Reply by onestone December 26, 20042004-12-26
The basics of INS are simple. You start at point X. You measure 
orientation in 3D, using FGM's or gyro's. You measure acceleration
with 
accelerometers in 3D, you obviously must know the calibration factors 
for all sensors. Having taken the acceleration readings you must 
normalise them against the calibration data. You do the same with the 
orientation sensors. Once you know the 3 orientation vectors (assuming 
you have DC accelerometers) you can compensate the g readings against 
that axis orientaion, having first compensated for any axial 
misalignments of and between the two sensors. This will yield an 
absolute +/-g vector, the integral of this is velocity, the integral of 
that is distance. Repeat this and accumulate the results. Reading to 
reading is very precise, error becomes huge over time, so needs 
periodically correcting.

By using two separate INS systems you reduce the rate of error 
accumulation, and improve the overall accuracy.

That's really it for the basics. You can readily determine your location 
relative to point X, then relative to each subsequent reecorrection 
point. Bear in mind that you must account for the existing velocity and 
angles after each correction.

There are quite a few extra things I employ to improve the precision of 
my own systems, but they are my bread and butter, so I hope the little 
I've written is of use.

Al

Peter Grey wrote:

> At 10:14 PM 25/11/2004, you wrote:
> 
> Thanks Al,
> 
> I believe you have selected my Christmas reading subject - INS.
> 
> Peter
> 
> 
>>In air only, as part of a small UAV project. Rather than a pitot tube I
>>used INS. Actually my basic system uses a pair of INS systems each with
>>different sensor types. In this way I have two sources for orientation
>>for example, each with different error characteristics. The project
>>started from two sources, one of these was trackign a small sailing
>>boat. The rough seas, and low velocities make it very difficult to track
>>without the addition of GPS. At that time GPS was too large and
>>expensive to be viable.
>>
>>I certainly see no reason not to use INS underwater. There may be a drop
>>in accuracy WRT air, but, short of the gradiometric systems used by
>>nuclear subs  it would be hard to find a better way I suspect. And,
>>certainly, earlier subs ran on heading, speed, time, current
calculations.
>>
>>Al
>>
>>Peter Grey wrote:
>>
>>>At 09:35 AM 25/11/2004, you wrote:
>>>
>>>Al,
>>>
>>>Have you done any work on measuring velocity in air and water (birds
and
>>>animals)?
>>>
>>>Thanks
>>>
>>>
>>>Peter
>>>
>>>
>>>
>>>>The heated wire method is not  robust enough for my use, which
is
>>>>another reason the mini fan system was scrapped. Ultrasonics
have no
>>>>moving parts, no parts that can be readily damaged. Rain can be
>>>>compensated for, and snow can be blown away. This is a very tiny
thing
>>>>with one very specific purpose. I doubt that Dave will be flying
his
>>>>sailplanes in rain or snow either. In my case the sensor only
activates
>>>>for the very short periods it might be needed, perhaps a few
minutes
>>>>every day.
>>>>
>>>>Al
>>>>
>>>>rolf.freitag@rolf... wrote:
>>>>
>>>>
>>>>
>>>>>Hi,
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>Thanks for that Al. I'm going to build an
ultrasonic windspeed sensor
>>>>>>for use particularly with slope soaring RC sailplanes.
This should be
>>>>>>a much cheaper and sturdier version than the units that
have small
>>>>>>fans inside a housing.
>>>>>
>>>>>
>>>>>and what about rain, snow etc.?
>>>>>Do you shield the sensor or shut of during rain/snow?
>>>>>
>>>>>And wouldn't be a heatwire + temperature sensor
cheaper?
>>>>>
>>>>>Rolf
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>.
>>>>>
>>>>>
>>>>>Yahoo! Groups Links
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>>.
>>>>
>>>>
>>>>Yahoo! Groups Links
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>.
>>>
>>>
>>>Yahoo! Groups Links
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>.
>>
>>
>>Yahoo! Groups Links
>>
>>
>>
>>
> 
> 
> 
> 
> 
> 
> .
> 
>  
> Yahoo! Groups Links
> 
> 
> 
>  
> 
> 
> 
> 


Beginning Microcontrollers with the MSP430

Reply by onestone December 5, 20042004-12-05
Hi Vincenczo. This group does not allow file attachments. Files have to 
be separately uploaded to the files area.

