Author Topic: Sensor Timer HELP appreciated (Read 1338 times)

Erik Janssen · « **on:** November 21, 2009, 03:21:32 AM »

I have been working on a line tension controlled timer and proved that it helps to overcome difficult weather conditions. In order to continue development I need your help.
The details of this setup are published in the engineering section of this forum. There were a lot of readers and I got some very useful feedback from this.
My background is engineering and when I learned to do some programming we punched holes in cards. Today there are a lot of parts and tools available in robot shops. I use processor and sensor boards from www.pololu.com and AVR4Studio (free download) to write the software. I use the Baby Orangutan processor board and Analog sensor. Being a total novice in this technology it took me a year to learn enough to program my own timer.
I managed to make a movie in calm weather where you can actually hear the motor changing rpm. In order to improve we have to define first what we want.
It would be great help if you would look into the movie and advice me where you want the motor to kick in faster, later, more, less or not at all. You may also award me points to get an idea of which maneuvers are ok or need some rework. (my flying is a bit rusty)

From the feedback I will publish a summary on this forum and use it as a design spec for next year

John Witt · « **Reply #1 on:** November 21, 2009, 02:20:06 PM »

Eric, it's hard to tell what's going on exactly. Talk to Alan Hahn about his technique for synchronizing a data logger and video to correlate what maneuver you are doing with the logged RPM.

I would also layout a map of the desired power/thrust/rpm etc for the flight. You don't necessarily need to fly the whole pattern either, which would let you get in more flights (suggested to me by Igor)

Also, some notes about the airplane set-up/weather conditions/line length etc. Take more data than you think you will need.

Looks to me like you fly OK, certainly good enough for what you're trying to do, Worthwhile project I think. Also talk to Howard Rush.

John W

Alan Hahn · « **Reply #2 on:** November 21, 2009, 02:58:19 PM »

Erik,
Compared to me you are an expert class flier!

It was intriguing hearing the motor pick up typically around the zenith. My own feeling, looking at my velocity plots is that the best time to apply power is before the airspeed drops too much--at least in the calm conditions. This basically is in climbing with angles less than 45 degrees, where I see the airspeed dropping the fastest. If you do that, then the airspeed above 45 degrees will be better, because in that region, the thrust doesn't have to fight the full force of gravity (just due to the angle of flight).

Like I mentioned, the "trigger" for this could be done using a 3 axis accelerometer. Lets let axis "1" be radial (down the wingspan), axis "3" be in the direction of thrust, and axis "2" at right angles to both. With the airplane sitting on the ground, axis "2" would be vertical.

Just before entering the bottom corner of a square maneuver, a₁ ~3g, a₂~1g, and a₃~0.
in the corner itself, a₁ is still ~3g, but a₂>10g, and a₃ is negative (opposite thrust). This state is where I would begin to add power. When you come out of the corner, a₂ drops back ~0, a₁ is a little less than 3 g, and a₃is still slightly negative. When a₃ gets close to zero (not decelerating anymore, you could begin to back off the power.

This is just a suggestion, obviously pointed at square corners. I think it would be useful to actually fly a passive 3 axis accelerometer and datalogger to see what is going on during the entire pattern just to see if you can find unique enough info from the 3 axes to actually generate unique triggers at the right time (and not at the wrong time!!!).

I keep thinking about making that a winter project---I see the accelerometer I'd like to use (+16g on each axis, but the readout is digital, while the datalogger I was thinking about is primarily analog.

I think it might be possible to work out a calm weather algorithm which would work with your windy-weather algorithm. Of course in relatively calm conditions, you almost need nothing extra anyway--you can solve it with more passive fixes (like props).

However next year's primary direction has got to be improving my flying. Making all these recent videos has pointed out to me how much it s*cks.

Howard Rush · « **Reply #3 on:** November 21, 2009, 06:13:22 PM »

I agree about recording the signal first. I'm scheming something similar, but first I'll make a recording. Even before that, assume the accelerometer signal has a lot of noise on it, particularly at frequencies higher than the band of interest.

