LX100 focus problem is not uncommon

Started Jun 4, 2015 | Discussions
breivogel Senior Member • Posts: 2,403
LX100 focus problem is not uncommon

I have noted focus problems with the LX100 (using single spot AF). Essentially, the camera will significantly front focus even though the focus object is far away (100's of feet). This happens when the focus box is placed over a dark area with bright spots, in particular tree leaves with sky showing through them.

Two examples from a second LX100 that I obtained showing the same problem as the first. Center crop  (approx 1500X2000) of the image, from Raw file converted in ACR. This was shot at f1.8, to accentuate focus issues (and is equivalent to f4, full frame). Hyperfocal distance is calculated to be 18 feet with 12mm fl. Trees are 50+ feet away. First image shows good focus, the second is poor in the distance, better close in.

Based on this, I am returning both cameras and looking for something else (too bad because I really like them otherwise).

First image, the focus box was set over the foliage on the right - no sky was included.

Second image, focused on the center of the left tree, where light is showing through.

 breivogel's gear list:breivogel's gear list
Sony Cyber-shot DSC-RX10 Panasonic Lumix DMC-LX7 Sony RX100 III Olympus TG-5 Sony RX100 IV +19 more
Panasonic LX100
If you believe there are incorrect tags, please send us this post using our feedback form.
axlotl Senior Member • Posts: 2,038
Re: LX100 focus problem is uncommon
2

To send back a camera (2 cameras, even) on the basis of one unexpected focussing result  seems premature to me.

One of the reasons I keep using Panasonic cameras, (including the LX100) is their highly reliable and accurate single shot AF.

Comment: You really cannot say what is common or uncommon on the basis of one situational challenge with 2 cameras.

Suggestion: Get one of the LX100s back. Go out and make 500 photos, minimum, of all kinds of subjects in a variety of conditions.

Use 1 Area AF, adjust AF box size and position as required.

I expect you will find that 99% of photos will be focussed where you intended them to be. That is certainly my experience.

Go back and re-examine the ones which are off and try to discover why. Experiment some more.

Possibilities:

1. If the camera has trouble locking focus on the part of the subject inside the AF box it will sometimes grab something bright and/or contrasty just outside the box, like, maybe the top of the railing in your example.  The fix for this is usually to make the AF box a bit smaller and/or move it a bit.

2. Contrast detect AF systems  generally have a problem focussing on multiple small bright lights, as on a christmas tree for instance. Presumably the flare around each light throws off the AF system. The fix for this is to focus on something different at about the same distance.  Maybe this phenomenon is present to some extent in your example. However I have made thousands of photos in similar situations with no AF problems.

Good luck

Andrew

OP breivogel Senior Member • Posts: 2,403
Re: LX100 focus problem is uncommon

axlotl wrote:

To send back a camera (2 cameras, even) on the basis of one unexpected focussing result seems premature to me.

One of the reasons I keep using Panasonic cameras, (including the LX100) is their highly reliable and accurate single shot AF.

Comment: You really cannot say what is common or uncommon on the basis of one situational challenge with 2 cameras.

Suggestion: Get one of the LX100s back. Go out and make 500 photos, minimum, of all kinds of subjects in a variety of conditions.

Use 1 Area AF, adjust AF box size and position as required.

I expect you will find that 99% of photos will be focussed where you intended them to be. That is certainly my experience.

Go back and re-examine the ones which are off and try to discover why. Experiment some more.

Possibilities:

1. If the camera has trouble locking focus on the part of the subject inside the AF box it will sometimes grab something bright and/or contrasty just outside the box, like, maybe the top of the railing in your example. The fix for this is usually to make the AF box a bit smaller and/or move it a bit.

2. Contrast detect AF systems generally have a problem focussing on multiple small bright lights, as on a christmas tree for instance. Presumably the flare around each light throws off the AF system. The fix for this is to focus on something different at about the same distance. Maybe this phenomenon is present to some extent in your example. However I have made thousands of photos in similar situations with no AF problems.

Good luck

Andrew

This is not one unexpected focusing result.  Probably shot over 100 pix that show the same effect. I can generally get the (good) results I want, provided that I don't focus on the troublesome subject (i.e. light coming through trees.  My prior experience with the LX3,5,7 was that the AF was reliable. Have also used an RX10 on the  same setup and had no problem.

RE your point 1.  The railing was far outside the AF Box (which was set on the next to smallest size). I would hope that the camera would stay with the box in general.

