AWB vs. UniWB experiment: the data

....but I hope it fits your thread:

Exposure technique: A mode, combination of LV blinkies and post-shot histograms. Using EC, back off onset of LV blinkies one notch, shoot, check post-shot histograms. If histograms show pile on r.h. edge, reshoot down 1/3 EV. Go further, if necessary (usually not). Stop when histograms look good.

Number of shots: 2 for each.

Default processed jpegs and RawDigger histograms:

Evaluation of exposure: (-1/3 to -1/2)

Shot indoors including window on partly sunny day. From the linear histograms, comparing corresponding peaks in the green channel, the AutoWB is down 1/3 EV in the upper regions and 1/2 EV in the shadows.

Processed jpegs and evaluation:

I adjusted them in ACR. I opened one in PS and the other in Bridge so that I could have them both on the screen at the same time. I attempted to equalize the two as much as possible. I began with the AutoWB shot and did my thing with it. I then attempted to obtain the same 'thing' with the UniWB shot.

I used the AutoWB temperature/tint to set WB for the UniWB shot. I picked various points (highlights, mids, shadows) to equalize exposure and brightness. Where possible, I made equal adjustments but made individual adjustments where warranted. I attempted also to make the histograms of the two shots look as similar as possible.

Unlike the other pairs, no matter what I did I was unable to make the AutoWB shot look as good (at least in my estimation) as the UniWB shot. For some reason, it simply lacks "body." The WB adjustment produced better color in the "indoor" area than the AutoWB, and the shadows are better.
--
gollywop

Chris Noble wrote:

Thank you Jonas, very interesting. I will do the same test this weekend and post it. We should all add it to our "get to know your camera" routine!

Thank you Chris.
I'm sorry for forgetting about the topic title rule.

What type of camera?

This was with the E-M5

I note the halogen light. Can you repeat in daylight?

Yes. But it would be of little use. Moment... Yes, with the same setup but in daylight I get about the same exposure values from both the WB settings; like 1/20 for AutoWB and 1/15 for UniWB. The difference, around half a stop, is to small for me to really care about in most cases.

My summary, tell me if I am wrong:

You exposed 0.5 EV below blinkies in both cases; AWB was 1.25 EV below saturation, UniWB was 0.5 EV below saturation (Green channel dominant). That suggests that your in-camera UniWB JPGs match Raw very closely? Also 2/3 EV "improvement" going from AWB to UniWB using same exposure technique? I mention that because my "default" starting point with my G1 is UniWB, + 2/3 EV compensation, after a lot of trial-error. Maybe just a coincidence.

You got it right. As I have my camera set to half stops between every step (shutter speed and aperture) and I know I get good images that way I think about the difference as 1 stop. [b]Minor[ b], or partly, blinkies from areas i want not to be burned are OK.

With my method of "measuring" we can conclude 2/3 of a stop only.

Everything matches close enough to make me satisfied.

In daylight the difference between UniWB and AutoWB or DaylightWB decreases; you get the most out of UniWB in candlelight, or from bulbs or halogens (or, of course, when the light is mixed).

I leave the camera at UniWB all the time. There is no need to fiddle around with the WB setting. Here is a sample:

In some ways the image of the old caravan above is typical; mixed light, I didn't want the bright pars to be burned out, I wanted some details in the shadow and I think that is more important the bigger you print. So the camera is set to UniWB and blinkies and the, for me, correct exposure is set in a second.

The blinkies also make it possible to shoot wherever there is a lot of contrast. It's like having an automatically warning spot meter all over the image. Very handy.

Thank you for starting this thread and also for the kindness not saying my sample is boring... !

regards,

Jonas

Exposure technique: A mode, combination of LV blinkies and post-shot histograms. Using EC, back off onset of LV blinkies one notch, shoot, check post-shot histograms. If histograms show pile on r.h. edge, reshoot down 1/3 EV. Go further, if necessary (usually not). Stop when histograms look good.

Number of shots: 2 for each.

Default processed jpegs and RawDigger histograms:

Evaluation of exposure: (-2/3 EV)

Shot completely indoors on partly sunny day. Hard to judge upper regions from linear, but comparing shoulders of Logged histograms, AutoWB would seem to be down 2/3 EV. From the linear histograms, AutoWB would appear to be down 1/2 to 2/3 EV depending on channel.

Processed jpegs and evaluation:

I adjusted them in ACR. I opened one in PS and the other in Bridge so that I could have them both on the screen at the same time. I attempted to equalize the two as much as possible. I began with the AutoWB shot and did my thing with it. I then attempted to obtain the same 'thing' with the UniWB shot.

I used the AutoWB temperature/tint to set WB for the UniWB shot. I picked various points (highlights, mids, shadows) to equalize exposure and brightness. Where possible, I made equal adjustments but made individual adjustments where warranted. I attempted also to make the histograms of the two shots look as similar as possible.

These two are very similar (other than the fact that the cat moved his head between shots). The UniWB shot has slightly better shadow detail and structure, but the 2/3 EV difference produces nothing to get excited about.
--
gollywop

Exposure technique: A mode, combination of LV blinkies and post-shot histograms. Using EC, back off onset of LV blinkies one notch, shoot, check post-shot histograms. If histograms show pile on r.h. edge, reshoot down 1/3 EV. Go further, if necessary (usually not). Stop when histograms look good.

Number of shots: 1 shot for UniWB, 2 for AutoWB.

Default processed jpegs and RawDigger histograms:

Evaluation of exposure: (-2/3 EV)

Shot outdoors, partly sunny day. Results are clear as a bell: AutoWB is down 2/3 EV across the board.