Al

Vincenzo Izzo wrote:
> 
> See this Rain sensor we produce on ceramic.
> I'm attaching a simple schematic to test it; smarter
> applications can be done using a microcontroller to
> measure the capacitance , driving the heater and
> sensing the temperature of the ceramic substrate.
> 
> Vincenzo
> 
> 
> 
> 
>  --- onestone <onestone@ones...> ha scritto: 
> 
> ---------------------------------
> The basics of INS are simple. You start at point X.
> You measure 
> orientation in 3D, using FGM's or gyro's. You measure
> acceleration with 
> accelerometers in 3D, you obviously must know the
> calibration factors 
> for all sensors. Having taken the acceleration
> readings you must 
> normalise them against the calibration data. You do
> the same with the 
> orientation sensors. Once you know the 3 orientation
> vectors (assuming 
> you have DC accelerometers) you can compensate the g
> readings against 
> that axis orientaion, having first compensated for any
> axial 
> misalignments of and between the two sensors. This
> will yield an 
> absolute +/-g vector, the integral of this is
> velocity, the integral of 
> that is distance. Repeat this and accumulate the
> results. Reading to 
> reading is very precise, error becomes huge over time,
> so needs 
> periodically correcting.
> 
> By using two separate INS systems you reduce the rate
> of error 
> accumulation, and improve the overall accuracy.
> 
> That's really it for the basics. You can readily
> determine your location 
> relative to point X, then relative to each subsequent
> reecorrection 
> point. Bear in mind that you must account for the
> existing velocity and 
> angles after each correction.
> 
> There are quite a few extra things I employ to improve
> the precision of 
> my own systems, but they are my bread and butter, so I
> hope the little 
> I've written is of use.
> 
> Al
> 
> Peter Grey wrote:
> 
> 
>>At 10:14 PM 25/11/2004, you wrote:
>>
>>Thanks Al,
>>
>>I believe you have selected my Christmas reading
> 
> subject - INS.
> 
>>Peter
>>
>>
>>
>>>In air only, as part of a small UAV project. Rather
> 
> than a pitot tube I
> 
>>>used INS. Actually my basic system uses a pair of
> 
> INS systems each with
> 
>>>different sensor types. In this way I have two
> 
> sources for orientation
> 
>>>for example, each with different error
> 
> characteristics. The project
> 
>>>started from two sources, one of these was trackign
> 
> a small sailing
> 
>>>boat. The rough seas, and low velocities make it
> 
> very difficult to track
> 
>>>without the addition of GPS. At that time GPS was
> 
> too large and
> 
>>>expensive to be viable.
>>>
>>>I certainly see no reason not to use INS underwater.
> 
> There may be a drop
> 
>>>in accuracy WRT air, but, short of the gradiometric
> 
> systems used by
> 
>>>nuclear subs  it would be hard to find a better way
> 
> I suspect. And,
> 
>>>certainly, earlier subs ran on heading, speed, time,
> 
> current calculations.
> 
>>>Al
>>>
>>>Peter Grey wrote:
>>>
>>>
>>>>At 09:35 AM 25/11/2004, you wrote:
>>>>
>>>>Al,
>>>>
>>>>Have you done any work on measuring velocity in air
> 
> and water (birds and
> 
>>>>animals)?
>>>>
>>>>Thanks
>>>>
>>>>
>>>>Peter
>>>>
>>>>
>>>>
>>>>
>>>>>The heated wire method is not  robust enough for
> 
> my use, which is
> 
>>>>>another reason the mini fan system was scrapped.
> 
> Ultrasonics have no
> 
>>>>>moving parts, no parts that can be readily
> 
> damaged. Rain can be
> 
>>>>>compensated for, and snow can be blown away. This
> 
> is a very tiny thing
> 
>>>>>with one very specific purpose. I doubt that Dave
> 
> will be flying his
> 
>>>>>sailplanes in rain or snow either. In my case the
> 
> sensor only activates
> 
>>>>>for the very short periods it might be needed,
> 
> perhaps a few minutes
> 
>>>>>every day.
>>>>>
>>>>>Al
>>>>>
>>>>>rolf.freitag@rolf... wrote:
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>Hi,
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>>Thanks for that Al. I'm going to build an
> 
> ultrasonic windspeed sensor
> 
>>>>>>>for use particularly with slope soaring RC
> 
> sailplanes. This should be
> 
>>>>>>>a much cheaper and sturdier version than the
> 
> units that have small
> 
>>>>>>>fans inside a housing.
>>>>>>
>>>>>>
>>>>>>and what about rain, snow etc.?
>>>>>>Do you shield the sensor or shut of during
> 
> rain/snow?
> 
>>>>>>And wouldn't be a heatwire + temperature sensor
> 
> cheaper?
> 
>>>>>>Rolf
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>.
>>>>>>
>>>>>>
>>>>>>.
>  
> 
> 
> 		
> ___________________________________ 
> Nuovo Yahoo! Messenger: E' molto pidivertente: Audibles, Avatar,
Webcam, Giochi, Rubrica Scaricalo ora! 
> http://it.messenger.yahoo.it
> 
> 
> 
> 
> 
> 
> 
> 
> 
> .
> 
>  
> Yahoo! Groups Links
> 
> 
> 
>  
> 
> 
> 
> 


