SLT -1/3 stop?

[JJLMD]

DPReview in their "Conclusions" section of the "Sony SLT-a99 Review" allows one to compare its high ISO performance with other cameras in its class including the Nikon D800 as well as Canon 5D MkIII. The Sony is scored HIGHER than either despite DXO showing Sony's ISO score being markedly inferior at 1555 compared with Canon's at 2293 and Nikon's at 2853.

Anyone remember when Total Harmonic Distrtion (THD) was the holy grail measurement for high end audio? We now know that THD has very poor correlation to actual sound quality.

 

[EinsteinsGhost]

NAwlins Contrarian wrote:

re: "Even controlled tests reveal Sony's SLTs delivering better color accuracy than their Canon and Nikon counterparts."

Sorry, no. On those numbers, lower means more accurate. They represent deviations from perfect color accuracy. The very best (in terms of color accuracy) current digital cameras are about 6, 9 is not so good, and films start around 10 or 11. So on the whole Sony actually has measurably less accurate color than Nikon and/or Canon.

Now accurate color and pleasing color are not necessarily the same thing. You might subjectively like the color from one camera with an average Delta E of 9 better than the color from another camera with an average Delta E of 7. In old-school terms, some people like Velvia 50 better than Astia 100F--but Astia is more accurate.

I stand corrected. Now, RX100 looks better (at 6.6).

 

[Just Having Fun]

JJLMD wrote:

DPReview in their "Conclusions" section of the "Sony SLT-a99 Review" allows one to compare its high ISO performance with other cameras in its class including the Nikon D800 as well as Canon 5D MkIII. The Sony is scored HIGHER than either despite DXO showing Sony's ISO score being markedly inferior at 1555 compared with Canon's at 2293 and Nikon's at 2853.

Anyone remember when Total Harmonic Distrtion (THD) was the holy grail measurement for high end audio? We now know that THD has very poor correlation to actual sound quality.

In the actual review they say things like: "By ISO 6400 the A99 displays image smearing that is noticeably more aggressive than both its Canon and Nikon rivals."

The A99 does have IBIS and the A99 has a multi-shot noise reduction mode (any camera can do that in post), but no one has ever said the A99 can match the Canon or Nikon at high ISO performance.

 

[JJLMD]

Look for yourself. Go to Conclusions page of SLT-a99 review. Go to yellow bars graphs at bottom of page. They state specifically comparisons of yellow bars are valid for cameras in same class. Compare D800 and 5DmkIII, which are in the same class as the SLT-a99, for high ISO performance.

Moreover, in the text of the Conclusions section, they repeatedly praise the SLT-a99's "particularly" "excellent" high ISO performance.

What am I missing?

 

[JJLMD]

Sorry...meant green bars graph...

 

[TrojMacReady]

Just Having Fun wrote:

TrojMacReady wrote:

dpyy wrote:

saralecaire wrote:

I read many say that the SLT reduces light transmission by about 1/3 stop.

If so, how dome DXOmark has the A99 low light score at 1555 vs the D600 (which apparently has a similar sensor) at 2980? Closer to 1/2 stop?

It's quite an easy exercise to test for yourself. Just go over to the dp test pages and setting the A99 to 3200 and D600 at 6400. And you will see that the noise is very comparable.

In other words, at high ISO A99 is one full stop behind Nikons/Canons FF.

Closer to half a stop, since the D600, 5DMKIII and D800 are clearly noisier at ISO 6400 than the A99 at ISO 3200. Or ISO 12800 vs 6400.

After checking the whole scene it very, very close (but not quite equal to) a full stop. The Nikon D600 image at ISO6400 is better focused (sharper) than the A99 at ISO3200. That slight burring may be masking a little of the A99 noise. There are too many variables to be exact though.

In any case saying a full stop is close to reality than saying a 1/2 stop.

No, focus differences do not affect noise. And reality is what can be seen in those crops posted or even better, when taking T stop differencesout of the equation (same Sigma lens):

Loading…

www.dpreview.com

Anyone with eyes would agree that's closer to half a stop than a full stop.

 

[TrojMacReady]

On the whole when just looking at ooc jpegs. DXO measures colour response in their full sensor tests and here Sony cameras consistently show less colour blindness than any of its peers the past years.

