Entropius
Veteran Member
I don't think this is too bold a claim. Whether Olympus is using this technique is another story, but the technique is viable in theory. Perhaps Olympus actually is using it, too -- based on their new IS. (Keep reading!)
As has been said, the problem with CDAF on 4/3 lenses is that their motors are not designed for it. PDAF systems can estimate how far the lens needs to slew in order to be in focus. So they make a measurement, instruct the lens motors to move the glass to that location, and then repeat until focus is achieved. The motors don't have to do things like reverse direction or start and stop quickly.
The trouble with CDAF is that all you get are "better" and "worse". You don't know whether the lens is focused too far in front or too far behind, or how much, so you wind up rapidly jiggling the lens around to see if it gets better or worse. The motors in the phase-detect lenses are not designed for this.
But wait! What if we had another way to change the focus of the lens other than the lens AF motor, a way that could make very rapid changes? Then we could use that for the rapid jiggling that causes problems for lens AF motors, figure out which way to go, and then engage the lens AF.
As anyone who has used old lenses (or the ZD 70-300) knows, moving the lens further or closer to the sensor will change the focus. Most modern lenses don't move all the elements together just because they weigh too much, but -- in principle -- you can do it this way.* So, what if there were a way to move the sensor instead? It's lighter than the lens elements and easier to move, so we could jiggle the sensor around to test the waters, and then use the lens AF motor to make "coarse" focusing adjustments.
Well, we've had movable sensors since the E-510 -- this is how Olympus IBIS works. The piezoelectric elements that move the lens can obviously move it very quickly and precisely, and start and stop it on a dime. (To confirm this, just take the lens off, turn IS on, and take a long exposure. Move the camera around and watch the sensor scoot around.) But this implementation of IBIS only moves the sensor side-to-side, not in and out, so it can't be used to change focus. In principle, though, there would be nothing too hard about putting an additional set of piezoelectrics (or whatever they use) on to move the sensor along the Z-axis (in and out).
BUT, it's been hinted that the new OM-D has "five-axis IBIS". This means -- up and down, left and right, roll, pitch, and yaw. If the sensor can pitch and yaw, then there are already Z-axis actuators on it -- and we have all the mechanical bits needed to move it in and out, also.
The comment is in here: http://www.43rumors.com/ft5-first-olympus-om-d-leaked-image/#comments .
As a plus, this trick could be used for focus finetune on manual lenses, too! You'll have to manually get it close, but then the shifting sensor can be used to finetune the AF.
As has been said, the problem with CDAF on 4/3 lenses is that their motors are not designed for it. PDAF systems can estimate how far the lens needs to slew in order to be in focus. So they make a measurement, instruct the lens motors to move the glass to that location, and then repeat until focus is achieved. The motors don't have to do things like reverse direction or start and stop quickly.
The trouble with CDAF is that all you get are "better" and "worse". You don't know whether the lens is focused too far in front or too far behind, or how much, so you wind up rapidly jiggling the lens around to see if it gets better or worse. The motors in the phase-detect lenses are not designed for this.
But wait! What if we had another way to change the focus of the lens other than the lens AF motor, a way that could make very rapid changes? Then we could use that for the rapid jiggling that causes problems for lens AF motors, figure out which way to go, and then engage the lens AF.
As anyone who has used old lenses (or the ZD 70-300) knows, moving the lens further or closer to the sensor will change the focus. Most modern lenses don't move all the elements together just because they weigh too much, but -- in principle -- you can do it this way.* So, what if there were a way to move the sensor instead? It's lighter than the lens elements and easier to move, so we could jiggle the sensor around to test the waters, and then use the lens AF motor to make "coarse" focusing adjustments.
Well, we've had movable sensors since the E-510 -- this is how Olympus IBIS works. The piezoelectric elements that move the lens can obviously move it very quickly and precisely, and start and stop it on a dime. (To confirm this, just take the lens off, turn IS on, and take a long exposure. Move the camera around and watch the sensor scoot around.) But this implementation of IBIS only moves the sensor side-to-side, not in and out, so it can't be used to change focus. In principle, though, there would be nothing too hard about putting an additional set of piezoelectrics (or whatever they use) on to move the sensor along the Z-axis (in and out).
BUT, it's been hinted that the new OM-D has "five-axis IBIS". This means -- up and down, left and right, roll, pitch, and yaw. If the sensor can pitch and yaw, then there are already Z-axis actuators on it -- and we have all the mechanical bits needed to move it in and out, also.
The comment is in here: http://www.43rumors.com/ft5-first-olympus-om-d-leaked-image/#comments .
As a plus, this trick could be used for focus finetune on manual lenses, too! You'll have to manually get it close, but then the shifting sensor can be used to finetune the AF.
