DIY gyros, copilots and other angle/accelleration sensing toys!

[KreAture]

Hi fellow DIY'ers!

I thought I'd start a thread here to discuss different methods of sensing motion/accelleration and the angle something has to our planes common enemy; "the ground".

What do we know so far?

Gyros:
* spinning kind with sensors detecting forces on the mount of the spinning "thing"
* vibrating kind (used in piezo-gyros and MEMS based gyros) with sensors detecting forces on or feedback from the tuning-fork or disc.
+ Accurate
+ Fast
- Expensive
- Not too easy for DIY

Inclinometers:
* multi-sensor based boards/units available
+ Quite accurate
- Expensive
- Needs calibration


So, what can we make ourselves ?

Gravety based Angle-sensor:
I have one sensor on the works now that uses a small tube of liquid with an airbubble in it. On top it has two small metal pads covering the two outer 1/3'rds of the cylinders length. On bottom it has a long metal pad covering entire length. The sensor is ready by measuring the difference in capacitance between the two smaller upper pads in reference to the larger bottom pad. The liquid and bubble create a sort of hysteresis making it less prone to false readings tue to accelleradion/deaccelleration along the axis of the tube. I believe this to be a good way to do tilt/pitch sensing, but maby there are simpler ways?


Any other ways we can detect angle or accelleration? Please people, come with suggestions! Some of us here are certainly capable of making our own gyro-substitutes, we just need to figure out how to do it! :)

[JMP_blackfoot]

Is this thread about making one's own sensors ? :confused:

[KreAture]

Yes. It's about making stuff for use in helis and planes for stabelizing/controlling with either 100% DIY stuff, or comercial sensors in DIY circuits.

[wittekonijn]

hallo


i ame trying to build an compas that wil fit in my submarine

i wil post pictures if i have anny


raf

[Mr.RC-CAM]

For Pitch and Roll attitude sensing, the Infrared method is another solution. It is not perfect, but conveniently avoids all the inertia obstacles that you will experience when you use the other methods you mentioned.

The most public example of the IR method is seen in FMA's Co-Pilot model stabilizer. IR designs like the Co-Pilot utilize the earth/sky heat signature (it is not optical) to determine horizon data. The sensor they use is a very self contained device and can be adapted for other applications.

For example, a low cost R/C model Artificial Horizon Indicator project, that uses the IR method, is fully described by following the links here: http://www.rcgroups.com/forums/showthread.php?t=265153

For utmost accuracy I would say that the gyro methods work nicely. But do not underestimate the difficulty in using them. The inertia issues will require that you employ very complex designs to achieve accurate results.

RC-CAM

[KreAture]

Yep RC-CAM. That's why I have been exploring the differential capacitance tilt-sensor.
It has a slower response that may serve our use well.

About the IR-sensor. I browed through your design and see that you use the FMA device. Thus the project is less interesting. That sensor is unavailable at all hobbyshops here, and if it was available it would be seriously overpriced anyways...

Back to the issue...
So, one can look at sky and ground and find the horizon :)
How about a rotating IR-sensor? Or a series of em in a 90 deg section of a circle?

By reading X sensors and looking at the signals, a PIC might be programmed to find the spot where it goes darker... ???

[KreAture]

Oh, and I suspect the FMA sensor uses two single line cmos sensors.
Theese would read in a strip of pixels (512 or 1024 of em!) and the information would be transferred to other circuits as a analog signal.

Such a sensor would cost me $47 a piece, so it's not a viable alternative.

[Mr.RC-CAM]

The FMA sensor uses IR thermopiles. You would need two per axis. The are about $15 each in single quantities. You might be able to use the sensors from surplus infrared motion detectors to reduce costs.

RC-CAM

[KreAture]

I think my differential capacitive idea is easier and cheaper then.
All I need is a thin-walled tube I can fill with water and some aluminum or copper tape.

[Mr.RC-CAM]

Sounds like a fun project. And no doubt low cost (air bubbles are cheap).

I suspect you will see some of the inertia issues, even with the suspension acting as a damping media. But the air bubble mass is low, so perhaps that will mask the issue. Typcial worse case is usually situations like long slow banks.

If you expect to do realtime model attitude control, you will probably need sensor response times that will require the bubble to be extremely mobile. If it is for basic model telemetry then performance requirements are definately not critical.

RC-CAM

[Ron van Sommeren]

Welkom Raf a.ka. Wittekonijn(huh?????)

Wittekonijn (Dutch) = white rabbit (Grace Slick, Jefferson Airplane???)

Groeten ;) Ron

[KreAture]

RC-CAM, the size of the bubble relative to tube diameter and the fluid viscosity defines the responsetime.

I did not invent this type of sensor. NASA did many years ago. They apparantly even use the sensor today. Maby to give some sort of reference to the electrical systems?

[stumax]

Kreature, how about a laser ring gyro? Surely you could handle making one of those! All you need is a loop of optic fiber, a single shot laser pulse and a very accurate time measuring method. I think it's about the only way you'll do away with inertial effects. Otherwise, how about putting a sensor on each corner of the aircraft to measure the earth's electric field strength - any variation will indicate a pitch or roll attitude. Sure, the field changes depending where on earth you are, but in the small envelope of space we will fly in I guess we can assume it's constant enough.

Stu.

[XJet]

The problem with gravity-based sensors for aircraft work is that they're also affected by the G-forces that a flying model can generate.

