INS gyro alternative or enhancement

Integrating a gyroscope (with a high pass filter to get rid of drift, and a
low pass filter to get rid of noise), and a accelerometer (with a low cut
off low pass filter) with a kalman filter will do the job?

But as I said before acceleration can be spikey or gradual, the gravity
vector it gives may be a little off

Terence


		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com

"Does using a high pass filter for the gyroscope helps getting rid of
the
drift?"

No, let me put it this way,  you know a resistor is not a really just a
resistor, but it also is a capacitor and inductor due to imperfections
in the resistor construction and wire leads, likewise a single axis
gyro is not really just a single axis rate sensor , but can be thought
of as a sensor combination consisting of multi axis gyros, multi axis
accelerometers, temperature sensor, time sensor all in one unit with
the outputs all combined to form a single output.  No amount of
filtering will get rid of the drift, the output of a single axis gyro
is dependent on the rates in all three axis, the accelerations in all
three axis, the temperature, and the amount of time the gyro has been
operating (among other things). The less drift you want, the more you
have to calibrate the "other" sensor outputs out.

steve wrote:

> No, let me put it this way,  you know a resistor is not a really just a
> resistor, but it also is a capacitor and inductor due to imperfections
> in the resistor construction and wire leads, likewise a single axis
> gyro is not really just a single axis rate sensor , but can be thought
> of as a sensor combination consisting of multi axis gyros, multi axis
> accelerometers, temperature sensor, time sensor all in one unit with
> the outputs all combined to form a single output.  No amount of
> filtering will get rid of the drift, the output of a single axis gyro
> is dependent on the rates in all three axis, the accelerations in all
> three axis, the temperature, and the amount of time the gyro has been
> operating (among other things). The less drift you want, the more you
> have to calibrate the "other" sensor outputs out.

I remember reading about laser gyros which have laser beams going two
directions around a ring, either with mirrors or though a fiber optic
cable.  The rotation is sensed through changes in the interference 
fringes between the two paths.  At high rotation rates they work fine,
but at lower rates they can phase lock due to the imperfections in the
mirrors or fiber.  The result is that the signal goes to zero when
it should not be zero.

There is also a piezoelectric based gyro that works using the coriolis
force.  A piezo crystal vibrates in one axis, say X, and a piezo 
detector detects vibrations in the Y direction.  If the device rotates 
around the Z axis the path won't be pure X anymore, with a Y component 
proportional to the rotation rate, and relative phase depending on the 
direction of rotation.  It is likely that some drift will need to be 
nulled out if the two aren't exactly perpendicular.  I have heard that 
this one is used in car navigation systems with GPS correction.

-- glen

glen herrmannsfeldt wrote:
> steve wrote:
> 
>> No, let me put it this way,  you know a resistor is not a really just a
>> resistor, but it also is a capacitor and inductor due to imperfections
>> in the resistor construction and wire leads, likewise a single axis
>> gyro is not really just a single axis rate sensor , but can be thought
>> of as a sensor combination consisting of multi axis gyros, multi axis
>> accelerometers, temperature sensor, time sensor all in one unit with
>> the outputs all combined to form a single output.  No amount of
>> filtering will get rid of the drift, the output of a single axis gyro
>> is dependent on the rates in all three axis, the accelerations in all
>> three axis, the temperature, and the amount of time the gyro has been
>> operating (among other things). The less drift you want, the more you
>> have to calibrate the "other" sensor outputs out.
> 
> 
> I remember reading about laser gyros which have laser beams going two
> directions around a ring, either with mirrors or though a fiber optic
> cable.  The rotation is sensed through changes in the interference 
> fringes between the two paths.  At high rotation rates they work fine,
> but at lower rates they can phase lock due to the imperfections in the
> mirrors or fiber.  The result is that the signal goes to zero when
> it should not be zero.

Laser ring gyros are way good and way expensive -- they're the standard 
for dead-reckoning style inertial navigation.

Fiber optic gyros (FOGs) range from almost as good (perhaps better) as 
free-space laser ring gyros down to not good at all.

