Win a Pentax Limited Lens: Raffle Now Live!

Hey there Frank.

Yes I agree with your points and I am intending to build a more accurate focus tester with a sensor parallel focussing surface,as well as trhe angled Ff-BF scale. A forum buddy has already sent me a pdf to print out a chart from an I'm going to make an MDF structure in the near future to glue the charts to.

However the tests I conducted in this thread were accurate enough to show the performance of my K20D versus my K-5. And the results align with my real world shootings.

cheers

Add mine to the list that has FF issues with 1.03. I see the issue at 4EV (as measured my my Sekonic light meter @ IOS 100) under tungsten lighting Some lenses are better than others. I'll also get different results shooting a test chart vs batteries vs random target. This is on a new camera received today with a build date of 2-11-2011. Is it any better than the previous 3 were with older FW? Perhaps a little, but I never expected to still see noticeable FF at 4 EV. If the FF existed at 2EV I probably wouldn't give it a second thought.

Now, I haven't checked any lenses for the need for AF adjust in daylight, so that could be part of the equation.

I plan to take the camera with me tomorrow to the rehearsal for a wedding I will be shooting on Sunday. I will see how it behaves in normal use indoor in a fairly dimly lit church. I will have my light meter with me to take ambient light readings as a reference point.

I'm now thinking that I should have held off a bit longer before jumping back to the K-5.

EDIT: After some additional poking around I found that I made a mistake in metering the scene and the FF is occurring at 2EV, not 4. I think this is more consistent with the observations of others. However, neither my K100DS or K-x front focused when shooting the same target with the same lens. Also the focus is dead on at 5EV natural light. Time to take a deep breath and see how the camera performs in normal use.

I've actually been using LV and the principle of Moire interference patterns to help me check focus accuracy, and I've found that it works pretty well: AF microadjustment for the 1Ds mark III, 1D Mk3, 5D Mk2, 7D (ignore how the page was written for Canon cameras, the method and moire interference is brand agnostic)

Firmware 1.03 was partially based on Falk's test data. Does this mean Falk has a perfectly-focussing camera now?

A quick glance at your tests suggests that

• it would have been possible to be more consistent. I wouldn't base a lot of conclusions on a test performed like this (note the varying angles (roll) of the chart). Regarding the shots of the non-focus chart objects: All bets are off what object the camera chose to focus on.
• your K20D has a tendency to back focus in daylight. Tungsten light makes it focus further in front. Its AF adjustment is set in a way that puts the focus point somewhat within the DOF in both scenarios but there is no denying that Tungsten light shifts the focus point.

The K-5 might still have a residual focus problem but judging from the tests -- to the extent they allow any conclusions to be made -- I'd say by choosing the corresponding AF adjustment for the K-5 you should be able to replicate the K20D behaviour.

P.S.: This is the first time I've seen actual shots from your K20D and I am no longer puzzled by your description of it being accurate at both daylight and Tungsten light. I believe it shows the standard K20D behaviour of front-focusing under Tungsten light combined with back-focus under daylight; a combination that could be considered to achieve incorrect, albeit acceptable focus in both situations.

1] I base my conclusions on shooting thousands of pictures, The tests here were just to confirm this, not to analyse absolute accurate measurements of each camera.

2] my K20D doesn't backfocus in daylight. All my lenses have different calibration settings in the K20D and none of them back focus

3] In fact my K20D starts to back focus in very low tungsten light with some lenses - around EV0


Not a chance. The behaviour of my K20D and K-5 is way different. My K20D is able to nail AF in Low light, while my K-5 cannot. Frankly if I could replicate the AF behaviour of my K20D I would be more than happy.

And lastly, as I've explained before the test was NOT a scientific study it was a comparison that is more than accurate enough to show the clear difference between the two cameras, even after applying FW1.03 on the K-5. As several people have indicated my K-5 is a duff copy in regards to the AF. My next step is to arrange a replacement. I may wait until I buy a second K-5 so that I am not without one for any period.

How can you say this? Your test images clearly show back focus. It's pretty bad on your third shot (FA50/1.4 in natural light). The K5 shows a problem, but your K20 isn't the gold standard you're making it out to be.

Once again, and tediously repeating myself... the tests were never supposed to be ultra accurate. DOF is always greater behind the focal point than in front of it. in any case focussing on shot four in tungsten light is no different than shot 3 in natural light for the K20D.