Reply by Vincenzo Izzo December 5, 20042004-12-05
See this Rain sensor we produce on ceramic.
I'm attaching a simple schematic to test it; smarter
applications can be done using a microcontroller to
measure the capacitance , driving the heater and
sensing the temperature of the ceramic substrate.

Vincenzo




 --- onestone <onestone@ones...> ha scritto: 

---------------------------------
The basics of INS are simple. You start at point X.
You measure 
orientation in 3D, using FGM's or gyro's. You measure
acceleration with 
accelerometers in 3D, you obviously must know the
calibration factors 
for all sensors. Having taken the acceleration
readings you must 
normalise them against the calibration data. You do
the same with the 
orientation sensors. Once you know the 3 orientation
vectors (assuming 
you have DC accelerometers) you can compensate the g
readings against 
that axis orientaion, having first compensated for any
axial 
misalignments of and between the two sensors. This
will yield an 
absolute +/-g vector, the integral of this is
velocity, the integral of 
that is distance. Repeat this and accumulate the
results. Reading to 
reading is very precise, error becomes huge over time,
so needs 
periodically correcting.

By using two separate INS systems you reduce the rate
of error 
accumulation, and improve the overall accuracy.

That's really it for the basics. You can readily
determine your location 
relative to point X, then relative to each subsequent
reecorrection 
point. Bear in mind that you must account for the
existing velocity and 
angles after each correction.

There are quite a few extra things I employ to improve
the precision of 
my own systems, but they are my bread and butter, so I
hope the little 
I've written is of use.

