Yes my question truly is that broad and vague :) I need to start getting smart on sensors and was hoping someone here had some books/websites to help me learn whats possible. I'm looking at trying to do low power (passive) sensing of I'm not sure what yet. Thinking about temperature, accelerometer, IR, seismic, acoustic, with some GPS thrown in for location and timing. Will take recommendation for other low powers I've left out. The plan is to network several together and see what can be learned through data aggregation of the various sensors in various locations. I've been looking at datasheets of the above but would like to back up to a bigger picture before I dive in here to find out what I'm overlooking or forgetting. Thanks for any help, Alan PS. Southern Oregon is my favorite having grown up there
Sensors
Started by ●November 22, 2004
Reply by ●November 22, 20042004-11-22
----- Original Message -----
From: "Alan Zubatch" <zubatch@zuba...>
To: <msp430@msp4...>
Sent: Monday, November 22, 2004 3:12 PM
Subject: [msp430] Sensors
>
> Yes my question truly is that broad and vague :)
>
> I need to start getting smart on sensors and was hoping someone here had
> some books/websites to help me learn whats possible. I'm looking at
trying
> to do low power (passive) sensing of I'm not sure what yet.
>
> Thinking about temperature, accelerometer, IR, seismic, acoustic, with
> some GPS thrown in for location and timing. Will take recommendation for
> other low powers I've left out.
>
> The plan is to network several together and see what can be learned
> through data aggregation of the various sensors in various locations.
>
> I've been looking at datasheets of the above but would like to back up
to
> a bigger picture before I dive in here to find out what I'm
overlooking or
> forgetting.
'Sensor fusion' is what you need. It's a difficult subject.
Leon
Reply by ●November 22, 20042004-11-22
With all respect to Leon I hate the term 'sensor fusion',
it's just
another meaningless piece of market hype to me, probably spawned in a
university somewhere to appear to be something new in an effort to raise
research funds. Another pet hate was mechatronics, basically what I've
been doing most of my working life. Changing the names now just confuses
my aging brain.
Any way, to the subject matter. There really is no end to sensors, and
little limit to what you can do with them. I have experimented and used
many different types. Here are some sources, and possible uses:_
Piezo. Simple sensors that respond to pressure or bending from
Measurement Systems. They also do resistive pressure sensors. Add a
small weight and you get a crude accelerometer. They can be used as
touch pads, switches, or force sensors.
Simple capacitive. Serena instruments had a very tiny, simple plate
capacitor sensor about the size of a button made from two conductive
plates with an internal 'O' ring, that was a polymer of specific
desnsity such that you created a simple capacitive load cell. One in
each corner of a glass plate gave you a quite decent drawing pad or
touch screen. I don' think they are doing this still, but I built a few
of my own using a simple oscillator output from a 1mm square 555. They
worked very nicely as low resolution load cells.
Accelerometers. I use Analog devices and MEMSIC. MEMSIC are a bit
cheaper, claim slightly lower noise and better absolute sensitivity, but
have a much lower bandwidth. I use these for everything from motion
analysis, to microphones, security, navigation systems, vibration
sensors, etc etc.
GMR magnetic sensors. I tried the KMZ51 from Philips, but the HMC1053 3
axis sensor from Homeywell is much easier to use, although it was a bit
expensive when it foirst came out. Great for 3 axis compasses and
navigation systems, also useful for earth field magnetometery and
sensing proximity of objects. 8 of these orthogonally aligned in a large
cube make an extremely sensitive magnetic gradiometer.
Flux Gate Magnetometers. For small and affordable you can't go past
Speake & Co in the UK. The FGM-1 is a very neat device, and the FGM-2 a
dual axis part. VERY sensitive, simpel to use, power in, ground and a
square wave out. Does require a fairly high clock speed on the micro to
get the best out of them. I used these in my original MANIC project in
the second INS system, but when I shrunk MANIC I built my own FGM on a
3D 5mm former, using very tiny op amp oscillators.