The ultimate gravity sensor is the human inner ear -- but it's far to unreliable to be the sole source of navigation which is why full-sized pilots can't fly in pitch-black or white-out conditions without instruments. Without instruments, pilots in zero-visibilty conditions relying on "the seat of their pants" quickly become disoriented and lose control.

For example -- it *is* possible to pull 1.5Gs continuously throughout a loop - even though you will be inverted for 50% of the time. If you were blindfolded you wouldn't know that you were upside down -- so how would a gravity-based sensor?

Many, many years ago we built free-flight models with pendulum-controlled elevators. These were very simple devices that worked on the assumption that a pendulum would always hang vertically. By attaching a pushrod to the pendulum and running that pushrod to the elevator, the airplane would (supposedly) be far more stable in the pitch plane.

As the plane pitched nose-up, the pendulum would remain vertical (which meant it would actually swing backwards and, since the pushrod was connected to the other side of the pivot point, it would pull on the pushrod, thus applying down-elevator.

Similarly, when the nose went down the pendulum would remain vertical (relative to the outside world) and thus would swing forwards relative to the fuselage - thus causing up elevator to be applied.

*IF* you could set them up correctly these planes flew surprisingly well but 9 times out of ten you'd get it wrong and they'd see-saw up and down quite violently until they smacked into the ground.

The other problem was that if turbulence caused the model to roll into a steep turn, the centrifugal forces (G forces) generated would cause the elevator to stay fixed in a neutral position -- even though the nose of the model was falling rapidly into a spiral.

As far as the gravity-based sensor (the pendulum) was concerned, the model was still flying level.

Just why did these free-flight models need extra pitch stability -- usually because they were scale models with very small tail-areas relative to the wing-area.

[OmegaDot]

I seem to recall some NASA papers from the mid 80's that discussed the differential field measurement method for roll control. Also, some NASA papers on a homebrew flux gate magnetometer were made public (built around a toriod core). I saw these presentations at a forum on autopilots at the EAA Oshkosh Fly-In.

Another angular rate sensor type used the "bending" of an air jet when the body carrying the orifice was yawed. The jet deflection was detected by differential temperatures of self heating filiments placed on either side of the stream (or all 4 quadrants). The "filiments" were miniature "grain of wheat" light bulbs with the glass broken off! The air stream was inducd by pulling a partial vacuum on the sensor body and bleeding air into the body through the orifice. The vacuum pump was a speaker diaphram pushing/pulling through a venturi.

[KreAture]

Well, this sensor I am talking about is NOT for aerobatics.
It is for flying level... I thought I made that clear?

[JMP_blackfoot]

The electrostatic autopilot was developed by Maynard Hill and Ben Givens around 1972. The system is described in :
"Introducing electrostatic autopilots"
AIAA Astronautic Journal of Aeronautics Vol. 10, No. 11, November 1972, p. 22
and :
"Electrostatic autopilots"
Flying Models, February 1973, p.20, No. 431

More on the amazing achievements of Maynard Hill :
http://www.modelaircraft.org/museum/bio/Hill.pdf

[KreAture]

How about a link to some actual technical info on that :)

[JMP_blackfoot]

How about a link to some actual technical info on that :)
Yes, that's what I haven't found yet. But I do have a paper copy of the article. I'll look for a link some more.
Best I've found so far : http://www.charlesriverrc.org/articles/asfwpp/lelke_alb2electronics.htm

[Hovertime]

Any developments since ?

[KreAture]

Mostly I've found expensive sensors not worth playing with.

[ZAGNUT]

The FMA sensor uses IR thermopiles. You would need two per axis. The are about $15 each in single quantities.

RC-CAM

i would be very interested in a source for these sensors as i would like to do a permanent co-pilot installation but don't want to destroy the stock sensor. the op-amp is no problem but for the sensor i haven't come up with anything in all my searches.


thanks,
dave

[Mr.RC-CAM]

I would be very interested in a source for these sensors as i would like to do a permanent co-pilot installation but don't want to destroy the stock sensor.
Digi-Key sells Melexis thermopile sensors. However, it is MUCH more cost effective to buy the ready made 2-axis sensor from FMA. Because, at $15 a sensor (you need four), plus the instrumentation amp and misc parts, you will spend more to make your own.

RC-CAM

[Hovertime]

Too bad you bubble leveler idea did not work out, sounds simple and cheap.

Another thought - mercury sensors? (if such are available)

On the other hand I doubt that any gravity based sensors could be successfully used in an aircraft, even for sedate flights.

[KreAture]

Didn't work ?

I haven't had time to test that yet.

[william541]

Has anyone tried to use the electrolytic dual axis tilt sensors for pitch and roll of a "slow moving" aircraft?

http://www.spectronsensors.com/tilt.html

look towards bottom of page. You need to excite them with pulses to prevent plating, but it may work. No idea on the cost, but they shouldn't be too expensive.

-Bill-

[william541]

OK, here is a crazy idea with several draw backs, but it would be cheap if it can be made to work.

The idea is to use clear lens LEDs with 4-5 LEDs pointing towards the ground and 4-5 pointing towards the sky.

Since LEDs act as small photo cells, perhaps the horizon could be detected based on light differential.

Drawbacks:

1) would not work at night
2) may have trouble with snow or tin building/ glass reflections

This works great to make solar panels track the sun. Maybe the system could use the sun and a real time clock/calendar in a micro to make the calculations?

Just a thought... -Bill-

[wowo44]

Same idea, but uses infrared temperature sensors to detect the horizon:

www.ctie.monash.edu.au/hargrave/horizon_sensing_autopilot.pdf

[Mr.RC-CAM]

uses infrared temperature sensors to detect the horizon:That is the basis for the FMA Co-Pilot stabilization system.