Alan Hahn · « **Reply #4 on:** November 21, 2009, 08:12:12 PM »

Quote from: Howard Rush on November 21, 2009, 06:13:22 PM

I agree about recording the signal first. I'm scheming something similar, but first I'll make a recording. Even before that, assume the accelerometer signal has a lot of noise on it, particularly at frequencies higher than the band of interest.

That's a good point.

Probably find a new way to balance the prop/motor (~150-160Hz).

If I was doing onboard motor corrections, I'd want a low pass filter that would eliminate frequencies above about 50Hz or so (but still keep the details of interest below this limit). The accelerometer I'm looking at has a configurable digital filter built in, so it should work fine. Problem is learning to code the data logger to handle I2C communication. If I could though, then I could also read out my airspeed sensor and altimeter.

Erik Janssen · « **Reply #5 on:** November 22, 2009, 06:50:45 AM »

Thanks for the compliment of my flying. In 1984 I finished 22nd in Chicopee and Jimmy Casale 4th, four weeks later we met at the Circle Burners competition where Windy Urtknowski let me fly his .60 Reno Sweeper. That model was a lot better than mine. I made Jimmy swet and finished 2nd. So what you see is a lot of airplane quality and a little pilot. The plane used for the experiments made it 21st in Shanghai 1994 and was originally powered by an Aldrich .46 ABC.

In this post I am searching for the functional design specifications about what we want a sensor to do, not primarily the technical issues involved. Sometimes it helps to shoot on an example.

I solved the filtering issue by taking 10.000 analog samples per second and an average value every 512 samples. This allows me to have 22 adjustments per second.

Dean Pappas · « **Reply #6 on:** November 22, 2009, 08:26:52 PM »

Hi Erik,
This film clip is a nice companion to the very impressive one where you fly in the North Sea gale winds!
Of course, the sound can be misleading on the streaming video, but several things occur to me.
1) the power addition sounds very good for both the vertical * and hourglass.
2) it happens a bit late almost everywhere else, especially in the consecutive rounds.
Does this agree with what you feel at the handle?
I think it agrees with your observations that the rounds get wide if the "kick-in" altitude is too low.
3) there seems to be a compromise between the tops of the under 45 maneuvers and the edges of the overhead 8.
Do you feel this during the transition through upwind in moderate wind?
4) I didn't see or hear obvious evidence of descending too fast from above 45.
Once again, how does it feel with the handle in your hand. (Oh yes, the rust is falling away very nicely)

So how quiet is the accelerometer output in flight? Is there any hope of adding a bandwidth-limited derivative term into the proportional one? I'm a hardware guy myself, but I do lots of collaboration on algorithms with my firmware counterparts. I often write difference equations or write filter/control routines in psuedo-code. The last time I wrote serious code was in 1979 in Assembly for a 6800, using a single-step and register dump as a debugger. It convinced me to be a hardware guy! I ought to force myself to endure the same learning curve that you are: my hat is off to you.

talk to you soon,
Dean

Alan Hahn · « **Reply #7 on:** November 22, 2009, 09:13:51 PM »

Quote from: Erik Janssen on November 22, 2009, 06:50:45 AM

Thanks for the compliment of my flying. In 1984 I finished 22nd in Chicopee and Jimmy Casale 4th, four weeks later we met at the Circle Burners competition where Windy Urtknowski let me fly his .60 Reno Sweeper. That model was a lot better than mine. I made Jimmy swet and finished 2nd. So what you see is a lot of airplane quality and a little pilot. The plane used for the experiments made it 21st in Shanghai 1994 and was originally powered by an Aldrich .46 ABC.

In this post I am searching for the functional design specifications about what we want a sensor to do, not primarily the technical issues involved. Sometimes it helps to shoot on an example.

I solved the filtering issue by taking 10.000 analog samples per second and an average value every 512 samples. This allows me to have 22 adjustments per second.

I guess I'm not too sure what you are asking.

From my simple calculations, the airspeed starts dropping as soon as you point the nose up. So soon after the nose is pointed up is the time to add power, not at the top when the speed has decayed to almost its minimum. That's my opinion of course, and as far as I know, you and Igor are the only two who have really tried to do something active to alleviate the tension at or near the top of the circle. So you two are the experts here, the rest of us are only conjectoring.