Al

Peter Grey wrote:

> At 10:14 PM 25/11/2004, you wrote:
> 
> Thanks Al,
> 
> I believe you have selected my Christmas reading
subject - INS.
> 
> Peter
> 
> 
>>In air only, as part of a small UAV project. Rather
than a pitot tube I
>>used INS. Actually my basic system uses a pair
of
INS systems each with
>>different sensor types. In this way I have two
sources for orientation
>>for example, each with different error
characteristics. The project
>>started from two sources, one of these was
trackign
a small sailing
>>boat. The rough seas, and low velocities make
it
very difficult to track
>>without the addition of GPS. At that time GPS
was
too large and
>>expensive to be viable.
>>
>>I certainly see no reason not to use INS underwater.
There may be a drop
>>in accuracy WRT air, but, short of the
gradiometric
systems used by
>>nuclear subs  it would be hard to find a better
way
I suspect. And,
>>certainly, earlier subs ran on heading, speed,
time,
current calculations.
>>
>>Al
>>
>>Peter Grey wrote:
>>
>>>At 09:35 AM 25/11/2004, you wrote:
>>>
>>>Al,
>>>
>>>Have you done any work on measuring velocity in air
and water (birds and
>>>animals)?
>>>
>>>Thanks
>>>
>>>
>>>Peter
>>>
>>>
>>>
>>>>The heated wire method is not  robust enough for
my use, which is
>>>>another reason the mini fan system was
scrapped.
Ultrasonics have no
>>>>moving parts, no parts that can be
readily
damaged. Rain can be
>>>>compensated for, and snow can be blown
away. This
is a very tiny thing
>>>>with one very specific purpose. I doubt
that Dave
will be flying his
>>>>sailplanes in rain or snow either. In
my case the
sensor only activates
>>>>for the very short periods it might be
needed,
perhaps a few minutes
>>>>every day.
>>>>
>>>>Al
>>>>
>>>>rolf.freitag@rolf... wrote:
>>>>
>>>>
>>>>
>>>>>Hi,
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>Thanks for that Al. I'm going to build an
ultrasonic windspeed sensor
>>>>>>for use particularly with slope
soaring RC
sailplanes. This should be
>>>>>>a much cheaper and sturdier
version than the
units that have small
>>>>>>fans inside a housing.
>>>>>
>>>>>
>>>>>and what about rain, snow etc.?
>>>>>Do you shield the sensor or shut of during
rain/snow?
>>>>>
>>>>>And wouldn't be a heatwire + temperature sensor
cheaper?
>>>>>
>>>>>Rolf
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>.
>>>>>
>>>>>
>>>>>.
 


		
___________________________________ 
Nuovo Yahoo! Messenger: E' molto pidivertente: Audibles, Avatar, Webcam,
Giochi, Rubrica Scaricalo ora! 
http://it.messenger.yahoo.it







Reply by onestone November 26, 20042004-11-26
I use an equivalent of 24, 18 and 15.5 bits. I guess the difference is 
time to travel. Time being the enemy of INS. In the rough ground case 
this was around 8 minutes.

Al

Matthias Weingart wrote:
> On Fri, Nov 26, 2004 at 10:04:28PM +1030, onestone wrote:
> 
> 
>>determine the likelihood of any given result. In field trials with a 
>>slowish moving vehicle over a very rough climb or descent, over 
>>potholed/rocky ground, the actual positional error after 1km of linear 
>>travel (ie point A to point B is exactly 1km in a straight line, 
>>ignoring X /Y/ offsets from the origin was under 5m in all 3 axes. With 
> 
> 
> Interesting: your 5m / 1000m are 0.5%, and my 0.01mm / 2mm are also 0.5%.
> What resolution do you use for the accelerometers, 16bits?
> 
>         Matthias
> 
> 
> 
> .
> 
>  
> Yahoo! Groups Links
> 
> 
> 
>  
> 
> 
> 
> 


Reply by Matthias Weingart November 26, 20042004-11-26
On Fri, Nov 26, 2004 at 10:04:28PM +1030, onestone wrote:

> determine the likelihood of any given result. In
field trials with a 
> slowish moving vehicle over a very rough climb or descent, over 
> potholed/rocky ground, the actual positional error after 1km of linear 
> travel (ie point A to point B is exactly 1km in a straight line, 
> ignoring X /Y/ offsets from the origin was under 5m in all 3 axes. With 

Interesting: your 5m / 1000m are 0.5%, and my 0.01mm / 2mm are also 0.5%.
What resolution do you use for the accelerometers, 16bits?

