The UAV Devboard is back in stock at SparkFun, with a new design.The reason for the design change is that Analog Devices no longer makes the gyro that we were using on the previous board, so we have switched the design to the LISY300AL. Paul Bizard and I have thoroughly tested the new board on fixed-wing aircraft, the LISY300ALs work just great with DCM.We have modified all of our existing firmware to be able to run on either board, and will make sure that all future firmware will also work on either board. If you have written firmware for the UAV dev board, you will only have to make a few minor changes to run it on the new one. The two board designs are nearly identical. The only differences are:1. The new board uses LISY300AL gyros instead of the ADXRS401, and the 6g range instead of the 1.5g range for the accelerometers. This will allow you to provide aerobatic control without saturating the sensors.2. The gyros, the accelerometers, and the A/D reference voltage are all tied to the 3.3 volt regulated supply. This will totally eliminate drift of gyros and accelerometers due to supply voltage changes.Each firmware project now has both a "green" version for the previous board and a "red" version for the new board.
I made a design with the LISY in it myself, but decided not to make a prototype because of these warnings I read. Personally, I can not think of a reason why the higher sample rate would make any difference, mainly because the vibrating mass is still vibrating at a frequency in the same range as some of our engines. The German experts even tried it with complex analog filtering, but it did not help.
The standard answer from ST is:
---
There's no way to avoid this issue.
4kHz is the resonant frequency of the internal masses. If you excite them, the internal MEMS will get crazy and the output will no longer be reliable.
With other devices (other manufacturers' gyros) resonant frequencies may be very different (from few kHz to several kHz) and hence, you may have not seen the device becoming crazy.
---
The best choice seems to be the AD ones (300 degrees/s), and the melexis ones being a close 2nd best (you still seem to get what you pay for...).
So I would really advise you to do some serious testing in different aircraft/engine setups, to get a better view on this problem.
Unfortunately, I was not able the find the "gyro comparison" document I once encountered.
Basically, what Rob said is right. There are plenty of 3 axis accelerometers, we are using one of those, but I have not seen a suitable 3 axis gyro just yet that can compete with three LISY300s for price and performance.
An IMU needs both accelerometers and gyros.
I have been doing some more testing on the prototype board that SparkFun made for me with the LISY300 gyros. They are performing great. With the new board, you will be able to turn your plane at 300 degrees per second without saturating the gyros, and pull 6 gs without saturating the accelerometers. I am getting closer to one of my goals of pulling a high acceleration, continous, constant altitude, highly banked turn under IMU control.
@David
Those links are not for gyros. I've seen that they now have 2 axis gyros integrated on one chip, but have yet to see a 3 axis one. It will be nice when they do finally come out with a high quality 3 axis gyro IC, at the right price.
I want to do more testing before I give you a definite yes or no about using the LISY300 for helicopters. Tom's information raises a concern, but my recent tests are encouraging that it may simply be a matter of aliasing.
I have a helicopter. I will put using it to do vibration testing of the LISY300 on my list of things to do, though I am not sure when I will get to it. There is a lot on my plate just now. In the meantime, I suggest that you consider whatever gyros that Tom recommends.
First of all, I want to thank Tom for alerting me to a possible issue with the LISY300 gyros that I was able to fix in firmware.
I took a closer look at the prototype board that I have with the LISY300 gyros in it. I noticed that if I tapped the gyros, DCM would "jump". Not good. Not a good sign at all. It would seem that there might be a problem with vibration....
I then took a look at my older board, with the ADXRS401. I tapped the gyros on it, everything was fine. So, at first glance, it would appear that the LISY300 is sensitive to vibration....
But then I thought about it some more...
It finally occurred to me that maybe the LISY300 simply had a higher frequency response than the ADXRS401, either because of the way that it is built, or because of the way the analog filters are designed.
If that is the case, all you have to do is raise the sampling rate high enough, and integrate!!
So I did. I raised the sampling rate from 500 Hz to 5000 Hz per channel. No problem for the dsPIC30F4011.
