Antiservo

[Howard Rush]

This is a weird subject; I don't know which topic to put it in.  I'm looking for something like an RC servo, but backward.  I want to rotate the shaft and have an RC pulse train matching the shaft position come out the wires.  Does such a thing exist?

[pipemakermike]

Hi Howard
It sound as though you are looking for a rotary shaft encoder.  These are a common component but you might need some electronics to translate the pulses into RCspeak

[Igor Burger]

"servo tester" with resistor trimmer (not electronic like jetibox) ... depends if you can live with the drag of its pot.

if not, then optical encoder as mentioned ... or hall sensors from modern R/C transmitters

[Curare]

yeah, I would have thought a hall effect sensor would be the best (no drag)

[James_Mynes]

I don't know if the output is compatible with what you're looking for, but isn't that what the steering knob on an RC car transmitter does?

[Mike Anderson]

Clancy Arnold's device changes the pulse width from 1.0 msec. to 2.0 msec. as a pot rotates (or slides) from one end to the other.  The pulse RATE is 20 hz.  As mentioned, an RC servo tester is an identical circuit, but in a nice box.

[Tim Wescott]

This came out of a phone conversation I had with Howard last night, where I suggested that he make such a critter by taking the guts out of a servo (including the drive gears and motor) and putting in a circuit board that would do the job.  We're kind of looking for something with the mounting convenience of a servo, but with the information flow going backwards.

[Larry Renger]

Of what are you trying to detect the position?   

[Doug Moon]

Of what are you trying to detect the position?   

Bellcrank maybe??  Please do tell.....

[Randy Cuberly]

This came out of a phone conversation I had with Howard last night, where I suggested that he make such a critter by taking the guts out of a servo (including the drive gears and motor) and putting in a circuit board that would do the job.  We're kind of looking for something with the mounting convenience of a servo, but with the information flow going backwards.

Something like this maybe:  https://www.dynapar.com/content.aspx?id=125

Might be a little bigger than you want!

Randy Cuberly

[Howard Rush]

Something like this maybe:  https://www.dynapar.com/content.aspx?id=125

Might be a little bigger than you want!

4 grams is OK.  A single push pin seems to be a flaky way to mount it.  I'd use at least three.

[Howard Rush]

Bellcrank maybe??  Please do tell.....

Bellcrank or anything hooked up to the flaps.  It's not for control, Heaven forbid.  It's to see how far the flaps move so we can see if we're getting our money's worth out of the Igor linkage.  The reason for the RC output is that that's what the TUT records.  A flap movie might suffice, but flight test data might be nicer. 

The servo tester or RC car knob driven by its own pushrod sounds good.  A little friction is probably OK for experimental use. 

[Randy Cuberly]

4 grams is OK.  A single push pin seems to be a flaky way to mount it.  I'd use at least three.

HHHmmmmmm...

Randy C.

[Andrew Hathaway]

How about the optical sensor and notched wheel from an oldschool PC mouse?  It's light, compact, zero drag, cheap and plentiful, etc you'd just have to deal with receiving the output signal.

[Howard Rush]

How about the optical sensor and notched wheel from an oldschool PC mouse?  It's light, compact, zero drag, cheap and plentiful, etc you'd just have to deal with receiving the output signal.

Good idea.  However, dealing with the output signal would take a real electronic person, not an impostor like me who doesn't know his Thevenins from his Nortons.  

[Brett Buck]

How about the optical sensor and notched wheel from an oldschool PC mouse?  It's light, compact, zero drag, cheap and plentiful, etc you'd just have to deal with receiving the output signal.

   I think the servo wants a fixed-frequency PWM pulse train, where the duty cycle corresponds to the position. I think it's monopolar, so 0= hard over one way, 100% equals hard over the other way, 50%=center. Run that through a low-pass filter and you get a voltage proportional to the duty cycle, which drives the servo loop. I think you are only get a string of pulses out of the encoder, you would need something to count them, and I don't think it is what a servo needs.

   Of course, someplace in a transmitter it probably generates the duty cycle from a pot position, so some fraction of a transmitter would probably serve the purpose.
 
    Brett

[Tim Wescott]

  I think the servo wants a fixed-frequency PWM pulse train, where the duty cycle corresponds to the position. I think it's monopolar, so 0= hard over one way, 100% equals hard over the other way, 50%=center.