Speaking of IR horizon sensing, I have decided to offer a few more "MAHI" artificial horizon boards. This device overlays fly-to markers on live downlinked video. The project's history and the latest PIC/Board offer is found here: http://www.rc-cam.com/forum/index.php?showtopic=294 . If interested, please post on the ongoing thread using the provided link.

RC-CAM

[william541]

Same idea, but uses infrared temperature sensors to detect the horizon:

www.ctie.monash.edu.au/hargrave/horizon_sensing_autopilot.pdf

Yes, I see that... Especially when reading the posts above from RC-CAM. He did a nice job with a video interface to display pitch and roll information.

I suppose a person might be able to use infrared LEDs, but I have never tested their photovoltaic response to see if they are more responsive to infrared or visible light.

Outside of a strap down gyro, I'm not sure what can be done for full aerobatic craft. Slow craft should be able to use a tilt sensor of some type though.

Fun to think about for sure... Bill

[wowo44]

Also related is the Paparazzi (http://www.nongnu.org/paparazzi) project: A fully autonomous UAV that basically uses a clone of the copilot sensor head for its pitch and roll data. It's pretty neat.

[RolfPW]

Hi, new idea?
I've slaughtered a logitech optical mouse. The sensor is a HP HDNS 2000 (1.2g). It contains a signal processor and a image sensor array. If i point this sensor (without any lens system) towards a window or in direction of a lamp, both quadrature outputs toggles. I'm actually searching for a useable lens, witch hopefully will improve perfomance. If it works, it may be a very accurate sensor module.

Gruss rolf

[KreAture]

Err, doesn't such a sensor only have two outputs ?
Wouldn't that limit your sensory input to when you are ON horizon or when you pass the horizon up or down? The real sensors used for such stuff has a long row of pixels to determine where horizon is, and not just if you are on it or passing it.

[sesat]

Kreature, two single pixel IR sensors can be used for horizon sensing if you point them facing away from each other.


Ram.

[RolfPW]

Kreature,
sorry HDNS 2000 was not the right one. Back home i see HDNS 2051 is it. 2 of them can be found in a dual optical mouse from Logitech. You can find the spec here:
http://cp.literature.agilent.com/litweb/pdf/5988-8477EN.pdf
This devices have two quadrature outputs for each axis (x, y) and one PS2 style serial interface.
The HDNS supply's motion information of two axis either accelerating or deccelerating and as decribed in the lower section of the spec the pixel array can be dumped via PS2.
With a combination of that three informations it may be possible to hold an aircraft in a stable flight. Testing with the optics are in progress. And please keep in mind: it's only an idea! (pooh must go back to school, learning english)

rolf

PS by dumping the array, nick angle tracking may be possible too with only one sensor array

[thanhTran]

Very interesting thread.

Has any one used this sensor before?

http://www.parallax.com/detail.asp?product_id=28017

It's pretty reasonable price and has two axis. It's also "air bubble type"

I was thinking about using this sensor for my self-stablizing flying machine. It's like a gravity based and momentum based sensor from what I understood, but I think it will not have the problem like the pendulum. It will be affected by temperature changed though. I don't know if it fast enough for flying as the mass of tiny air bubble is too small :). I think parallax use this sensor for their walking robots :)

Thanh

[Mr.RC-CAM]

Has any one used this sensor before? I used the Memsic 2125 in my HeadTrack-R project. It works as advertised. Some photos & info: http://groups.msn.com/rccam/fpvunavuavprojects.msnw?albumlist=2

RC-CAM

[william541]

The Memsic accelerometers look like really neat parts, I'm still reading up on the spec sheets and AP notes.

I like the idea of the industry standard parts such as the MXR7210G/M as they tend to be lower cost and probably have a second source as well. The web price at the Memsic web site was $9.30 ea. Not too bad... I also like the ratiometric parts as you can make the excitation and the A-D reference track each other.

I suppose a person could use one of these devices for a slow craft. I'm not sure if a 3-axis accelerometer would do any better in a faster craft without the use of a gyro stabilized platform to mount them on. I don't know the answer to this one at the moment. Maybe time to revisit the strap down innertial nav system notes I read somewhere several years back. This may be getting a bit complicated for a RC project.

If this would work, you could tie the accelerometers into your compass to tilt compensate it. I have used the KMZ52 by Philips in a compass design. With a 10-bit A-D, you can get a bit better than 1 degree resolution. Perhaps do both on the same board.

Maybe a look into ultralight avionics on the market might yeild some more possibilities. -Bill-

[william541]

I used the Memsic 2125 in my HeadTrack-R project. It works as advertised. Some photos & info: http://groups.msn.com/rccam/fpvunavuavprojects.msnw?albumlist=2

RC-CAM

I read a little about your HeadTrack-R, were you planning on re-posting it on your RC-CAM site? I couldn't locate much detail on the MSN forum. Looks like you already tilt compensated your compass... I had not seen flux sensors like that in a few years, kind of neat to see them again.

Bill

[Mr.RC-CAM]

I read a little about your HeadTrack-R, were you planning on re-posting it on your RC-CAM site? The Memsics worked well but the digital compass did not meet my needs (the link has all the details about the problems it gave me). If I ever come across some affordable & painless technology that offers reliable heading operation then I will update the project. At this point the only worthy solutions are budget busters.