Now I can bang away at the gyros as hard as I can, everything is fine. I will still have to do vibration testing, but the fact that I could make the "glitching" go away by raising the sampling frequency is very encouraging.
It appears that at the lower sampling rate I was getting aliasing, because the frequency response of the gyro is higher than 500 Hz. When I tapped the gyro, it must have produced ringing at some frequencies that were multiples of 500 Hz. Raising the sampling rate to 5000 Hz and integrating solved the problem. My roll-pitch-yaw demo works great on the new board now.
Tom, by any chance do you know what sampling rates they used in the German forums? I think it would make a big difference. Gyro manufacturers usually have specific recommendations for their particular gyros to avoid aliasing with the vibration frequency that they use.
In any case, I want to thank Tom very much for alerting me to the issue so that I could fix it before I released the new board and firmware.
Based on Tom's note, I am not encouraged about using the LISY300 gyros in a situation where there is a lot of vibration, but I am not so sure they could not be used until I do some testing myself, for a couple of reasons:
1. Some algorithms are better than others at rejecting errors caused by vibration. DCM is basically a pure integrator, and will greatly attenuate errors in the lower kHz region. Also, the drift compensation portion of the algorithm will attenutate the error caused by aliasing when the vibrations of the motor match the vibrations of the gyro. Unless the gyro is resonating to saturation, DCM might be able to work with the LISY300 in this situation.
2. You might be able to mount the board to attenuate the vibrations.
Though it is a long shot. If I were you, I would be cautious about using the LISY300 if there is a lot of vibration.
The LISY300 is not fit for quadrocopters. On the German UAV forums they have been comparing every gyro on the market, and the LISY300 was considered the poorest performer, mainly due to its vibrating mass being in the lower kHz region, similar to the vibrations caused by the motor.
@William, do you think the LISY300 gyros would perform well enough to stabilize a helicopter or quadrotor? (Obviously I'd have to add magnetometer support back into the estimator algorithm).
@Hugo,
I am aware of the ADXRS610. I tried out the LISY300 instead because of the price. I had seen some good reports on it. I am hoping that SparkFun will pass along some of the savings, and lower the price of the board. At any rate, I have asked them to do that.
I did evaluate the LISY300 before deciding to use it. SparkFun made a prototype board for me to test before going into full production. I tested it and was very pleased with the results. The performance of my firmware in the new board is actually better than the previous board: I can rotate 4 times faster without saturating the gyros.
So, I am not sure why some folks have not seen the same good performance that I have seen. There could be any number of reasons, ranging from the algorithms that were used, to ESD problems, to noise pickup.
At the moment, I am waiting for the first board in the new production run. When I get it, I will test it thoroughly before releasing the new design.
Comments
I made a design with the LISY in it myself, but decided not to make a prototype because of these warnings I read. Personally, I can not think of a reason why the higher sample rate would make any difference, mainly because the vibrating mass is still vibrating at a frequency in the same range as some of our engines. The German experts even tried it with complex analog filtering, but it did not help.
The standard answer from ST is:
---
There's no way to avoid this issue.
4kHz is the resonant frequency of the internal masses. If you excite them, the internal MEMS will get crazy and the output will no longer be reliable.
With other devices (other manufacturers' gyros) resonant frequencies may be very different (from few kHz to several kHz) and hence, you may have not seen the device becoming crazy.
---
The best choice seems to be the AD ones (300 degrees/s), and the melexis ones being a close 2nd best (you still seem to get what you pay for...).
So I would really advise you to do some serious testing in different aircraft/engine setups, to get a better view on this problem.
Unfortunately, I was not able the find the "gyro comparison" document I once encountered.
Best regards,
Tom
Basically, what Rob said is right. There are plenty of 3 axis accelerometers, we are using one of those, but I have not seen a suitable 3 axis gyro just yet that can compete with three LISY300s for price and performance.
An IMU needs both accelerometers and gyros.
I have been doing some more testing on the prototype board that SparkFun made for me with the LISY300 gyros. They are performing great. With the new board, you will be able to turn your plane at 300 degrees per second without saturating the gyros, and pull 6 gs without saturating the accelerometers. I am getting closer to one of my goals of pulling a high acceleration, continous, constant altitude, highly banked turn under IMU control.
