Thursday, June 26, 2008

K20D Exposure Inaccuracy and Inconsistency

As we all know, the DPR review of the K20D has been out.

Whilst as usual, typical DPR review does not measurbate AF speed, AF accuracy, system time lag and exposure accuracy which IMO are the most important performance aspects for a DSLR, the sample photo gallery of the review does give us some hints, on the exposure accuracy of the Pentax DSLR system, with their current "flagship":-

By doing some simple and basic statistical anlaysis, we can actually get much more useful information than what DPR have already told us:-

1. There are 12 samples out of the total 36 did not require exposure compensation. So, the hit rate is only 1/3, or 33.3%;

2. For the rest of the 24 samples (2/3 of he total samples) that required manual exposure compensation, 16, i.e., 16/36 or 44.4%, had +ve exposure compensation applied (so as to avoid underexposure);

3. Whereas for the other 8, i.e., 8/36 or 22.2%, -ve exposure compensation was made (so as to avoid overexposure);

4. The mean (average) value for the +ve compensation is +0.59EV with a standard deviation of +/- 0.21EV. The mostly applied (mode) compensation value is +0.7EV ;

5. The mean value for the -ve compensation is -0.55EV with a standard deviation of +/-0.17EV. The mostly applied compensation value is -0.7EV;

Well, as I always emphasize, we need to compare other results from the same (gang of) measurbator(s) for those measurbation reports published at the measurbation sites, so as to make things meaningful and for fair comparison as well so that we could interpret the results sensibly. By using the same methodology above, here we can get a summary of the results for the K20D, 450D (Canon), D300 (Nikon), A350 (Sony) and E-420 (Olympus), which are all the latest current DSLRs as most recently reviewed by the DPR. I tabulate them as follows:-

Camera (DPR Sample Gallery Link):Pentax K20DCanon 450DNikon D300Sony A350Olympus E-420
Number of Samples36
Correct Exposure Rate33.3%44.4%75%80.6%80%
Underexposure Rate44.4%2.8%5.6%2.8%2.5%
Underexposure Compensation Amount (Mode)+0.7EV+0.7EV (one sample only)+0.3EV and +0.7EV (two only)+0.3EV (one only)+0.3EV (one only)
Underexposure Compensation Amount (Mean)+0.59EV+0.7EV (one only)+0.5EV+0.3EV (one only)+0.3EV (one only)
Underexposure Compensation (Standard Deviation to Mean)+/-0.21EV0EV (of course)+/-0.2EV0EV (of course)0EV (of course)
Overexposure Rate22.2%52.8%19.4%16.7%17.5%
Overexposure Compensation Amount (Mode)-0.7EV-0.7EV-0.7EV-0.7EV-0.3EV
Overexposure Compensation Amount (Mean)-0.55EV-0.6EV-0.59EV-0.73EV-0.36EV
Overexposure Compensation (Standard Deviation to Mean)+/-0.17EV+/-0.27EV+/-0.18EV+/-0.24EV+/-0.14EV

Remark for Above:
Green colour means the Best; Red colour means the Worst.

So, here comes the further analysis and observations for the above:-

1. The most unreliable DSLR in the exposure accuracy department is the Pentax K20D;

2. The King of Underexposure belongs to the Pentax! The K20D underexposes pictures in an unbelievable large percentage of 44.4%, which is very far more than the underexposure failure rate of all other DSLRs under test, which are all able to keep a very low percentage of underexposure;

3. The amount of underexposure for the Pentax is in a great amount of 0.7EV, also, in addition to the huge number of underexposed shots recorded (as they could be if not compensated);

4. The King of Overexposure belongs to the 450D, which has also an unbelieveable high rate of 52.8% (!?) of overexposure recorded (as they could be if not compensated). Whilst many Pentax fanboys always talked about the overexposure issue of the entry level Canon DSLRs just in order to defend the infamous underexposure problem of the Pentax DSLRs, I actually found that kind of defences is silly and pointless. Does a tends-to-overexpose-most 450D would turn a tends-to-underexpose-most K20D into a correctly exposing DSLR with their unsubstantiated argument? This stupidity just recalls my memory of my primary school classmates who told my teachers about the wrong acts of other classmates when they were caught of doing something wrong! ;-)

5. Surprisingly, the most reliable exposure system and DSLR outthere is the Olympus E-420. Despite the larger number of sample photos posted in the gallery. It nearly didn't underexpose any photos nor the amount of underexposure is significant (0.3EV only and for one sample only, out of 40). It did not overexpose more also, most of the time only a -0.3EV compensation is required (for those 7 overposed shots (as it could have been), there are 6 out of 40 applied -0.3EV only). I am really impressed. Well Done, Olympus!

6. In fact, the exposure system of Sony and Nikon are also very good to good. It is just that the Olympus is so excellent that they cannot be the best in this Shootout.

