Previous news story    Next news story

Canon EOS 70D review gains test scene samples

By dpreview staff on Sep 12, 2013 at 07:50 GMT
Buy on GearShopFrom $1,099.008 deals

We're working towards completing our review of the Canon EOS 70D and have been shooting our test scene with a production camera. We've published the test shots, including downloadable Raw files of both the daylight and low light scene. The EOS 70D review will also be one of the few chances to see our outgoing test scene and our new, more challenging, more informative scene alongside one another. Click through to see how it performs.

Now we have a production-standard EOS 70D we'll be forging ahead to get the review completed as soon as we can. In the meantime, here's a chance to see in close detail and in challenging lighting, how Canon's Dual AF Pixel sensor behaves.

566
I own it
239
I want it
40
I had it
Discuss in the forums

Comments

Total comments: 79
microstudio
By microstudio (1 month ago)

I have 70d canon , very nice in video , autofocus work greate, but the lens 18-135 give me many erors when i take photo. I want to send the lens for calibrate . It's posible to use sigma 18-35 mm 1.8 for video whit autofocus? www.nunti-de-lux.ro. Somebody tested the sigma whit autofocus?
Thanks

0 upvotes
lynmay
By lynmay (7 months ago)

Interesting test photo subject matter, but it doesn't work for historical comparisons. I think the black background doesn't work for that reason.

Wonder if that is the point. Less and less true difference between the new cameras and the old when it comes to image quality. I guess if you can't see the difference in similar images then they can convince you to buy the newest, latest, greatest.

Bells and whistles are great, but if there is not a significant difference in image quality then I guess they have to come up with some other way to get your money.

0 upvotes
Franz Kerschbaum
By Franz Kerschbaum (7 months ago)

I love to see this old Austrian 20 Schilling bill in your test setup! I was using it as a kid...

0 upvotes
Matt
By Matt (7 months ago)

Can you guys do AF tests with moving subjects? Honestly the differences in static image quality between todays top cameras seem so small that its alsmot not worth pages of review anyways.

3 upvotes
gingerbaker
By gingerbaker (7 months ago)

I agree!

Could be easy, too. A basketball suspended from a string in the ceiling, brought back to standardized positions (different heights to produce different velocities when released) using standardized lighting. Camera on continuous focus and drive at standardized distance, starts shooting when ball is released. Count the number of in-focus shots.

Ball could be released so that one could assess various angles relative to the camera - straight-on toward/away from camera; 45 degrees vector relative to camera.

dpr could glue small printing labels on the ball with different size fonts to standardize the focus accuracy rankings.

dpr would need a room with standardized lighting, 1 basketball or soccer ball, 1 string, 1 stepladder, 1 piece of wood mounted on a sturdy tripod to hold ball in position, 1 set of alignment marks on the floor to position the tripod.

Comment edited 6 minutes after posting
1 upvote
yabokkie
By yabokkie (7 months ago)

the ball may move too fast for AF tracking, unless the ceiling is very high like the third floor of a stadium.

0 upvotes
yabokkie
By yabokkie (7 months ago)

the AF tracking is intentionally designed to be slow for that's the way we do with naked eyes. it can be distracked easily which is under expectation.

Comment edited 2 times, last edit 9 minutes after posting
0 upvotes
peevee1
By peevee1 (7 months ago)

Look at JPEG test @ ISO25,600 in low light. D70/D7100 - f/5.6, 1/80s, X-Pro1 f/5.6 1/30s. Almost 1.5 stops difference. Come on DPR, some quality control in your tests is badly needed. And why AutoWB (with "preserve warm tones" no less ON!! sheesh!) for some cameras and manual WB for others?

Discrepancies like these make the tests useless. Good thing you only need to retest (yes, retest!) 10 cameras or so.

3 upvotes
yabokkie
By yabokkie (7 months ago)

sorry to say but better ignore everything > 6400.
it'll be digital gain at ultra high ISOs anyway.

