll find out, a 0 or 180 degree phase relationship between input and output was desirable in addition to maximizing the output signal swing.

Equipped with a signal generator (my computer) and an amplifier, I was able to sweep through a range of frequencies up to 10 kHz while wiggling the gyroscope and monitoring the input-output response.  Coincidentally, the largest magnitude response and suitable phase shift were both observed at about 1 kHz, without capacitive tuning. It was time to build a computer-free circuit:

Here I have modified the typical op-amp based Wien bridge oscillator to provide a higher current output. Resistors R(1,2,3,4,10) and diodes D(1,2) adjust the output voltage swing and prevent op-amp saturation. C5 provides the bulk of the output feedback, while R7 is to help with initial start-up. Resistors R(8,9) relieve the BJTs of excessive power dissipation by sharing some of the waste. Did I really need to say all that? Am I rambling on too much?

The following screenshot shows the waveforms produced by this circuit while powering the primary pickoff coil:

The visible periodic glitch is caused by a combination of the pickoff inductance and the limited speed at which the circuit can react to the dead-zone caused by the base-emitter junctions of the transistors. A snubber would alleviate this issue.

Step 3: Demodulating the output.

In order to report both magnitude and direction information of angular velocity, some kind of phase sensitive rectification of the pickoff output is required. This can be achieved by using a couple of bipolar transistors and a differential amplifier to work with the synchronized input and output waveforms of the pickoff:

(Adapted from [5])

By studying this circuit, it becomes apparent that for the best results, i.e. maximum sensitivity, both the pickoff windings must be either in phase, or 180 degrees out of phase. A 90 degree separation is worst, but can be corrected for by using a lead or lag compensator on the pickoff primary sampling input. To show why an out of phase relationship is undesirable, compare in the following illustration, a normalized pickoff input (red), pickoff output (blue), and how the rectified difference (filled in green) affects the average output magnitude (purple):

Referring back to the schematic, it is also worth noting that because BJTs have been used here for the rectification, there will be some dead-zone around the waveform zero crossings, so the pickoff oscillator need not have perfect zero-crossing characteristics – something I hadn’t thought of at the time. The rest of the circuit is just offset nulling, some gain and some filtering.

Results

The output bandwidth of the demodulator is limited to about 3 Hz, so the time scale of the following graph covers 10 seconds to show the relevant information. Here I’ve tried to rotate the gyro by hand at a constant rate back and forth 180 degrees while pausing in between turns:

From the rate output, it looks as though I’ve consumed too much coffee, but after integrating the output to give angular position, you can see that I didn’t do too badly after all. If this isn’t convincing enough, perhaps the following video will give a better idea of the response:

Final words?

Getting the GG400’ mechanical gyro to provide a useful, legible output is not a trivial task. While focusing on the basics, many design aspects such as stability, efficiency, and elegance were pushed aside. The motor controller works well in a test environment but suffers from drift, while the pickoff oscillator and demodulator could be combined and simplified. Maybe one day I’ll improve on these things, but for now I’m satisfied and I hope you are too.

References

1. US Dynamics Corporation, “Rate Gyroscopes: 475 Series”, US Dynamics Corporation, December 18, 2006, [Online]. Available: http://www.usdynamicscorp.com/literature/general/475_series1.pdf [Accessed: February, 2010].
2. Goldmine Electronics, “SALE! Missile Rate Gyroscope”, Goldmine Electronics, http://www.goldmine-elec-products.com/prodinfo.asp?number=G17270 [Accessed: February, 2010].
3. US Dynamics Corporation, “US Dynamics Model 475 Series Rate Gyroscope Technical Brief” (USD-AN-001), US Dynamics Corporation, December, 2005, [Online]. http://www.usdynamics.org/whatsnew/news/USD%20News%20-%20US%20Dynamics%20Rate%20Gyroscope%20AN-0011.pdf [Accessed: February, 2010].
4. US Dynamics Corporation, “Spinning Mass Mechanical Gyroscopes” (USD-AN-005), US Dynamics Corporation, August, 2006, [Online]. http://www.usdynamicscorp.com/literature/general/AN-005%20USD%20Spinning%20Mass%20Gyroscopes.pdf [Accessed: February, 2010].
5. US Dynamics Corporation, “Basic Output Generator Excitation and Demodulation Requirements for US Dynamics Miniature Gyroscopes” (USD-AN-004), US Dynamics Corporation, August, 2006, [Online]. http://www.usdynamicscorp.com/literature/general/AN-004%20USD%20Pickoff%20IO.pdf [Accessed: February, 2010]