RE your point 2.  I think that this may be the issue in this case. Another reviewer on Imaging Resouce had the LX fail to AF  with a night time street shot. This could be explained as the lights greatly exceed the exposure highlight level, but this should not be the case when focusing on trees against the sky.

Among your 1000's of photos, have you ever  focused on a similar subject, and inspected the photo at 100% ? Maybe you sample is better, but I have used two cameras with widely different serial numbers with the same result. I suspect that many folks have not either noticed or looked closely for the problem. Somehow I spotted it on the first time out to my standard photo testing vista.

 breivogel's gear list:breivogel's gear list
Sony Cyber-shot DSC-RX10 Panasonic Lumix DMC-LX7 Sony RX100 III Olympus TG-5 Sony RX100 IV +19 more
LucaPCP Contributing Member • Posts: 831
Re: LX100 focus problem is not uncommon
1

I am doing some experiments -- I also had some occasional focus problems.

With the "pinpoint focus", which is much slower, I haven't had any problems to date, and the images are very very sharp.

This is a problem that might be fixable with a software update.

I will post more as I discover more about the problem.

-- hide signature --

Luca

 LucaPCP's gear list:LucaPCP's gear list
Panasonic Lumix DMC-LX7 Panasonic LX100 Ricoh GR II Ricoh GR III Olympus E-M1 +1 more
LesK Regular Member • Posts: 370
Re: LX100 focus problem is not uncommon
4

Such an easy problem to manage (my LX100 does the same thing), with a minor focus/recompose, that it is hardly an issue once aware of it (for me). Your loss. Good luck finding the perfect camera.

OP breivogel Senior Member • Posts: 2,403
Pinpoint Focus helps!
1

As per suggestions, I tried pinpoint focus, versus single area (next to smallest AF box). It really makes a big difference.  The question is: is there a downside to doing this?  I am used to using single point, fixed size  AF on my DSLR (D750), so I am used to this (in a low contrast scene, I just focus on an available detail  in the FOV and recompose).

Overall scene. Focus is in the very center, on the middle tree inside, where bits of sky show thru.

100% crop from center, using single area AF. 9 out of 10 images like this - out of focus.

100% crop from center, using pinpoint AF. 13 out of 13 images were in focus.

 breivogel's gear list:breivogel's gear list
Sony Cyber-shot DSC-RX10 Panasonic Lumix DMC-LX7 Sony RX100 III Olympus TG-5 Sony RX100 IV +19 more
LucaPCP Contributing Member • Posts: 831
Re: Pinpoint Focus helps!

Good, thanks for this confirmation!  I am going to use pinpoint focus for landscape photos.

I think the more general behavior is also fixable via software update if Panasonic decides to release one, but I am glad to have this option.

-- hide signature --

Luca

 LucaPCP's gear list:LucaPCP's gear list
Panasonic Lumix DMC-LX7 Panasonic LX100 Ricoh GR II Ricoh GR III Olympus E-M1 +1 more
LucaPCP Contributing Member • Posts: 831
Re: Pinpoint Focus helps!

And thanks so much for reporting the issue.  I also had some photos of far-away subjects (my classical test scenario, the houses in my street) in which the houses were not sharp, but a bit randomly so, and I was trying to figure out whether it was a lens problem or what.  Now I will simply use pinpoint focus for landscape photos, which for me is a rather minor inconvenience, and I can have more confidence that the results will be sharp.

-- hide signature --

Luca

 LucaPCP's gear list:LucaPCP's gear list
Panasonic Lumix DMC-LX7 Panasonic LX100 Ricoh GR II Ricoh GR III Olympus E-M1 +1 more
l_objectif
l_objectif Forum Pro • Posts: 21,401
Re: Pinpoint Focus helps!
4

LucaPCP wrote:

Good, thanks for this confirmation! I am going to use pinpoint focus for landscape photos.

Also, you should not use F/1.7 for landscape!...

 l_objectif's gear list:l_objectif's gear list
Leica D-Lux (Typ 109) Leica D-LUX 3 Leica D-LUX 4 Leica D-LUX 5 Nikon D300 +12 more
particleman78 Regular Member • Posts: 211
Re: Pinpoint Focus helps!

I think the main drawback of pinpoint focus is that it is slower than the other modes.  So it is not ideal for shooting action,

breivogel wrote:

As per suggestions, I tried pinpoint focus, versus single area (next to smallest AF box). It really makes a big difference. The question is: is there a downside to doing this? I am used to using single point, fixed size AF on my DSLR (D750), so I am used to this (in a low contrast scene, I just focus on an available detail in the FOV and recompose).

