I have been searching for a good 3-axis accelerometer. As usual Phidgets has one, and it is great. Now I need to measure rotations as well, how about a 3 axis piezo-electric gyro, with a USB interface.

Or better, both together with a USB interface. That would make sure that the axes are aligned in the factory. Great device for robots or anything that is trying to measure movements. 1d versions would be a great related product. You could build cool Wii sticks with them or lots of interesting interface devices.

Another two cents, I think that was my good idea for the week.

Now it is gone.

Adam

It's a good idea. How much drift in the heading are you willing to accept? (degrees/hour)

Chester

Dimensions:
Carrier board: 2 x 2" (51x51mm)
Overall height: 0.9" (23mm)
Weight: 29g (without battery)

Specifications:
• Three gyro readings (150-300 degree/s max rate)
• Three Accelerometers/Tilt readings (±6g)


Are the specs of the board, thatI was looking at to go on the back side of a wheel controller. The board in question is not very densely packed, and I imagine can be made 1/2 the size with little work. The height would remain similar since you need to have two 2-axis gyros perpendicular. However, smaller would be cool.

Current gyro manufacturing sheet (Analog Devices):
Range +/- 80 to +/- 320 °/s
Sensitivity 0.018 °/s/LSB
TypicalBand Width (kHz) 0.05kHz
Noise Density (°/s/rtHz) 0.05
Nonlinearity 0.1% of FS
Temp Sensor Yes
Voltage Reference Yes
Voltage Supply (V) 4.75 to 5.25
Supply Current (max) 18mA
Temp Range (°C) -40 to 85

These seem to have reasonable drift, and it is pretty good with a bias signal to recalibrate. Drift, would be good to be <0.1%. Something that could be kalman filtered out. Would be an interesting feature if the system had a checking gyro/accelerometer, that could smell drift because it was cross-coupled. [Not sure if that would work, but it would be cool if it could essentially steer back to some baseline when it seemed to get out of kilter. Isn't drift directly related to the input voltage's stability?]

I would really dig a 6DOF system. It would be great in one package.

Sorry, if I was not precise enough in my answer about the drift. It should be better than the SparkFun stuff. AD has much better gyros for similar $.

A

Pick the largest drift rate you would be willing to accept.

0.25 Degrees / Minute
1 Degrees / Minute
5 Degrees / Minute
20 Degrees / Minute
60 Degrees / Minute

How often are you willing to stop moving and allow the bias to be measured and calibrated out?

Chester

Hey guys,

Between 0.25>1deg/min would be ok. I make controllers and visualization package for UAV and semi-autonomous platforms. For use on the UAVs especially, can't stop, unless I can filter out the motion during hover by just saying... It did not happen. That could be a possibilty, just drop the data after a reset bias flag. Would that be suitable instead a stop?

For the controllers it is possible that they have a rest on a table requirement, but if there was a "soft" reset it could be a really cool thing. Again, I am not sure it would work.

Adam

The problem with Gyros is the drift gets worse over time. If you can do a dead-stop (hover is not good enough) for several seconds, every ten minutes, then you can counteract the drift and hold it to less than 1 degee/minute.

The problem we at Phidgets have is that a gyro is an extremely useful and simultaneously limited device. Many people are not used to dealing with error that accumulates continuously. It gets much worse once you begin integrating acceleration -> velocity -> position.

Chester

Yeah, i know about the unrecoverable drift.. This is why I was hunting for something more precise than the SparkFun stuff. I have been monkeying with them. It is not super suitable for attitude correction for a helicopter. I am trying to to estimate a correction signal, but the bias is not linear so the kalman can not really get rid of it.

The flight times are around 1.25h. So I would need to keep the set down requirement to a minimum. Is there an in between?

I am using the cheaper (but they are not throw away cheap) gyros for my headset controller. However, the tough one is still the UAVs. Let me know if you guys have any ideas. Can they be paired and reversed? That way it is a (Gyro- (-Gyro))/2 situation? That way they would tend to filter each other. I kept thinking of countering the gyro signal with the accelerometer's signals. Since you could get 2-axis tilt from it, and use it to sober the drift. Would that be a suitable correcting function?

I really like you guys' stuff, and it would make my code simpler if it was all Phidgets based.

The problem with the sparkfun modules is MEMS gyros need an incredibly good Analog-Digital converter, with all that entails - power supply conditioning, rock solid references. To be reasonably stable over time and position, you need a several-point temperature compensation, and a 3D characterization to compensate for the effects of gravity on the angular rate output.

I love gyros. We've made several prototypes at Phidgets, and I would really like to see what our users could do.

I think the best system is a GPS + 3 axis gyro + 3 axis accelerometer, all welded together with a kalman filter. An application that was generally active could probably knock out the gyro drift with the GPS position. That said, I know little about kalman filters.

Phidgets will make a PhidgetGyroscope eventually. For now, there is no production planned.

Chester

Hi Chester,

Thanks for the attempt. I have been playing with the spark fun stuff and the AD stuff. I have been using with some success a kalmann filter to integrate the rotations and from a separate (phidgets in this case) 3 axis accelerometer and another filter running.. help the drift,by saying... hey you are way out of bounds on position via gps so reset to here. Then cross-checking the angles from the gyros to say "hey pinhead gyro boy... what do you mean that you are x deg off. gravity is that way." I then write a signal bias during a steady state/hover exercise so that the new voltages are this angle then proceed. It seems to work, but I have only a few flights.

I keep wanting to have this cool wear a headset, so you can have the flight controls visible below the headset. That way it is like a real plane, flying first person with real controls. You can also do some cool switching from UAV to UAV or UAV to UGV. All dirst person, at the moment for local legal compliance all first person flights are done with a spotter.

Let me know if you have any plans or ideas, happy to work with you guys on it. I am sure you have my contact details I would be interested in 6DOF IMUs for real applications.

I know we have been talking about this for a while now. Since I am using an 888 for the ADC. There is no reason why we could not use a 10b ADC. Would be awesome to not have to stitch together the different pieces by hand.

A nice 3 axis 300deg/s gyro and your 3-10g gyro would be great. They would be easily aligned. Simple driver code could even do basic kalman filtering that would rule. We have things working on flying uav's. This would be easy to setup a self bias correcting driver in one go.

Adam

Any updates on gyros?