Those links are not for gyros. I've seen that they now have 2 axis gyros integrated on one chip, but have yet to see a 3 axis one. It will be nice when they do finally come out with a high quality 3 axis gyro IC, at the right price.
I want to do more testing before I give you a definite yes or no about using the LISY300 for helicopters. Tom's information raises a concern, but my recent tests are encouraging that it may simply be a matter of aliasing.
I have a helicopter. I will put using it to do vibration testing of the LISY300 on my list of things to do, though I am not sure when I will get to it. There is a lot on my plate just now. In the meantime, I suggest that you consider whatever gyros that Tom recommends.
Bill
First of all, I want to thank Tom for alerting me to a possible issue with the LISY300 gyros that I was able to fix in firmware.
I took a closer look at the prototype board that I have with the LISY300 gyros in it. I noticed that if I tapped the gyros, DCM would "jump". Not good. Not a good sign at all. It would seem that there might be a problem with vibration....
I then took a look at my older board, with the ADXRS401. I tapped the gyros on it, everything was fine. So, at first glance, it would appear that the LISY300 is sensitive to vibration....
But then I thought about it some more...
It finally occurred to me that maybe the LISY300 simply had a higher frequency response than the ADXRS401, either because of the way that it is built, or because of the way the analog filters are designed.
If that is the case, all you have to do is raise the sampling rate high enough, and integrate!!
So I did. I raised the sampling rate from 500 Hz to 5000 Hz per channel. No problem for the dsPIC30F4011.
Now I can bang away at the gyros as hard as I can, everything is fine. I will still have to do vibration testing, but the fact that I could make the "glitching" go away by raising the sampling frequency is very encouraging.
It appears that at the lower sampling rate I was getting aliasing, because the frequency response of the gyro is higher than 500 Hz. When I tapped the gyro, it must have produced ringing at some frequencies that were multiples of 500 Hz. Raising the sampling rate to 5000 Hz and integrating solved the problem. My roll-pitch-yaw demo works great on the new board now.
Tom, by any chance do you know what sampling rates they used in the German forums? I think it would make a big difference. Gyro manufacturers usually have specific recommendations for their particular gyros to avoid aliasing with the vibration frequency that they use.
In any case, I want to thank Tom very much for alerting me to the issue so that I could fix it before I released the new board and firmware.
Thank you.
Bill Premerlani
Based on Tom's note, I am not encouraged about using the LISY300 gyros in a situation where there is a lot of vibration, but I am not so sure they could not be used until I do some testing myself, for a couple of reasons:
1. Some algorithms are better than others at rejecting errors caused by vibration. DCM is basically a pure integrator, and will greatly attenuate errors in the lower kHz region. Also, the drift compensation portion of the algorithm will attenutate the error caused by aliasing when the vibrations of the motor match the vibrations of the gyro. Unless the gyro is resonating to saturation, DCM might be able to work with the LISY300 in this situation.
2. You might be able to mount the board to attenuate the vibrations.
Though it is a long shot. If I were you, I would be cautious about using the LISY300 if there is a lot of vibration.
What gyro do you recommend?
Bill
Tom
I am aware of the ADXRS610. I tried out the LISY300 instead because of the price. I had seen some good reports on it. I am hoping that SparkFun will pass along some of the savings, and lower the price of the board. At any rate, I have asked them to do that.
I did evaluate the LISY300 before deciding to use it. SparkFun made a prototype board for me to test before going into full production. I tested it and was very pleased with the results. The performance of my firmware in the new board is actually better than the previous board: I can rotate 4 times faster without saturating the gyros.
So, I am not sure why some folks have not seen the same good performance that I have seen. There could be any number of reasons, ranging from the algorithms that were used, to ESD problems, to noise pickup.
At the moment, I am waiting for the first board in the new production run. When I get it, I will test it thoroughly before releasing the new design.
-
1
-
2
-
3
-
4
of 4 Next