I have a hard time understanding that this should be remotely close to the accuracy, speed and stability of a MEMS rate gyro. At those wavelengths, even the slightest variation in pitch will produce large errors.
Care to explain why bias is of any practical relevance when you are only interested in the rate (to stabilize yaw), with a MEMS rate gyro?
Because the bias varies with time and is indistinguishable from a nearly constant yaw rate. There’s no fixed and independent “zero rate” signal or output to know when you’re not turning. MEMS gyros are extremely good at determining short term changes in yaw rate but terrible at giving you an absolute value.
At those wavelengths, even the slightest variation in pitch will produce large errors.
I’ve never heard of differential GPS being used in an aircraft that way, and there are easier ways to achieve the same goal, but it would probably work, We don’t care about noise or short term errors caused by a change in pitch. We combine the short term rate change accuracy of the MEMS output with the short-term noisy but long-term stable GPS measurement.
The best analogy with which you’re familiar is probably using a magnetic compass to reset a heading indicator. The HSI is short-term stable and good in turns but it drifts with time. The sprit compass is reliable hour to hour, but hard to make an accurate reading from in turbulence. An even closer analogy is a gyroscopic HSI with a flux gate magnetometer to drive the corrections.
One thing which may be relevant is that in jets you definitely want really low long-term yaw – because yaw wastes fuel.
I recall being in the cockpit of a DC10 (pre 9/11 obviously) and they said there is an accelerometer with a 0.001G resolution, sensing along the longitudinal axis of the aircraft, and they use this to tweak the yaw to get the maximum speed in cruise.
In light GA, people routinely fly with tons of yaw and all kinds of other stuff. Half the planes are badly rigged anyway, with one flap hanging down more than the other, which gets compensated by bending aileron trim tabs, etc. The yaw is measured by a primitive ball and most pilots don’t pay much attention to that anyway.
If you have accurate fuel gauges, when flying alone (or otherwise unbalanced e.g. 3 people total) you should run down the LH tank by a certain amount, so the two ailerons are equal. Then trim the rudder for a centred yaw ball. This should produce another 1-3kt.
On a well equipped jet, one can achieve wings-level i.e. zero roll (in cruise) fairly easily just by sensing the gravity vector, and anyway the drag curve is very flat around the zero-roll region. And pitch is whatever it needs to be for the flying conditions. But there is no way, with accelerometers of any kind, to measure long-term yaw, yet any long-term yaw will be costing you fuel.
The only other way I can think of doing it would be to measure the air pressure difference between two static air vents, on opposite sides of the hull. However that won’t generate much of a signal around the zero-yaw region…
So I can see why they used differential GPS to get the aircraft flying really straight.
Well, I just wonder about how a pair of differential? GPS work, since they presumably can replace a yaw rate gyro. I would think a GPS of any kind would be much more long term than any MEMS rate gyro.
Peter wrote:
The only other way I can think of doing it would be to measure the air pressure difference between two static air vents, on opposite sides of the hull. However that won’t generate much of a signal around the zero-yaw region…
If it can’t generate much of a signal, then the difference in fuel consumption can’t be much either. I would think perhaps the problem is that no aircraft is 100% straight and is rigged to 100% accuracy. So the yaw with the least fuel consumption does not need to be at exact zero degree yaw through the air. With one propeller, this will also make things more complicated.
Because the bias varies with time and is indistinguishable from a nearly constant yaw rate.
I don’t understand the logic. If the yaw rate is constant (no angular acceleration), there is nothing for the yaw damper to do, ignoring sideslip control which is a separate function to dutch roll damping and doesn’t use the gyro anyway.
If the aircraft is in a constant rate turn, you don’t want the yaw damper trying to prevent your turn, so it will not respond to that condition except deal with sideslip.
So there will be a high-pass filter on the gyro signal, and the bias is not an issue.
Indeed, I suspect that Embraer (who will surely be familiar with about 80 years of history of yaw damping) have done this for removing a steady-state yaw, rather than for removing yaw oscillation which is easily done with a cheap gyro, and while at it they used the same solution for removing yaw oscillation (software costs nothing) 
This isn’t complicated, and whether anyone understands or agrees with it or not, it’s still true. An AHRS has to provide an attitude solution for the aircraft. It can’t do that with MEMS alone, it needs external steady references to calibrate and remain calibrated during flight. If those external references were steady and smooth enough then there’d be no need for MEMS gyros. But they’re not. But MEMS gyros alone won’t do it.
Your AI uses gravity as an external reference; your Heading Indicator uses the compass as an external reference. For reasons already explained by me that’s not sufficient for a MEMS system so GPS is used to fill the gap.
Referring specifically to a yaw damper: it doesn’t merely" damp yaw" – it’s part of a whole system that adjusts the rudder to keep the ball in the middle. Quite obviously it can’t do that if if thinks the aircraft is continuously rotating about all three axes at the same time because the gyros aren’t mulled properly.
Of course if you want to pony up for three ring laser gyros, you can dispense with GPS. MEMS just ain’t that good.
dnj wrote:
If the yaw rate is constant (no angular acceleration), there is nothing for the yaw damper to do, ignoring sideslip control which is a separate function to dutch roll damping and doesn’t use the gyro anyway.
I’m not rated for it, but IIRC, yaw damper on the Phenom 300 does both and I don’t recall any degraded mode where it would do just one of those. If any input the yaw damper needs becomes invalid, then it disengages.
A yaw damper is effectively two separate systems.
One damps the Dutch roll (oscillatory) movement, and requires yaw rate information but not attitude.
The other is sideslip control (ball in the middle) and by definition you only need a single accelerometer as an input for this. No gyros and definitely no GPS. Some simple yaw dampers only do this without doing the Dutch roll bit.
So the Phenom GPS failure issue can only apply to the former, and I still don’t understand why this would ever need an attitude solution.
Peter, what do you mean by “steady state yaw”?
