A900 ISO 320: Part II

[douglasf13]

So, I decided to start a new thread in regards to the A900's ISO 200 vs. ISO 320 questions (since the last was so long,) and below are a few of my own examples. Andrey's examples in the last thread, http://forum.getdpi.com/forum/showpost.php?p=103668&postcount=50, showed the difference in deep shadows between the two settings, and mine to seem to show similar results.

Since some questioned Andrey on using the same exposure for his test, I shot two scenes. In the first scene, the ISO 200 has approximately 2/3EV more light hitting the sensor than the ISO 320 shot. In the second scene, both ISOs are exposed identically, like Andrey's example (I do agree with Andrey that testing the second way makes more sense.) Now, these shots are not nearly controlled as much as Andrey's, being handheld with windowlight, but I took the pics within seconds of each other, and I don't believe that it should affect the outcome much. Had I known I was going to show these to other people, I would have brought out the tripod.

These were processed in LR3 b2 with everything zeroed out except exposure and brightness. All NR and sharpening were turned off, too, and a linear tone curve was used. The links to the RAWs are provided near the bottom of the post so you can test with your converter.


Example 1:

These pics need about 4 or so stops of exposure compensation in the converter. Yeah, I know that's pretty underexposed, but the point is to demonstrate what bringing up deep shadows would look like.

ISO 200 at f4 1/350, 20% of original size (red box indicates approximate crops)

ISO 320 at f4 1/500, 20% of original size

Crop 1, ISO 200 (crops are 100%)

Crop 1, ISO 320

Crop 2, ISO 200

Crop 2, ISO 320

Crop 3, ISO 200

Crop 3, ISO 320

Notice how all three of the ISO 200 crops exhibit purple color shift and blotching. Sure, the ISO 320 is noisier, but that's to be expected...especially since it received 2/3 EV less light to the sensor.



Example 2:

These pics are ridiculously underexposed in order to exacerbate the issue. It took 5 or so stops of EV compensation to get them to this level, so this is just to show what's going on deep in the trenches. I don't even need to post crops, because the difference is so obvious.

ISO 200 at f2.8 1/125, 20% of original size

ISO 320 at f2.8 1/125, 20% of original size

Clearly, there is color/blotching issues all over the place in the ISO 200 shot. The rug in the hallway (top right) is purple!


RAWs for all four files found here:

http://www.mediafire.com/?sharekey=7d1e99da3f2d1bb608f8df73f2072ed68b25338b297f6d31ea4ac78345cbe4ce

http://www.mediafire.com/?emzwjr3mhmw

http://www.mediafire.com/?wjib0onn0td


Overall, I don't think my sentiment between ISO 200 vs. ISO 320 has changed. Although ISO 320 does have a bit more noise overall than ISO 200, ISO 200 really messes up deep shadows. So, if capturing a large dynamic range and boosting dark shadows is a part of your workflow, ISO 320 seems the way to go.

What do you think? I'd love to see more examples posted.

 

[hardloaf]

Ok, I couldn't look at that LR noise and green shadows So I've processed both 200 and 320 shots in RPP and here is small 100% screenshot. Should mention that they also underexposed for about 4 stops, top is 200, bottom 320.

 

[douglasf13]

Thanks, I was hoping you would. Since I'm new to RPP, I figured I should let you handle processing the raws in that converter What's with the green in the LR3 version?

The purple blobbing is really obvious!


p.s. i did mention that they were around 4 stops underexposed, i believe.

 

[hardloaf]

Thanks, I was hoping you would. Since I'm new to RPP, I figured I should let you handle processing the raws in that converter What's with the green in the LR3 version?

The purple blobbing is really obvious!


p.s. i did mention that they were around 4 stops underexposed, i believe.

You did, but it wasn't clear to me if all of them or only last 2 were underexposed

In any case practical meaning of all this is that if you shoot 250 and lower don't count on anything below 7 stops with spot on exposure - it's going to be blotchy, clipped and ugly colored. On 320 you can go couple of stops lower and still have something remotely usable.
Considering that meters of all modern DSLRs are underexposing for 0.7 stop at least, including A900, and that people tend to underexpose even more, add victims of ETTR to the risk group - they can easily get in troubles even below 5 1/2 - 6 stops.

 

[Engreeks1]

In any case practical meaning of all this is that if you shoot 250 and lower don't count on anything below 7 stops with spot on exposure - it's going to be blotchy, clipped and ugly colored.

On 320 you can go couple of stops lower and still have something remotely usable.
Considering that meters of all modern DSLRs are underexposing for 0.7 stop at least, including A900, and that people tend to underexpose even more, add victims of ETTR to the risk group - they can easily get in troubles even below 5 1/2 - 6 stops.

