Thread Tools
May 10, 2009, 08:58 AM
Registered User
kbosak's Avatar
Quote:
Originally Posted by dmgoedde
You are hilarious Jack! I hope you could hear through my mumbling voice that I stated that I purposefully reversed pitch and roll on the display to make it track the IMU just for the video!
Seriously, I saw at least 3 amateur FPV HUDs with reversed roll in Poland. We have a lenghty thread on one of the Polish forums with ppl arguing which one is better This is like a crowd of IMU zombies claiming that accelerometer should be enough for wing levelling.
Sign up now
to remove ads between posts
May 10, 2009, 09:05 AM
Registered User
kbosak's Avatar
Quote:
Originally Posted by dmgoedde
TThe REALLY funny thing to me is accelerometers in an IMU give REAL noisy data in flight... a lot noisier than thermopiles!!! IMHO thermopiles are a much better sensor than accelerometers.
Oh cmon lenghtly educated talk ended like this. Of course they are noisy. So what. No IMU implementer cares about temporary value of any accelerometer, as nobody implementing thermopile leveller cares about number of photons per square meter per second (oh yes they do fluctuate as hell, not to mention about their energy being highly dependend on the actual wavelength, while the wavelenght changes alot if we are looking at the sky or downwards, but then come reflections from metalic objects like roofs and Russian Woodpecker emitter whatradically changes the spectrum, in short, it is funny that given the number of scattered light thermopiles give any reasonnable data at all ).
May 10, 2009, 09:25 AM
Gaftopher
Gary Mortimer's Avatar
Surely there can only be one way for a HUD to work??

Anything that moves away from the norm is just plain silly.

Will confuse the heck out of real pilots!
May 10, 2009, 10:21 AM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by biloxthecat
June....... hmmmmmmmmmmmmmmmmmmmmmmmmmmm that's not far after the end of May for us here in Australia! What to do..........
Nooo sorry. New product = another iteration of export classification. US/Canada only by June. Elsewhere at least 2 months after that, best case.
May 10, 2009, 10:43 AM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by kbosak
Oh cmon lenghtly educated talk ended like this. Of course they are noisy. So what. No IMU implementer cares about temporary value of any accelerometer, as nobody implementing thermopile leveller cares about number of photons per square meter per second (oh yes they do fluctuate as hell, not to mention about their energy being highly dependend on the actual wavelength, while the wavelenght changes alot if we are looking at the sky or downwards, but then come reflections from metalic objects like roofs and Russian Woodpecker emitter whatradically changes the spectrum, in short, it is funny that given the number of scattered light thermopiles give any reasonnable data at all ).
The range of energies doesn't vary that much... wavelength range from 3um to about 15um, so that's a 5x range of energies. Not even 1 order of magnitude! The thermopiles are in differential pairs so of course we don't care about certain things that vary between Earth and Sky, just so long as there is a gradient.

My point was that without data filtering, you CAN trust single measurements of thermopile. By nature they are relatively low bandpass devices and not noisy. Not true at all with Accels as you pointed out. Heck, in my implementation of IMU I'm employing such tricks as 16X oversample of each ADC channel but I leave it in 16X format (12 bit ADC + 4 more bits) without a shift right by 4 bits to get it back down to 12 bits. I just use it as a 16 bit value. On top of that I'm recycling tricks that take last 4 samples and estimate a cleaner current accel estimate based on the trend. I leave that in another 16X format by virtue of how I set that up. Now we are up to a 20 bit value. Sure there is noise, but I assume it's Gaussian or at least some sort of symmetric distribution around the true value. Next we take cleaned raw accel for X, Y, and Z axes, and subtract out centripetal forces (Y and Z) and linear acceleration of the fore/aft axis (X accel), and viola finally there is a pretty clean attitude-only component for X, Y, and Z accel axes. Nontheless this is still a LOT noiser than the data from gyro axes. So I take the last state estimate and push it forward in time considering the 3 axis rotations and delta-T since last iteration using my in-house developed strapdown equations, and I have a new estimate for XYZ attitude components of the model based only on gyro data. Finally mix in a very tiny amount of accel data to keep the model converged on reality. Using the cleaned attitude-only components of Accel data keeps everything sane. This IMU actually will stay converged for 15 seconds through complex motions with 5000:1 predict to measurement update ratio, however that is only on a bench test. Continual complex motions in flight require much heavier weighting towards accel measure updates.... between 100:1 and 500:1. Still pretty tiny luckily.