 

[Just Having Fun]

Nice try. The d600 not only looks clearer and sharper but has almost a full stop less noise. No wonder Sony is dumping the SLTs and starting a fire sale on the current models.
The future models will be much better and closer to Nikon.

 

[NAwlins Contrarian]

re: "Anyone remember when Total Harmonic Distrtion (THD) was the holy grail measurement for high end audio? We now know that THD has very poor correlation to actual sound quality."

I agree with you that there were times when audio manufacturers liked to tout THD. As certain types of amps came on the market, they could produce very low rated THD, which sounded impressive--to the not-too-knowledgeable. Like with so many things, figures don't (usually / totally) lie, but liars figure. The issues with THD and some camera parallels, IMO:

(1) THD is only one relevant parameter affecting overall sound quality, other important ones including frequency response, noise, and wow / flutter (remember them, with analog sources?!). Kind of like with digital cameras, you can focus on resolution or noise, but to ignore the other is to ignore something very important--and the same goes for dynamic range, color accuracy, etc.

(2) Sometimes there are major diminishing returns. While I have no doubt that many can at least at times hear an improvement going from, say, 2% THD to 0.2%, the specs at least kept getting lower. Is 0.05% THD audibly better than 0.1% THD? I doubt it! Kind of like cramming 16 MP onto a 1/2.3-inch sensor for a compact digicam. Over and above hard limits like lens performance, what percent of compact digicam shots are shot other than hand-held? I suspect that is a much more important limit on recorded detail than, say, the difference between 8 MP and 16 MP. But again, the marketers push the number that sounds impressive, even if in the real world it isn't too important (at least past some point).

 

[ultimitsu]

EinsteinsGhost wrote:

saralecaire wrote:

I read many say that the SLT reduces light transmission by about 1/3 stop.

If so, how dome DXOmark has the A99 low light score at 1555 vs the D600 (which apparently has a similar sensor) at 2980? Closer to 1/2 stop?

DXOmark's rating favors sensors that are tweaked to expose to the right, saturate early and consequently have a higher DXO ISO rating which ultimately affects low light score, and every other score. A consequence to it is that, with quick saturation sensors, chance of blowing highlights goes up. This is also a reason why A900 appears to be as good as A99 but in fact it isn't.

If you shoot with ETTR, you're technically doing what Nikon does. For all practical reasons, you're looking at a difference of only a half stop (between D600/RX1 and A99) but in exchange, you're getting full time AF/tracking in both stills and video modes.

You are confused over the concept of ISO.

ISO is a standard for output image brightness at a given luminance/light level at a fixed shutterspeed and aperture and focal length.

This concept is completely unrelated to metering or exposure choice of the camera.

Basically, when we point D600 and A99 at the same scene, using the same lens, setting shutter speed and aperture at the same value, take the picture, then we brighten that picture to the same brightness and we say these are ISO 100 images.

A99 will of course receive less light because SLT absorbs certain amount of light, thus it will require more brightening to bring the brightness level of final image up to ISO standard, of course that extra brightening will cause more noisy, less DR, less colour depth and less SNR.

But, of course, because it did not receive as much light, it is further away from saturation point. But that is not because it has a deeper well, it is simply because it is less efficient at collection light.

 

[ultimitsu]

EinsteinsGhost wrote:

Wally626 wrote: Why the DXO ISO ratings are scoring the new Sony cameras so low I do not know.

That is primarily because of the way DXO measures ISO and then uses it for other measurements. And this applies differently to RX1 than to A99.

ISO is an final output brightness standard. DXO measures how close is an camera's RAW at a given ISO setting to the ISO standard brightness. That however has no influence to ultimate scores.

For example when comparing DR at ISO 800, we are talking about the actual ISO 800 standard. at this ISO standard, A99's image would have 10.7 stop DR, you can obtain it by shooting ISO 800 setting, which would generate a RAW image with iso 439's brightness, and brighten it by 0.9 stop to ISO 800's brightness, or you could shoot at ISO setting of 1250, which would generate a RAW image with brightness of ISO 686, you can then brighten that by 0.3 stop to ISO 800's brightness, but in each case your SNR, DR, colour depth will probably be the same.

If sensor is tuned to, or by design, saturate early, it will have a lower DXO ISO. And as a result, all other measurements are impacted since DXO does not use exposure ISO for these ratings (whereas people assume they are one and the same).

The flaw in this statement had been addressed previously.

