PDA

View Full Version : Gyro Filter


MikeDubreuil
01-21-2003, 10:21 AM
It it possible to filter the output of the gyro? It is mounted to the base of a "Demo-Bot" and often gets interference from normal turning. Does anyone have information on how effective an RC low pass filter would be?

rbayer
01-21-2003, 05:07 PM
It really depends on what scaling factor you use in your filter. InnovationFirst has a whitepaper on this, so you can at least find an example there. In theory, experimenting with the coefficient should give you a fairly good reading once you find the ideal settings.

--Rob

Al Skierkiewicz
01-21-2003, 09:57 PM
The documentation on the gyro and interface circuitry includes a filter. If you read carefully and follow the circuit, you will see the recommendation of a 47HZ high pass filter and a 185 HZ low pass filter is part of the interface amplifier. In the car example it is mentioned that the gyro does exhibit some output from the up down motions of the vehicle without these filters in place. Still filters are chosen for specific applications, a head mounted sensor for interactive game play may require a 0.1Hz to 10Hz filter to eliminate small high frequency head movements from affecting the needed head turning data. (from BEI/Systron data)

MikeDubreuil
01-21-2003, 10:28 PM
Great, Thanks....

I did not check the new gyro's box. Will I find BEI's filter information in there?

Rbayer, I did not see the whitepaper you are referring to on the IF website.
- Mike

Lloyd Burns
01-25-2003, 01:21 AM
Check the CD search feature - use "gyro filter" as the search parameters . The second oldest result I got was the Techokats' .pdf on using the gyro with a software filter to balance on the ramp.

-=-=-=-=-

The big problem with the gyro, if it is used for dead reckoning, is that it goes non-linear at turn rates less than 90 degrees/second. This is the turn rate of your forearm as you raise it from horizontal to vertical over a whole second ("one steamboat"). If your robot turns only that fast, it will appear quite sedate. If it turns faster, your robot can't tell where it is heading.

FotoPlasma
01-25-2003, 04:35 AM
Originally posted by Lloyd Burns
The big problem with the gyro, if it is used for dead reckoning, is that it goes non-linear at turn rates less than 90 degrees/second. This is the turn rate of your forearm as you raise it from horizontal to vertical over a whole second ("one steamboat"). If your robot turns only that fast, it will appear quite sedate. If it turns faster, your robot can't tell where it is heading.

It's actually worse than that.

The GyroChip which we got in the kit, for a few seasons, until 2003, actually had a maximum detectable rate of 64deg/sec, after which it's maxed out. This year's GyroChip is only marginally better at 75deg/sec.

Basically, if you're planning on having a quick drivetrain, and using it to its full potential, you're not going to be able to use the GyroChip to dictate your rate of turn, without some tweaking.

references:
http://www.systron.com/prodinfo/AQRSSpc.html
http://www.systron.com/GyroChip%20Theory.pdf (page 5)

crazycliffy
02-04-2003, 02:55 PM
This is like a bandpass filter with cutoffs to allow
ranges within the 47Hz and 185Hz frequencies.

Where did you get these numbers from? I look at
the www.systron.com mfg website, and they don't
spec these numbers.

Are these included with the actual gyro for 2003?
I don't have the 2003 gyro yet, as I'm doing some
prototyping work with the 2002 model.

Also I did a search using chief delphi (CD) && google.com
and didn't get the numbers you spec'd.

thanks for any clarification.


The documentation on the gyro and interface circuitry includes a filter. If you read carefully and follow the circuit, you will see the recommendation of a 47HZ high pass filter and a 185 HZ low pass filter is part of the interface amplifier. In the car example it is mentioned that the gyro does exhibit some output from the up down motions of the vehicle without these filters in place. Still filters are chosen for specific applications, a head mounted sensor for interactive game play may require a 0.1Hz to 10Hz filter to eliminate small high frequency head movements from affecting the needed head turning data. (from BEI/Systron data

Al Skierkiewicz
02-04-2003, 04:26 PM
It took some searching on the BEI site to find application notes and some other sources. Each application is going to require different implementations so just be aware that the device needs some trade offs to be useful. Many teams have successfully implemented them in the past.