I'm not making my K20D out to be some kind of Gold standard. I'm just saying My K-5 focusses correctly down at around EV2 where the K-5 does not.

Your test shots speak a different language.
Try oversharpening the test shots like crazy (maybe also boosting the contrast); this may help to better show where the DOF is and how it shifts and/or is not centred as it is supposed to be. I hope this might help.

No one asks for the rigour of a scientific study. But even "quick n' dirty" tests require a minimum level of discipline to make them useful. Judging from the variations in your test shots regarding the roll of the chart (which seem to do the K-5 a disfavour by emphasising any FF there might be), I get the impression that more care is necessary to allow reliable judgement about differences of the magnitude you are observing.

I'd be honestly delighted if your K-5 were a duff copy (because it would mean that the normal behaviour is better), but I wouldn't bet on it. I'm honestly looking forward to seeing your evaluation of the second K-5. I think it makes sense to hold off sending the first one back until you have verified that it is indeed a duff copy.

In practical terms an even distribution is pretty much what one should aim for, given the distances these test shots are made with. From my AF adjustment hints:

"The distribution of the DOF before and beyond the focus point approaches 1:1 the closer you get to the subject. The often quoted 1:3 ratio only applies for a subject at a third of the hyperfocal distance."

Maybe this is part of the reason why we are looking at the charts differently?

I have to agree that I got a similar impression as adestefan. Before I saw your test shots I also wondered what special K20D you might have and was concerned that the K-5 doesn't come close. Now I'm not concerned anymore.

Of course, I'd prefer if you would have come back and declared the K-5 to be flawless with the 1.03 update and maybe there is still a slight problem with it. But I no longer wonder how on earth your K20D could outperform the K-5.

I've taken enough ( thousands of ) photos on both cameras now to know that my K20D's AF performs much better than my K-5.

Even on the day I got my K-5 I was mesing around with it and snapped a license plate lit vy my car headlights and was shocked to notice it was way out of focus. A scene my K-20D would never fail to not focus. Since isntalling 1.03 I've confirmed shooting a license plate in simlar conditions has not improved. LV AF nails it every time


I did hold off sending it back very patiently because of the rumours of the FW update. Now that I've installed the 1,03 update and confirmed, sadly, it hasn't fided the fault then soon I shall send my K-5 back for a replacement. Im just waiting to get the funds to buy a 2nd body before sending this one back so I don't have to be without a K-5. becuase depite the fault I love the camera to bits

Yes I could repeat my tests and make them more accurate, but there's no point. The tests were accurate enough to confirm that my K-5 has a front focussing problem compared to my K20D, and that the 1.03 update hasn't fixed it.

Not that I want to defend "bad" testing setups, but did you guys see how far FF the k5 shot at the end was? Nitpicking about the k20 being a tiny bit backfocused in daylight won't change how far off the k5 shot was...

On a separate note, it's not supremely obvious to me that the k5 MUST be better than a k20 in low light AF. It's possible that it could be worse, just because of some other tradeoff made to improve AF speed or accuracy or whatever in bright light. It should work to specifications though, and it certainly seems that this sample is worse than that of many others on this forum.

Thanks sewebster - The tiny bit of back focussing the K20D seems to exhibit in these pics is also there in the low light tungsten shot as well. But the nitpickers chose to ignore that.

There's always more DOF behind than in front of the focal point. I think it's 2/3:1/3 ratio IIRC. And I set my lens calibrations with that in mind.

Anyway I've already stated it wasn't a laboratory test and was more than accurate enough to show up the deficiencies of 'my' K-5 compared to my K20D



Yup. Thanks - Rep to you!

All well and good but there might be alternative explanations rather than "My K-5 doesn't hold a candle to my K20D". However, you don't seem to be interested in exploring views that differ from your preconception and I lost interest trying to help.

You think.

I may repeat that I consider the possibility that your conclusion is correct but your test shots and your omission of following up with something more substantial than just repeats of the same arguments over and over again have instilled enough doubt in me that I no longer think that the firmware update failed to deliver. It was to be expected that there will be a residual problem in certain situations but I don't expect the K-5 to be worse then the K20D, provided both are calibrated with the same intent (e.g,. hitting a compromise in different lighting). Again, I cannot be sure but your tests and arguments do nothing to persuade me to join you in your interpretation.


