View Full Version : EF ring USM lenses - the FTM differential

Doug Kerr
December 23rd, 2010, 10:02 AM
All Canon EF-series lenses with ring-type USM AF drive have full-time manual focusing (FTM). On a few, this is done electrically, but on most it is done mechanically. The key is a differential mechanism . I thought it might be worthwhile to explain the principal of these and its application to these lenses.

Functionally, the purpose of the differential is to take two rotary inputs - the output of the USM motor itself, and the movement of the manual focusing ring, and sum them algebraically, using the sum to control the position of the focusing cam barrel (which actually moves some or all lens elements to change the focus).

We'll sneak up on this gradually.

Figure 1 illustrates the basic principle involved, in an abstract context:

Figure 1. Differential mechanism principle

Here, the "mechanism" is just a simple lever, and we think in terms of very small movements. The vertical positions of the left and right ends of the bar, x and z, can be thought of as two inputs to the differential (although any two of its "nodes" can be thought of as inputs). We consider the vertical position of node y to be the output of the differential. Its movement is half the sum of the movements of the input nodes, x and z.

Note for future references that if we are not making any movement of one of the inputs, it must be held stationary. If we just let node x loose when we do not mean to move it, then if there is any "load" on the output (we are using it to move something, for example) then a movement of node z (as an input) will just cause node x (supposedly an "input") to move in the opposite direction, presumably not what we want in this outlook. (This is called "reaction" movement.)

In figure 2 we see an actual mechanism that will exemplify this principle, in this example for linear movement:

Figure 2. Wheel carrier differential mechanism

We see three movable (sliding) plates, two for our inputs and one, sandwiched between them, for our output. The output plate carries a wheel in a small axle, residing in a hole in the plate.

On the top and bottom of the sandwich are two pressure/drag plates, not able to slide. The bottom one is fixed to the "chassis", and the top one is pressed down by a compression spring. The result is twofold:

The two input plates are pressed against the differential wheel.

There is frictional drag on both input plates (which avoids their moving if we should turn loose of them when not wishing to explicitly change their positions).

The diameter of the wheel currently running from its points of contact with the two input plates corresponds to the lever in figure 1.

Thus we see that the movement of the output plate is just the average of the movements of the two input plates. Said another way, it is half the sum of their movements.

Now we will bend this setup into a circle to match the way it is deployed in a ring USM EF lens. In figure 3, to help us properly visualize this, we see the "output ring" (the successor to the "output plate" in figure 2).

Figure 3. Output ring

Here, the output plate becomes a ring, and there are actually three output wheels (mainly to provide for a stable relationship between the layers of the sandwich when they all become rings).

We also see in cross-section the focusing cam barrel itself, passing through the center of the output ring. The output ring is liked to it by a "drive key" (sort of a "drag link") so that if the ring rotates, the barrel will rotate as well, while eliminating the possibility of "binding" that might arise if the barrel was just rigidly attached to the interior of the ring (since each member rotates in its own "bearings").

In figure 4 we see the whole rig (schematically, of course).

Figure 4. FTM differential scheme

We are looking down on the whole system, and we see only parts of the rings, including the part of the output wing where one of the output wheels is at the moment (right on top) the output ring.

On the top we see the stator of the USM motor, which does not rotate. (it is actually more complicated in profile than shown, but that is of no consequence to the issue here, and I didn't care to draw it just now!)

The upper input ring is in fact the rotor of the USM motor, the part that rotates. We note that when the motor is not energized, its rotor is essentially braked by friction between the stator and the rotor, under the influence of the spring (pressure between the two is needed to make the motor work anyway), so we need no further friction arrangement to eliminate "reaction" movement of the rotor when we move the other input ring.

The lower input ring is driven by the manual focus ring (typically by way of a pin on the inside of the ring - there is a gap between two parts of the "outer barrel" that allows the pin to pass through and move continuously around).

Here, we do need a frictional brake on that input ring, which is provided by the fixed pressure/drag plate against which it rides axially.

We see that the spring does three things:

Holds the USM motor rotor against the stator, required for the motor to operate, and as well to provide "braking" of the rotor when the motor is not being driven.

Holds the two input rings tightly against the three output wheels, so the differential drive action will take place as we previously described.

Holds the lower input ring against the pressure plate to provide the frictional drag just mentioned.

As mentioned above, the output ring drives the focusing cam barrel by way of a key that accommodates slight differences in the "orbits" of the two members so that no binding will occur.

And the music comes out here!

Note that if we turn the manual focusing ring until the focusing cam barrel reaches the end of its travel, forced further movement of the ring will cause the output wheels to "skid" against the input ring, accommodating the overtravel. Thus no separate relief clutch is needed.

Neat, wot?

Best regards,


Ben Rubinstein
December 23rd, 2010, 12:38 PM
You lost me with the words 'and sum them algebraically', kudos though for an explanation far above my head. It is very comforting when moving through this life to know that there are so many intelligent, intellectual and academic individuals out there to shore up my shortcomings.

Doug Kerr
December 23rd, 2010, 01:43 PM
You lost me with the words 'and sum them algebraically'
Sorry - that just means, "add them together, taking their directions into account".

Thanks for the compliments.

Best regards,


Andrew Stannard
December 23rd, 2010, 01:50 PM
Hi Doug,

Thanks for this interesting article. The engineer in me has always been curious as to how the full manual focusing worked internally, so thanks for the explanation.

So a couple of questions....

1. Does your 50mm 1.4 use a different mechanism, if not then what of the above explanation is not working correctly?

2. Where do you source this information? Reverse engineering? That seems like an expensive option if you couldn't then re-engineer!


Doug Kerr
December 23rd, 2010, 06:06 PM
Hi, Andrew,

Hi Doug,

Thanks for this interesting article. The engineer in me has always been curious as to how the full manual focusing worked internally, so thanks for the explanation.

So a couple of questions....

1. Does your 50mm 1.4 use a different mechanism, if not then what of the above explanation is not working correctly?

We do not know the mechanism of the EF 50mm f/1.4 USM. We should by this time next week (I have an AF drive assembly coming for reverse engineering - only USD 70.00.)

It may use a differential, following the same concept but with a different implementation. Or it may be a clutch system, quite different.

Once I know that, then the source of the anomalous behavior will be much easier to divine.

I am also getting a flood of test results and observations from dpr guys.

2. Where do you source this information?

It comes from many things, in this case much of it from the Canon EF Lens Work III book (with a lot of reading between the lines), worked together with the results of over 60 years of of dealing with (and reading about) mechanisms of many kinds.

[quote]Reverse engineering? That seems like an expensive option if you couldn't then re-engineer!

Some reverse engineering is non-invasive, some invasive, some destructive.

Best regards,