Erik Janssen · « **Reply #8 on:** November 23, 2009, 01:34:00 AM »

Dean
1, thanks
2, I have put it late so it does not mess up my loops, putting the switching point lower will make the loops wider than high, especially the bottom loop of the vertical eight.
3, true
4, feels soft when recovering, my wingover starts to be ok
At the handle I have no points where panic rules, all feels quite secure as line tension is everywhere. It feels funny in the vertical eight, when crossing over to the top loop I start doubting if it comes and then the power is there. In the overhead eight it feels fast overhead as we fly tailwind.
I introduced a derivative term and from the noticeable slower motor sound I expected the down slope to be slower. PID has not changed since 1954, so if last value – actual value > 0 then the plane picks up speed and we may shut the throttle down. But this will only help if the model bleeds off speed fast enough and it does not. My plane goes up in 2,7 seconds in the vertical eight and comes down in 2,4 so my next plane will have more wing area and less weight.

Alan,
Be a judge,

A data logger will tell me the variations in centrifugal forces but will NOT tell me if a loop is round. I am looking for functional stories like:

Applying power in the hourglass in wind seems a good solution it looks to come on a little late and don't you think it goes down a bit fast.

From these stories we gather requirements. Suppose we want to reduce the rpm after the 3rd corner of the hourglass we can figure out how to do that. More important is to know if this will work. So maybe someone may step in for a small experiment. Put a RC receiver in the model, make a movie of the wingover, fly a second one and manually go full throttle when the model goes up and throttle back at the top of the zenith. If this gives smooth transitions try to fly an hourglass the same way. If this works we can pick up the challenge to build something that recognises the 3rd corner of the hourglass, if it doesn't work we have to figure out how to brake an electric motor, then it is useless to work on the controller.

So we gather stories, sort out the requirements, test the assumptions and then build it.

At 393 readers of my previous posts there must be more opinions, please react.

Your calculations are right, but there is something that you cannot hear. Governor puts instantly extra power on as soon as the nose points up and the prop starts to lose airspeed, power goes up from 400 to 600 Watts, when the headwind is encountered above the 45 degree line governor cuts the power and the sensor opens the throttle again to 600 Watts that is what you hear.

Paul Walker · « **Reply #9 on:** November 25, 2009, 01:41:53 PM »

I too have had this conversation with Kim, about varying the power as a function of line tension. We have also considered accelerometers as a means to regulate the speed.

However, after flying electric for 3 years now, I am not convinced that this is necessary, or desirable. Clearly it IS something we CAN do, but do we need to? I have convinced myself that I don't want it. Maybe it's the fact that I'm getting older, or maybe something else, but I sure like the absolute constant speed of my electric systems. It makes flying the pattern more precisely, easier.

My systems are overkill by others standards. However, I have flown them in wind I would not fly a gas plane in, and it powered through more than adequately. Even in those conditions, it maintained a "relative' constant speed (not that anything can be constant in a very strong wind).

I understand the attraction, and I will monitor your efforts closely. I watched your video, and it looked (sounded) to me like the power came on too late. Yes, I know that's where the line tension is minimal, but that speed up would really bother me. I wish you the best in your efforts. Some day we could all be running devices you pioneered!

For me, it's more to be gained by practicing (training) more, at this time.

Keep us informed.

Paul Walker

Howard Rush · « **Reply #10 on:** November 25, 2009, 09:57:40 PM »

Quote from: Paul Walker on November 25, 2009, 01:41:53 PM

However, after flying electric for 3 years now, I am not convinced that this is necessary, or desirable. Clearly it IS something we CAN do, but do we need to? I have convinced myself that i don't want it. Maybe it the fact that I'm getting older, or something else, but I sure like the absolute constant speed of my electric systems. It makes flying the pattern more precisely, easier.

My systems are overkill by others standards. However, I have flown them in wind I would not fly a gas plane in, and it powered through more than adequately. Even in those conditions, it maintained a "relative' constant speed (not that anything can be constant in a very strong wind).

Winning by 30 points seems to have made you complacent.