I noticed but I refuse to give this single shot a lot of weight given the apparent sloppiness in the conduct of the "test". Note that it could also be lens related. It is just impossible to say what the reason for this one outlier is. The one thing we don't need as a reaction to a new firmware are messages of the kind "I did a quick-n'-dirty / rapid / good-enough test and found that ...". I would have been interested to dig down to the truth with Smeggypants but he apparently has no interest in that.

If you read my previous posts, you'll notice that I included exactly that possibility, but questioned the alleged ability of the K20D to be immune against Tungsten lighting. Well, even Smeggypant's copy isn't immune, so one less miracle to wonder about.


That explains a few things. You didn't take the quote from my "AF adjustments hints" seriously, but maybe you'll find another source more convincing:

"In the macro regime the DOF distribution (front:rear) is 1:1 and a landscape captured at hyperfocus has a distribution of 1:∞. So where does the holy 1:2 fit in?"

I suggest that you put the parameters of your test setup into a DOF calculator that shows the DOF distribution. I predict that you'll get nowhere near the DOF distribution you are expecting.

What you are continuing to ignore is the FACT that I have taken many thousands of pics with my K20D with several different lenses and in all kinds of light levels, which I have calibrated and I get sharp results.

I am quite capable of discerning wheteher my K-5 is capable or NOT of performing as well as my K20D.

The test I uploaded was more than accurate enough to show that my K-5 still has front focussing after installing 1.03 and has no where near the accuracy of the K20D under the same circumstances.


Yes we do. And to accuse me of not dfigging down to the truth is offfensive. I've already found the truth regarding my bodies. For the squillionth time I know my cameras and I know my K-5 is not behaving to spec an falls way short of the AF performance of my K20D


This isn't a court of law where I am trying to convince you of the truth.


Well it's focussing abhility is the same at EV2 on both natural and tungsten light. My test pics show that. I've already pointed that out to yo so why are you still ignoring that? If you continue to ignore my test results and take them in the context


Frankly I couldn't care less. You are selectivelty nitpicking on my test results, and ignoring the fact I've taken thousands of real world pictures with these cameras and no when something is wrong. It's like my results go against your dogma so you are selectively makign arguments to avoid having to concede that my K-5 front focusses where My K20D does not.

Lopok at it again ....

K20D mails it. the K-5 does not




FOR THE LAST TIME THE DIFFERENCES IN LOW LIGHT AF BETWEEN THE CAMS FAR OUTWEIGHS ANY SLIGHT INACCURACIES DUE TO THE TESTING PROCEDURE

Hi Smeggy,

I normally set my K20 AF adjust to a somewhat BF condition in daylight so that it will stay within the DOF in lower tungsten light, which causes a move to FF.

As long as the image stays at an acceptable focus as the light changes, I see nothing wrong with that. In fact, setting the camera to be perfectly centered in one lighting condition or another removes a significant portion of the available range of the AF system from bright daylight to dim tungsten.

It is the only possible work-around for the earlier cameras that slide to FF in low tungsten light other than changing the AF adjust every time you go from inside to outside or vice versa. Otherwise, you have to set it for the conditions you shoot in most often.

I suppose a balance of color temp and light level could be calculated somehow and the AF point set in the center under a balanced condition, but most of us have no easy way to do that.

The trick is to not set the daylight AF calibration point so deep that you get BF in more blue daylight conditions. Field shots to confirm are the best way to tell this, and if your K20 is holding acceptable focus from daylight to dim tungsten in the field, the BF cal is working for you.

On another note, if you have Photoshop, try this technique to easily see where the DOF begins and ends on your test shots (I took the liberty of borrowing one of your test shots to illustrate, I hope that you do not mind):



The steps are:

Open the image

Desaturate to B&W

On the Filter menu select "Stylize" and then "emboss"

Play with the angle to get the best level of embossing (this will depend upon where the main sources of light were when you shot it).

Leave the height at 3 pixels and adjust the amount to make the lines look as clear as possible.

Adjust contrast if the image was a bit soft (and the FA 50 @ f1.4 is always a bit soft!).

EDITED TO ADD: This technique is obviously not at all needed on the Sigma test you posted above

Ray