Comment edited 2 times, last edit 10 minutes after posting
0 upvotes
Richard Butler
By Richard Butler (7 months ago)

All cameras should be on Auto WB in the low light test - if you've spotted some that aren't, we'll change them.

Also, if a manufacturer sets 'Preserve Warm Tones' as the default, then it's reasonable to assume that it'll be a minority of people who turn it off - so we leave it at the default setting.

A bug in our system meant the X-Pro1 shot is showing the wrong metadata - the ISO 25,600 shot was taken at 1/60th of a second, F5.6. I'll try to find out how that happened.

Comment edited 8 minutes after posting
0 upvotes
DonSantos
By DonSantos (7 months ago)

The x-pro1 daylight was shot at 1/2000 f/7.1 6400iso
The 5d mark III daylight was shot at 1/200 f/5.6 6400iso

The x-pro1 tungsten was shot at 1/15 f/7.1 6400iso
The 5d mark III daylight was shot at 1/15 f/5.6 6400sio

Richard. Is it really fair to make iso comparisons that are 2/3 stop difference?

3 upvotes
JackM
By JackM (7 months ago)

But you can't even spell 70D.

0 upvotes
peevee1
By peevee1 (7 months ago)

"All cameras should be on Auto WB in the low light test - if you've spotted some that aren't, we'll change them."

Some? For example, low light, Canon 700D (right there in your default comparison), JPEG, ISO 6400 - manual WB. D7100 JPEG, ISO 6400 - manual WB. Again, right there in your default comparison. Same thing at every other ISO for these two.

3 upvotes
Richard Butler
By Richard Butler (7 months ago)

That's a metadata problem - the images themselves were Auto white balance - hence the strong yellow tinge to them all.

0 upvotes
DonSantos
By DonSantos (7 months ago)

You have a new test scene! This is time to use new testing procedures.

Instead of pinning aperture and iso only you also need to pin shutter speed.

After you pin shutter speed then normalize the image with gain +/-. This is the only way to assure proper iso comparisions. It's not a fair comparison that the e-m1 gets 2/3 stop more light than the 70d.

4 upvotes
ET2
By ET2 (7 months ago)

Different camera companies apply different tonal curve. This won't work well for jpeg.

Comment edited 58 seconds after posting
0 upvotes
yabokkie
By yabokkie (7 months ago)

JPEGs may lose some quality but error will be smaller and the result more comparable on a more level ground.

0 upvotes
peevee1
By peevee1 (7 months ago)

Actually, aperture/shutter-speed should be DoF-normalized.

0 upvotes
yabokkie
By yabokkie (7 months ago)

> DoF-normalized

then it will be difficult to compare at the same ISO
should use equivalent ISO which is simple area ratio.

Comment edited 1 minute after posting
1 upvote
igor_s
By igor_s (7 months ago)

DonSantos, pinning the shutter speed or whatsoever makes sense only if you know that the firmware reports the really applied values. The reported shutter speeds may be incorrect just like the ISO values often are.

0 upvotes
yabokkie
By yabokkie (7 months ago)

nothing is reliable, shutter or aperture (and transmission), but ISO is not something we care. we can do good tests without knowing real values of ISO settings of a camera, though we can know what they are if other parts of the test are done correctly.

0 upvotes
bwoodahl
By bwoodahl (7 months ago)

Folks, look at the old man's eyes "pencil drawing" at 400 / RAW with incandescent lighting. 5DMk3 > D7100 > LumixGX7 > 70D. To me, the 70D is not even close.

http://s8.postimg.org/3qefu0l51/Screen_Shot_2013_09_12_at_3_21_41_PM.png

0 upvotes
lem12
By lem12 (7 months ago)

Would be much helpful to compare more cameras at this new test scene.

0 upvotes
ZAnton
By ZAnton (7 months ago)

Agree. Put a Nikon D600 as a Nikon's FF benchmark.

