The best design for a remote follow focus unit.

My company carries out development of things like this all the time. The biggest problem is the mechanical coupling of a motor drive to the focus. You can lenses with motorised focus built in, but they don't provide and feedback as to the position.

I would recommend using a PIC12F675 for the processors at both ends. They are very cheap and have a built in 4MHz oscillator, so no need for crystals just +5v and 0v. The position signal can be provided by a potentiometer (10k) connected between 5v and ground. The output from this can be fed in to one of the pics ADCs (10-bit so plenty of resolution). Connect another pin from the pic to a small AM transmitted. These are low-cost (£4) and usually have 4 pins; +5v,

0v, data, antenna. These can be difficult to get going as there is little on the net about their use. People rave on about how you must use Manchester encoding for any data you want to send. This isn't strictly true as you can get away with just normal 9600 baud serial data but you need to know a few tricks to keep the DC component zero. So get you pic to take an adc reading and transmit it with a 10ms delay.

On the other end you will need the pic to read the current position of the lens (use a 10k pot connected to one of the adc lines). Get the pic to compare the old position with the new one and decide which direction and how fast to turn the motor. On the motor side I wouldn't suggest using a stepper motor as these are a bit more involved on the control side and cannot develop as much torque as a similar sized normal motor. For development you might want to look at using a hacked servo (plenty of information on the net about this procedure). To control the direction and speed of a hacked servo you simply need to provide a 1-2ms pulse (every 20ms). This works out nicely as you can receive data (via an AM receiver), perform a comparison of the new and old position then produce your 1-2ms pulse, then wait for the next set of data.

Best regards,

Geoffrey

Thanks for the input Geoffrey.

I'm wondering if you know of any example circuits on the net that show how to interface the PIC with an AM transciever?

I have been doing a lot of research since I made that last post, I'm afraid I disagree about the potentiometer and the hacked servo. I don't see how this would be better then a stepper motor on the motor side and a incremental encoder (similar to a computer mouse.) The first PIC can count movement on the encoder and send a packet of data to the second PIC corresponding to the current location of the knob, probably a 16bit integer. The second PIC simply moves the stepper the appropriate number of steps to match it.

I made a more diagrams, take a look:

Ian Bloom

Mike Keesling sent this email to me:

Before you start out with the hardware you want use, think about the task at hand.

What focal length lenses do you intend to use? What f- stop ? Just some old standard mount stuff, or some nice B-speeds? Want to run a zoom too? what's it's focal length? Look at your D.O.F. tables, and consider the granularity of your positional information.

Do some research on how much lag is perceptable between command and response? Think about the sustained data rate you require and the range you really need. Think about the band you choose, and who else is going to be on it every time you roll camera.

Then after all that, then think about whether you can afford not to find an old Preston system or a brand new Bartech, as opposed to running through tail ends and reshooting every time you catch a radio spike, or worse, blowing the shot, and not knowing it until you get to post.

Trust me, the amount of engineering that goes into something like this is astonishing.

If this is for your own personal use, you may get by with a modified Futaba radio. Pull out the joystick, put in a potentiometer, drive your lenses with some nice futaba servos. Try Vantec for modified aircraft radios for ground frequencies.

If you are trying to build a product for sale, you have a very long and painful road ahead.

Mike

In response to Mike's message:

This is definitely for my own personal use. Though I may build them custom for other filmmakers that I know, this would not be a commercial venture.

As far as the lenses go: Right now I am trying to use this device on an Arri SR 2 camera. It would mostly be used on prime lenses. All of the lenses that I use, with the exception of the zoom lense have a gear ring for the focus. This gear ring attaches a device called a "follow focus" that slides onto two bars underneath the lense. This device basically transfers the motion of the focus ring to a knob on the side of the camera. (Making it easy for a camera assistant to adjust the focus during a shot.) See this diagram below:

(On some browsers you'll need to copy this link into your address bar instead of just clicking the link. Some quirk with geocities.)

Since most professional film cameras and some pro-video cameras come standard with a follow focus, the exact lense and camera are not that important. Pretty much all follow focuses have a square hole in the center of the knob which is perfect for adapting a motor to.

As far as radio spikes go, I'm not worried about that, since the entire device should be digital, and the RF link should include steps for packetization and error checking. A spike in the radio signal should not be capable of doing anything but interrupting communication momentarily.

Resolution and lag time are the real issue that I'm wondering about. The resolution of the device is related to the lag time, in the sense that, if we use stepper motors, a more precise increment is most likely going to mean a slower motor. We want the device to be able to match as much as possible the mobility that a camera assistant would have directly in his hand. Using servos could be faster but it would require some sort of feedback mechanism (another rotary encoder).

Here is the latest draft of the design: (On some browsers you'll need to copy this link into your address bar instead of just clicking the link. Some quirk with geocities.)

I'd be interested if someone knows how to judge, how fast a given stepper motor can turn and how much torque it has.

Thanks, Ian Bloom

Mike Keesling sent this email to me:

Before you start out with the hardware you want use, think about the task at hand.

What focal length lenses do you intend to use? What f- stop ? Just some old standard mount stuff, or some nice B-speeds? Want to run a zoom too? what's it's focal length? Look at your D.O.F. tables, and consider the granularity of your positional information.

Do some research on how much lag is perceptable between command and response? Think about the sustained data rate you require and the range you really need. Think about the band you choose, and who else is going to be on it every time you roll camera.

Then after all that, then think about whether you can afford not to find an old Preston system or a brand new Bartech, as opposed to running through tail ends and reshooting every time you catch a radio spike, or worse, blowing the shot, and not knowing it until you get to post.

Trust me, the amount of engineering that goes into something like this is astonishing.

If this is for your own personal use, you may get by with a modified Futaba radio. Pull out the joystick, put in a potentiometer, drive your lenses with some nice futaba servos. Try Vantec for modified aircraft radios for ground frequencies.

If you are trying to build a product for sale, you have a very long and painful road ahead.

Mike

Here's the thing,

As a struggling cinematographer, I can't afford to buy a proffessional follow focus unit even used, nor can I afford expensive analog boards. So currently the price of this machine as I've designed it is around $200, (see the current idea

'll need to copy this into your browser bar, a link won't do it.

Now here is a thought experiment, take you mouse and smoothly drag it from the left of the screen to the right (assuming you don't have a laser mouse) thats an incremental encoder (and a very cheap one) giving you somewhere in the neighborhood of 288 increments of resolution per complete rotation of the mouse ball. Gear it up and you could get much more. Now take a look at the focus ring on an average prime lense and imagine how many degrees of resolution are neccessary for the focus puller to get good focus. Not that many if the subject is close to camera, a lot more if the subject is far away, but the farther away they get the more Depth of Field you have. So imagine that you need maybe 4096 increments around the entire circumference of the focus ring to get good focus, there is no reason why you couldn't get that or even more with incremental encoders.

As far as DC servos go. I believe you, but I'd like to here why it's better than a stepper?

Thanks for your help, Ian Bloom