Actually the servo expects a pulse train, with pulses ranging from 1ms to 2ms, at a nominal frame rate of 50Hz.  For most transmitters 1ms is stick back and 2ms is stick forward, although Fred Cronenwett can cite a counterexample.  The frame rate can vary somewhat.

You used to be able to test 72MHz transmitters by putting your TV on an unoccupied channel around 4, and looking at the bright bands moving around.

(Servo workings put at the end of the post)

I think you are only get a string of pulses out of the encoder, you would need something to count them, and I don't think it is what a servo needs.

Yes, you would need to count pulses, and to know where center (or some other absolute position) is you'd need an index pulse which mice don't have.  Moreover, the angular resolution of the wheel depicted is much worse than the accuracy attainable with a pot or magnetic device.

  Of course, someplace in a transmitter it probably generates the duty cycle from a pot position, so some fraction of a transmitter would probably serve the purpose.

Modern transmitters run the pots straight into an analog input on a microcontroller, which does all the rest in code.  It would be hard to whack the thing in two.

The appropriate circuit to use with a potentiometer input would be a 555 timer chip of one flavor or another connected as an astable multivibrator.  I could have designed the circuit in the time it's taken to respond with this post.

   I think the servo wants a fixed-frequency PWM pulse train, where the duty cycle corresponds to the position. I think it's monopolar, so 0= hard over one way, 100% equals hard over the other way, 50%=center. Run that through a low-pass filter and you get a voltage proportional to the duty cycle, which drives the servo loop.

In an analog servo the pulse is stretched, and compared to a pulse that is generated by the servo amplifier, and whose duration depends on the position of the pot wiper (and, hence, the servo position).  If the pulses match the motor is powered down; if the pulses differ then the motor is powered in the correct direction to make the difference less.  One consequence of this is that the control loop sampling rate is dependent on the transmitter's frame rate.  Another consequence is that if you use too high of a frame rate the servo will misbehave badly.  Alas, the TUT's frame rate is slightly over 120Hz.

In a digital servo the pulse width is measured and fed as a command to the controller, which operates at a fixed sampling rate which is considerably higher than the pulse frame rate.  This is but one of the reasons that digital servos work better.

[Tim Wescott]

Here's the circuit to work with a gutted servo.  I'm not sure what the timing vs. pot rotation would be -- I'm pretty sure you wouldn't get a full 1ms to 2ms pulse width variation from a 90-degree servo arm rotation.

[Tim Wescott]

I just sacrificed a vintage ACE R/C micro-servo to science (well, the rest of it -- the motor was already gone).  With the gear train removed, the action is quite free except for a light amount of binding at one end of the travel.  I suspect that a cheap ball-bearing servo would have imperceptible friction if you didn't munge it up taking the drive gears out.

I don't know if I can wedge an anti-servo circuit in there using through-hole parts, but the board layout is 16.5 x 9mm, which should fit into a wide range of servo bodies.

[Dean Pappas]

Hi Howard, Hi Tim,
Sounds like a neat little experiment you propose, there: after all why put the breakpoint in the variable elevator:flap ratio in a place that doesn't accomplish what you want?
may I suggest the pirated servo mechanicals and pot as Tim described with the pilfered application circuit suggested on page 28 of the LTC6992 datasheet, as found in the Linear Tech website <www.linear.com>

I am terribly interested in the results!

take care,
  Dean Pappas.

[frank williams]

Here is one that I have used.  You can replace the pot to suit needs.  Its a kit, cheap, small, and it works.
http://www.hansenhobbies.com/products/rcelkits/sckit/
Frank

[Tim Wescott]

Here is one that I have used.  You can replace the pot to suit needs.  Its a kit, cheap, small, and it works.
http://www.hansenhobbies.com/products/rcelkits/sckit/
Frank

You may have to jigger component values, but that should work.  It's a bit big, so it may have to go on the outside of the servo.

This circuit works, and fits in that ACE R/C "micro" servo when it's built up with through-hole parts on a bit of prototyping board -- with the possible exception of the 470nF cap you could build it out of parts from Radio Shack.  (The diode is a bit weird because I'm using a common surface-mount part in my schematic-capture program, pursuant to laying the thing out with surface-mount parts.  I used a common small-signal through-hole diode for the prototype).

[Tim Wescott]

Antiservo in action.  I used an old Royal servo for the mechanicals.