        Matthias

Reply by onestone November 26, 20042004-11-26
Whoops, not quite so simple. For orientation my post was confusing. If 
you are usign FGM's (Flux Gate Magnetometers, you can guage the exact 
orientation of your object vs the earth field. You will need to know 
where on the earths surface you are, and the angle of the earths 
magnetic field in that region.  This is quite simple. The accelerometers 
DC mode, while the unit is static can be used to cross check, and 
perhaps refine this. If you are using gyros, you will need to use the 
accelerometer DC mode to guage the starting orientation, and the gyros 
to track the changes.

Al

Peter Grey wrote:
> At 11:01 AM 26-12-04, you wrote:
> 
> Thanks Al, A good simple answer even I can understand!
> 
> Peter
> 
> 
>>The basics of INS are simple. You start at point X. You measure
>>orientation in 3D, using FGM's or gyro's. You measure
acceleration with
>>accelerometers in 3D, you obviously must know the calibration factors
>>for all sensors. Having taken the acceleration readings you must
>>normalise them against the calibration data. You do the same with the
>>orientation sensors. Once you know the 3 orientation vectors (assuming
>>you have DC accelerometers) you can compensate the g readings against
>>that axis orientaion, having first compensated for any axial
>>misalignments of and between the two sensors. This will yield an
>>absolute +/-g vector, the integral of this is velocity, the integral
of
>>that is distance. Repeat this and accumulate the results. Reading to
>>reading is very precise, error becomes huge over time, so needs
>>periodically correcting.
>>
>>By using two separate INS systems you reduce the rate of error
>>accumulation, and improve the overall accuracy.
>>
>>That's really it for the basics. You can readily determine your
location
>>relative to point X, then relative to each subsequent reecorrection
>>point. Bear in mind that you must account for the existing velocity and
>>angles after each correction.
>>
>>There are quite a few extra things I employ to improve the precision of
>>my own systems, but they are my bread and butter, so I hope the little
>>I've written is of use.
>>
>>Al
>>
>>Peter Grey wrote:
>>
>>
>>>At 10:14 PM 25/11/2004, you wrote:
>>>
>>>Thanks Al,
>>>
>>>I believe you have selected my Christmas reading subject - INS.
>>>
>>>Peter
>>>
>>>
>>>
>>>>In air only, as part of a small UAV project. Rather than a pitot
tube I
>>>>used INS. Actually my basic system uses a pair of INS systems
each with
>>>>different sensor types. In this way I have two sources for
orientation
>>>>for example, each with different error characteristics. The
project
>>>>started from two sources, one of these was trackign a small
sailing
>>>>boat. The rough seas, and low velocities make it very difficult
to track
>>>>without the addition of GPS. At that time GPS was too large and
>>>>expensive to be viable.
>>>>
>>>>I certainly see no reason not to use INS underwater. There may
be a drop
>>>>in accuracy WRT air, but, short of the gradiometric systems used
by
>>>>nuclear subs  it would be hard to find a better way I suspect.
And,
>>>>certainly, earlier subs ran on heading, speed, time, current
calculations.
>>>>
>>>>Al
>>>>
>>>>Peter Grey wrote:
>>>>
>>>>
>>>>>At 09:35 AM 25/11/2004, you wrote:
>>>>>
>>>>>Al,
>>>>>
>>>>>Have you done any work on measuring velocity in air and
water (birds and
>>>>>animals)?
>>>>>
>>>>>Thanks
>>>>>
>>>>>
>>>>>Peter
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>The heated wire method is not  robust enough for my use,
which is
>>>>>>another reason the mini fan system was scrapped.
Ultrasonics have no
>>>>>>moving parts, no parts that can be readily damaged. Rain
can be
>>>>>>compensated for, and snow can be blown away. This is a
very tiny thing
>>>>>>with one very specific purpose. I doubt that Dave will
be flying his
>>>>>>sailplanes in rain or snow either. In my case the sensor
only activates
>>>>>>for the very short periods it might be needed, perhaps a
few minutes
>>>>>>every day.
>>>>>>
>>>>>>Al
>>>>>>
>>>>>>rolf.freitag@rolf... wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>>Hi,
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>>Thanks for that Al. I'm going to build an
ultrasonic windspeed sensor
>>>>>>>>for use particularly with slope soaring RC
sailplanes. This should be
>>>>>>>>a much cheaper and sturdier version than the
units that have small
>>>>>>>>fans inside a housing.
>>>>>>>
>>>>>>>
>>>>>>>and what about rain, snow etc.?
>>>>>>>Do you shield the sensor or shut of during
rain/snow?
>>>>>>>
>>>>>>>And wouldn't be a heatwire + temperature sensor
cheaper?
>>>>>>>
>>>>>>>Rolf
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>.
>>>>>>>
>>>>>>>
>>>>>>>Yahoo! Groups Links
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>.
>>>>>>
>>>>>>
>>>>>>Yahoo! Groups Links
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>.
>>>>>
>>>>>
>>>>>Yahoo! Groups Links
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>>.
>>>>
>>>>
>>>>Yahoo! Groups Links
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>.
>>>
>>>
>>>Yahoo! Groups Links
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>.
>>
>>
>>Yahoo! Groups Links
>>
>>
>>
>>
> 
> 
> 
> 
> 
> 
> .
> 
>  
> Yahoo! Groups Links
> 
> 
> 
>  
> 
> 
> 
> 