GPS. If you want an intergal antenna you can't go past the GH80D from
furuno, decent price and respectably good unit. But for small size, and
the fastest TTFF (time to First Fix) the latest FS-Oncore from Freescale
is absolutely unbeatable, and the smallest commercially available GPS
I've seen. You will need to add an antenna. Yageo are a great source,
but getting low volume is like gettign blood out of a stone.
Pressure sensors. I still like the motorola parts, but Sensym are also
good. I tend to use whichever has the physically smallest unit at the
time, and am biased towards 3V parts.
Temperature. Nat Semi probably have the most accurate over all, although
it varies all the time. The LM92 is nice standard unit, although emails
from them suggest this is superceded by an even better part. For simple,
and lots of them the DS18S20 1 wire interface part from Maxim/Dallas is
simple to use and quite reasonable in performance.
Humidity. Sensirion have the best around in my opinion, and they
integrate a temperature sensor.
Optical:_
For straight IR sensing I like Fairchildfor their sensitivity. For a
wide range of colour specific sensors, linear array sensors etc I think
TAOS have a great range.
Chemical. TAOS have a quirky, but very effective chemical sensor using
an optical reference waveguide.But see the next entry.
Surface Plasmon Resonance. Ti call this SPREETA. It is an optical sensor
that measures the change in refractivity, and reflectivity through a
gold half mirror in a small module. The moduel has a light source and a
linear array of 256 sensors. The device can be used for sensing
chemicals, specific metals, proteins etc etc. Simply by conditioning the
external window surface with an antigen that the target protein will
bind to.
Wind. MANIC has a tiny 4 point ultrasonic wind speed and ddirection
sensor with no movingparts. You can uild these yourself easily.
Gyros. I have been wathcin Irvine SEnsors, nor Microsensors with their
2mm square coriolis sensor for several years, but their quotes are
US$150 each, and, considering the spec is not as good as that of the
ADSR300 from analog devices I use the latter part. It costs 20% of the
price. NOT mil spec but I use them again in my INS system, and the
accuracy is certainly good enough for autonav in a small UAV. ]
Proximity. QPROX make a whole range of capacitive type proximity
devices. If they still make it the QT9701 is well worth a look.
Well that's about it for a starting point. I'm sure I've missed a
few,
but they can wait until later.
Al
Alan Zubatch wrote:
> Yes my question truly is that broad and vague :)
>
> I need to start getting smart on sensors and was hoping someone here had
some books/websites to help me learn whats possible. I'm looking at trying
to do low power (passive) sensing of I'm not sure what yet.
>
> Thinking about temperature, accelerometer, IR, seismic, acoustic, with some
GPS thrown in for location and timing. Will take recommendation for other low
powers I've left out.
>
> The plan is to network several together and see what can be learned through
data aggregation of the various sensors in various locations.
>
> I've been looking at datasheets of the above but would like to back up
to a bigger picture before I dive in here to find out what I'm overlooking
or forgetting.
>
> Thanks for any help,
>
> Alan
>
>
> PS. Southern Oregon is my favorite having grown up there
>
>
>
>
>
>
> .
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
Reply by ●November 22, 20042004-11-22
On Tue, 23 Nov 2004 02:29:30 +1030, you wrote:
>With all respect to Leon I hate the term
'sensor fusion', it's just
>another meaningless piece of market hype to me, probably spawned in a
>university somewhere to appear to be something new in an effort to raise
> research funds. Another pet hate was mechatronics, basically what
I've
>been doing most of my working life. Changing the names now just confuses
>my aging brain.
Actually, from personal experience working with _certain_ US projects, it's
very
real. What I know of it uses Kalman or Bucy-Kalman techniques. Basically, you
know a priori some information about your various sensor systems and their
errors and noise shapes and you also know some information about the targets
you
are tracking and their time-dependent abilities to impact the sensors (their
acceleration limits, for example) and you develop some "optimal"
filtering of
each individual 'system' and then develop the resulting error
covariance matrix.
