Which IMU?

[Matthew Mucker]

Those of you who have worked with any of the 6DOF or 9DOF chips that are on the market: which ones do you like and which do you not, and why? I'm realizing (again) that there's not as much time as I think until AVC and I need to settle on parts selection really quickly.

[Michael Shimniok]

I have been using the minimu-9 at first in 9dof later just 2d gyro. The board has used various generations of STM gyros L3G4200D and later. They are nicely resistant to rover vibration.

Michael

On Feb 27, 2015 7:23 AM, "Matthew Mucker" <matthe...@gmail.com> wrote:

Those of you who have worked with any of the 6DOF or 9DOF chips that are on the market: which ones do you like and which do you not, and why? I'm realizing (again) that there's not as much time as I think until AVC and I need to settle on parts selection really quickly.

--
You received this message because you are subscribed to the Google Groups "diyrovers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diyrovers+...@googlegroups.com.
To post to this group, send email to diyr...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/d12227b4-7544-4357-ae00-57d3ef89e82b%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Ted Meyers]

I second the minimu-9, it is cheap, easy, available, and robust.  I also mostly use it for the gyros.  At some point I'd like to add an AHRS along with some bump detection and recovery.

Interesting note on the gyros: the highest resolution setting by far works best on my rovers (they don't max out the deg/sec unless bumped hard) and I had to calibrate the gyro rates to get good results -- the nominal rate is typically off by 3-4%. 

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CANtmtVkienyELosiqhy2T3NirYB%2BZAOCPKx%2B9wv1nnF4cVB45Q%40mail.gmail.com.

[jesse brockmann]

I'm also on board with in minimu-9.  $19.95 for the V3 is a great deal.     Am going to look at the MPU 9250 as well this year, see if it offers any better results.

JesseJay

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CADnZWKNZJsPAbR7-%2BB0QNjPCJgzmB3qAVyb7jf6XaRwGn%3DQ6Rw%40mail.gmail.com.

[Jon Watte]

The MPU-6050 is only $5 on ebay...

I'm using it, but I don't have anything to compare against.

Sincerely,

jw

Sincerely,

Jon Watte


--
"I find that the harder I work, the more luck I seem to have." -- Thomas Jefferson

To view this discussion on the web visit https://groups.google.com/d/msgid/diyrovers/CANqFZg2T9jm6tbStJ5ejpkK9Y%3Dw5Ba3Q2Xg6LDuCi0VeF4rwHA%40mail.gmail.com.

[Thomas Roell]

I do like the MPU-6050. Why ? Because it has Accel/Gyro arriving at pretty much the same rate, which can be up to 1000Hz.

I do NOT like the MPU-9150 to much. The embedded AK8975 is 2.5x less accurate/precises than a HMC5883L/HMC5983 (which can give you 0.5 degrees). For dead reckoning that may be an issue. For me it is as while the rover is stationary the heading will be derived from the magnetometer.

I do NOT like the MPU-9250 at all. The AK8963 is somewhat better, but still sucks in terms of resolution. I believe 1.5 degrees should be possible with it, but it's rather noisy ... The Accel/Gyro look good on paper, except that they now take insanely long for filtering. Hence you needed to do that on the host again ... So no advantage over the MPU-6050/MPU-9150 combo.

I do NOT like the ST based products up to the latest generation (the ones they announced, but not ship yet). With a LGD20H/LSM303D (like the Pololu IMUs) or the LSM9DS0 Accel and Gyro have different rates, which makes the effective rate you can get into an EKF or DCM really only 400Hz (2 times oversampling). Even worse, ST does not specify the delay their lowpass/highpass filters cause. That effect of the delay was something observable for me with the MPU-6050 if cranked up all the way into the 10ms range.

Personally I am back to a GY-86/GY-87 setup (MPU-6050+HMC5883L, BMP180 or MS5611), or simply using a RY835AI GPS/IMU combo. Latter one includes a NEO-M8N GPS/GLONASS setup with a 35mm passive patch antenna, MPU-6050+HMC5983 and a BMP180, all for $50.

- Thomas

[Wayne Holder]

I've also used several iterations of the Mini-IMU including, most recently, version 3, and found it worthy.  The only caveat is that you have to calibrate (offset) the magnetometer axes to get reliable readings.

Wayne

On Fri, Feb 27, 2015 at 6:22 AM, Matthew Mucker <matthe...@gmail.com> wrote:

Those of you who have worked with any of the 6DOF or 9DOF chips that are on the market: which ones do you like and which do you not, and why? I'm realizing (again) that there's not as much time as I think until AVC and I need to settle on parts selection really quickly.

--

[Thomas Roell]

I was no precise enough. Bad engineering habit.

IMHO the selection of the components depends upon what you want to do:

(1) You want to use a Accel/Gyro setup with an EKF. In that case you want to 2 things. One is to have Accel and Gyro arrive at the same time, and the other one is that you want to know the time between 2 samples. Ideally you want to have the data timestamped. The INT line of the MPU-6050 gives you the time stamp more or less (MPU-9150 and MPU-9250 as well). However that can be cumbersome, and in reality the dT matters,and with a 1000Hz sample rate that is known within bounds. Not knowing the when introduces a 1ms uncertainty, which with a 10m/s vehicle is equivalent to 1cm. Not a biggy IMHO, given all the other factors involved. Now giving up the same sample time of Accel/Gyro essentially halfs you sample rate, as you need to consider the worst case (well, it's trickier mathematically, but the 2x factor is close enough). With a L3GD20/LSM303 combo an 800Hz sampling this gives a 2.5ms uncertainty, or 2.5cm. So still not a biggy, but really not ideal.

(2) You want to use a split setup, where Gyro/WheelTick is fed into the EKF, and Accel/Magnetometer are used for a tilt compensated compass while the rover stands still. In that case you don't really care about the accel/magnetometer sample rate. The gyro is what matters, and using a L3GD20 with 800Hz without a timestamp still gives you 1.25ms uncertainty, or 1.25cm for the 10m/s rover... Anyway, for that scenario the L3GD20/LSM303 combo trumps the MPU-9150 simply because it has a way better magnetometer. It should be equivalent to a MPU-6050/HMC5883L combo.

The newer IMUs (MPU-9250, LSM303C, LSM9D1 and such) kind of suck on the magnetometer side as they are build for being put into cell-phones where there are possible huge magnetic distortions. So they have a bigger range there, but less precision. Also they tend to have a bigger focus on size and power consumption, with the end result that built in filtering is slower. Thus the older generation parts might be actually more accurate/precise overall.

Now there is a L3GD20/LSM303D/BMP180 module called GY-89 available for around $10 from the usual suspects ... I will give that perhaps a try when I get around ...

One more comment regarding the L3GD20. It has this builtin FIFO, and a status bit that says when the FIFO is empty. Thus you could have some calibration code that polls this bit to synchronize a MCU local timer with the L3GD20 ... That could provide a 0.1ms accurate timestamp ... But it's kind of nasty code, at least doing the same with MPU-6050 was ;-)

- Thomas