Reply by onestone November 26, 20042004-11-26
The positional error achievable over 10 seconds, say, at high speed, is 
less than 0.25% oft he total disatnce travelled, noise is an issue, but 
if the noise is uniformly distributed it should null over time, however 
I use various other methods, like combinations of all 12 sensors to 
determine the likelihood of any given result. In field trials with a 
slowish moving vehicle over a very rough climb or descent, over 
potholed/rocky ground, the actual positional error after 1km of linear 
travel (ie point A to point B is exactly 1km in a straight line, 
ignoring X /Y/ offsets from the origin was under 5m in all 3 axes. With 
a fastish moving vehicle over easier terrain the error was slightly over 
half of this.

The caveat is that I am not simply using conventional INS techniques, 
although the GPS was inactive in all of the above tests.

Even the popular 'drag' sensors achieve accuracy in th area of 1%, and

they are typically uncompensated accelerometers, crude in comparison.

There are many situations that trip the system up, but they are being 
tackled slowly.

Al

Matthias Weingart wrote:
> On Fri, Nov 26, 2004 at 06:52:31PM +0800, Peter Grey wrote:
> 
> 
>>>misalignments of and between the two sensors. This will yield an
>>>absolute +/-g vector, the integral of this is velocity, the
integral of
>>>that is distance. Repeat this and accumulate the results. Reading to
>>>reading is very precise, error becomes huge over time, so needs
>>>periodically correcting.
> 
> 
> Onestone,
> 
> how large is the distance error over time (e.g. 1 second) you could reach?
> Because of the nature of double integration the errors are incresing with
at
> least time^2 (or much worse). I made a distance sensor for a short shock
> movement and get 10um error after 20ms in worst case. I think the noise
> of the accelerometers is the biggest problem. I also tried it for longer
> times, but I got totally useless results.
> 
> M.
> 
> 
> 
> .
> 
>  
> Yahoo! Groups Links
> 
> 
> 
>  
> 
> 
> 
> 


Reply by Matthias Weingart November 26, 20042004-11-26
On Fri, Nov 26, 2004 at 06:52:31PM +0800, Peter Grey wrote:

> >misalignments of and between the two sensors.
This will yield an
> >absolute +/-g vector, the integral of this is velocity, the
integral of
> >that is distance. Repeat this and accumulate the results. Reading to
> >reading is very precise, error becomes huge over time, so needs
> >periodically correcting.