 

[ultimitsu]

EinsteinsGhost wrote:

But, DXO quotes its other ratings based on sensor ISO.

This sentence has no logical connection to the sentence below.

So, if a sensor that can absorb a lot of white light before reaching saturation, will have lower sensitivity to white light,

This makes no sense. the deeper the well result in high DR, nothing else. The only thing lowers sensitivity is lower efficiency on the sensor, or if there is a translucent glass absorbing some light before it hits the sensor.

which gets it lower sensor ISO which affects its low-light noise rating because, not exposure but sensor ISO is used to quote the measurements.

As per above, lower "sensor iso" is a product of bad efficiency of light gathering ability, it has nothing to do with well depth. A very sensitive sensor - one with very high QE, could very possibly have a very shallow well and gets blown quickly.

Whereas, out in the field, a person doesn't select sensor ISO, rather an ISO that computes to a proper exposure.

ISO is not part of exposure. the only three things have a play in exposure are:

1. scene luminance
2. aperture
3. shutter time

ISO is the standard of brightness of the final image.

And even if one chooses to shoot ISO-less (as I occasionally do), we're still exposing for a particular exposure ISO. For example, I use ISO 400 or 800 to shoot NBA action, at 1/500-1/640s, f/2.8. This is 1-2 stops underexposed, compared to ISO 1600,

They are 1 or 2 stops darker than ISO 1600, but the exposure are exactly the same - 1/500s F2.8.

but without any consideration to DXO ISO.

What is DXO ISO?

 

[ultimitsu]

EinsteinsGhost wrote:

Allan Olesen wrote:

EinsteinsGhost wrote:

Allan Olesen wrote:

Try this thought experiment:

1. You set your camera at ISO 1600, illuminate the sensor with a given light intensity and measure the SNR.
2. You set your camera at ISO 6400, illuminate the sensor with the same light intensity and measure the SNR.
3. SNR result turns out to be the same in both situations.

How does ISO influence on the measurement if the result is the same regardless of chosen ISO?

This behaviour is the reality for most cameras.

SNR will not stay the same.

Yes, it will. And this fact is really important for your understanding of the topic.

SNR will NOT be the same at ISO 1600 compared to ISO 6400.

That are the same because the lighting, ss, aperture are the same. ISO is merely output brightness, its SNR does not change.

But then, why even talk in those terms if exposure ISO (AKA Camera ISO AKA Manufacturer ISO) does not matter? Perhaps we should only talk about sensor ISO? And does that change?

None of them natter because the lighting, ss, aperture are the same.

You don't use light. You capture light. You either use Auto ISO, or Manual ISO.

I think this is the root of your confusion... it is understandable because many people have this confusion and that is because manufactures played a big part in this whole misunderstanding.

the iso you select on your camera really has very little to do with ISO standard. they are simply a bad way to label in-camera amplification.

The only exception from this is situations where I have more light than I need at base ISO. But that situation is not relevant in a discussion regarding high ISO noise.

In fact, it is very much relevant, if you make a case for exposure ISO being meaningless.

That just made no sense to me at all.

 

[JJLMD]

Popular Photography's noise data would tell us that Fujifilms's X-E1 should have poor high ISO performance but every subjective reviewer finds their high ISO performance to be class leading. Ditto Sony RX1's noise data.

My point is that too many posters above cite DXO but DXO's 3 metrics to rank sensors are simply that: 3 metrics. Is there a THD in there? Maybe. Maybe not. Subjective evaluation of actual image quality by experienced observers is IMHO more meaningful. DPReview cites the Sony SLT-a99's "particularly" "excellent" high ISO performance. In their green bar graph, it ranks higher than the Nikon D800 and Canon 5DmkIII for high ISO performance.

Bottom line? The Sony SLT-a99's low light capabilities rock! End of discussion.

 

[ultimitsu]

JJLMD wrote:

Popular Photography's noise data would tell us that Fujifilms's X-E1 should have poor high ISO performance but every subjective reviewer finds their high ISO performance to be class leading. Ditto Sony RX1's noise data.

High iso noise performance is more about detail, DR, and colour retention than amount of noise.

My point is that too many posters above cite DXO but DXO's 3 metrics to rank sensors are simply that: 3 metrics.

I did not see one person citing DXO's overall sensor rank, let alone "too many". DXO's DR and high iso measurement is often cited, I do not inappropriateness in them.