Ironically, the above shootout comparison breaks the myth of "Pentax DSLRs preserve highlights best" (Really?) as particular Pentaxians have been insisting on and strongly believe like the Gospel! The K20D is obviously not preserving highlight well neither, which just comes in second after the Canon 450D, for the risk of burning highlights. But do bear in mind that the K20D has also a super great tendency of underexposure (which no others will), which simply implies it is just a very inferior exposure system with very poor consistency afterall!

I know that some people would think or just to argue that the sample size is just "too small" (as always, when the test results are not so favourable to *them*. But how many samples are those guys need? There are 4 x 36 + 40 sample photos plus at least five officially provided bodies with various different lenses tested already). Or, those brand-blinded fan boys are consistently accusing that the reviewer(s) of DPR are being biased or even they don't know how to use the Pentax DSLRs (really funny! In fact the DPR crew are the real camera experts! They are just professional gear heads with far more technial knowledge about cameras than any Average Joe - they earn a living by testing cameras afterall!).

Anyway, I know that those people would still not be convinced but would start to bash Phil Askey and his employees. SO, I would point you to an expert Pentax user who is so official (and maybe too official! ;-)) and authoritative for you to believe. Inspect all the EXIF data of the sample photos in the following links:-

See, think and judge yourself. I won't explain and talk furthermore (as I have no more to say)!

In fact, the wonky metering and exposure system of Pentax DSLRs has been one of the biggest annoyances to many Pentaxians (as who have encountered the problem and reported or complained about it for years). The users have been simply wasting too much time in fiddling around with the exposure compensation values, just hoping to get the best exposure but in the end either photo opportunities were missed or just not-so-ideal exposures were resulted.

If Pentax still do not improve their system which is just with such a low hit rate and has great underexposure tendency but yet actually does not preserve highlights well neither, I really doubt how they could compete (or simply just to survive).

I think technically there must be something wrong inside their metering/exposure system in their DSLRs (and/or digital lenses) which only Pentax themselves would know (for why their this metering/exposure system is inferior). But anyway I still hope that they can improve and will improve, in the near future, and in the coming next models (although I have been rather disappointed and desperate for long already).

Read Also:- Underexposure Tendency of K10D and K100D


Anonymous said...

Rice, the most important point for a DSLR is imo the image quality - go out and take some shots like i do. You will see that there is only the NIKON D3, CANON 5D/1DS, LEICA M8 and the PENTAX K20D in terms of excellent IQ with max. details on the market. AND the pro´s use RAW and give their photos the perfect finish - so where is the problem? Always bashing PENTAX from your side is so boring...

RiceHigh said...

The problem is with a K20D, we'll probably get the same 44% of underexposed shots whereas with a D300, say, the user will get only about 5% of underexposed pics for the same scenes and subjects he takes for his shots.

Also, if you really believe that RAW is the magical tools for curing wrong exposures, you'll be totally wrong. Haven't you read this?

Well, it seems that you even forget about the tremendous extra time spent in doing the RAW correction and processing.

Chris Tham said...

Thanks. Your post made me feel glad I bought a Pentax rather than a Canon. I far prefer having a camera that "underexposes" rather than "overexposes."

Having said that though, so far I have shot over 10,000 photos on my camera, and I haven't seen many significant instances of underexposure or overexposure. In fact, I just processed a recent shoot with over 1000 photos and almost all (>990) were taken in 0EV and they seem correctly exposed to me, judging by histogram readings.

Maybe I'm just lucky?

But then, I shoot in Raw mode exclusively.

Read his blog at your own risk said...

I can't even begin to explain how much garbage is in this blog post.
Wouldn't even want to waste my time on this.
To anyone that really takes this garbage serious, I feel very sorry for you.

Anonymous said...

As we all know, the DPR review of the K20D has been out.

Whilst as usual, typical RiceHigh blog does not test AF speed, system time lag and exposure accuracy which he heard from others are the most important performance aspects for a DSLR, this ricehigh blog does give us some hints, of the BLOG accuracy of the Pentax DSLR system.

These are the results:

0% is actually true or even remotely measured by RH.
10% are just a cry for attention
30% is for filling the time in this special hospital
60% is for: i love to read my post over and over again.


Anonymous said...

The variance sucks, but at least the K20 is doing the right thing WRT raw images: If in doubt, underexpose.

Just add a modern metering and AF system, please. Oh, and a sensor with less readout noise.

What highlights? said...

Seems Phil is also a "fanboy"
There's really not a lot of headroom in K20D files that have been over-exposed (though this isn't usually a problem given the K20D's metering, which if anything tends to expose for the highlights).
Maybe you can write to Phil and tell him of your geni-ass discoveries.

Anonymous said...

Hey Ricehigh
You are seriously full of shit.

Anonymous said...

Why do you bother talking about Pentax? Since it's obvious you don't like Pentax or you work for Conakon.