These can then be combined in objective ways to develop an overall "sensor
fusion" that tracks the target better than any one sensor system alone can
achieve.
Jon
Reply by ●November 22, 20042004-11-22
On Mon, 22 Nov 2004 09:11:56 -0800, I wrote:
>and then develop the resulting error covariance
matrix.
That part is developed in real-time, from real sensor data, and then
'fused'
with similar error covariances from other sensor systems in real-time.
Jon
Reply by ●November 22, 20042004-11-22
----- Original Message -----
From: "onestone" <onestone@ones...>
To: <msp430@msp4...>
Sent: Monday, November 22, 2004 3:59 PM
Subject: Re: [msp430] Sensors
>
> With all respect to Leon I hate the term 'sensor fusion',
it's just
> another meaningless piece of market hype to me, probably spawned in a
> university somewhere to appear to be something new in an effort to raise
> research funds. Another pet hate was mechatronics, basically what
I've
> been doing most of my working life. Changing the names now just confuses
> my aging brain.
I thought he was asking about what to do with the data from several
different types of sensor, like an intrusion detection system where one
might have sound, IR, ground vibration, etc. and there is a requirement to
put all the data together and come to a decision about whether it was a
false alarm or not. I don't know of a better term than 'sensor
fusion' for
that sort of activity. The military have similar problems detecting and
identifying targets.
Leon
Reply by ●November 22, 20042004-11-22
Thanks for all the input. Data aggregation/sensor fusion is indeed the end goal to detect......something. Detection, classification, tracking, and prediction are where we'd likely go once we get data. Problem is most of those will be dependent on types of sensors.No real requirements yet so I'd like it to be as flexible as possible. Low power passive/duty cycled active distance sensor recommendations or a book that covers such if anyone knows of one. Take it easy on the universities, I resemble that remark {8 ?) Sometimes when you're doing something new the "sponsor" doesn't see it as new unless you give it a new name that they can understand :( mumble..mumble....managers with little to no technical background....mumble...mumble Alan ----- Original Message ----- From: Leon Heller To: msp430@msp4... Sent: Monday, November 22, 2004 12:25 PM Subject: Re: [msp430] Sensors ----- Original Message ----- From: "onestone" <onestone@ones...> To: <msp430@msp4...> Sent: Monday, November 22, 2004 3:59 PM Subject: Re: [msp430] Sensors > > With all respect to Leon I hate the term 'sensor fusion', it's just > another meaningless piece of market hype to me, probably spawned in a > university somewhere to appear to be something new in an effort to raise > research funds. Another pet hate was mechatronics, basically what I've > been doing most of my working life. Changing the names now just confuses > my aging brain. I thought he was asking about what to do with the data from several different types of sensor, like an intrusion detection system where one might have sound, IR, ground vibration, etc. and there is a requirement to put all the data together and come to a decision about whether it was a false alarm or not. I don't know of a better term than 'sensor fusion' for that sort of activity. The military have similar problems detecting and identifying targets. Leon . ------ .
Reply by ●November 22, 20042004-11-22
----- Original Message -----
From: "Jonathan Kirwan" <jkirwan@jkir...>
To: <msp430@msp4...>
Sent: Monday, November 22, 2004 5:11 PM
Subject: Re: [msp430] Sensors
>
> On Tue, 23 Nov 2004 02:29:30 +1030, you wrote:
>
>>With all respect to Leon I hate the term 'sensor fusion',
it's just
>>another meaningless piece of market hype to me, probably spawned in a
>>university somewhere to appear to be something new in an effort to raise
>> research funds. Another pet hate was mechatronics, basically what
I've
>>been doing most of my working life. Changing the names now just confuses
>>my aging brain.
