Basic (and less basic) stuff about servos, radios, pulse width and frame rates

[UPGl2AYDD]

You two should work for Nasa. Seriously.

[lankey]

Originally Posted by UPGl2AYDD

You two should work for Nasa. Seriously.

Think you'll find that would be the European Space Agency Mike!!

They don't sell much continental beer the other side of the pond.

[Ber60]

It is the beer that made us do these things the only thing is that ESA don't know it (yet)

[jperkosk]

Originally Posted by UPGl2AYDD

You two should work for Nasa. Seriously.

Naaah, NASA and ESA are low-tech pussies of yesteryear, they would spend $100,000 and never have gotten a Belt to fly decently!

[redbird300]

I have been doing those measurements in post 20 again, with another multimeter that I had to test anyway. I must congratulate Erwin, his previous readings from his oscilloscope were spot on, considering the possible tolerances. Not that I doubted that.

Here are the results. The multimeter reads Hz and %, but I already converted those values to milliseconds in the table below. One thing that worries me a bit is that the servotester has a very high frame rate in the 760µs mode: 561 Hertz. I expected more something like 300-400 Hertz. Otherwise all looks normal.





And here you see how I did it, don't mind the battery and the BEC, they are only there to power the tester.
This is the first measurement, 66.82 Hertz, the frame rate in the 1520µs mode.

[mnwizard]

Excellent work Raf!!


Let's all drop Bugster a PM an nominate this thread to stickydom

[AnTi Taco]

so with a 600 size heli that uses bell cranks from servos to swash what kind of torque is needed? i got 132oz now is that good enough?

[redbird300]

While I was at it, I decided to do an extra, more interesting measurement:

Let's see if this DX8 fast "11 ms mode" is really true (instead of the normal 22 ms mode), and no marketing hype, lol. Pretty sure it isn't, just kidding. Hooked up the multimeter to the elevator channel of the RX, with the TX in 11 ms mode.

Guess what, the frame rate is exactly 90.9 Hz = 11 ms. Pulse width is of course 1.5 ms at mid stick.

Measured the min and max pulse width also, by holding low and high stick while measuring. And after that, I set the dual rates at 125% in the radio, which is the maximum, to see how much it would influence the pulse width. Here are the results, look for yourself how much difference there is:

[redbird300]

Originally Posted by AnTi Taco

so with a 600 size heli that uses bell cranks from servos to swash what kind of torque is needed? i got 132oz now is that good enough?

Sorry, but I don't know of any people that managed to measure the real torque of servos in a heli in flight, not saying it would be totally impossible, but didn't find any data yet. Not that I searched hard, BTW. You can be pretty sure though that most standard size heli servos (for cyclic) will do the job. If in doubt, just look what others are using on the same heli, I guess that's what we all do ?

And with that question, you're making me thinking....hard...lol.

[Tomstoy2]

It's a good question, when is enough, enough?

You can argue specs until you're blue in the face, but it all comes down to what happens in flight.
I'm sure you can bolt the bird down, solder an adapter and do a test under load, but that is a worse case scenario. In the air, servo load will be less as the heli 'gives' is position up.
Example: in flight a servo will see full load initially then rapidly decay, whereas with the heli tied down the load will be constant.
Pretty much what you want to read anyway.
I think it would make an interresting test. Not only load rating but consumption under load.

[Jonnyheli]

What you would need is a very good 3D pilot that can throw a heli 'round like a rag doll... Get a device that is small enough to be strapped onto the heli, that will somehow measure the torque during the flight. But saves the data in a log so you can see the results at the end of the flight

I don't know how easy this would be to achieve... I have no clue about this sort of stuff...

-Jonny

[redbird300]

I have a little idea about how to know the real torque in real flight conditions, but it can not be measured directly.

Let's first assume we have a static benchtest of a servo, where we accurately measure the torque with a good dynamometer, and note the according currents. We would have a table like this: 100 grams of torque = 0.1 Amps / 200 grams of torque = 0.2 Amps / etc... up to the max stall torque. The numbers are only examples of course.

Second, we would need to fly the heli, with an onboard (fast) datalogger in the power line of the servos, only measuring the real time currents in other words.

After the flight, we could read how much current the servos took at any moment, and compare this with the values in the table, which would give us the torque that the servos developped. Just a first idea, might need to think a little more about this, but it is doable.

[redbird300]

About this resolution discussion we had before, and servo accuracy:

It is very unlikely to feel any real-life difference between 1024 and 2048 bit resolution transmitters, at least when it purely considers resolution on cyclic servos. (Not talking about 11 or 22 ms frame rates now, this could perhaps make a noticeable difference.)

