>> >The 324's performance could be improved on (thus the 3260). >> >> I assume there are lots of other opamps that could be used, also. >> But it should be low-noise, low power types, I suppose. >You need sufficient GBW at 50 kHz to get rid of a certain residual signal >from the sensor. You will see. I suppose the sensor will have some noise at around 25 kHz, as that is the resonance frequency of the sensor, but what is the 50 kHz you refer to? Harmonics? >> I'm not sure of how much resolution I will need on the ADC, and I'm not >Sensor output ~= 0.3..0.8 mV / (degree * second) Oh. In the datasheet I've seen it says 1.11 mV/deg/sec. That means if the rotational speed changes from zero to 1 degree per second the sensor will change its output with 1.11 mV, right? If I use the MAX 187, I will get 1 mV resolution over the range 0 to 4.096 V. That should suffice.... >> sure how to do the mixing with the incoming signal from the receiver. >> Does it sound ok to kind of overlay the gyro signal on the receiver signal? >> With adjustable gain, of course. >Oh ! I had thought that you'd done this before... (interfacing to PWM >RC servo circuits). You need to do some reading here. Mixing with the >signal is DEFINITELY not the way to go. I have worked with R/C servos and PICs before. I once made a robot this way. What I don't know is how to let both the gyro and the pilot have control over the tail. I'm not sure if you've misunderstood me a little, so I'll explain a bit better what I had in mind. The receiver will send a signal to the gyro. Lets say a 1.5 mS signal. If we assume the heli is at zero rotation at this point and a gust of wind tries to turn the heli, then the gyro will add an error correction signal of for instance 0.2 mS to the receiver signal before sending it to the servo. The amount of error correction will depend on both gyro gain and how fast the heli is turned by the wind. So, when I say mixing I mean adding and subtracting to the signal in the PIC, not analog mixing of the signals. Does it still sound like I'm totally lost here? >Also, there are some implications to make a closed loop servo (hint: the >stability of a loop with gain and respectable delays and dead-bands built >in). Yes, but I suppose I will have to experiment a little with this... >> I saw an example amplifier circuit on the Murata website, but they used both >> a highpass and a lowpass filter on the signal, thus eliminating the DC >> component. >Look again, I am quite sure that the DC component is not eliminated unless >the circuit is meant to do that (i.e. rate of rate gyro == rpm >stabilizer). I have no idea what it's meant to do, but since they use an RC highpass filter immediately on the sensor output, the DC component is eliminated. The filter has a cutoff frequency of approx. 0.3 Hz. >> >Yes. You can even make the chopper turn slowly if the radio signal is >> >lost... which is a very good idea in view of the price of the bird imho. >> >> But how would you determine for how long it should turn? >Until the ground stops it or radio liaison is re-established, whichever >happens first... ;) This assumes that the heli knows how to auto-rotate >w/o help (ok, skip the flare, use a shovel instead, but it's still >better than a *sieve* and *magnifiying glass*). Have you seen a heli that's auto-rotating on its own? And in an auto rotation most model helis will have the tail rotor disengaged anyway, so the gyro has no control over things. Some helis can be steered with the tail in an auto-rotaion, but if so, you would definately not want to spend a lot of precious energy on pirouetting on your way down. >> What about an initial zero read routine in the gyro software? It would >> of course have to assume that the heli is not rotating when it is first >> powered, but that is usually the case, so.... >You can do that. Then that's what I will do. >> Why would you want to integrate the signal? Isn't this kind of a >> closed-loop thing? You apply opposite tail rotor until the rotation >> stops? (Or until a preset deadband is reached) >I was thinking of a safety feature here. If the total degrees over a >certain time is too large, assume that the sensor died and turn off servo >loop. Aha. That's not a bad idea, although it will have to have quite a long time constant, since pirouetting is a very common thing to play with... -Oyvind P.S: I can take this discussion private, if people think it's a bandwidth waste.