DIY imu - sparkfun - idea with redundancy

[kbosak]

Hi,
Just to look for reasonnable and quite representative prices,
at sparkfun,
we can get IMU
IMU 5 Degrees of Freedom fro 110USD,
it requires ADC and microcontroller but otherwise it's reasonnably priced.
Problem: for airplane/helo, it would miss one rate sensor (yaw?).
plus, its Gyro Sensor IDG-300 has 500deg/s max range :eek: what is unlikely to happen except in a spin.
My idea is to add 3x MLX90609, 60USD each per ready board (no tiny legs soldering),
neccesarily in 75 degree/sec in order to have more precise readings.
The result:
3*60+110USD=290USD what is less than ADIS16350 or ADIS16355 (350-500USD?),
but has the advantage of having accelerometers whose drifts could cancel out (OK it will need Kalman).

Do you think this is good idea,
and is ADXL330 with +/-3g range reasonnable or we need something more precise,
or maybe you know cheap Accelerometer with digital out? :confused:

We are talking here about either a set of ready PCB boards, or DIP elements, or a single integrated SMD element comprising 6DOFs.
Please no BGA/FPGA proposals.

[jetblackaircra]

even though the airplane will not continuously rotate at 500 deg/s, turbulence might cause a transient (momentary) movement in excess of 500 deg/s for a short period of time. This would saturate the sensor and your attitude fix would then be invalid.

On the flip side, usually the high rate gyros will have poor resolution unless you are using a 12 or 16 bit ADC.

Both of these issues are why a filter is required. I would recommend getting the high rate gyros and using a micro fast enough to run the kalman filter code.

Mike

[Jack Crossfire]

Began thinking breaking apart a trio of real cheap rate gyros like the PG-03 is probably a better move than anything from Sparkfun. 300`/sec wasn't sensitive enough. The biggest factor in drift was stable reference voltage.

[jetblackaircra]

Began thinking breaking apart a trio of real cheap rate gyros like the PG-03 is probably a better move than anything from Sparkfun. 300`/sec wasn't sensitive enough. The biggest factor in drift was stable reference voltage.

Why use the analog gyros?

http://www.analog.com/en/prod/0,,764_801_ADIS16255%2C00.html

These are fully calibrated and temperature compensated. Give a digital output over SPI. they're like 75 bucks each on an eval board. expensive but very nice gyro.

If you go cheap you're going to get a mediocre IMU. But, if you've just got a park flyer you want to see fly straight and level and maybe do some lazy turns, maybe something cheaper will work.

[Jack Crossfire]

Began to suspect better gyros can be found at lower prices by breaking apart commercial products because they are made in mass quantities, the RC manufacturers have exclusive agreements with gyro manufacturers not to produce parts for individual sale, most of their price is the gyro, and the stuff from Sparkfun has done a lot worse than commercial heading hold gyros. A world class IMU rivaling fiber optic can probably be built out of 3 GY611's.

[jetblackaircra]

That may be, but those will run you close to 4 bones a piece. That's 12 hundred just on gyros. Then you need accelerometers, etc..... Go with a $75 calibrated and temp compensated digital gyro and be under $500 for the whole system.

Full scale certified EFIS (electronic flight instrument system) units use similar gyros. If you spec out the sensor properly and write your code properly you shouldn't have any trouble coming up with a robust inexpensive (Under $1000) 6DOF IMU which will work in almost any aircraft. For crying outloud, www.dynonavionics.com has an experimental aircraft EFIS for like $2500 with a sunlight readable LCD and a bunch of other bells and whistles. You know they're marking it up close to 100% or more. I guarantee you if you tear one open it's got a set of gyros that cost no more than $200 per unit in small quantities.

[kbosak]

Began thinking breaking apart a trio of real cheap rate gyros like the PG-03 is probably a better move than anything from Sparkfun. 300`/sec wasn't sensitive enough. The biggest factor in drift was stable reference voltage.
Exactly because of your blog (excellent lecture!) I wanted to use 75deg/s gyros while retaining 300deg/s gyros, remembering about wind gusts.

[kbosak]

Why use the analog gyros?
...because they come with sparkfun 5DOF IMU for 109USD.
you add multichannel 16 bit ADC (20USD?) ADR02 (a few USD?) and fly.
if you want 6 DOF, you add 1-3 precise digitals, say from Melexis as I proposed. And at the end you still end cheaper than 'assembled' 6DOF imu of any seller I have found, if we talk about reasonnable quality sensors that actually worked for somebody in the hobby.

[kbosak]

On the flip side, usually the high rate gyros will have poor resolution unless you are using a 12 or 16 bit ADC.

Both of these issues are why a filter is required. I would recommend getting the high rate gyros and using a micro fast enough to run the kalman filter code.

99% agree, but small detail: we cannot extract information 'that is not there'. No filter will extract information from sensors that CAN BE simply too noisy.
I am suspecting that no matter what range you gyro you choose, the MEMS that is inside has noise level that will show up at LSB around and 10-12bit (this is why output from assembled, digital gyros are 12-14 bit - the choice is not by coincidence).
In other words I want to use 16bit external ADC, but I dont think I get more valuable data than 12bits from those gyros, so the choice between 500deg/s or 75 deg/s is like gaining 2-3 bits. In practical scale, 500deg gyro at 12bits is expected to be not better around null reading than 75deg sampled at 10bits by uP.