[RolfPW]

Last night i have attached a lens, recycled from a CDROM at that sensor (#32, #35). I'm thinking it's not the optimal solution, but i'm on the right way. In the morning i demonstrated the setup to one guy, who works alongside me. He suggested me to make a small video and some pictures of the stuff and post it here. He will host the files for a while at his site, until mine is up.

a video of the slightly working apparatus:
http://home.versanet.de/~b-konze/image/MouseEye.AVI

a closeup of the pixel array (field aperture + polarizing filter):
http://home.versanet.de/~b-konze/image/Pixelarray.jpg

one dead, opened sensor:
http://home.versanet.de/~b-konze/image/brokenEye.jpg

Research will go on (heading lock or. . .)

rolf

codec not found: http://www.sm-art.hu/link/codec/wmv9VCMsetup.exe

[william541]

If I ever come across some affordable & painless technology that offers reliable heading operation then I will update the project. At this point the only worthy solutions are budget busters.

I would be willing to collaborate on a compass design if 0.5 to 1 degree resolution would be adequate. I have never tilt compensated one, so I'm not up to speed on the algorithm to do so. The known working design I have in mind requires 2 channels of a 10-bit A-D. The heading calculations are done in firmware all with integer arithmetic thanks to a programmer buddies efforts.

It should be possible to come up with a single board that can perform pitch, roll and heading information. Perhaps with differential and absolute pressure sensors, you could also throw in air speed and pressure altitude.

I really like the idea of having the attitude information on the video, and perhaps it could also be encoded into the closed caption data... I think it was around line 27 in NTSC video. This would allow you to go full screen with the video and still have a means to capture data during the flight all with the same transmitter.

One additional note:

I'm not sure how well a magnetic compass gets along with the high current wiring on some electric RC craft. Usually you try and stay at least 12" away from wires carrying more than one amp. The hard iron calibration should be able deal with the motor and servo magnets I would think.

-Bill-

[Mr.RC-CAM]

I would be willing to collaborate on a compass design if 0.5 to 1 degree resolution would be adequate.That would be plenty for my headtracker.

Pitch and Roll detection using inertia sensors (i.e., Memsic) would also require gyros and Kalman filtering if the setup was mounted on a model. However, the IR sensor method avoids that sort of complication. But, a properly designed Kalman based solution is more precise.

A compass on my MAHI project would be very useful. If you come up with something that you wish to share then count me in. :)

RC-CAM

[ZAGNUT]

y'all might find something of interest in Gordon Anderson's "magicbox" project. basic explanation can be found under "projects" and more technical info including source code is in the download section of his site, www.mstar2k.com


dave

[KreAture]

I really like the pixel-array a lot better than the IR thermo sensors.
They give a much better indication of the surroundings.
Both have a big problem with my typical flying spots though as there are buildings, mountains and stuff blocking the horizon up to 600ft some places...

I want a sensor that feels the movements relative to "the fabric space"! :p

[RolfPW]

Basic evaluation of the pixel-array on my testbed has been closed. Here are some results:
around 24 main clicks on a 90 degree right turn
mostly the same on the turn back to initial angle
same result on a the left side turn
As one quadrature sequence consists 4 states, so the resolution must be 4 times better.
Next i'll make a PCB, must select a suitable controller. . .
Maybe some weeks or month later i can post better results.
rolf

[KreAture]

These pixel-arrays would probably give some resolution to the sensoring:
http://www.elfa.se/pdf/75/07520786.pdf (512 pixels)
http://www.elfa.se/pdf/75/07520778.pdf (1024 pixels)
Both are linear strip image sensors.

Smaller one can do very fast readouts where all pixels-values are held in a buffer for simultaneous readout. The 1024 one samples each pixel in turn.

Note: Theese sensors do NOT output digital data, they are controlled digitally but output a analog charge value for each pixel wich you would need to do a ADC on...

Pretty cool huh ?

[KreAture]

sesat
You are ofcource right, but such a setup requires calibration to certain extents and will be easily fooled by buildings and stuff in the way.
The pixel-strips are a lot more powerfull for locating a horizon but won't help on the obscured horizon issue...

Again, I wish we could sense the gravity of earth without being sensitive to motion :(
Unfortunately we can't tell the difference between accelleration and gravity. (Atleast not according to Einstein...)

[Mr.RC-CAM]

Those sensors are sensitive to visible light. You will obtain more reliable results with moving up to the IR spectrum. The Earth's heat signature has proven to work better for horizon detection than visible light methods. Even NASA has used it for spacecraft re-entry.

I guess I'm not sure why you have dismissed the IR thermopile sensor sort of method (like the FMA CoPilot uses). It has more resolution than a 1024 pixel array and it works with little fuss. If cost is the problem then there might be surplus fire detectors out there that you can rob parts from.

RC-CAM

[KreAture]

No no no...

I never said I dismissed it. Also I never said anything about not using a IR-filter on the 512 or 1024 pixel sensors...

You would get a immediate fix on both sides of horizon with the strip-sensors btw and they have a LOT higher resolution than a simple thermopile. Each of the pixels are analog and would look at a different angle of sky.

Anyway, you need to be abowe obstructions before this stuff works. Spaceship reentry is a bit higher than I intend to fly. My fav spot is between a BIG builting and a small row of trees...

[Mr.RC-CAM]

The sensitivity of those two strip-sensors is very low in the IR region. Even with a filter on it, the practical resolution of any single sensor will be low due to the reduced light/dark voltage differential. In contrast, the IR thermopile sensors are optimized to react to true IR heat and are extremely sensitive at the longer wavelengths. Fractional centigrade values can be measured with them.