Nearly all, if not all, available laser gyros have a mechanical or 
electro-optical "dither" added in that creates an apparent oscillatory 
motion that breaks up the phase locking.
> 
> There is also a piezoelectric based gyro that works using the coriolis
> force.  A piezo crystal vibrates in one axis, say X, and a piezo 
> detector detects vibrations in the Y direction.  If the device rotates 
> around the Z axis the path won't be pure X anymore, with a Y component 
> proportional to the rotation rate, and relative phase depending on the 
> direction of rotation.  It is likely that some drift will need to be 
> nulled out if the two aren't exactly perpendicular.  I have heard that 
> this one is used in car navigation systems with GPS correction.
> 
These piezo gyros can either be made from quartz or from silicon.  The 
silicon ones are what the OP is working with.  Quartz gyros are better, 
but more expensive.  The less expensive FOGs pick up at about the 
performance (and price point) of the best quartz sensor.

You left out mechanical gyros, which aren't dead by any means.  While 
mechanical gyros such as the Northrup Grumman G2000 do have drawbacks 
compared to FOGs and other solid-state gyros, they can be very small for 
the degree of accuracy that they offer compared to the competition, and 
they are very quiet at low frequencies -- FOGs and piezo gyros tend to 
have a noise spectrum that is flat in rate and therefore a Wiener 
process in position.  The good mechanical gyros have significantly less 
rate noise.

-------------------------------------------
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

glen herrmannsfeldt wrote:

   ...

> There is also a piezoelectric based gyro that works using the coriolis
> force.  A piezo crystal vibrates in one axis, say X, and a piezo 
> detector detects vibrations in the Y direction.  If the device rotates 
> around the Z axis the path won't be pure X anymore, with a Y component 
> proportional to the rotation rate, and relative phase depending on the 
> direction of rotation.  It is likely that some drift will need to be 
> nulled out if the two aren't exactly perpendicular.  I have heard that 
> this one is used in car navigation systems with GPS correction.

Presumably, that's how a housefly's halteres work. The analogy is 
complete if you consider nerve impulses generated by pressure to be 
piezoelectric in origin.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Have anyone experience with a compass?
I tried honeywell, but I don't really have good experience with them.
There is another one from PNI MicroMag3, but I do not know how good they
are. Yes Yes, manufacturers (honeywell) says how accurate, how good they
are in the specs, but in reality, they suck big time. I was getting like
30 (maybe bigger or smaller due to nonlinearity) degress error. They don't
have a high resolution at all. Either the value stays what it has or change
greatly. 

Sorry guys, I really cannot afford those fancy gyroscope. I think I need
something that has a bound. After integrating, it can drift to infinity. 

I was thinking of feeding the acceleration orientation if the magnitude is
close to 9.8. Then at least, it will not drift as bad. 

These guys did some really good thesis on INS by the way
http://epubl.luth.se/1402-1617/2000/081/LTU-EX-00081-SE.pdf
http://www.casde.iitb.ac.in/Publications/pdfdoc-2004/vikas-ddp.pdf

Terence
		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com

"If you find a way to
rival the accuracies of the units above at a fraction of the price
please let me know -- I know a few companies that'll buy them. "

I suspect the only way to significantly drop the price would be to
develop a highly automated calibration system to reduce the labor
involved

glen herrmannsfeldt wrote:
> There is also a piezoelectric based gyro that works using the coriolis
> force.  A piezo crystal vibrates in one axis, say X, and a piezo
> detector detects vibrations in the Y direction.  If the device rotates
> around the Z axis the path won't be pure X anymore, with a Y component
> proportional to the rotation rate, and relative phase depending on the
> direction of rotation.  It is likely that some drift will need to be
> nulled out if the two aren't exactly perpendicular.  I have heard that
> this one is used in car navigation systems with GPS correction.

Been there... I wrote my BSc project over the "wine glass gyro",
so called because the vibrating body resembled a wine glass.
The piezoelectric elements were mounted every 45 degrees around the
rim of the "glass", the axis of rotation being the symmetry axis
of the "glass". In the application we were concerned with, it was
not so much a device to stabilize the platform, as a detector to
estimate the order of magnitude on which the platform rotated.

My partner and I made just about every possible mistake and blunder
in the book (and probably invented some on our own as well) but we
did find that the mechanical resonance frequency in the metal (brass)
depended a lot on temperature. GPS correction was not an issue at
the time, but the "wine glass gyro" might just work for some
applications.

As for me, the main outcome was that hardware electronics was not my
cup of tea. Or glass of wine, as it were.

Rune

So any smart way to do auto-calibration?
Terence
		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com

other then building a customized all-in-one calibration system for a
specific application, no