Hi Hardloaf - what do you mean when you say "don't count on anything below 7 stops"? You mean don't count on being able to get clean details from parts of the images that are 7 stops below middle? What's middle?

Also I people using ETTR are over exposing, not under exposing so I'm not following your last paragraph about ETTR.

Thanks!

 

[hardloaf]

Hi Hardloaf - what do you mean when you say "don't count on anything below 7 stops"? You mean don't count on being able to get clean details from parts of the images that are 7 stops below middle? What's middle?

I'm counting from saturation point, i.e. 7 stops of total usable dynamic range.

Also I people using ETTR are over exposing, not under exposing so I'm not following your last paragraph about ETTR.

Thanks!

When people using ETTR they either over or under exposing. With this approach it's almost impossible to expose properly on base and close ISOs. The most typical problem is 'save too much of highlights', i.e. underexpose critical areas - that's what I meant by victims. Moving important areas too close to the saturation point, overexposing, happens far less frequently, it's not as bad in this regard and may even save you some shadows.

 

[Bill_Green]

Also I people using ETTR are over exposing, not under exposing so I'm not following your last paragraph about ETTR.

Thanks!

Hopefully Andrey will jump in if I get this wrong. ETTR is an acronym that gets tossed around a lot but means different things to different people. One meaning, and I think this is what Andrey is talking about, is you slide the histogram to the right, but make sure not to clip anything. The result is you protect the highlights at the expense of everything else. What you are talking about, and I think what most people think of when they talk about ETTR is moving the histogram to the right for the sole purpose of protecting the shadows. This can be an effective technique, but it comes at the risk of oversaturating colors on the highlight end.

 

[douglasf13]

Well put, Bill, although I believe colors actually get washed out more when midtones are exposed too far to the right.

I think, ultimately, most people view ETTR as "exposing as far right as possible without clipping highlights, because it makes the image less noisy." Like you mentioned, using that technique without being careful often leads to underexposed midtones ( because one may be trying not to overexpose something like chrome reflections, etc,) or overexposed midtones with bad color. Ultimately, it seems Andrey's advice to pretty much ignore ETTR with most modern cameras seems to be on point.

 

[hardloaf]

Hopefully Andrey will jump in if I get this wrong. ETTR is an acronym that gets tossed around a lot but means different things to different people. One meaning, and I think this is what Andrey is talking about, is you slide the histogram to the right, but make sure not to clip anything. The result is you protect the highlights at the expense of everything else. What you are talking about, and I think what most people think of when they talk about ETTR is moving the histogram to the right for the sole purpose of protecting the shadows. This can be an effective technique, but it comes at the risk of oversaturating colors on the highlight end.

Bill, basic idea of ETTR was in using the whole sensor range for picture and keeping it as "to the right" as possible, i.e. avoiding blank space on the right side of histogram. People usually take it as 'don't clip at any cost, but keep it as close to the clipping as you can'. The whole concept though is completely wrong because it shifts photographer's attention from primary subject of picture and it's placement on the sensor range to brightest part of the scene. As a result with ETTR in most of cases you'll end up placing critical parts of image exactly where you shouldn't, i.e. in highlights affected by flare and get washed colors and no details or in shadows producing noisy picture and preserving irrelevant highlights.
As they say - "For every human problem, there is a neat, simple solution; and it is always wrong." ETTR is exactly the case Exposing should be mindful, trusting your camera meter works better, if you understand what is actually going on is perfect.

 

[hornblade]

Thanks for posting these shots, definitely helps illustrate the 320 vs 200 difference. It's been a source of confusion for me, so I really appreciate these examples.

I'm still a little confused though about the right way to apply this knowledge to my shooting. I focus on landscapes, and deal with high dynamic range scenes either by using graduated ND filters or bracketing and blending multiple exposures. Consequently, while processing, I don't often feel the need to pull up the shadows. Either the shadows are supposed to be dark and I can leave them as is, or I have another exposure to blend in where that part of the image is correctly exposed.

My question is -- does the blotchiness only crop up when post processing and pushing exposure like this? I haven't noticed anything like this in my ISO 200 shots, but it might be because I don't bump up the exposure in this way. I realize that in this case the post processing is done to illustrate the problems w/ the originally captured data, but if this isn't regularly encountered when processing images, I'd much rather be able to use ISO 200 worry free.