Using the strip sensor arrays you mentioned might offer an advantage in a model stabilization app in that they could help eliminate the calibration step (a curved sensor would be good). But from what I experienced, four thermopiles arranged as a 2-axis sensor array would probably work beyond your expectations.

Of course your idea may indeed be the Holy Grail of horizon sensing, so ignoring me is probably best. :)

[sesat]

A linear thermopile array would seem ideal for what you guys are talking about. Perkin Elmer makes an 8 element of 5 degree fov each. I have no idea how to get my hands on them.

Anyone know where to get inexpensive (<$20) thermopiles with narrow fields of view (<25 degrees)? The Melexis samples are 100 degrees :eek:


Ram

[KreAture]

Mr.RC-CAM
The gws app did not work in my application... It uses the IR method so I am looking for alternatives. For other uses it may be fine.

The problem is, as I said the building. It blocks everything, making the plane flip wing up thinking it needs to find the horizon again.

[Mr.RC-CAM]

The problem is, as I said the building. It blocks everything, making the plane flip wing up thinking it needs to find the horizon again.I understand. Using a visual or thermal defined horizon reference is probably not an option for you. Inertia sensors with gyro compensation is probably your best bet. Or a new flying site. :)

[william541]

A compass on my MAHI project would be very useful. If you come up with something that you wish to share then count me in. :)


Drop me an e-mail with your contact info and I'll gather up some information for you. Maybe we can work it into something useful for the RC hobby. -Bill-

[william541]

On the thermopile issue, even the inexpensive motion sensors in alarm system use them. The lenses are also inexpensive and come in a variety of types to cover different "look" angles. The lens spreads the detection area into numerous infrared fingers or zones.

I can see where uneven trees and buildings might pose a problem for infrared, especially at lower altitudes. I'm inclined to agree with the gyro stabilized accelerometers for applications like this.

Not sure if there is an inexpensive (<$15 US) source for micro gyros.

I still think an electrolytic tilt sensor would work fine for lumbering aerial photography work.

[KreAture]

The cheap gyros I got (PM-03 or something) are $30 or so...
For my heli I now use a HLG200 and I have two PM-03 (or something) lying hre for other stuff.

For ultralight and diy it doesn't do much good though.

[william541]

I just ran across this open source project for UAV Attitude Heading & Reference System (AHRS). Loads of good information and opportunities to get involved and add to the project.

They are scavenging parts from a gyroscopic mouse and are testing the system out on a RC helicopter.

http://autopilot.sourceforge.net/ahrs.html


I looked for the data sheet of the Murata ENC-05 piezo gyro, but only found the ENC-03M as a current part.

http://murata.com/sensor/index.html

From what I have read around the net, the ENC-05 can be bought on eBay as part of a gyroscopic mouse... If Murata does not list the ENC-05, are they obsolete or is this an OEM special order part number?

At this same murata link, there are infrared sensors and fresnel lenses if anybody is interested.

-Bill-

[william541]

In thinking about all that has been cussed and discussed here, I'm starting to see possibilities with the array idea from KreAture, Rolf and sesat(RAM). I had to take a couple of pictures to illustrate my thoughts.

The idea is to read the array which can be in the visual spectrum or preferably the IR spectrum (say 10nm). Then take all pixel intensities and assign them a level of 0 to say 16 based on "brightness". Then from this data, find a horizontal or diagnal line in the image to represent the horizon.

The sensor(s) mounted and looking sideways would determine the pitch and the sensors mounted and looking fore and/or aft would determine roll. You could do this with 2 such arrays, but it would work better with 4 as you could average the scene on opposite sides of the craft.

In the attached "simulated IR" images, you can see several places where the horizon angle can be seen. Even with buildings or trees (The sample has a tree on the left and more on the horizon below a cold sky), there should always be a horizontal reference in the image. If after each scan of a sensor, the image was scaled (weighted from 0-16), it would automatically compensate for image intensity and self calibrate on the fly. This would take a fast micro or a slow micro tied to an inexpensive DSP or FPGA used to crunch the data.

This is kind of like the visual navigation and inspection systems only lower resolution, just enough to find the horizon. If you kept track of the pitch and roll values, you could tell when you were upside down or right side up.

Just a thought, -Bill-

[william541]

I ran across this paper from two folks at Monash University in Austrailia. They are using the onboard nose camera and electronics to determine the artifical horizon similar to what is expressed above.

In case you are interested, here is a link to their paper.

http://www-ist.massey.ac.nz/icara2004/files/Papers/Paper59_ICARA2004_339_344.pdf

-Bill-

P.S. RC-CAM, I have the compass information when you are ready.

[mrbaseballny]

I've neen following this thread with some enthusiasm as I am designing my own "flight Computer/Co-pilot" that would be placed in my Harrier for pilot assistance in keeping a steady and level hover. I have been researching tilt sensors;as I am not concerned with G-forces, it will be only used while hovering. I'm also trying to keep weight to a bare minimum.I have seen several different models that might be of use to you as well for your project.

1.) http://www.althensensors.com/products/index3.php?MGID=18&ID=159(see first picture below)

2.) http://www.jameco.com/webapp/wcs/stores/servlet/ProductDisplay?langId=-1&storeId=10001&catalogId=10001&productId=130269

3.)http://www.sensorsmag.com/articles/0500/120/main.shtml

Although the first one is slightly too big for my use, but it might be used for a larger plane.