Onestone,

how large is the distance error over time (e.g. 1 second) you could reach?
Because of the nature of double integration the errors are incresing with at
least time^2 (or much worse). I made a distance sensor for a short shock
movement and get 10um error after 20ms in worst case. I think the noise
of the accelerometers is the biggest problem. I also tried it for longer
times, but I got totally useless results.

M.

Reply by Peter Grey November 26, 20042004-11-26
At 11:01 AM 26-12-04, you wrote:

Thanks Al, A good simple answer even I can understand!

Peter

>The basics of INS are simple. You start at point X.
You measure
>orientation in 3D, using FGM's or gyro's. You measure acceleration
with
>accelerometers in 3D, you obviously must know the calibration factors
>for all sensors. Having taken the acceleration readings you must
>normalise them against the calibration data. You do the same with the
>orientation sensors. Once you know the 3 orientation vectors (assuming
>you have DC accelerometers) you can compensate the g readings against
>that axis orientaion, having first compensated for any axial
>misalignments of and between the two sensors. This will yield an
>absolute +/-g vector, the integral of this is velocity, the integral of
>that is distance. Repeat this and accumulate the results. Reading to
>reading is very precise, error becomes huge over time, so needs
>periodically correcting.
>
>By using two separate INS systems you reduce the rate of error
>accumulation, and improve the overall accuracy.
>
>That's really it for the basics. You can readily determine your
location
>relative to point X, then relative to each subsequent reecorrection
>point. Bear in mind that you must account for the existing velocity and
>angles after each correction.
>
>There are quite a few extra things I employ to improve the precision of
>my own systems, but they are my bread and butter, so I hope the little
>I've written is of use.
>
>Al
>
>Peter Grey wrote:
>
> > At 10:14 PM 25/11/2004, you wrote:
> >
> > Thanks Al,
> >
> > I believe you have selected my Christmas reading subject - INS.
> >
> > Peter
> >
> >
> >>In air only, as part of a small UAV project. Rather than a pitot
tube I
> >>used INS. Actually my basic system uses a pair of INS systems each
with
> >>different sensor types. In this way I have two sources for
orientation
> >>for example, each with different error characteristics. The project
> >>started from two sources, one of these was trackign a small sailing
> >>boat. The rough seas, and low velocities make it very difficult to
track
> >>without the addition of GPS. At that time GPS was too large and
> >>expensive to be viable.
> >>
> >>I certainly see no reason not to use INS underwater. There may be a
drop
> >>in accuracy WRT air, but, short of the gradiometric systems used by
> >>nuclear subs  it would be hard to find a better way I suspect. And,
> >>certainly, earlier subs ran on heading, speed, time, current
calculations.
> >>
> >>Al
> >>
> >>Peter Grey wrote:
> >>
> >>>At 09:35 AM 25/11/2004, you wrote:
> >>>
> >>>Al,
> >>>
> >>>Have you done any work on measuring velocity in air and water
(birds and
> >>>animals)?
> >>>
> >>>Thanks
> >>>
> >>>
> >>>Peter
> >>>
> >>>
> >>>
> >>>>The heated wire method is not  robust enough for my use,
which is
> >>>>another reason the mini fan system was scrapped.
Ultrasonics have no
> >>>>moving parts, no parts that can be readily damaged. Rain
can be
> >>>>compensated for, and snow can be blown away. This is a very
tiny thing
> >>>>with one very specific purpose. I doubt that Dave will be
flying his
> >>>>sailplanes in rain or snow either. In my case the sensor
only activates
> >>>>for the very short periods it might be needed, perhaps a
few minutes
> >>>>every day.
> >>>>
> >>>>Al
> >>>>
> >>>>rolf.freitag@rolf... wrote:
> >>>>
> >>>>
> >>>>
> >>>>>Hi,
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>>Thanks for that Al. I'm going to build an
ultrasonic windspeed sensor
> >>>>>>for use particularly with slope soaring RC
sailplanes. This should be
> >>>>>>a much cheaper and sturdier version than the units
that have small
> >>>>>>fans inside a housing.
> >>>>>
> >>>>>
> >>>>>and what about rain, snow etc.?
> >>>>>Do you shield the sensor or shut of during rain/snow?
> >>>>>
> >>>>>And wouldn't be a heatwire + temperature sensor
cheaper?
> >>>>>
> >>>>>Rolf
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>.
> >>>>>
> >>>>>
> >>>>>Yahoo! Groups Links
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>.
> >>>>
> >>>>
> >>>>Yahoo! Groups Links
> >>>>
> >>>>
> >>>>
> >>>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>.
> >>>
> >>>
> >>>Yahoo! Groups Links
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>
> >>
> >>
> >>
> >>.
> >>
> >>
> >>Yahoo! Groups Links
> >>
> >>
> >>
> >>
> >
> >
> >
> >
> >
> >
> > .
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
> >
> >
> >
> >
>
>
>
>
>.
>
>
>Yahoo! Groups Links
>
>
>
>