0 upvotes
Zerg2905
By Zerg2905 (7 months ago)

This makes me want the 70D sensor in a lighter "package", e.g. in the probable EOS 750D. For a possible / probable 7D mk II, an even better sensor is needed (or at least this is what I would like). My opinion. Cheers! :)

0 upvotes
marike6
By marike6 (7 months ago)

I have a feeling that Canon is going to pull out all the stops with the 7D Mk II and almost force Nikon to respond. Not sure what sensor it will have (70D or something new) but it should be a very nice camera.

0 upvotes
Peiasdf
By Peiasdf (7 months ago)

The new test scene is so hard to use. I am so used to the old one that the new one leaves me lost on where I should use to compare cameras. The watch face, egypt on the globe, paperclip, battery, old lady's face and label on bottle are all gone.....

5 upvotes
Richard Butler
By Richard Butler (7 months ago)

This is something we thought long and hard about.

In the end the decision was made not to include any of the elements from the old scene, since they would be tiny in the new scene and hence not comparable.

In time, we hope that we'll all (us included) learn where to look for the important differences between cameras.

0 upvotes
yabokkie
By yabokkie (7 months ago)

would appreciate PH not be used as a standard for it makes comparison of different aspect ratios hard.

when we use the comparison tool, we see a fraction of image and not the aspect ratio which is really irrelevant. it's not field of view but magnification.

it may be called "area magnification" that a subject projected on a sensor occupies a fixed fraction of area regardless of the size and shape of sensor (be it circular or triangular).

or simply "any subject gets same number of pixels on sensors of same pixel count." the opposite will look strange.

then the test target may look like a cross of 3 rectangles of exactly the same area for 4:3, 3:2, and 16:9 sensors. the fully overlaped center area be used for sensor comparisons and non-overlaped parts are needed for lens tests.

Comment edited 3 minutes after posting
0 upvotes
Richard Butler
By Richard Butler (7 months ago)

@yabokkie - it's a good idea, but it's too late to completely redesign the chart. In fairness we've been inviting comments about the chart for nearly a year and this is the first time we've seen this suggestion.

Comment edited 3 minutes after posting
0 upvotes
yabokkie
By yabokkie (7 months ago)

currently you use 100% of 4:3 against 88.9% of 3:2 (area) which get 0.17 stops lower performance (= log2(88.9%), I arbitrarily think the total error should be less than 1/6 stops and this consumes all of it, for nothing).

think about the tests you are going to do in the following years from medium format to smart phones ...

0 upvotes
plasnu
By plasnu (7 months ago)

D800, please.

I noticed that the lens on 70D is horrible compare to the E-M1 lens. They do look about the same at the center, but at the corner, E-M1 look much better.

0 upvotes
marike6
By marike6 (7 months ago)

You may want to check that again.

The 70D and EM-1 corners look almost indentical, but in the center and borders, it's clear the 70D is showing better per-pixel sharpness in RAW. Labels on the Vinegar, playing cards, Schilling bank note, tubes of paint labels, spools of thread, two pencils drawings all show higher acutance, more detail on the 70D image.

3 upvotes
plasnu
By plasnu (7 months ago)

OK, you;'re right. My mistake.

0 upvotes
Richard Butler
By Richard Butler (7 months ago)

D800 is coming soon.

0 upvotes
kecajkerugo
By kecajkerugo (7 months ago)

for JPEG shooters this may not be true! Oly show up with some great results across the picture. The difference though is really secondary importance in "normal" shooting which is amazing and just show that m4/3 are "equally" capable as DSLRs to the level that one can chose a camera per something else than pure image quality.

0 upvotes
peevee1
By peevee1 (7 months ago)

What are you looking at? JPEG ISO 100, replace 700d with E-M1, put the rectangle at extreme lower left corner. Now look at the lower right corner of the crop. First (lowest) set of lines is merged on both. But one up is distinctive on E-M1, merged on 70d. In RAW it is still merged, only now drowns in CA on both 70D and D7100 shots.

0 upvotes
marike6
By marike6 (7 months ago)

The 70D and D7100 RAWs seem to lead most of the other crop sensor cameras in sharpness.