Just me thoughts:)

[mrbaseballny]

Can anyone tell me if this looks like it will work?

thanks in advance

[Mr.RC-CAM]

Can anyone tell me if this looks like it will work?All that is missing is about a 1000 lines of assy code. :)

Some ideas for you:

* The PIC16F627 has a very stable internal 4Mhz osc (1% accuracy). So, delete the external xtal.

* Delete the MCLR parts. Instead, use the Power up timer and brownout features (work just as well). This will keep MCLR free for easier ICSP programming (you'll eventually want to support that).

* When you route your PCB board, pay attention to the ADXL ground paths (treat the ADXL like it was an analog component).

* If you move the programming jumper over to Port B, you can omit the 47K pullup resistor. Port B has programmable pullups for you to use.

* The PIC16F627 is only 1K code space. From my experience, this project will easily exceed that. Use the 2K part (F628).

* Your PIC symbol shows Vdd, but Vss is missing.

[william541]

I'm guessing these servos are not under radio control as well?

In looking at the datasheet and info on the ADXL202, looks like the 14pin package may be on the way out with the 8 pin LCC being more "favored", but if you already have a source, your good to go.

You may want to think about extra filtering on the VCC to the accelerometer since your probably after minimum tilt signals during a hover. The data sheet shows recommendations for doing this.

I have not interfaced synchros before, but I always try and buffer the output of a micro if the signals are going off-board. Or as a minimum, decouple them for RF with a .01 and/or .001uF cap. 10 cents worth of caps can prevent problems later.

RC-Cam and others have synchro experiance and can let you know what is required for stable operation and long life.

-Bill-

[mrbaseballny]

For those interested I found that on this website http://www.ringolake.com/pic_proj/servo/accel/accel.html, but I was wondering if it would fit my application(why reinvent the wheel when someone else has done it for you:) I would like to be able to modify this schematic to control the servos while its hovering and be able to take inputs from the transmitter. I figured I would start here and modify what I needed to make it fit:)

I have the assembly code in a notepad document that he wrote as well, I'll have to look it over as I'm new to this type of programming:(

Mr. RC-Cam, William 541 thanks for the input!

BTW: I see that analog sells the ADXL202 on their website here: http://www.analog.com/en/prod/0,,ADXL202,00.html or at sparkfun here:http://www.sparkfun.com/shop/index.php?shop=1&cart=204091&cat=71&

[sesat]

Been thinking about pitch/roll sensing in urban areas or below the treeline: a triad of of ultrasonic rangefinders. Like the 11meter Devantech SRF08.

Ram.

[KreAture]

That would limit useable areas to flat places...

[william541]

a triad of of ultrasonic rangefinders.
Ram.

I was thinking about that too only using infrared laser range finders, but I couldn't figure out how to avoid issues with sloping terrain not to mention buildings and vehicles getting under one leg of the sensor array.

But it could be used for an altimeter. Or even for terrain avoidance if it was pointing forward...

-Bill-

[sesat]

Friends, how about a suntracker?

Ram.

[sesat]

I did find that Perkin Elmer 8 pixel linear thermopile array I was looking for. Devantech UK kits it with a I2C interface for nearly $100 a piece. Four of them (for pitch & roll) are more than I can afford.

A linear array on a swivel panning back and forth could be used to hunt for thermals.


Ram.

[william541]

Friends, how about a suntracker?

Ram.

You may be able to use sun detection as a compass of sorts. Pitch and roll would be a bit tricky though.

Has anybody found a CMOS image sensor similar to the Mitsubishi M64285FP?

I think this chip is no longer in production, but it was a B/W 32X32 CMOS imager array that had built in edge detection and serial output of the data all in a 10-pin chip. Still looking into sensors to detect the horizon angle with a visual means.

Looks like the quad thermopile still has the edge for simplicity and cost so far for pitch and roll measurements. I'm not sure if there will be a simple solution for use around buildings and trees though.

Still thinking about it ... It's a tough one. -Bill-

[RolfPW]

M64282FP may be an alternative. It's a 128X123 sensor with edge detection and serial output. You can find it in a gameboy camera. Have collected some of them at ebay (very cheap).
I have used this chip in some projects and can say: It's a cool piece of silicon! If you average all 4 pixel in x and y direction, you can get nearly the same resolution (32x32). If you can't find the spec., i can PM it to you.

rolf
ps. what's the hack, how can i post a picture

[william541]

Rolf,

Seems like all the Mitsubishi sensors are all obsolete now, too bad...

I like your idea of the optical mouse sensor by Agilent. Seems that the sensor chips can be had for under $6 US. The ADNS-2620 looks interesting in that it is a small form factor (8-pin staggered dip). You can read the pixel data from the 18X18 array by reading/writing the pixel data register. You just write the first pixel address and it will automitaclly increment to the next pixel on successive read cycles.

Not sure what kind of lens you would need to focus at a distance. The sensor has a small pinhole that would take some good alignment with the optics.

Post back if you do some more tests with these little sensors...

-Bill-

[sesat]

Does anyone have experience with the CMU vision computer?


Ram.

[sesat]

Does anyone know where I could get those little cameras that go in mobile phones?

Sparkfun used to sell a CMOS imager on its own untill recently, Telit enforced their requirement that Nathan sell them only bundled with the GSM module.

32x32 imager would be crude for direction sensing, far too crude for pitch sensing.


Ram.