 

[hardloaf]

My question is -- does the blotchiness only crop up when post processing and pushing exposure like this? I haven't noticed anything like this in my ISO 200 shots, but it might be because I don't bump up the exposure in this way. I realize that in this case the post processing is done to illustrate the problems w/ the originally captured data, but if this isn't regularly encountered when processing images, I'd much rather be able to use ISO 200 worry free.

If you have deep shadows on a picture the blotches are going to be there on ISO 200. To decide if the whole problem affects your shooting style just make couple of properly exposed shots with nice deep shadows with some details there on 200 and 320 and process them the same way as you would normally do. This way you'll have two shots to compare and you'll see if there is anything to worry about. The problem may pop up as mis-coloration for example - that's what I usually see, kind of like dirty looking brown shadows.
Hope this helps

Andrey

 

[vtube]

The problem may pop up as mis-coloration for example - that's what I usually see, kind of like dirty looking brown shadows.

Is there any difference between cRAW and RAW?

Need to test myselft, but maybe blotches are related to raw compression.

--
Andrea Olivotto
http://www.andreaolivotto.com

 

[hardloaf]

Is there any difference between cRAW and RAW?

Need to test myselft, but maybe blotches are related to raw compression.

--
Andrea Olivotto
http://www.andreaolivotto.com

There is a difference between cRaw and plain Raw - cRaw is a lossy compression and I never use it simply because I don't want any additional processing steps in camera.
Regarding blotches - they have nothing to do with compression, you'll have them in both modes.

Regards,

 

[Dave_Anderson]

Has anyone taken a second look at this issue, directly comparing firmware v1 vs v2? There are bits of information cropping up that would seem to indicate that Sony has changed the game a bit, for example this blog post. That's not really a controlled test and I didn't think to investigate before/after myself since there was no indication from Sony that the update would impact IQ.

 

[pegelli]

OK, something of a controlled test with the A850 and my old but trusty 50/1.7 at f4.

All converted with LR 3.3 at my default settings, WB daylight and blacks at 0 (so no noise hidden in the pitch black :grin. Firmware on the camera is V2 (never did this with V1, so cannot compare). All are 1:1 crops (so 1 camera pixel = 1 screen pixel)

Btw, I repeated the test, at the same distance and settings, with my A700:

With this test I didn't see the shadow blotchiness at ISO 200 in either camera, and the absolute noise is following normal expectations (i.e less exposure results in more noise). Maybe V2 fixed the problem, or doing this test at -3EV isn't "dark" enough yet.

Comments welcome, as I'm trying to learn as much as possible from continueing this discussion.

 

[edwardkaraa]

Has anyone taken a second look at this issue, directly comparing firmware v1 vs v2? There are bits of information cropping up that would seem to indicate that Sony has changed the game a bit, for example this blog post. That's not really a controlled test and I didn't think to investigate before/after myself since there was no indication from Sony that the update would impact IQ.

There may be a difference in the way the camera handles noise in jpgs, more chroma noise reduction and less luminance noise reduction, but I'm almost sure the raw files are the same. I can't see any difference before and after.

 

[philip_pj]

I would like to ask: how frequently do A900/A700 users pull up *3 plus full stops* in post processing, and should one have expectations that differ from the images shown here? That seems totally at odds with standard photographic practice.

 

[douglasf13]

In practice, it isn't neccesarily about pulling up entire images 3 stops. It is about exposing scenes with large dynamic ranges and pulling up shadow detail. So, in essence, it is about maximizing the total usable DR of the camera.

Granted, we're getting closer and closer to the point of ISO-less cameras, so the shooter can leave the camera at a single ISO and any gain can be done in the raw converter.

 

[Dave_Anderson]

Thanks Pegelli -- I could have done that but didn't see much point without the "before".

Ed, I'm not seeing any real differences myself(yet) but I've been busy with other things and not shooting much in the last couple of weeks. If there is a change only to the JPEG engine(the blog post above doesn't say, RAW or JPEG) then it's certain I'll never benefit.

Philip, the +3 adjustments above are intended to take a subtle problem and "bring it into the light", so to speak. IMHO it's not at all unusual to apply a stop or two of "fill light" and at these levels the blotchiness as seen in the second example in the OP can be a real problem. When this happens there is little choice other than limit the DR of the output, and sometimes that's the difference between a keeper and trash.

Douglas, if RAW were truly RAW we might be there already. The notion that more light on the sensor(200 vs 320) results in poorer shadow detail is a real head-scratcher.

 

[douglasf13]

...The notion that more light on the sensor(200 vs 320) results in poorer shadow detail is a real head-scratcher.

From what I understand, it actually isn't about the amount of light, but, rather, how the ADCs respond to a bit more gain.