Erik Janssen · « **Reply #11 on:** November 26, 2009, 02:07:27 AM »

The only substitute for practice is more practice. Great to see that you can judge it from such a small movie.

I would not use it if my setup got me in the top 10 either, practice got me from Chicopee 22, Landres 17 and Shanghai 21st. Competition limits your ability to work on developments. Without the pressure of competitions I have my hands free for development. I only fly a few to see where I am standing.

I agree with your remark that the sensor kicks in a bit late, this is done to preserve the roundness of the loops and an area where work is to be done. The sensor is intended to help in difficult areas, governor creates constant airspeed and that is best for the model to fly in.

I enclosed an example of power output and where to look for improvement and a drawing of where the power is given by governor and my sensor.

I rely on governor for the low squares, this puts extra power on the prop as soon as we have loss in speed. The 400 Watts in this example is a bit low, on a 12x6 APC-E I would expect 480 Watts. A 12,5x6 Graupner CAM, Hyperion 4200 CL, CC45 and AXI 2826/10 will give you 580 Watts. This improves maintaining airspeed in low corners.

A power train is a chain of components, the weakest one will limit the output. I can give you some suggestions where to look.

Your batteries can be the limiting factor, I once had Chinese batteries that were unable to deliver more power. You can put two packs in parallel and see if amps go up.
The ESC is may be a bit slow, a kontronik Jazz gives 500 Watts in the configuration where a CC45 gives 580 Watts. Just swap controllers and make a flight.
Maybe you are flying a bit slow, the faster you fly the higher the peak output in the wingover. (roughly 20 Watts more for every 0,1 sec laptime gain, Igor flies 5,0)
For a 780 rpm/volt motor a 12x6 seems small, I would suggest to use a bigger prop unless the rpm starts to drop to much. We tried a 13x6,5 on the AXI and found out the motor lost 570rpm in a wingover. On a Spitz 3020 with 1040 rpm/volt the 12x6 Graupner was the maximum we could use.

TDM · « **Reply #12 on:** November 30, 2009, 08:03:42 AM »

Erik this is awesome.

I was also looking in to the Basic Stamp in a parallel similar project but I think if you introduce some of the following ideas it will make your system even better. 3 d accelerometer will determine when the maneuver is over by sensing the horizontal laps between the maneuvers which will open the door to set individual power levels for each maneuver. This will enable us to customize the over all flight in good weather wind high versus low altitude. Now these are some adds on that can be very useful. Instead of having one output for one motor have several servo outputs for gear for trim tab or Rabe rudder. You can have fly by wire Rabe with impute from bellcrank and you can have adjustable trim tab to counteract the engine torque when it changes RPM and also how about the retracts?
What i envisioned is an Excel type file where you impute the parameters like Engine 1,2,3,4, power levels for each maneuver. Maybe have a take off set up a landing setup Including gear power level ramp up after so many seconds for landing keep idle so many seconds after gear down for those cool power landings. A set up for Rabe ruder that reads deflection from bellcrank and sends the appropriate impute to a rudder servo. A set up for trim tab that gets the impute from the motor power. and some general parameters like a gain for the top of the maneuvers flight time etc.
Maybe a separate page for Scale flight.
But the idea is once you set the parameters you can download to the plane and go fly.

Ok this is my dream. By the way you may get away with no timer functions because the system will know exactly when the pattern is over and it will go to landing mode.

A system like this would be killer and you can save the setups for future use in a small lap top. Anyways this is my personal goal.

Erik Janssen · « **Reply #13 on:** November 30, 2009, 08:30:48 AM »

If you go to www.pololu.com and select the LV168 robot controller you get a headstart. The Demo3 software handles the LCD display, the ADC conversion and the pwm output. At www.AVRfreaks.com you can find tutorials how to put a value in the eeprom memory.

The 6g sensor is connected to port 6 and I modified the ADC initialisation from PC7 to PC6 and started working on my logic.

I constructed my system with the baby orangutan and plug in the buttons and LCD identical to the LV168 configuration. I can change my parameters at the flying field using my controller. In the engineering section there is another lead with more tech detail and pictures.