Why ? Well, I found some detailled reports that clearly proved such a thing, and I checked them myself (can't help it, I was born like that ), to come to the same conclusion in the end:

How ? I connected a frequency meter again to the elevator output of my AR8000 receiver (2048 bit / 11 ms), and made the smallest possible elevator stick movements . I repeated this a whole lot of times, to try to get the smallest movement possible, and at the same time I read the difference in pulse width that came with it.

Guess what, I needed about 5 µseconds of pulse width at least for the smallest possible elevator stick movement !

Knowing that pulse width can vary in this case from 1.11 ms to 1.89 ms, for a total of 0.78 ms or 780 µseconds, we can calculate the number of steps (or movements) my hands can do over the whole stick range: 780 / 5 = 156 steps. In other words, I simply can’t put the stick in more positions from the low end to the high end !

It is clear now that having 1024 or 2048 possible positions isn’t of much use, as we humans are too physically limited to enjoy that kind of resolution. Remember, talking about cyclic servos alone, in a flybarred setup. A flybarless controller might very well be able to do better, as well as a good gyro. Anyway, it was a nice experiment, and I feel retarded now.

I rest my case, your honor.

[redbird300]

Oh, and also noticed in the specs that a good analogue servo like a HS-65 for example has a 4 µs deadband, and the digital HS-5065 has a 2 µs deadband. Just some examples, very high end digital ones might be slightly better. But that means that those servos won't even move for the first 4 or 2 µs of change in pulse width. Meaning that my lousy 5 µs movements were perhaps not that bad after all, lol.

[knutenf]

Thanks for your amasing thougts and insight Redbird, its just incredible, even I understand and learn a little bit from your thoughs and splended explanations. You might even have found a way to measure a electronic limit to human response versus electronics. I hope this wount end in a exlucion test for people who wants to fly Belt CP choppers though. I`m afraid I would be excluded then. Anyway: Happy new years to you and to all helifreaks from all countries, youre all great.

[redbird300]

Originally Posted by redbird300

I have been doing those measurements in post 20 again, with another multimeter that I had to test anyway. I must congratulate Erwin, his previous readings from his oscilloscope were spot on, considering the possible tolerances. Not that I doubted that.

Here are the results. The multimeter reads Hz and %, but I already converted those values to milliseconds in the table below. One thing that worries me a bit is that the servotester has a very high frame rate in the 760µs mode: 561 Hertz. I expected more something like 300-400 Hertz. Otherwise all looks normal.

And now for some graphical views of the same data, with the latest little tool that the mailman dropped in my box today, must be worlds smallest oscilloscope, lol. That's one of those typical small bottles of threadlocker beside it, as a reference.



Only measured the 1520µs position of the servotester, with the knob turned completely to the left (closed) and after that completely to the right (open). The results are the same values as before, but it is nice to see the waveform change, and to be able to save it to disk.





"Freq" is the frame rate, 1/66.8 = a frame width of about 15 ms.
Delta T = pulse width, varies from 900µs to 2240µs (6 to 15%).

[mnwizard]

As usual Raf, outstanding work

[redbird300]

One more measurement, while I’m playing with it:

I did put the servotester on “automatic”, this means it will make any connected servo go from one side to the other and back, repeatedly without stopping. The speed is adjusted with the black knob on the tester.

In other words, you’ll see the signal on the little scope that a servo would receive when you push the corresponding stick on your TX forward and pull it backward again all the time, to the limits. Note that the frequency (= frame rate) doesn’t change of course as said before, only the duty cycle ( = pulse width).

Just for nerdy fun, lol. On the other hand, it gives a nice visual insight of how things work inside your radio gear:


[ame]http://www.youtube.com/watch?v=eSYM3Plk75M[/ame]

[CyprusDave]

Very cool, I want one! More interesting picture than half the TV programs I get on the the big screen in the living room

DSO Nano V2, $89 - that's the one, right? Some other cool (OK geeky) stuff on their web site.

EDIT: still looking at the SeeedStudio web site. I'd say this is the display used in the DX6i TX
http://www.seeedstudio.com/depot/gra...?cPath=163_164
OK, I'm sure a lot of LCD displays look alike, but having just had the display out of my TX, I'm sure this is the one.

Why did I have the display out? New thread starting soon

[redbird300]

Yes, that’s the one Dave. This is a single channel scope, but it works nicely for low frequency stuff. There is an alternative and much better firmware available on the SeeedStudio forum, written by user “BenF”. Version 3.43 now. I flashed it right away when I got it, and it’s just perfect.

In case you might be interested, there will also be a faster quad channel version available in a few months, but it will cost more of course. Look for “DSO QUAD”, it’s also on the same site, only a beta version for now.