Autofocus experiment on manual at "W"

[Mike Sanchez]

Because I am, from time to time, experiencing out of focus outcomes I've been running a series of experiments on my autofocus system while focus is set at manual.

Very simple experiments.

1) I set the camera at full telephoto, the switch on the lens barrel to MAN and then, in good lighting point at something with both horizontal and vertical bars and press, for 5 seconds, the PUSH AUTO button.

2) I set the camera at full wide angle, the switch on the lens barrel to MAN and then, in good lighting point at something with h and v bars and press, for 5 seconds, the PUSH AUTO button.

When one releases the PUSH AUTO button the camera reports out to the screen the camera's perception of the distance to the object. An indirect, quantitative assessment of focus.

Comparing the two distance numbers, for the same object, provides a look at repeatability both within a particular zoom and from W to T.

When I am set at full telephoto I get very good repeatability of distance measurements. For example.....at full telephoto I repeatedly, after multiple releases of PUSH AUTO, got a distance of 1.4 m.

HOWEVER, when I am set at full wide angle, my repeatability of measurement drops drastically. For the same object, ostensibly at 1.4 m at fully telephoto, when set at wide angle, after PUSH AUTO, I get anything between 0.8m and 3.5m.

Unless others on this board get the same type of result and somehow the depth of field at wide angle is being factored in to the measurement.....I think I have a problem with my autofocus system.

Can someone run the same experiment please on an object in the 2 m distance range? I would be very interested in the distance measurements following PUSH AUTO (5 seconds) for WIDE angle and full telephoto.

Thanks.

Call me a bit worried at this point.

[Boyd Ostroff]

I generally shoot stage shows in manual focus mode and use the "push auto" button to focus as I go. On a wide shot I really don't think it matters all that much since there's so much depth of field. But I get very inconsistent results zoomed in at full also. I assume this may have to do with low light conditions, and exactly what the autofocus picks up on.

But regardless, I doubt there's anything wrong with your camera. I think autofocus bases its results on something like image contrast.... maybe someone with more tech knowledge can explain this?

[David Korb]

Worry all ya want...im seeing the exact same issue here...even tryed different ae modes...i tryed it at 1.2 meters and ranged from 2.4 to 5 meters

[Jan Roovers]

This is normal.
( I was specialised in Optimal Digital Stochastic Controlengineering)
In wideangle the optimalisation algorithm is not fed with new information within a wider area where the picture is sharp and can and will hold there.
(The algorithm will use decreasing steps till the steps will not show different results.)
The result will be maximum sharp at the end.

But Boyd is also right that it depends also on what the lens sees in the middle. That inroduces another compromis.

For optimal results Sony's manual writes that "Push Auto" is best used in tele also for wideanglepictures, which i always do.

Alltogether the best results are reached in manual in tele. So my procedure is Push Auto and manually adhust further.
When I use wideangle only I just go to tele, Push Auto and go back to wide. (The manual adjusting will not visibly improve sharpness here.)

(This advice is not 100% applicable when using a wideangle adapter because of cromatic abberations that sometimes can be introduced by the convertor. The algorithm can be cheated here. If tele is not necessary avoid max tele in my procedure here and go back to maxtele minus 25% for adjusting sharpness.)

Good luck.

And don't worry about your cam!

[Julian Luttrell]

Hi,

I have found that the "distance" reported by the camera (PDX10) when using manual focus is just so much garbage. I can focus on something 3m away and have the camera tell me it's anything from 1.2 to 10m (IIRC).

My advice is to ignore the numbers.

Regards,

Julian

[Mike Sanchez]

First, let me say thank you for all of the replies.

I would like to say that one significant benefit of purchasing a camera like PDX10 or TRV950 is this site and the body of knowledgeable users this site provides access to. I am grateful to all for sharing.

Boyd.....thanks.....I was guessing that depth of field was large and the numbers did not matter much....but...wanted to check.

David......thanks for taking the time to run the experiment even in different AE modes.....I've stopped worrying for the moment (but will undoubtedly find something else to wonder about as I learn my way through this video maze).

Jan.....I too am a controls engineer (morphed through too many hours of work into an image processing engineer at a large US company (not sure we are allowed to say company names on this site)). But....I've worked only on deterministic control in the presence of noise.....have not built stochastic models into the controls themselves. Sounds like you have a very interesting job. Thanks for the reply and the explanation on the optimization routine. I probably understood completely only the part about using tele, focussing, and zooming back to wide though.... :-). This is what I plan to do from now on. Thanks for the reply.

Julian....thanks.....advice hereby taken.

Thanks to all. Again.....one of the benefits of the TRV is this site and this knowledge reservoir.

[Boyd Ostroff]

<<<-- Originally posted by Mike Sanchez :I too am a controls engineer (morphed through too many hours of work into an image processing engineer at a large US company (not sure we are allowed to say company names on this site) -->>>

Well since your location is listed as Rochester, I'll hazard a guess it might be either Kodak or Xerox (can't imagine why we wouldn't be able to say that here, many of us list our jobs in our "signatures" )

FWIW, I still have a place in the country near Lake Ontario, about an hour East of Rochester. Beautiful area.... in the SUMMER! ;-)

[Mike Sanchez]

Boyd,

The Finger Lakes area is truly beautiful, especially in summer.

Winter can be beautiful.....but it is long.

Fortunately, to make up for the long winters, we have very high NY state taxes..... !.

My signature contains current title. I am one of the image content developers for the Xerox DocuSP high speed digital color controller. A digital front end for Xerox high speed color printers.

One significant reason for migrating to the 3 CCD architecture. More accurate color......