The LV168 weights 20 gram without display, the same as the baby-O with the bearer plate I use.

Software is unlimited, just use your imagination. The rule book is quite explicit about moving aerodynamic surfaces, that is not allowed.

Alan Hahn · « **Reply #14 on:** November 30, 2009, 08:57:36 AM »

Erik,
I was looking at your Eagletree data plot trying to figure it out when it finally dawned on me that it was my plane and my plot!

Just one comment, I don't see any evidence in the rpm sagging where you show it---that area is actually just after my 4 leaf clover (the vertical pullout is the peak just past 340 s).

I do agree that the rpm seems to slowly decrease (in a linear straight line) as a function of time. At the beginning, the center of the rpm trace is about 7950 rpm, while at the end it is at ~7900 rpm. However I don't really see any load induced sags, which leads me to believe I am ok on the battery. I am not sure what exactly is going on with the rpm decrease (all of 50 rpm!), whether it is real and due to something in the ESC, or if it is fake and something in the Eagletree. I need to run my new ICE controller and see what it is claiming how close I am to hitting 100% throttle near the end of the flight.

Have you posted a link to a site that gives info on your micro? I think you have, but I'm not sure where it is.

Thanks, Alan

Erik Janssen · « **Reply #15 on:** November 30, 2009, 09:11:29 AM »

I thought it was fun to show you your own graph, I was waiting on your reaction...... You are sitting on gold without knowing.

The purple line consists of variations in rpm, with the bnb logger I can zoom into the results and then you see the variations, an AXI loses 300 rpm under load, a Spitz 3020 even more, a kontronik Jazz is a disaster, cold batteries make it worse.

We managed to throw out 584 Watts with an AXI and a Graupner CAM 12,5 x 6 in the wingover where your setup delivers 400...... Using a 13x6,5 caused saturation with negative results. So carefully working on all the elements in your power train can add considerable power to go vertical in low corners. Try two 4S batteries in parallel. Just optimising governor without sensor. (you see loops, wingover and squares)

The pololu site is crammed with information, have fun reading.

TDM · « **Reply #16 on:** December 02, 2009, 07:27:18 AM »

How do companies use 3axis accelerometers and 3 axis gyros to determine the speed of a plane? They are rather light but the software algorithms must be crazy to try to go to. If we could do that it would be awesome. What do you say?

Alan Hahn · « **Reply #17 on:** December 02, 2009, 07:43:39 AM »

Quote from: Dorin Morosanu on December 02, 2009, 07:27:18 AM

How do companies use 3axis accelerometers and 3 axis gyros to determine the speed of a plane? They are rather light but the software algorithms must be crazy to try to go to. If we could do that it would be awesome. What do you say?

Well of course in principle, if you measure all accelerations (translational and rotational) accurately, and know the initial position and velocity, you can compute your position --just simple kinematics. Any errors will cause the predicted and actual positions to diverge. I'd rather rely on a GPS to be honest, or some other waypoint check.

TDM · « **Reply #18 on:** December 02, 2009, 10:28:52 AM »

True dead reckoning induces errors all the time.

Howard Rush · « **Reply #19 on:** December 02, 2009, 03:05:59 PM »

Quote from: Dorin Morosanu on December 02, 2009, 07:27:18 AM

How do companies use 3axis accelerometers and 3 axis gyros to determine the speed of a plane? They are rather light but the software algorithms must be crazy to try to go to. If we could do that it would be awesome. What do you say?

It takes really good accelerometers and really good gyros. There is some calculation involved. Google "Inertial navigation system". These systems are still big and expensive, but who knows what will happen five years from now?

One of the interesting things about an inertial navigation system is that when you turn it on while it's stationary on the ground, without any external information it knows the latitude at its location, just by feeling the earth turning and sensing which way is up.

Alan Hahn · « **Reply #20 on:** December 02, 2009, 09:05:36 PM »

Back many years ago, when I was a grad student, I had a buddy who was doing some work ---not sure exactly what--but it included tracking and correcting glitches in gyros. He mentioned that it had something to do with the navigation system in subs. I guess if you have 10⁹ bucks to solve problems, you can get it done!

A little more than my budget.