[RolfPW]

Bill/william541,
after a look at the ADNS-2620 spec. i think the SQUAL, maximum, minimum and pixel-sum functions may be useful for determination of the sensor output quality and X and Y displacement information are given too.
If i compare the Agilent chips---->
HDNS2000/2051:
+ quadrature encoder outputs for either x and y direction
+ we can ignore the SPI and use only quad. information
+ 2 of them can be slaughtered from one mouse
- no quality information ( if exposed to clear, blue sky or under gray weather conditions they tell us nothing )
- 2X form factor of the ADNS-2620
ADNS-2620:
+ quality information
+ small form factor
+ quality information ( but what to do under poor conditions ? )
- we must implement the SPI protocol

What's about the optics? As you have posted, there is only a small pinhole and the sensor array is small too, i think you must use a lense with a short focal distance. On the practical side, i use lenses from CDROM-drives. ( only 1 or 2 of 8 lenses i have tested will do )

good luck rolf
ps. Philips KMZ51/52 may be helper devices, if you can place them apart from high current traces.

[william541]

Rolf,

In the interest of cost, what about using a pinhole lens? Some of the real cheap digital cameras use a pinhole lens and they have the added benefit of a great depth of focus, which is what would be needed in this application.

Some experiments would need to be done to determine how far to place the pinhole from the sensor array to get the optimum field of view for detection purposes.

I see that the ADNS-2620 and others have an ALC of sorts where it adjusts the shutter values until it receives proper exposure. Since these are usually used with LED illumination, I have no idea how well they would work with ambient light conditions.

Clearly some experiments would need to be done to see if it holds any promise at all. It has been questioned that a 32X32 array may not be enough resolution to determine attitude angles; these sensors are even smaller, 18X18 pixels. Honestly I don't know what the minimum resolution requirement would be for useful measurements.

With the SPI interface, at least you have access to all the registers and possibilities of one micro talking to several sensors without the need for massive I/O requirements.

Have Fun, Bill

[RolfPW]

Bill,
i have used pinhole lenses (0.4.. 0.8 mm) on HDNS2000/2051 with no response from the chips. I think, we must expose that sort of chips with adequate illumination.
Happy Easter!
rolf

[william541]

Rolf,

Well that is not encouraging news, I was hoping they would be a bit more sensitive than that. I guess we need to look into other low resolution B/W pinhole CMOS sensors...

I was just thinking that if the device did not have enough resolution for pitch and roll, it could be pointed down and give ground speed and drift information. A high altitude mouse of sorts!

I've been meaning to ask, is that a Zebra Finch that you use for your icon picture?

Regards, Bill

[RolfPW]

Bill,
yes it's 'NeNe' a Zebra Finch. Hi is the leader of a squadron with 6 members.
good luck
rolf

[william541]

Does anyone know where I could get those little cameras that go in mobile phones? Ram.

Try Omnivision, they have a good selection of sensors.

http://www.ovt.com/

These are also used in a mini camera modules like the following

http://www.quasarelectronics.com/c-cam2a.htm

Speaking of Omnivision, the OV810 might hold some promise in the RC aircraft attitude application. I see the cost is $1.50 in die form, so not too useful for us at the moment. I dropped them an e-mail concerning this application, we'll see what they say. It is so new that all I can find are the press releases through a google search.

-Bill-

P.S. Rolf, nice to hear about your Zebra squadron! We had one around our place for a while, I think someones pet got loose, he was hanging around with some house finches for several weeks.

[radiohound]

Here is the Melexis Thermopile sensor at more affordable cost for low quantities. http://www.scalerobotics.com/ecommerce/os/catalog/

[sesat]

I love you Radiohound!!!

Looks like I'm not going through with my Roithner order.

Ram.

[radiohound]

Here is the schematic from the Paparazzi UAV project on how to use the melexis sensors. Voltage supply is 3.2 volts. A level aircraft gives an analog voltage of 1.6 volts.

More info at: http://www.uavs.net
and http://www.scalerobotics.com/store

Walter

[william541]

Hi Walter,

That is a pretty slick way to null out the two thermopile sensors connected in series, saves an Op-Amp that way. I would guess the thermopile pairs would have to be well matched to do this.

Question though, since they are not using the internal thermistor anyway, why do they connect the thermopile VSS to the VCC/2 instead of connecting to ground? This makes the case of the thermopile energised and increases the chances leakage that could throw the VCC/2 off and slew the output of the circuit.

Just curious? -Bill-

[william541]

This is just too fun a thread to let die down. I have been mulling over an idea for attitude measurement and I know I'm missing something but it just hasn't clicked in my mind yet.

The idea is to use a 3-axis solid state magnetometer (a 3-axis compass if you will). The easiest way to picture this working is at the North pole. Up North, the magnetic vector points straight down to the ground. I picture an airplane as a flat piece of paper , then push a pencil straight through the center of the paper pointing down to represent the earths main magnetic vector. Now tilt the paper nose up. This now forms an acute angle between the magnetic vector and the nose. Same holds true if you tilt the paper left or right. It does not tell you direction, but wouldn't it tell you the pitch and roll attitude?

Now if we move to North America, the magnetic vector has about a 60 degree tilt, still pointing down. If we know this angle before hand, couldn't we still calculate aircraft attitude?

The idea would be to use the 3-axis magnetometer to find the strongest magnetic vector in three dimensions and then compare this angle to the aircrafts horizontal surfaces that the magnetometer is mounted to. In the 60 degree example, as long as the main magnetic vector was 60 degrees in relation to the horizontal plane of the aircraft, we would be level.

The one drawback I can see is if the aircraft was tilted 60 degrees (North America example) so as to be perpendicular to the magnetic vector, then we wouldnt know if we were pointed up, down, or at a bank angle. We would just know we are 60 degrees off level in some direction.

