Good servomotors?

I need a motor that can control speed very precisely under a moderate but fairly constant torque, to within a couple percent at least. To give you an idea what torque is needed, it's going to be used to pull photographic film through a 35mm camera as the rest of the camera rotates (for panoramas). The motor needs to be accurately controllable between 5RPM and 70RPM.

Steppers are easy to implement but I'm worried that common 1.8*/200 per revolution steppers like this (Pololu - Stepper Motor: Bipolar, 200 Steps/Rev, 35×26mm, 7.4V, 0.28 A/Phase) will have too 'course' of stepping action and cause vibration or 'notchy' film motion. I could use a stepper and gear it down several times using rubber belts and pulleys, but I'm having a hard time finding rubber belts and pulleys online. Does anyone have any input on how smooth these things turn against a load at say 50rpm? I could add a flywheel I guess.

I could also use a servomotor, but ready-made high-torque servos are hard to find. I could use a regular gear motor like this (Pololu - 100:1 Micro Metal Gearmotor LP 6V) and somehow put an encoder on it to make it a servo, but I have tried that before and it was hard to control except for in a narrow speed range.

Is this for a standard frame-by-frame sort of application, or more like a continuous, Alpa Roto-style slit aperture panoramic camera?

It's a continuous-motion setup. I have already built a prototype using a 60rpm gear motor and achieved encouraging result. My prototype operated in completely 'open loop' or 'dead reckoning' mode which quickly lost sync. If I could find a way to add an encoder to my current gearmotor I could reuse it, but that will be hard, I think. I would rather buy a motor with an encoder on it. I want controllability from 5 to 70rpm, and the ability to get feedback from the motor, at least, so that I can servo-lock it to the camera motion. A stepper would be way easier, but I don't think it would be smooth enough.

I can't imagine why you wouldn't directly couple the film sprockets to the pan rotation.

Do you mean with hardware gearing? There are a couple reasons. One, the sync between the two speeds needs to be matched, and I don't think you could build a mechanical linkage that can be 'trimmed' enough. You would have to get lucky and have the two shaft speeds end up integer multiples of each other. Second, it's desirable to have the thing work with multiple lenses--much easier with software. Third, the way I have it set up now, the speed of the winding motor actually has to change as the camera pans because the effective diameter of the takeup spool increases as film winds on. This is not a problem in software as my prototype has proven.

Third, the way I have it set up now, the speed of the winding motor actually has to change as the camera pans because the effective diameter of the takeup spool

Why not do it the way tape decks do it- the take-up reel is on a slipping clutch, and the capstan provides the constant velocity? (which is also how the Roto did it, IIRC)

Because it would be difficult; it would amount to making my own camera from scratch. Right now I'm using a 35mm camera body. The lens mount and shutter come in handy. Although your idea is not a bad one, but I would still need a precise servomotor to drive the capstan, and would then have to make a camera and a film gate and a capstan. But I have shown that the approach I'm using can also work, and I only need a servomotor.

Can anyone think of anything that would have servomotors I could rob? Like VCR or tape decks, or other mechanisms? Or does anyone have good ideas how to add an encoder to a basic gearmotor? I tried gluing a piece of PCB to the little stub that sticks out the back of the motor, which spins much faster than the output shaft, and sensing it with IR and hall sensors, but I could never get reliable detection for some reason; I think it was RF from the motor brushes.

Because it would be difficult; it would amount to making my own camera from scratch

I had to go an look to remind myself, but that's how still film cameras are constructed too - the framing is controlled by the sprocket, not the take-up reel.

So that when you advance to the next frame, it goes exactly 8 perforations, and not too far (wasting film) or too short (overlapping images).

You seem to be confused. I will not be shooting 'frames' of anything. The film will be continuously advancing at the same speed as the image is moving across the film gate, shooting a continuous exposure...there are no 'frames'. This is how the photofinish cameras at horse race tracks work. Or at least how they worked back when they shot film. The film moves through the film gate continuously at the same speed as the horses ride by.

you are using any kind of conventional camera body, it MUST have at least one proper sprocket that "meters" the film.

Sorry; I've already proven you wrong. My prototype works fine using the film rewind spool; I just need to improve on the open-loop control by adding an encoder to the gearmotor. It's not very versatile just running the gear motor in open-loop PWM mode with no way to know how fast it's turning. I can empirically tune it in, but it's a fiddly process. There is no need to use sprockets for this; I think a capstan would be perfect but then I would have to build my own camera. In summary, using the film rewind spool on the 35mm body works fine, I just need a high-torque servomotor, or some way to add an encoder to the gearmotor I have now. I tried building my own encoder by adding stuff to the motor stub shaft and failed. Robotics hardware seems to be dominated by steppers and I can't find any servomotors...just steppers and plain old gear motors, as if nobody ever had the need for a high-torque, small servomotor. Actually just now I found these: Pololu - 70:1 Metal Gearmotor 37Dx54L mm 12V with 64 CPR Encoder (No End Cap) which look exactly what I need, but man...$40. Maybe I should give up and try to use a geared-down stepper, but I'm concerned that the 'stepping' action will not be smooth enough.

Not clear what is the benefit of using old-fashioned chemical film? That is a dying (nearly dead?) 19th century technology.

As a photographic artist, that specializes in making silver prints, I am amused at your attitude that chemical film is 'old-fashioned'--and the implication that anything old is therefore worthless. Silver halides are beautiful in ways that matrices of numbers never can be.

I'm having real difficulty following this now - the camera has a mechanism designed specifically to draw the film with the correct tension and at a steady rate, but you choose instead to use a mechanism that is simply designed to take up slack, and allows slippage.

Am I reading this correctly?

I cringe at the thought of using "rim drive" for something like your application.

Who said anything about 'rim drive'. 'Rim drive' to me implies a capstan-like friction drive. Using the film rewind spool is nothing like that. There's no way it can slip; the film is taped directly to the spool.

the camera has a mechanism designed specifically to draw the film with the correct tension and at a steady rate

The camera has no such mechanism. The sprocket drives in 35mm cameras, and in movie cameras for that matter, are not designed to smoothly pull the film through the gate at a steady rate. They are designed to advance the film to the next frame. I'm not shooting frames, so the current mechanism is worthless to me. Modifying the camera to allow use of the sprocket with continuous film motion would require me to completely gut the internal workings of the camera and make a new drive interface with the film drive sprocket. I might as well make my own camera. It's easier to simply couple to the film rewind spool. This way, I don't even have to modify the camera! The only disadvantage is that the diameter of the spool increases as film winds on, but that is easy to fix in software. The rewind spool speed simply decreases linearly with platform rotation; finding the correct fudge factor is easy. In fact my back-of-the-envelope calculations, measuring the film thickness with calipers, were pretty much right on.

See pictures below.

Schematic of the rewind spool and film:

Picture of the mask in the film gate:

Picture of the coupling I made for the rewind knob:

Picture of the above coupling meshing with the rewind spool:

This works fine; the only problem is the open-loop control with the plain gearmotor, which actually works but only in a fiddly way. I've already taken pictures with this setup folks. I already found the motor I need at Pololu. I'll just buy one of these:

and once I get it reconfigured with that motor motor I'll be rockin'. It's not really a servo motor because I'll have to implement my own PID/servo loop in software, but it's basically what I was trying to do last time with making my own encoder...but hopefully this one will work.