The X-Pro1 and E-M1 seem similar, and a bit softer in comparison to the 70D and D7100. I do miss the Lira note of the old test scene, but if you look at the engraved bank note or the hair, for example, it's pretty easy to judge RAW detail.

2 upvotes
autoy
By autoy (7 months ago)

To be fair, X-Trans does not fare very well in ACR with default settings, but try Aperture or Capture One and it does indeed match the sharpness of the mentioned systems.

0 upvotes
Rbrt
By Rbrt (7 months ago)

The comparison choices are too limited. The 6D should be in the list.

0 upvotes
yabokkie
By yabokkie (7 months ago)

would rather prefer it's done slowly carefully first,
than a great volume of work done in the wrong way.

3 upvotes
Rbrt
By Rbrt (7 months ago)

If you're saying the others are coming, then I agree!

0 upvotes
yabokkie
By yabokkie (7 months ago)

I'm saying that what we get now cannot be used to compare image quality. DPReview need a review of their test design and do again.

0 upvotes
Kelcey Smith
By Kelcey Smith (7 months ago)

Don't worry it's coming! We are working to add a backlog of cameras but we wanted to release current cameras in a timely manner as we work through that list.

3 upvotes
steelhead3
By steelhead3 (7 months ago)

I hope that use the best 50mm that each system has when you re shoot (the new 50 1.4 CZ from sony rather than the old 50). When doing resolution tests, lens and camera work together; to test camera only the sigma 50 would be a comparison tool if it was available in the mount tested.

0 upvotes
Streetlight
By Streetlight (7 months ago)

I'm not an expert at these kinds of tests, but I wonder how much these comparison tests are more a comparison test of lenses than the the camera. Certainly for the same camera make with the same lens mount one can use the same lens and see differences in the rest of the hardware. All bets are off when comparing different makes when exactly the same lens can't be used across them.

0 upvotes
Dimit
By Dimit (7 months ago)

Gendlemen of DPR,with all due respect,your comparison tool for the low light IQ is definately wrong.How can it be that e.g.at 25600 iso the best of all the existing database gear to be the RX100II (OK,after 5Diii)???
Should pay attention to the exposures equalization algorithm,or what?
Just a feedback in good faith..

1 upvote
ET2
By ET2 (7 months ago)

RX100II doesn't have ISO 25600, so you are looking at wrong ISOs, probably.

0 upvotes
kecajkerugo
By kecajkerugo (7 months ago)

yet another good (but bulky) DSLR from Canon.
Taking ocasion I am trying to attract your attention to the Oly E-M1 latest machine and want you to see that the high ISO OF LATEST M4/3 SENSOR IS RIGHT ON PAIR WITH THE BEST DSLR LIKE THIS ONE.

4 upvotes
Jogger
By Jogger (7 months ago)

lol, this bulky DSLR from Canon will sell 10-15x more than the Oly EM1 ... has Oly ever made a profit from m43?? no.

8 upvotes
Chez Wimpy
By Chez Wimpy (7 months ago)

All the more reason to bring attention to it.

1 upvote
lol101
By lol101 (7 months ago)

It remains to be seen if Oly still plays with its ISO ratings to deliver this result.

If the exposure value (ie the amount of light reaching the sensor) has to be twice as much as other cameras for a given ISO, then it's not really a performance but more likely that you should compare ISO 6400 from Oly to ISO 3200 of other manufacturers.

1 upvote
turbsy
By turbsy (7 months ago)

Why do you feel the need to attract attention to the oly camera? Do you get payed by Olympus? or just a fanboy?

5 upvotes
Dave Oddie
By Dave Oddie (7 months ago)

"If the exposure value (ie the amount of light reaching the sensor) has to be twice as much as other cameras for a given ISO, then it's not really a performance but more likely that you should compare ISO 6400 from Oly to ISO 3200 of other manufacturers."

Why? The exposure value at a given ISO gives you an EV number. So for example at ISO 100, EV 13 which is a typical daylight scene, clouds but bright and no shadows should mean the camera sets an exposure of 1/125 and F8.