I know I must be missing something else but I'm too tired to think of what it might be.... Any ideas out there?

-Bill-

[Mr.RC-CAM]

Now if we move to North America, the magnetic vector has about a 60 degree tilt, still pointing down. If we know this angle before hand, couldn't we still calculate aircraft attitude?I can share my experience with measuring the earth's magnetic fields for attitude sensing. Back in the 80's I created a digital compass for robotics applications. It was a Flux Gate design. Its main purpose was for traditional compass heading navigation. But it also had a mode where it could be used to demonstrate if the sensor was off-axis (tilted).

A DOS PC demo was created that showed the sensed fields as a graphic component. Once calibrated and perfectly level, the graphic was a single dot inside a big sphere. As the sensor was tilted, the dot would grow in length, essentially becoming a line, with the length representing how much off-axis the sensor was. The direction the line tilted indicated which way the sensor was tilted. It was very interesting to watch. And it nearly worked.

The problem was that the earth's mag fields can change radically as you move geographically. If the sensor was calibrated for one area, then moved a few hundred yards, the display would sometimes be biased by the local magnetic field environment. I believe this error is referred to as magnetic inclination. From what I hear, there are areas where this effect is very intense (due to iron deposits).

The demo was a crowd pleaser. But, in the grand scheme of things, it really wasn't much use for anything more than a bit of entertainment. It was fun to hold the sensor in your hand and balance it until you got the tilting line to shrink into a tight little dot.

RC-CAM

[william541]

The problem was that the earth's mag fields can change radically as you move geographically. If the sensor was calibrated for one area, then moved a few hundred yards, the display would sometimes be biased by the local magnetic field environment. I believe this error is referred to as magnetic inclination. From what I hear, there are areas where this effect is very intense (due to iron deposits).RC-CAM

I sure wish I cold have seen that demonstration, it does sound like fun. The inclination angle is indeed what I was trying to measure with the instrument. I was hoping that the inclination angle would be as stable as the declination angle is for a given flying area and perhaps only the field intensity would fluctuate. Do you recall how many degrees of inclination change you saw for a given area?

Sounds like some experiments might be in order. But like you say, it may work for one area and not others. I'm hoping I live in a worse case area. The ground here is almost all basalt (lava). I have read that geologists sometimes use a magnetometer to locate lava tubes by mapping an area and looking at the changes in intensity. I do not know if they take into account any changes in the inclination angle. They also measure the angle of ferrous crystal structures in lava to use as a history of the earths magnetic field change over time.

-Bill-

[Mr.RC-CAM]

Do you recall how many degrees of inclination change you saw for a given area?Between the office and my home (about 8 miles), the error was probably a dozen degrees or so. I suspect that each locality will be a total wild card.

RC-CAM

[william541]

Wow, that is more than the inclination charts show for a distance of several thousand miles in North America even if you travel North/South.

Wonder what all was going on and if there was a way to minimize it? I sure know that if you are inside a metal building that the electronic compass really gets messed up compared to readings outside the building. -Bill-

[Mr.RC-CAM]

Oops. I should have explained the design a bit more. My comment was based on how it was displayed on the demo screen. The tilt values (and errors) that were seen had been expanded by the system gain. Lower gain would not have allowed the tilt function to work well. The absolute magnetic error value was never known (not a purpose of the system).

Edit: BTW, from what I recall, local deviations can cause several degrees of real error. There are some regions where the magnetic issues become very severe due to local iron reserves and other environmental gotcha's. I think this is why attitude sensing, by way of measuring the earth's fields, has not been very popular. But perhaps for a controlled area the idea would work fine.

RC-CAM

[william541]

Well, perhaps not all is not lost on the idea. I'll run some tests and push them through a spread sheet to see if it warrants further investigation.

-Bill-

[Mr.RC-CAM]

Just in case you missed it, I edited while you were posting. So, I re-edited and added "edit" to the new text.

The local site specific characteristics is often not found on the global maps. So, my gut feeling is that you shouldn't use the published maps for your test data. Instead, travel the area you will be flying at and measure the actual field strengths along the expected flight paths.

RC-CAM

[lostrider]

For bubbles, you can use lighter gas, in tube.
when its under pressure(liquid mode), liquidity and bubble are perfect.

[radiohound]

Here is a two axis tilt sensor which uses thermopiles to determine the heat of ground vs air/space. It gives out an analog voltage for two axis. It is available here: http://www.scalerobotics.com/store/catalog for about $50

[jhdgkss]

Hi there iam working on this same problem and i think i have cracked it and it all rellies on useing pic chip and an accelerometer. The programing is slow in learning but i am makeing good progress.

[KreAture]

I suggest anyone wanting to program for a microcontroller and know anything about C, to go to http://www.avrfreaks.net and search for the free tool suite WinAVR to use with AVRStudio. The result is free C-compiler and development suite for AVR type chips with a simulator/debugger. In addition to being very easy to learn, you can make a programmer for AVR chips using only 3 resistors and a 25-pin connector for your printerport. Worth a look, for sure.

[Quacker]

What's a printer port?

[Gary Warner]

What's a printer port?

LOL :rolleyes:

I see we have a new computer.... :D

History 101...

Printer Port:
adj. - computer port
a receptacle on the back of older computers that has numerous pins, used to connect the printer to a computer with a cable consisting of no less that 12 wires and as many as 25 wires (but no one knows exactly how many wires were used). Also known as a total waste of gold plating used for all the pins. See USB.

Gary
--