OP breivogel Senior Member • Posts: 2,403
Re: Pinpoint Focus helps!

Pinpoint is fairly fast in good light, though slower then single area. In dim light, it is unable to achieve focus on lower contrast objects, while the single area is really good . I think one needs to look at the scene and decide what will work best. Pinpoint for landscape for sure. Maybe also when a lot of  small bright light sources are in the image.

Would be nice if panasonic would address this, but their record on firmware updates is not encouraging.

 breivogel's gear list:breivogel's gear list
Sony Cyber-shot DSC-RX10 Panasonic Lumix DMC-LX7 Sony RX100 III Olympus TG-5 Sony RX100 IV +19 more
Ed B
Ed B Forum Pro • Posts: 11,064
Re: Pinpoint Focus helps!
1

l_objectif wrote:

LucaPCP wrote:

Good, thanks for this confirmation! I am going to use pinpoint focus for landscape photos.

Also, you should not use F/1.7 for landscape!...

Bingo!!!!!!!!

I wondered when someone would comment on one of his big problems.

Can't imagine why anyone would use a 1.7 aperture for anything other than extremely low light or portrait photography.

I suppose people who like blurred backgrounds and sometimes blurred foregrounds would use it for other things but for landscapes, f8.0 might be a good starting place.

l_objectif
l_objectif Forum Pro • Posts: 21,401
Re: Pinpoint Focus helps!

Ed B wrote:

l_objectif wrote:

LucaPCP wrote:

Good, thanks for this confirmation! I am going to use pinpoint focus for landscape photos.

Also, you should not use F/1.7 for landscape!...

Bingo!!!!!!!!

I wondered when someone would comment on one of his big problems.

Can't imagine why anyone would use a 1.7 aperture for anything other than extremely low light or portrait photography.

I suppose people who like blurred backgrounds and sometimes blurred foregrounds would use it for other things but for landscapes, f8.0 might be a good starting place.

I have D-Lux, LX 100 sibling, and use 1.7 only for low light and when, in some situations, I wish to have a blurry background!... With this lens, you get the best result for landscape with 5.6; and, a larger area focusing, of course!

 l_objectif's gear list:l_objectif's gear list
Leica D-Lux (Typ 109) Leica D-LUX 3 Leica D-LUX 4 Leica D-LUX 5 Nikon D300 +12 more
OP breivogel Senior Member • Posts: 2,403
why f1.8?
1

The point to using F1.8 was to accentuate the focus accuracy issue. One could conceivably set the camera to F8, and leave it focused to infinity focus all the time and not use AF at all (at a  singe F.L, o fcourse). However, it is good to  know that the camera has an serious issue at F1.8, which one might use under lower light conditions.

It is also worth noting, that from my tests, the AF at F8 is also affected by the focus inaccuracy (to a lesser degree, but noticeable), and that you get sharper photos using the Pinpoint AF here as well.

I would guess that the lens  is sharpest  around F4. and F8 is getting into diffraction territory (equivalent to F18, full frame).

 breivogel's gear list:breivogel's gear list
Sony Cyber-shot DSC-RX10 Panasonic Lumix DMC-LX7 Sony RX100 III Olympus TG-5 Sony RX100 IV +19 more
LucaPCP Contributing Member • Posts: 831
Re: why f1.8?
1

Tonight I was taking street photos at dusk with pinpoint focus, and the results are incredibly sharp.  Using pinpoint focus in light that required 1/20s f/2 ISO 1600 was not a problem.

-- hide signature --

Luca

 LucaPCP's gear list:LucaPCP's gear list
Panasonic Lumix DMC-LX7 Panasonic LX100 Ricoh GR II Ricoh GR III Olympus E-M1 +1 more
Detail Man
Detail Man Forum Pro • Posts: 17,171
Re: Differences related to CDAF System Focus Area Sizes
1

breivogel wrote:

As per suggestions, I tried pinpoint focus, versus single area (next to smallest AF box). It really makes a big difference. The question is: is there a downside to doing this? I am used to using single point, fixed size AF on my DSLR (D750), so I am used to this (in a low contrast scene, I just focus on an available detail in the FOV and recompose).

Overall scene. Focus is in the very center, on the middle tree inside, where bits of sky show thru.

100% crop from center, using single area AF. 9 out of 10 images like this - out of focus.

100% crop from center, using pinpoint AF. 13 out of 13 images were in focus.

.