If that is what you get when you meter this scene with an Oly why would you NOT compare the output at this ISO with the ISO 100 output from any other camera assuming the other cameras also correctly metered the scene to mean1/125 at F8?

If the Oly set an exposure of 1/60 @ F8 @ ISO 100 for an EV 13 scene you many have a point but I bet it doesn't.

Comment edited 2 minutes after posting
1 upvote
Ferling
By Ferling (7 months ago)

First. Some of us continue to stick with a particular vendor because we have a lot of money invested in lenses.

Second. What may be bulky to you, is a necessity to others, including myself. Of course, Canon/Nikon offers smaller solutions that will accept those same lenses.

So, I don't think a few features or even an edge in IQ/ISO would be compelling enough when one considers the total cost of conversion.

Comment edited 44 seconds after posting
0 upvotes
BeaniePic
By BeaniePic (7 months ago)

I don't know what you mean by bulky. This size camera is great in the hand to use. It fits well with easy access to the controls. Smaller CSC type cameras make it hard (For Me and some other photographers) to use. Access to functions have a way to go on smaller cameras. (My and people I speak to views anyway)

1 upvote
rfsIII
By rfsIII (7 months ago)

For me, it's not the size in the hand so much as it is the weight on the shoulder at the end of the day. It's also the bulk of the camera bag—it's a pain schlepping a big bag of full-size DSLR lenses and bodies through airports and in among crowds. When I'm out and about I want to feel like a photographer, not a donkey. If I were a studio photographer, it wouldn't be an issue.

1 upvote
Ayoh
By Ayoh (7 months ago)

These comparisons does not seem fair as the exposure differs between cameras. For example look at this option:

Canon 5D Mark III and Olympus E-M1
RAW
ISO 25600
LOW LIGHT

The exposure parameters are:
Canon - 1/60, f7.1
Olympus - 1/50, f5.6

In the comparison EM-1 has similar noise to 5D under all sizes(full/print/web). The amount of light received by the camera sensors must have been different between the cameras to get this result.

1 upvote
Henry M. Hertz
By Henry M. Hertz (7 months ago)

yeah i noticed that dpreview is not very good at making things equal for all tested cameras a long time ago.

i wonder why they don´t test all canons at f16 or f22 and nikons at f8.

Comment edited 2 times, last edit 2 minutes after posting
0 upvotes
le_alain
By le_alain (7 months ago)

Or ISO 25600 is not equal for some camera makers
;)

4 upvotes
hippo84
By hippo84 (7 months ago)

http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Olympus/PEN-EPM1 - Olympus is not very honest in marking ISO. For example, ISO12800 set on EPM1 is only ISO7693. That's why we see the difference in camera settings.

2 upvotes
Greg VdB
By Greg VdB (7 months ago)

Le-alain and hippo84 are right of course. Also, different lenses have different optimal settings, and since these are camera reviews I'm sure the apertures selected are those giving the optimal results to not make the outcome depending on the lens instead.

PS: Henry, I hope you do realize that f16-22 gives significantly worse sharpness figures than f8 due to diffraction? Not everything is down to a conspiracy, and dpreview reviewers generally know very well what they are doing...

2 upvotes
TrojMacReady
By TrojMacReady (7 months ago)

Do not confuse the DXOmark ISO measurements with indicated camera ISO's. The latter is what directly affects comparisons seen here, the DXO ISO's do not necessarily reflect metering or even "cheating". It's only a way to normalize output for RAW comparison purposes.

The top of the "noise" page from the final review, will usually tell you how camera ISO's compare to one given ISO standard ( ISO 12232:2006).

Comment edited 2 times, last edit 2 minutes after posting
3 upvotes
yabokkie
By yabokkie (7 months ago)

better all tests shot at the same shutter speed for high shutter speeds will have large error and low ones more noise.

also we are likely to shoot high ISOs at relatively low shutter speeds (like hand shake limited) and low ones at very fast speeds, but most tests do the opposite.