Subjective evaluation of actual image quality by experienced observers is IMHO more meaningful.

Who are they and what is the objective standard of "experienced"?

DPReview cites the Sony SLT-a99's "particularly" "excellent" high ISO performance. In their green bar graph, it ranks higher than the Nikon D800 and Canon 5DmkIII for high ISO performance.

Prime example of why "Subjective evaluation of actual image quality by experienced observers" is completely worthless. Even DPR's very own studio test shots show both 5D3 and D800 are superior in high ISO.

Bottom line? The Sony SLT-a99's low light capabilities rock! End of discussion.

But the exact thing that OP is not asking and this thread is nto discussing, isnt it?

 

[TrojMacReady]

Just Having Fun wrote:

Nice try. The d600 not only looks clearer and sharper

Like I said, anyone with eyes. And since the A99 shot is clearly sharper, you don't seem to qualify to pass any relative judgements worth listening to.

But with your history, that didn't come unexpected.

 

[JJLMD]

ultimitsu wrote:

JJLMD wrote:

Popular Photography's noise data would tell us that Fujifilms's X-E1 should have poor high ISO performance but every subjective reviewer finds their high ISO performance to be class leading. Ditto Sony RX1's noise data.

High iso noise performance is more about detail, DR, and colour retention than amount of noise.

My point is that too many posters above cite DXO but DXO's 3 metrics to rank sensors are simply that: 3 metrics.

I did not see one person citing DXO's overall sensor rank, let alone "too many". DXO's DR and high iso measurement is often cited, I do not inappropriateness in them.

Subjective evaluation of actual image quality by experienced observers is IMHO more meaningful.

Who are they and what is the objective standard of "experienced"?

DPReview cites the Sony SLT-a99's "particularly" "excellent" high ISO performance. In their green bar graph, it ranks higher than the Nikon D800 and Canon 5DmkIII for high ISO performance.

Prime example of why "Subjective evaluation of actual image quality by experienced observers" is completely worthless. Even DPR's very own studio test shots show both 5D3 and D800 are superior in high ISO.

Bottom line? The Sony SLT-a99's low light capabilities rock! End of discussion.

But the exact thing that OP is not asking and this thread is nto discussing, isnt it?

Good point about distinction between high ISO performance and noise.

The original poster cites the Sony's DXO ISO score and asks about its relationship to SLT technology. It sounded to me like the original poster was assessing its low light capabilities and to what degree it is hampered by SLT technology.

 

[JJLMD]

If it helps, I have the RX1 as well. It has the same sensor as the SLT-a99 without the SLT. With identical f-stop and shutter speed, the RX1's ISO will set at 4000 whereas the SLT-a99 will set at 5000.

 

[Mark K]

Wally626 wrote: Why the DXO ISO ratings are scoring the new Sony cameras so low I do not know.

That is primarily because of the way DXO measures ISO and then uses it for other measurements. And this applies differently to RX1 than to A99.

If sensor is tuned to, or by design, saturate early, it will have a lower DXO ISO. And as a result, all other measurements are impacted since DXO does not use exposure ISO for these ratings (whereas people assume they are one and the same).

It is very simple.....SLT does affect optical performance substantially

 

[Allan Olesen]

EinsteinsGhost wrote:

Allan Olesen wrote:

EinsteinsGhost wrote:

Allan Olesen wrote:

Try this thought experiment:

1. You set your camera at ISO 1600, illuminate the sensor with a given light intensity and measure the SNR.
2. You set your camera at ISO 6400, illuminate the sensor with the same light intensity and measure the SNR.
3. SNR result turns out to be the same in both situations.

How does ISO influence on the measurement if the result is the same regardless of chosen ISO?

This behaviour is the reality for most cameras.

SNR will not stay the same.

Yes, it will. And this fact is really important for your understanding of the topic.

SNR will NOT be the same at ISO 1600 compared to ISO 6400.

I don't see any point in continuing this yes/no discussion.

Take two raw photos. One with ISO 6400, "correctly" exposed. One with ISO 1600, with the same exposure (same shutter speed, aperture and scene lighting) as the first photo.

Increase brightness of the second photo 2 stops in PP (using software which doesn't make optimization behind the scenes). Disable all noise reduction for both photos. Result: They will have the same amount of noise. And they will have the same brightness = amount of signal.

So your SNR is the same in the two photos.