The information below is of a general technical analytical nature - and does not purport to render judgments as to whether your LX100 should "stay or go". What it shows is not different between cameras, and speaks to the efficay and overall performance of CDAF systems.

Downloaded your posted JPG (the first image displayed above). Cropped a 142x142 pixel square from close to center of the image (height and width equal to 10% of image-height). From that crop, I also made 71x71 pixel square from close to center of the crop (height and width equal to 5% of image-height). Performed the same operations on both sized crops.

Linearized the RGB channels (removing the sRGB gamma-correction), and converted the images to greyscale - simulating what the CDAF servo-system (roughly, approximatey) "sees".

Note (in the histogram statistics displayed below) the average (mean) greyscale value of the smaller (71x71 pixel) crop is approximately twice that of the larger (142x142 pixel) crop:

142x142 Square Pixel Focus Area

71x71 Square Pixel Focus Area

.

Had a look in the spatial frequency domain, as well. The area under the curve (the mean) of the spatial and the spatial frequency data is the same - so normalization scaling was performed in order to reflect that after transforming to the spatial frequency domain (via Fast Fourier Transform, or FFT).

Note (in the 2-D FFTs displayed below) that the distribution of the spatial frequency image-data within the smaller (71x71 pixel) crop appears to be better defined at the lower spatial frequencies (which exist around the center of the diagram) relative to the higher spatial frequencies (which exist outward toward the edges of the diagram) as compared to the image-data within the larger (142x142 pixel) crop (where the spatial frequency image-data appears less concentrated at low frequencies, and spread-out more evenly across higher frequencies), affecting servo-system accuracy.

256x256 FFT of 142x142 Square Pixel Focus Area

.

Normalized 128x128 FFT of 71x71 Square Pixel Focus Area

.

Regardless of the specific implementation of a CDAF servo-system, it will have an easier time estimating focus-distance when the average value of the sampled image-data is higher in value (resulting in a higher signal/noise ratio), and has a spatial frequency content that is more concentrated at lower (as opposed to being spread out towards higher) spatial frequencies.

Downsides of having to use small/spot focus-areas involves requirements that the EVF/LED display have adequate visble resolution, and that the camera (as well as subject-matter being focused on) remain more positionally stable in order to place and maintain the smaller focus-area as intended.

DM

axlotl Senior Member • Posts: 2,038
Re: Confirmation on the tree issue.

Center crop

Full frame

I was initially skeptical about breivogel,s post but here is confirmation. The LX100 has trouble focussing on fine foliage backlit. The AF box was in the center of the frame, at default size. I have also found in the past that it is unreliable with multiple small light sources as in Christmas decorations.  I have found other MILCs with CDAF also have this problem. The solution is to be aware of the issue and in the case, for instance of the picture above, focus on a part of the trees without the backlight light shining through.

The issue is not about backlighting as such in which the camera usually focusses with no trouble, but the multiple small bright points of light.

Overall I find the camera focusses correctly almost all the time. I have not had any problems with most subjects.

Andrew

yellodog Senior Member • Posts: 2,304
Re: Confirmation on the tree issue.

I for one can't reproduce that effect but I always prefocus to check what the camera has focused on before shooting. Of course you'll have to take my word for it that what is in focus is what I intended to focus on

Detail Man
Detail Man Forum Pro • Posts: 17,171
Re: Differences related to LX100 CDAF System Performance Tests
2

Detail Man wrote:

breivogel wrote:

As per suggestions, I tried pinpoint focus, versus single area (next to smallest AF box). It really makes a big difference. The question is: is there a downside to doing this? I am used to using single point, fixed size AF on my DSLR (D750), so I am used to this (in a low contrast scene, I just focus on an available detail in the FOV and recompose).

Overall scene. Focus is in the very center, on the middle tree inside, where bits of sky show thru.

100% crop from center, using single area AF. 9 out of 10 images like this - out of focus.

100% crop from center, using pinpoint AF. 13 out of 13 images were in focus.

.

The information below is of a general technical analytical nature - and does not purport to render judgments as to whether your LX100 should "stay or go". What it shows is not different between cameras, and speaks to the efficay and overall performance of CDAF systems.

Downloaded your posted JPG (the first image displayed above). Cropped a 142x142 pixel square from close to center of the image (height and width equal to 10% of image-height). From that crop, I also made 71x71 pixel square from close to center of the crop (height and width equal to 5% of image-height). Performed the same operations on both sized crops.

Linearized the RGB channels (removing the sRGB gamma-correction), and converted the images to greyscale - simulating what the CDAF servo-system (roughly, approximatey) "sees".