Processed jpegs and evaluation:

I adjusted them in ACR. I opened one in PS and the other in Bridge so that I could have them both on the screen at the same time. I attempted to equalize the two as much as possible. I began with the AutoWB shot and did my thing with it. I then attempted to obtain the same 'thing' with the UniWB shot.

I used the AutoWB temperature/tint to set WB for the UniWB shot. I picked various points (highlights, mids, shadows) to equalize exposure and brightness. Where possible, I made equal adjustments but made individual adjustments where warranted. I attempted also to make the histograms of the two shots look as similar as possible.

Here the differences in the shadows are clearly in favor of UniWB. In general, this is not 'startling.' But the differences at the base of the foliage, particularly just above the corner of the stone wall, are pretty strong.
--
gollywop

Exposure technique: A mode, I began by trying to use the LV blinkies, but it brecame immediately clear that they were going to be misleading for this scene and lead to significant overexposure. So I began each shot with EC at 0, and checked the post-shot histograms. I then altered EC as necessary (downward in this case) till I got acceptable post-shot histograms.

Number of shots: 1 shot for UniWB, 2 for AutoWB.

Default processed jpegs and RawDigger histograms:

Evaluation of exposure: (-1/3 EV)

indoors at night, incandescent light. AutoWB is down 1/3 in green highlights, down 1/2 in red mids, and a little over 1/2 in the blue shaddows.

Processed jpegs and evaluation:

I adjusted them in ACR. I opened one in PS and the other in Bridge so that I could have them both on the screen at the same time. I attempted to equalize the two as much as possible. I began with the AutoWB shot and did my thing with it. I then attempted to obtain the same 'thing' with the UniWB shot.

I used the AutoWB temperature/tint to set WB for the UniWB shot. I picked various points (highlights, mids, shadows) to equalize exposure and brightness. Where possible, I made equal adjustments but made individual adjustments where warranted. I attempted also to make the histograms of the two shots look as similar as possible.

These final jpegs are very similar. The UniWB shot has slightly better detail in the wood grains and imperfections. It also has slightly better shadow detail.

Exposure technique: M mode, I began by trying to use the LV blinkies with A mode, but with AutoWB the EC bottomed out at -3 and the blinkies kept going. So I switched to M mode at the same aperture, but increased shutter speed till the blinkies stopped. I shot and checked post-shot histograms. I continued to increase shutter speed till the post-shot histograms were appropriate.

Number of shots: 3 shots for UniWB, 2 for AutoWB. Note: In practice I would most likely have gotten the UniWB shot in 2 tries. Once I saw the opening post-shot histogram I knew I'd want two reductions, but, because of this test, I played it conservatively.

Default processed jpegs and RawDigger histograms:

Evaluation of exposure: (-1/3 EV)

indoors at night, shot to include part of light. AutoWB down 1/3 EV across the board.

Processed jpegs and evaluation:

I adjusted them in ACR. I opened one in PS and the other in Bridge so that I could have them both on the screen at the same time. I attempted to equalize the two as much as possible. I began with the AutoWB shot and did my thing with it. I then attempted to obtain the same 'thing' with the UniWB shot.

I used the AutoWB temperature/tint to set WB for the UniWB shot. I picked various points (highlights, mids, shadows) to equalize exposure and brightness. Where possible, I made equal adjustments but made individual adjustments where warranted. I attempted also to make the histograms of the two shots look as similar as possible.

I was surprised how close these two shots were. Again you can see better shadow detail with the UniWB shot, but it's not an overwhelming difference. The AutoWB shot also has more chroma noise, which is best seen in the creamy bottom to the small milk jug. The only way you can really compare these shots is to download them and put them either in ACR/LR or PS together. Since they were taken with a tripod, the a/b toggling will be good.
--
gollywop

Chris Noble wrote:

Can you blow up a deep shadow area so we can compare the noise?

Chris, please, enough! I've already spent too much time on all this. Please have Vlad PM me the instructions for the drop box, and I'll put the ORFs in there for anyone to examine.

You make a claim to the forum at large, but are unwilling to provide a simple crop? And you expect people to download full-res files? Strange.

gollywop wrote:

Chris Noble wrote:

Can you blow up a deep shadow area so we can compare the noise?

Chris, please, enough! I've already spent too much time on all this. Please have Vlad PM me the instructions for the drop box, and I'll put the ORFs in there for anyone to examine.

Vlad S wrote:

gollywop wrote:

Here the differences in the shadows are clearly in favor of UniWB. In general, this is not 'startling.' But the differences at the base of the foliage, particularly just above the corner of the stone wall, are pretty strong.

I still would like to see the 100% crops of the area that you are talking about, because I can't see the difference in the scaled down versions. And yes, I do have a calibrated, true 8 bit, IPS monitor.

Ok, I relented, but just for these. I'm on vacation, and this is becoming like a job.

Chris Noble wrote:

Can you blow up a deep shadow area so we can compare the noise?

As I said to Vlad, I'll do these, but please, no more.

Vlad S wrote:

gollywop wrote:

Ok, I relented, but just for these. I'm on vacation, and this is becoming like a job.

Thank you for posting these. I appreciate your taking time from your vacation (no sarcasm), but I guess your passion keeps you in this forum...

I can barely tell that the images are ever so slightly different, but I can't really pick which one is better. I think that UniWB has a little more contrast in the shadows, but the AutoWB has a little more detail. But if I did not see the file names I really would not be able to tell which is which.

Anyone else care to comment on the differences between these crops?

The best thing you can do is download both images (largest size) into ACR or LR. Then you can a/b them with the up/down arrows at full screen (press space bar).