Heli altitude hold

[gjestico]

Hello,
I would like to come up with a device to hold a helicopter at a fixed altitude for AP use. Heli is about 10 lbs and electric.
Thinking to use an accurate baro sensor such as the scp1000, but that may not do the job as on the bench it fluctuates +- 2 feet. What if an accelerometer was added to fine tune the reading, IE detect when the direction of movement (rising, sinking) has changed.
Should be simpler than an all-out autopilot as the collective pitch channel can be modulated which has a direct effect on the desired result - gaining/loosing altitude. This is a non-ccpm setup so there is a single collective pitch servo.
I have a freind who is a whiz with micros and programming, but has never done this sort of thing. I am the gearhead who builds the mech parts of the helicopter :)
Any suggestions ?

Greg

[ThaiskyDigital]

Try RCAP3+Altitude hold , we are now happy with airplane, and we going to try with heli soon.

Or if you can try HeliCommand from Geraman it is good too but just cost too much.

[kd7ost]

I think that 2 feet +/- is pretty darned good. Due to the dynamics of flight, lift and sink in air, response curve of your collective etc, it would be hard to get the accuracy you're looking for. I will tell you I'm not a heli pilot so please take this with a grain of salt. But I do fly planes with altitude holds and it's typical to see at least 25 feet of up and down movement within the set altitude. Unless it's dead calm and you trim and adjust throttle and elevator dead on. A little breeze and all bets are off.

You have to remember that a pilot is anticipatory and can react before things go awry.

An autopilot is reactive. It doesn't create input changes until the craft starts to go off course. You need a healthy amount of exponential response too. You want a lot of response if the craft is way off course or altitude. But only a little or slight response when it's barely off your setting.

You might do better (depending on altitude and ground surface) to try to hack a laser distance measuring device.

If you do use a BP sensor under a Helo, you might have to run some fuel line out the tail boom a ways to get out of the direct rotor wash or tail rotor wash. That will put errors in for sure. The sensor needs to be pretty well baffled from the wash.

[gjestico]

we already have the baro sensor mounted on the heli, with data downlink to a computer on the ground. I intitially thought the rotor wash etc would screw up the reading but it seems to be stable. have done testing at 30-50 feet and the altitude reading is very stable. At that altitude the pilot can still see altitude changes in the heli and correct, so the results jive. The problem is at 100+ feet or so looking at the bottom of the heli it gets more difficult to judge climb/fall. Pilot can react to the called-out altitude reading but theres a lag in reaction etc.
what I had intended was to use the accellerometer to simulate that "seat of the pants" feel one gets in a real aircraft at the slightest onset of pos or neg G's.
Got to get altitude hold within a few feet for this app.
PID loop tuning will take care of the reaction if the error is large or small, I think.
I dont think the helicommand has an altitude control system, does it ? its an pitch/roll stabilizer only ?
Greg

[kd7ost]

Cool, You've proven things out already. Good luck. It sounds like a lot of fun. But I'm a hardware junky like you. I can build fairly well. But I need someone else to write code.

Good luck

Dan

[ThaiskyDigital]

I dont think the helicommand has an altitude control system, does it ? its an pitch/roll stabilizer only ?
Greg


Yes ,helicommand have altitude control which is quite good, the bad point is that it infared which sent to ground can work just for 3 meter.

[gjestico]

this job needs alt hold at 400 feet ;)
Luckily my friend is even smarter than I am so I guess we will figure it out in time.

[ThaiskyDigital]

Good News Plese Let Me Buy One If Possible.

[ssozonoff]

Hi gjestico,

Please let me know how you get on. I have plans to do the same thing, maybe we could combine forces or something.

I also looked at this http://www.u-nav.com/circuitboards/alt3.html but have not tried it.

Also found this: http://alai.h3m.com/~s0350672/althold.htm

I have a couple of scp1000 and I am looking at this sensor as well http://www.intersema.ch/site/technical/ms5534.php

Thanks,
Serge

[gjestico]

Update:
So far using the basic code we have come up with the heli will hold altitude to within a band of about +- 10 feet. Accelerometer has not been used in the control loop yet we have just used it in the telemetry data in oder to see what kind of forces the loop is subjecting the heli to.
One problem I believe is that in the high res mode, the sensor updates 1.8 times per sec. Resolution whould be around 4" at this setting (theretical). There is a faster update option on the sensor but that will reduce the sensor resolution by over half. I think if there was a way to use the accel data as well as the pressure data in the loop that a better position hold would work. We are not sure how to use both devices as inputs to the control loop.
Any ideas from the code wizards would be greatly appreciated .

[AZ ChopperCam]

I'm anxious to see how this works out. I need altitude hold to within +/- 8' and I need it now! I wish Spartan would finally get around to an altitude hold for the AP.

[gjestico]

I suspect Angelos has discovered there are 100 guys waiting to buy his gyro for every one guy who wants any AP stuff. His gyro is the shiznit (I have one) they are always sold out . Im on a waiting list for my 2nd one. He most likely has his hands full building gyros.
Lets hope he remembers his loyal AP followers !.

[dmgoedde]

OK, I'm almost becoming notorious for saying this... but.....

For my AttoPilot development, I have created a real-time data fitting routine with the SCP1000. If you are inclined, there is enough detail here to re-create it yourself. You take the lower res (15bit) pressure data at 9Hz, and by doing real-time least-squares fit of the last 1 second of data (N=10 data points, from t=0 to t=-1/9s, t=-2/9s, etc... back to t=-1.0s), Then using some discrete algabraic equations (that can be found in Microsoft Excel under Help Search for "LINEST" function), you determine Y intercept at t=0 (right now) and a slope of the best fit line. The X-values of 0 going back to -1 seconds are chosen so that Y-intercept at time=0 is the current data point. Y axis units are Pascal. Slope dP/dt is in Pascals/second instantaneous rate.

With my routine I get both very fast response (9Hz update) yet about 17bit resolution on pressure, the best of both worlds! The fit routine is not influenced heavily by a single rogue data point (a "flyer", no pun intended), and a side benefit of the fit results is slope of the line (dP/dt). This allows more sophisticated PID control because rather than slow 1.8Hz data rate, and slow point-to-point estimation of rate of error change to calculate the D term, you get 9Hz best-fit dP/dt. You can feed the 9Hz estimate of dP/dt right into calculating the D term. This tends towards something I dare say looks like the human "proactive" flight control, rather than a "reactive" control that kd7ost warned about.

I would recommend to use a sampling tube that is outside of the turbulence cause by down-wash. Unfortunately this would be a 1m tube or so, but I think that would help your accuracy of alt hold as well. The tube could be a flexible piece of fuel tubing running back the tail boom, terminated by a piece of plastic tube that extends about 30cm behind the tail rotor.

So my "notorius" offer is that I can make a special-purpose module for you that is a spin-off of my AttoPilot development. I have code in-hand, experience with this hardware, and would add a 2nd input channel to have real-time gain control just like FMA CoPilot uses. For the heck of it, I would sell a few of these 100% at cost, no profit. I just LOVE doing this stuff. If you don't like it, return it for a full refund. You can contact me here or at dmgoedde@gmail.com

Dean

[AZ ChopperCam]

PM sent!

[arrio]

PM sent.

[Crash Pilot]

PM sent to email address