Reply by Peter Grey November 25, 20042004-11-25
At 10:14 PM 25/11/2004, you wrote:

Thanks Al,

I believe you have selected my Christmas reading subject - INS.

Peter

>In air only, as part of a small UAV project. Rather
than a pitot tube I
>used INS. Actually my basic system uses a pair of INS systems each with
>different sensor types. In this way I have two sources for orientation
>for example, each with different error characteristics. The project
>started from two sources, one of these was trackign a small sailing
>boat. The rough seas, and low velocities make it very difficult to track
>without the addition of GPS. At that time GPS was too large and
>expensive to be viable.
>
>I certainly see no reason not to use INS underwater. There may be a drop
>in accuracy WRT air, but, short of the gradiometric systems used by
>nuclear subs  it would be hard to find a better way I suspect. And,
>certainly, earlier subs ran on heading, speed, time, current calculations.
>
>Al
>
>Peter Grey wrote:
> > At 09:35 AM 25/11/2004, you wrote:
> >
> > Al,
> >
> > Have you done any work on measuring velocity in air and water (birds
and
> > animals)?
> >
> > Thanks
> >
> >
> > Peter
> >
> >
> >>The heated wire method is not  robust enough for my use, which is
> >>another reason the mini fan system was scrapped. Ultrasonics have
no
> >>moving parts, no parts that can be readily damaged. Rain can be
> >>compensated for, and snow can be blown away. This is a very tiny
thing
> >>with one very specific purpose. I doubt that Dave will be flying
his
> >>sailplanes in rain or snow either. In my case the sensor only
activates
> >>for the very short periods it might be needed, perhaps a few
minutes
> >>every day.
> >>
> >>Al
> >>
> >>rolf.freitag@rolf... wrote:
> >>
> >>
> >>>Hi,
> >>>
> >>>
> >>>
> >>>>Thanks for that Al. I'm going to build an ultrasonic
windspeed sensor
> >>>>for use particularly with slope soaring RC sailplanes. This
should be
> >>>>a much cheaper and sturdier version than the units that
have small
> >>>>fans inside a housing.
> >>>
> >>>
> >>>and what about rain, snow etc.?
> >>>Do you shield the sensor or shut of during rain/snow?
> >>>
> >>>And wouldn't be a heatwire + temperature sensor cheaper?
> >>>
> >>>Rolf
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>.
> >>>
> >>>
> >>>Yahoo! Groups Links
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>
> >>
> >>
> >>
> >>.
> >>
> >>
> >>Yahoo! Groups Links
> >>
> >>
> >>
> >>
> >
> >
> >
> >
> >
> >
> > .
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
> >
> >
> >
> >
>
>
>
>
>.
>
>
>Yahoo! Groups Links
>
>
>
>