Note (in the histogram statistics displayed below) the average (mean) greyscale value of the smaller (71x71 pixel) crop is approximately twice that of the larger (142x142 pixel) crop:

142x142 Square Pixel Focus Area

71x71 Square Pixel Focus Area

.

Had a look in the spatial frequency domain, as well. The area under the curve (the mean) of the spatial and the spatial frequency data is the same - so normalization scaling was performed in order to reflect that after transforming to the spatial frequency domain (via Fast Fourier Transform, or FFT).

Note (in the 2-D FFTs displayed below) that the distribution of the spatial frequency image-data within the smaller (71x71 pixel) crop appears to be better defined at the lower spatial frequencies (which exist around the center of the diagram) relative to the higher spatial frequencies (which exist outward toward the edges of the diagram) as compared to the image-data within the larger (142x142 pixel) crop (where the spatial frequency image-data appears less concentrated at low frequencies, and spread-out more evenly across higher frequencies), affecting servo-system accuracy.

256x256 FFT of 142x142 Square Pixel Focus Area

.

Normalized 128x128 FFT of 71x71 Square Pixel Focus Area

.

Regardless of the specific implementation of a CDAF servo-system, it will have an easier time estimating focus-distance when the average value of the sampled image-data is higher in value (resulting in a higher signal/noise ratio), and has a spatial frequency content that is more concentrated at lower (as opposed to being spread out towards higher) spatial frequencies.

Downsides of having to use small/spot focus-areas involves requirements that the EVF/LED display have adequate visble resolution, and that the camera (as well as subject-matter being focused on) remain more positionally stable in order to place and maintain the smaller focus-area as intended.

.

In addition to the information discussed above (which relates to differences in the spatial frequency distribution of energy in the spectrum in the above example case between my best approximated crops simulating the LX100 "single area" and "pinpoint" AF focus-areas) ...

... and not here professing to encompass/diagnose the complicated totality of the (proprietary) CDAF system performances of the various (Panasonic and Sony) camera models discussed ...

... I have identified what I think may be a significant difference between breivogel's (recorded RAW) example (the "overall scene" image posted above) and axlotl's (presumably recorded JPG) example (posted here), both of which demonstrated AF accuracy problems - and M H S's (recorded RAW) example (posted here), which did not opine AF accuracy problems - although to my eyes a noticable difference exists between the (unidentified as to which is which "single area" and "pinpoint" AF focus-area) images that M H S posted, and which may (possibly) relate to the "spatial frequency distribution" issue.

From close inspection of the JPG RGB channel tone values within the small lighted areas of breivogel's as well as axlotl's posted example images, it appears very likely that (some/all) of the image-sensor photosites of interest involved were "saturated" (non-linearly "clipped") - such that the associated CDAF system sampled RAW ADU values remained constant, providing no change in image-data to the CDAF system in the course of it's "hunting" for a maximal contrast difference as a function of focus-distance. On the other hand, I found that in M H S's posted example, it appears that the image-sensor photosites involved within the small lighted areas were (likely) not "saturated" (non-linearly "clipped").

.

From reading about Panasonic's "Depth From Defocus" AF technology implemention in the GH4 and FZ1000 models, the most specific information about which was found here: it appears to me that while such a system topology may be helpful in the process of initially "ramping up" towards a desired lens-system focus-distance, the differences in terms of de-focus diameters on the image-sensor plane [at least in the case of a single thin-lens model, as is implemented and informatively displayed in the (free, stable, and non-invasive Windows OS only) cBlur utility, which can be downloaded here] when the CDAF system has neared a desired focus-distance appear very likely to be essentially insignificant - and thus are not likely to represent a meaningful role in the manifestation of this CDAF problem.

One characteristic noticed (which is probably not a significant factor here) is that the differences in image-sensor plane de-focus blur diameters are somewhat (to a relatively small extent) larger in magnitude in the area in front of the-focus distance (as compared to behind it).

.

When a CDAF system attempts to "servo" to a particular focus-distance associated with a maximum measured amount of contrast within the focus-area in the course of changing the lens-system focus-distance, it depends upon those changes resulting in some amount of change in the individual image-sensor photosite data. Any "saturated" (non-linearly "clipped", and thus unchanging in value) RAW ADU levels are not going to provide the changes in data that CDAF systems require in order to "servo" to a particular focus-distance.