0 upvotes
Richard Butler
By Richard Butler (7 months ago)

TrojMacReady is absolutely correct - DxO's figures do not show 'ISO cheating' - they acknowledge this on their page that explains the data.

The difference in exposure between the 5D Mark III and the E-M1 is 1/3EV which isn't a hugely significant difference in the grand scheme of things.

2 upvotes
yabokkie
By yabokkie (7 months ago)

think we have to expect the horizontal axis be exposure per ISO standard or we cannot use DxOMark charts to compare sensor performance.

Comment edited 41 seconds after posting
0 upvotes
Richard Butler
By Richard Butler (7 months ago)

@yabokkie - I think you're saying that DxO is right to do what it does (though your simplistic use of 'per ISO standard' risks being misleading). I'm not disputing that. If that isn't what you mean, please try to re-state your point.

What I am saying (and they also state) is that what DxO shows is not ISO cheating in any meaningful sense. Our test shots suggest the E-M1 may be around 1/3EV less sensitive than the EOS 5D Mark III, using the section of the ISO standard that manufacturers use (the one marked on cameras).

Comment edited 2 minutes after posting
0 upvotes
Chiemsee
By Chiemsee (7 months ago)

Greg VdB wrote: "and dpreview reviewers generally know very well what they are doing..."

How do you KNOW this?

I would say dpreview knows (obviously) quite a lot about photography. But unfortunately they seem not to be trained physicists/statisticians... The results here LOOK very reputalbe and professional. But in my opinion from a scientific point of view they are not.

I don't think this is tragic - it's just a sad thing that the LEADING websites on the web (concerning camera/sensor/lens tests) neither do have scientific standards nor scientific training in what they are doing (including DxO unfortunately :( ).

Don't get me wrong. I'm not saying everything is wrong here. I just feel that many people take results here as absolute and objective and apodictic truths.

The most obvious problem with the tests here is (of course) sample size. With a sample size of 1 (and often minute differences discussed) you couldn't publish any scientific paper anywhere without being laughed at...

Comment edited 53 seconds after posting
1 upvote
rfsIII
By rfsIII (7 months ago)

You have an excellent point, but I'm assuming that these guys only get one sample of a camera to test. So is there a scientific way for them to take that into account? Certainly there must be lots of other situations in science and engineering when you only get one shot at something, like when you do a flyby of a planet or or the thing you are testing is so expensive or rare that your budget can only afford one. So what would you do then? Help us out here....

0 upvotes
yabokkie
By yabokkie (7 months ago)

@R Butler,

I was actually thinking of an exposure meter when saying "per ISO standard." I would shoot all studio shots manually using shutter speeds suggested by a Sekonic.

JPEGs may lose quality if anyone tries to correct uneven brightness (but for image quality "bright" or "dark" really means SNR dBs instead of DNs or cd/m2).

we need a way to know the real exposure to decide where the chart should be positioned horizontally. DxOMark have their own standard and that's okay because our priority is to have a level ground (makers have their own standards, too. but Canon and Nikon are the same).

I assume all DxOMark charts at a certain ISO value have the same exposure to unit sensor area (may be shifted from ISO standard). I assume them do the thing right.

it has to be called cheating if the exposures are too different (how different is a subjective issue that we may want to decide by polls).

Comment edited 2 times, last edit 8 minutes after posting
0 upvotes
TrojMacReady
By TrojMacReady (7 months ago)

One of the problems with measuring RAW (binary) data rather than looking at RAW files opened in converters, is that for example you don't take into account possible default adjustments done under the hood in those converters for "correct exposure". The A99 for example was the first SLT to rely on this to compensate for the light "loss" due to the beamsplittter, a digital correction in the converter rather than a digital correction before writing the RAW file (simple multiplying of values or extra analogue sensor gain, which both means throwing away a bit of highlight headroom). In the DXO ISO chart, that means the A99 suddenly scores much lower than previous SLT cameras, which lead some to believe that it "cheats" when it has nothing to do with metering of visual exposures.

