[Brightcolours]

[quote name='photonius' timestamp='1285287166' post='3186']

I'm in the camp that at least with USM lenses, it shouldn't be the limitation of the lens.

Two things to remember: Compact cameras have smaller sensors, larger DOF, they do not need to focus as precisely as APS-C or FF, which need to be resolved much finer.

Second possibility: CCD versus CMOS, and their implementations.

With compact cameras they may actually only read subsets of the whole sensor for analysis (selected regions, or binned images - since DOF is sufficient)

With CMOS sensors, it seems until recently one had to read each whole image into the processor and analyse of course a whole series - processor and memory demanding. Now, there is the movie mode, reading only samples of the chip - 30fps. Alas, it's resolution limited, so probably not 100% suitable for fast lenses with tiny DOF.



Another issue is finding the initial right focus range. With compacts, due to high DOF, contrast detect has something to work with from the start usually. With many dSLR lenses (e.g. tele), a contrast detect may just see mush at the beginning and going through the whole focus range is much longer, due to the finer resolution needed because of the smaller DOF the lenses have to handle- if the AF knew that it could jump in larger steps, it would go faster of course.



In any case, it seems the difference is not that much nowadays. THe EF-S 60mm USM macro seems to focus quite fast on a 550D - not really that much worse than P&S, once it's in a range where the contrast detect has something to work with.



I think the trick will be to have a little lever with a phase detect system pop-in from the side to give initial fast focus.

[/quote]

No idea why you want to be in the camp of that it would matter, if a lens has an USM motor or not, because that just does not make sense at all. USM is nothing special, it is just a motor just like the other motors are. They get talked to by the camera in a similar fashion too. It is like saying a diesel engine and a wankel motor somehow need a different acceleration pedal. <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/blink.gif' class='bbc_emoticon' alt='' />



About CCD or CMOS: Anyone who ever used a compact camera with CCD knows there was LIVE VIEW. So no need to speculate about reading subsets, the whole image got read and converted for live view, and the area used for AF evaluation... got evaluated for contrast.

Does not matter whether your compact digital has CMOS or CCD.



About "finding focus range"... that is "phase detect AF talk".

Contrast detect AF does never have a "focus range" or an idea about focus, all it does is see if contrast increases, until it decreases again (then it knows focus must have been reached).



If you think the 550D focusses as fast as a compact digital... you must have not used newer compact digitals <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/wink.gif' class='bbc_emoticon' alt='' />.

Of course it is true that Canon has improved the speed of the algorithm a bit, compared to the earlier 450D for instance. But it still is... slow. Not comparable to what for instance Panasonic achieves with micro 4/3rds and micro 4/3rds lenses. And the 550D is not only a bit faster with an EF-S 60mm f2.8 macro, but also with a 35mm f2 for instance, without USM.

But that is it... just "a bit" faster. Nothing like contrast detect AF with dedicated lenses, or PD-AF.

[Guest]

[quote name='Brightcolours' timestamp='1285288305' post='3187']

No idea why you want to be in the camp of that it would matter, if a lens has an USM motor or not, because that just does not make sense at all. USM is nothing special, it is just a motor just like the other motors are. They get talked to by the camera in a similar fashion too. It is like saying a diesel engine and a wankel motor somehow need a different acceleration pedal. <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/blink.gif' class='bbc_emoticon' alt='' />

[/quote]



Sorry about the "camp", just wanted to express the idea that I thought todays lenses should be able to deal with the algorithm, however what could be limiting is motor speed. I.e. with contrast detect, you have a loop system, i.e. you measure, move, measure etc. A USM motor will usually be able to go through this faster than a micromotor.

About CCD or CMOS: Anyone who ever used a compact camera with CCD knows there was LIVE VIEW. So no need to speculate about reading subsets, the whole image got read and converted for live view, and the area used for AF evaluation... got evaluated for contrast.

Does not matter whether your compact digital has CMOS or CCD.

Well here I don't really know how it worked in older P&S cameras. And of course all this is live view. However, certainly a few years ago, processors were not that fast. So, if the contrast detect only needs to analyse a subset of the image, it will be much faster than a full say 4000x3000 image, as you say. So, probably in older P&S cameras only the area indicated by the AF marks was actually analysed. (Now you have face detect etc., so obviously the whole frame is looked at). However, If read-out speed was a limiting factor in older P&S cameras, it would have made sense to only read out the AF part. But I don't know if they ever did anything like that.

For example, with CCD cameras I used, it is very easy to go into a binning mode, which gives a smaller image that could be transferred much faster than the full image - but I don't know if that was ever used in P&S ccds.

About "finding focus range"... that is "phase detect AF talk".

Contrast detect AF does never have a "focus range" or an idea about focus, all it does is see if contrast increases, until it decreases again (then it knows focus must have been reached).

No, it's not "focus range". What I meant is this. With PHase dectct, there are situations when the sensor doesn't see anything (worst case: blue sky).

Likewise, with contrast detect, there will be similar situations where it won't see anything, and it has to search. Not being familiar with the algorithms, I sort of guess that perhaps at present they don't do so well yet, and there are more situations when they have to go through a tedious search than with phase detect.

Would actually be nice to a have a comparison to see which of the two locks on faster.

If you think the 550D focusses as fast as a compact digital... you must have not used newer compact digitals <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/wink.gif' class='bbc_emoticon' alt='' />.

Of course it is true that Canon has improved the speed of the algorithm a bit, compared to the earlier 450D for instance. But it still is... slow. Not comparable to what for instance Panasonic achieves with micro 4/3rds and micro 4/3rds lenses. And the 550D is not only a bit faster with an EF-S 60mm f2.8 macro, but also with a 35mm f2 for instance, without USM.

But that is it... just "a bit" faster. Nothing like contrast detect AF with dedicated lenses, or PD-AF.

Well, I didn't say they are the same.

Just tried it with a new Sony Hx5 and the Canon 550D /60mm USm, but just indoor light in a normal range, i.e. 5 meters versus 1m. ~ 1 sec versus 2 secs. A bit slower for the 60mm than I remembered.

But I tried also a 18-55 IS kit, and indeed, to my surprise, there wasn't much difference to the 60mm USM. With both lenses you sort of could see and hear the steps of the iterative process. But for the last steps, no improvement was seen on the screen. I suspect for the P&S it would already have stopped, because of larger DOF.