>
> Actually, from personal experience working with _certain_ US projects,
> it's very
> real. What I know of it uses Kalman or Bucy-Kalman techniques.
> Basically, you
> know a priori some information about your various sensor systems and their
> errors and noise shapes and you also know some information about the
> targets you
> are tracking and their time-dependent abilities to impact the sensors
> (their
> acceleration limits, for example) and you develop some "optimal"
filtering
> of
> each individual 'system' and then develop the resulting error
covariance
> matrix.
> These can then be combined in objective ways to develop an overall
"sensor
> fusion" that tracks the target better than any one sensor system alone
can
> achieve.
We used Kalman filter techniques where I used to work, using data from
multiple IR detection arrays, in a people counter for shopping malls, etc. I
didn't have anything to do with the tracking algorithms, but was
responsible
for converting the prototype floating-point code written in C on a fast
Pentium PC to C and assembler running on a little 33 MIPs fixed-point ADI
DSP system I designed. It was quite challenging. 8-)
Leon
Reply by ●November 22, 20042004-11-22
Exactly what MANIC does, except I don't only combine various sensors,
I
further combine various sensor groups. And, again, this is something
I've been doing for years without having to define a fancy name for it.
I've always felt it better, where precision is needed, to use more than
1 type of sensor.
Al
Jonathan Kirwan wrote:
> On Tue, 23 Nov 2004 02:29:30 +1030, you
wrote:
>
>
>>With all respect to Leon I hate the term 'sensor fusion',
it's just
>>another meaningless piece of market hype to me, probably spawned in a
>>university somewhere to appear to be something new in an effort to raise
>> research funds. Another pet hate was mechatronics, basically what
I've
>>been doing most of my working life. Changing the names now just confuses
>>my aging brain.
>
>
> Actually, from personal experience working with _certain_ US projects,
it's very
> real. What I know of it uses Kalman or Bucy-Kalman techniques. Basically,
you
> know a priori some information about your various sensor systems and their
> errors and noise shapes and you also know some information about the
targets you
> are tracking and their time-dependent abilities to impact the sensors
(their
> acceleration limits, for example) and you develop some "optimal"
filtering of
> each individual 'system' and then develop the resulting error
covariance matrix.
> These can then be combined in objective ways to develop an overall
"sensor
> fusion" that tracks the target better than any one sensor system alone
can
> achieve.
>
> Jon
>
>
>
> .
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
Reply by ●November 22, 20042004-11-22
My understanding was that he simply wanted to know of a whole range of
sensor recommendations, which I gave him, to learn how to use these
ones, then to figure out (much later on) how to combinen them to achieve
specific results. hence I answeed with a list of some of the sensors,
and sources that I use.
MANIC is a system of mine that is about matchbox sized. It now has
several configurations, the least of which has 18 sensors, GPS and a
single RF system. I've never needed a special term to convey its
meaning, or what it does, and most of the interest in it has come from
the military and emergency services.
Al
Leon Heller wrote:
> ----- Original Message -----
> From: "onestone" <onestone@ones...>
> To: <msp430@msp4...>
> Sent: Monday, November 22, 2004 3:59 PM
> Subject: Re: [msp430] Sensors
>
>
>
>>With all respect to Leon I hate the term 'sensor fusion',
it's just
>>another meaningless piece of market hype to me, probably spawned in a
>>university somewhere to appear to be something new in an effort to raise
>> research funds. Another pet hate was mechatronics, basically what
I've
>>been doing most of my working life. Changing the names now just confuses
>>my aging brain.
>
>
> I thought he was asking about what to do with the data from several
> different types of sensor, like an intrusion detection system where one
> might have sound, IR, ground vibration, etc. and there is a requirement to
> put all the data together and come to a decision about whether it was a
> false alarm or not. I don't know of a better term than 'sensor
fusion' for
> that sort of activity. The military have similar problems detecting and
> identifying targets.
>
> Leon
>
>
>
>
> .
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>