The X100 uses this above ISO 1600. Which can be witnessed in the DXO ISO chart as a "flatliner", even though the visual exposures are still affected above ISO 1600, both in RAW (supported converters) and jpegs.

Comment edited 3 times, last edit 7 minutes after posting
2 upvotes
yabokkie
By yabokkie (7 months ago)

DxOMark can let the camera decide shutter speed and offset the result horizontally but those images cannot be used in DPReview's tool for the exposures are all different.

since DxOMark's measured ISO values scatter over a lot of numbers I assume (hope) they do it the hard way with good accuracy.

Comment edited 8 minutes after posting
0 upvotes
Richard Butler
By Richard Butler (7 months ago)

@Chiemsee - None of us are physicists but several of us have science degrees and have a pretty good understanding of the scientific method.

Our tests are designed to be as fair, consistent and meaningful as possible, but they're not supposed to be definitive scientific tests - mainly because there isn't an obvious mechanism to pay for the amount of work that would require.

Our testing is planned in such a way as to make it as useful as possible. They're not the only tests that can be conducted - you'll see plenty of people who will insist that the only way we should do them are the way they propose.

We have a pretty good understanding of what our tests do and don't show and where the limitations of them are but, as I say, the aim is to be fair, consistent and photographically relevant.

0 upvotes
igor_s
By igor_s (7 months ago)

Dear Mr. Butler, to find out whether the claimed ISO values are correct, you need to know the real shutter speed and lens transmission at all relevant settings. Using these values and a standard lighting, you can calculate the real ISO values from the test images. The lens transmission is measured by DxOMark, but what about the shutter speed? If some firmware would show 1/60s instead of the real 1/30 s, you would "see" a 1 stop better ISO performance.

1 upvote
Richard Butler
By Richard Butler (7 months ago)

We'll look into those low-light images - that's a bigger discrepancy than I'd expect.

1 upvote
Chiemsee
By Chiemsee (7 months ago)

R Butler wrote: "We have a pretty good understanding of what our tests do"

Let's assume that's correct. Then still most of your readers don't have this understanding. And they use your tests in a, well let's say 'dangerous' way.

A lot of discussions concentrate on MINUTE variations at pixel level.

Without analyzing
- sample variation,
- RAW converter influence,
- interaction effects of certain lens/body combinations
quite a lot of the conclusions are, well let's put it in a simple but unfortunately true word: useless.

Again, discussing MINUTE variations based on sample size 1 tests without even given a ROUGH estimation of sigma are - scientificifally speaking - massively flawed. It could be regarded as intellectually dishonest to offer those test to the (often scientifically untrained) public without explicitly indicating the limitations... ;-/

0 upvotes
Chiemsee
By Chiemsee (7 months ago)

But that's not a dpreview specific problem but a problem of photo image test websites in general.

But let's think positive:

If engineers built bridges after doing material tests this way, you wouldn't want to drive across them any more... ;) With photo tests it seems to be o.k. Better this way than the other way around.

0 upvotes
yabokkie
By yabokkie (7 months ago)

I just looked at EXIF of some of the new test shots.
their exposure values suggest different conditions,
given all aperture, shutter, and ISO numbers are accurate,
which is of course not ture, but the calculation goes,
3136 lux for E-M1 and GX7
3164 lux for E-M5,
3781 lux for 5D3,
3920 lux for 70D and NEX-6,

the cameras will be on a level ground at these illuminance, or in other words, if the lighting in the studio provided less than 3920 lux for 70D and NEX-6, these two cameras should have performed worse and it's not fair for them.

0 upvotes
yabokkie
By yabokkie (7 months ago)

may be we need two values like 3000 lux for base ISO to 6400 and 0.3 lux for 6400 and up (I would prefer lower to avoid too high shutter speed).

then all the EVs will be fixed for each ISO and we can have better comparisons. this will be a good base for further more accurate tests.

0 upvotes
Total comments: 79