Videos will come as soon as I sort out a bad GPS reception issue. In Atto you can fully seperate stability and navigation issues by using the Assisted-RC mode to simply fly around like how a FMA Co-Pilot works. Once stability is tuned well, then the integrator moves on to navigation gains. In all my IMU flights the last 3 days I've gotten horrible GPS reception (barely 4 or 5 sats in lock on the ground, and frequently losing lock in flight with craft entering left loiter circle. Yesterday in Auto mode I could get a decent loiter circle but not worthy of a video. 2 weeks ago on the Spectra flying wing (different GPS module) all autonomous IMU flights were 100% smooth because I got 9-11 sats consistently with the psuedo-differential type lock (WAAS). In all cases the assisted-RC mode (IMU as an attitude reference) things are rock solid. I just have to put a ground plane under this GPS or add a ferrite toroid, and will have some excellent videos very soon.
May 10, 2009, 04:49 PM
ios
ios
Registered User
Quote:
Originally Posted by dmgoedde
Nooo sorry. New product = another iteration of export classification. US/Canada only by June. Elsewhere at least 2 months after that, best case.
Dean, are you shooting for one or two products? Will Attopilot v1.8 continue being sold after Attopilot IMU becomes available?
May 10, 2009, 06:33 PM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by ios
Dean, are you shooting for one or two products? Will Attopilot v1.8 continue being sold after Attopilot IMU becomes available?
Correct - two products. Thermopile and IMU. All my commercial interest was ok to evaluate the Atto performance based on good demonstrated results of the thermopile unit, but they want the "mil spec" performance of an IMU for attitude. Iterative models like used in IMU are up my alley already anyway, so it just took a bit of effort (and prodding) to push it over the top to something that works well. IMU not trivial, but far from impossible to implement.

BTW - 4" square copper plate under the GPS does indeed seem to help lock quite a bit!
May 10, 2009, 10:08 PM
Registered User
Quote:
Originally Posted by dmgoedde
Correct - two products. Thermopile and IMU. All my commercial interest was ok to evaluate the Atto performance based on good demonstrated results of the thermopile unit, but they want the "mil spec" performance of an IMU for attitude. Iterative models like used in IMU are up my alley already anyway, so it just took a bit of effort (and prodding) to push it over the top to something that works well. IMU not trivial, but far from impossible to implement.

BTW - 4" square copper plate under the GPS does indeed seem to help lock quite a bit!

Time to cut that square down until we start seeing GPS issues again, it would be nice to cut down some of the weight, not that it will be all that much but as you know once you get all the equipment on the bird it starts to get up on the wing loading. Every little bit counts......
May 11, 2009, 12:22 AM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by airmcn_3
Time to cut that square down until we start seeing GPS issues again, it would be nice to cut down some of the weight, not that it will be all that much but as you know once you get all the equipment on the bird it starts to get up on the wing loading. Every little bit counts......
This is a section of 1/32" PCB. Pretty lightweight. More of a size nuisance. I have to admit I saw this somewhere before (can't recall where). No electrical connection to GPS's ground, just double side taped to it. I have to guess the copper plate acts to somehow accumulate signal. It appears to be 1/2 wavelength.

11 sats in lock in under 5 seconds now. Good GPS lock is worth every gram of that copper plate.
May 11, 2009, 12:30 AM
Registered User
dmgoedde's Avatar
Thread OP
Before anyone squaks that I'm flying this without a COA or have an unfair advantage because of whatever... keep in mind this is a E-Flite Mini FuntanaX, and YES there is an IMU on this but the flight mode is assisted-RC, which is functionally equivalent to the FMA Co-Pilot. I'm in a remote-ish area, and this plane weighs but perhaps 1 kg. Altitude kept low and distance close. Police drive by and wave all the time. To them I'm just a dude flying an RC toy airplane on a dead end street.

The reasons for doing this demo:
1) This is a small no-dihedral 3D aerobatic plane, NOT an 8' Telemaster or other high-wing trainer. What's the point of demonstrating an IMU on a plane that is inherently stable and barely needs control input?
2) After code enhancements added 4 days ago, I am UNABLE to confuse this IMU no matter what I do in RC flight. 14 flights, since then, just cannot confuse it. The video shows inverted flight and spiralling dives in RC then passing control to Auto for recovery.

I recently heard a certain IMU called "rock solid" by the company that makes it. Rocks can be broken. I'm aiming for 'bullet proof' myself


(1 min 15 sec)
Last edited by dmgoedde; May 11, 2009 at 12:42 AM.
May 11, 2009, 12:49 AM
Registered Crasher
Bilox's Avatar
Quote:
Originally Posted by dmgoedde
Nooo sorry. New product = another iteration of export classification. US/Canada only by June. Elsewhere at least 2 months after that, best case.

Thermopiles it is then!!
May 11, 2009, 06:21 AM
Registered User
kbosak's Avatar
Quote:
Originally Posted by dmgoedde
I recently heard a certain IMU called "rock solid" by the company that makes it. Rocks can be broken. I'm aiming for 'bullet proof' myself
Excellent.
Did I see by eye something like 10-15 roll error that fades away in 2s? Of course nothing to worry about.
May 11, 2009, 07:47 AM
Registered User

Ground Plain


I've used aluminum tape for years on my Slowfly FPV to improve GPS performance.

About 1:25 into this video you see it on the Elevator surface. I mounted the gps antennae back their because this PIC,OSD and Video TX caused it lockon problems.

UAV Slowfly (1 min 34 sec)


Kisssys
May 11, 2009, 08:27 AM
Quote:
Originally Posted by dmgoedde
This is a section of 1/32" PCB. Pretty lightweight. More of a size nuisance. I have to admit I saw this somewhere before (can't recall where). No electrical connection to GPS's ground, just double side taped to it. I have to guess the copper plate acts to somehow accumulate signal.
You are correct. Your plate gives the exact same effect that wearing my aluminum helmet does: It collects cosmic energy and helps me think better all day. And the best part is it works on rainy days.
May 11, 2009, 08:34 AM
Gaftopher
Gary Mortimer's Avatar
Paint that grey and sell it to the military KISSYS


Quick Reply
Message:

Thread Tools