Why then would the alternative use of the LX100's "pinpoint" AF focus-area appear to result in improved results in such situations - where such "saturated" (non-linearly "clipped", and thus unchanging in value) RAW ADU levels (also) exist within the CDAF system analyzed image-data ? My best guess (given how little we know about the implementation of these proprietary CDAF systems) is that it (may) be the better defined (more concentrated at lower spatial frequencies) nature of the 2-D spatial frequency spectrums displayed in my post above.

.

As camera preview indicators reflect (simulated, or actual) in-camera JPG-engine results, and do not reliably reflect RAW channel ADU values, and since proprietary commercial RAW processors (for various reasons) cannot necessarily be trusted to report "saturated" (non-linearly "clipped") RAW channel levels, the best tool for making such assessments (including within user-selected areas of inspection only) is the excellent RawDigger application. A (free, and non-invasive) alternative is RAW Therapee 4.x (which includes RAW-level histogram viewing). One should, it seems, verify the RAW-level linearity of test shots where it comes to the image-sensor photosites included with a selected AF focus-area.

.

Given the above information discussed, it is my thought that (any) CDAF system (including the perhaps similar systems implemented in the GH4 and FZ1000, etc.) may well tend to exhibit decreased performance and reliablity in auto-focusing results in such situations (where the image-data utilized contains some/all non-linearly "clipped" photosite RAW levels appearing against a darker background) - as such non-linearity of image-data is a clear problem.

In general (there existing "no free lunch"), all such CDAF systems that seek to achieve faster AF performance are likely to [in some way(s)] exhibit reduced AF accuracy, and pursuits of high sensitivity to low light levels are likely to exhibit more operational vulnerability to the presence of (predominantly Photon Shot, as opposed to the in many cases lower level Read) image-noise existing in darker areas, by reducing the effective "signal/noise ratio" of the image-data that is sampled and analyzed by such CDAF systems.

.

A few technical notes:

Remember that (as far as I know) all CDAF camera systems utilize the minimum F-Number (maximum aperture diameter) of the lens-system when performing auto-focusing operation. Thus, the F-Number that the shot is recorded at is not a relevant factor in this matter of CDAF.

The Hyperfocal Distance associated with the LX100's wide-angle (10.9 mm) minimum Focal Length (when focused at infinity), an F-Number of F=1.7, and a 15 Micron Circle of Confusion (derived from dividing the 30 Micron CoC of a full-frame image-sensor for the case of viewing an 8"x10" sized image at 25 cm with 20/20 human visual acuity by the Micro 4/3 Crop Factor of 2) is around 4.66 Meters. If/when the LX100 is focusing near "infinity" (which the test shots posted came close to attempting to do), the "Near Focus Distance" associated with deriving "Depth of Field" measures exists near to that Hyperfocal Distance.

In a case where the LX100 CDAF system actually is, in fact, "hunting for focus" at distances (for mathematical purposes) approaching "infinity", (then) the "Near Focus Distance" would exist at (around) 15.29 Feet (4.66 Meters), which is approximately double that stated here:

In the LX100, shooting at 24mm (equivalent) - the depth of field is 8 feet to infinity at f1.7 ...

Assessments of "front focusing" where nearby (and thus larger in proportion within the image-frame) subject-matter within an image is relatively lacking in high spatial frequency content can be a "dicey business". Nevertheless, I do not disagree that the lens-system focus-distances in the posted example images demonstrating this problem do indeed appear to be very poorly defined.

.

I hope that perhaps my ("brand/model agnostic") thoughts may (possibly) lend some insights.

DM

OP breivogel Senior Member • Posts: 2,403
Re: Differences related to LX100 CDAF System Performance Tests

You have certainly gone into great and interesting detail. My assumption was that af worked by doing edge detection (e.g. Sobel or high pass),  measuring the peak to peak amplitude of such, and trying to adjust focus to maximize. These are fast routines to perform in real time.

I did verify that the bright spots (actually sky) peeking through leaves did not saturate under my shooting conditions. Of course, what the CDAF algorithm might do in selecting an appropriate exposure during the period of its operation is unknown.  Perhaps it does overexpose and thus clips the highlights, resulting in difficulty in determining accurate focus.  The fact that several other cameras did AF properly under the same conditions leads me to suspect that Panasonic is doing something wrong. One might have expected, given their long experience with mirror less cameras and the associated CDAF, that they would have got it right.

 breivogel's gear list:breivogel's gear list
Sony Cyber-shot DSC-RX10 Panasonic Lumix DMC-LX7 Sony RX100 III Olympus TG-5 Sony RX100 IV +19 more
Keyboard shortcuts:
FForum MMy threads