Anonymous said...

This underexposes might be the first series problems. Cause I might be the first or second one K20D owners in Finland and I noticed the same... I complain it, and they change my K20D for the new one. Since I don't have had any problems with exposure...

RiceHigh said...

It has been reported before for inconsistent quality (control) of the K20D by the Polish review site in their K20D review, it is mentioned that:-

"The very first K20D unit produced loud focusing noises, exposed badly, and was assembled badly -- it was replaced by the local distributor, and many issues were not present in the replacement... but some came back in the third K20D unit tested."

So, this really is a serious quality issue for what it reports. And as everybody including Pentax knows, the Dpreview review is very important to the sales of the K20D and there is not any mistake to be allowed. As such, it should easily be believed that the Dpreview K20D testing sample(s) should not be a bad one anyway. In contrast, it should be the one(s) which had been carefully calibrated and selected by Pentax. But yet, the revealed results are yet disappointed, even Dpreview writer does or does not specifically pointed out those, say, the reported ever hotty pixels problem (that they say they have never seen) and this exposure inconsistency issue.

AshlandOregon said...

I've had to send my K10D in for repair, so in the meantime I'm using a Canon 40D. Instantly I've seen a huge difference the accuracy of the metering of the 40D. A much higher percentage (it feels like 93%) of my shots are properly exposed. With my Pentax, I'm usually post processing just about every file for correct exposure. The Canon images don't seem to have the same "edge" as what I was getting from the Pentax, but as tonight—I also really had minimal processing for the 250+ photos I'd taken over the weekend.

The Canon also better integrates with their flash system compared to the sporatic metering I've been getting with the Pentax.

Anonymous said...

Do you ever take pictures in real live? This stuff is so boring.

RiceHigh said...

Of course, no Pentaxian will have any excitement to see these results. They are rather disappointing afterall. But, it is true.

Anonymous said...

Ricehighy, I have to say, that you are completely right. When those forum people on the pentax forum in DPreview are blaiming you, they show that they did not have normal eyes. I have a K20D and it is regarding exposure, a very bad dslr. Do never say this on a forum as you will be banned (as I was on the Ricoh forum when blaiming some stupid things on the GX100. I think the reason they act like that is very simple. They buy it, discover that it is not that good and than they need to find anything to justify why they bought in fact something that is not good at all. So instead that they should call you silly, they should blaim themselves.

Doug's Books said...

I have had my K20D for several months now. Underexposure and just getting any exposure at all have been major issues. It has been incredibly annoying to have augment nearly 40% of the pictures I take. At least underexposure is often fixable. But to spend hours adjusting exposure quality defeats the purpose of getting into photography as an amateur. I am not a happy camper and am puzzled by so many positive reviews of this camera.

RiceHigh said...

> I am not a happy camper and am puzzled by so many positive reviews of this camera.

I think most of the reviews we find on the net or printed magazines are mediocre and not much useful.

For the real stuff by more talent reviewers, most of the time they have some reasons behind to not tell the evils (owing to the conflicts of interests of their own behind).

Do read the end-user feedbacks (like yours) or reviewers who have no conflict of interest of any kind at all - those who write only for their interests just like Ken Rockwell and me. ;-)

Anyway, if the complaints are clearly described and with evidence, you can know the truth.

As for underexposure with Pentax DSLRs, it is just an evil that never gone! My new K-m is of no exception. In Pentax land, proper and accurate metering dated back only to the film SLR days. In fact, the MZ line of SLRs are quite accurate and consistent in metering. The 6/8-segment multi-pattern metering is very good.

Eddie said...

Well, I think you cannot make a statistic based on different photos taken in different lighting situations. Every camera needs exposure compensation in certain cases. That doesn't mean that it under or overexposes. Such statistic would be valid only if comparing the same scene from the exact same spot using a tripod, with the light not changing, same exposure settings and compare the photos for difference in exposure.

RiceHigh said...

As you have said, stats are just stats. By no means stats can be equal!

Even if you shot at the almost the same time and at the same place, lighting condition can still change and exposure (requirements) can be different.

All in all, those stats and analysis just give you an idea on how much (or less) exposure compensation is required with different DSLRs.

gybp said...

sometimes +2 ev comp. is not enough on my k20d....

Anonymous said...

So it isn't a camera for beginners...learn to use it and you will get amazing results. It is a more hands on approach and allows for a stronger knowledge of how your camera will work.

Anonymous said...

The 16 segment metering was notorious for underexposure on just about all the bodies I used. It was quite hopeless in low light too simply well under and by well over 2 stops at times.

The newer metering is def better in my view

Anonymous said...

Great fun. Love these Hilarious pieces.

I use manual and control my metering. I was taught by a pentax mx.


Post a Comment

Related Posts

Creative Commons License
RiceHigh's Pentax Blog by RiceHigh is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.