Why are UV highlights so tricky to expose properly?

[Doug A]

UV highlights often record non-linearly. I can take a shot and, upon review, the histogram might show exposure peaking near the middle. Add perhaps 1/3 to 2/3 steps more exposure and the histogram is close to or crashing into the right side. This is all done with EL-Nikkor closed down to the same aperture and camera on manual. Light appears the same. Is there something about UV light that causes this?

Thanks,

Doug A

[photoni]

if I understand the question... it's because the internal sensor reads the standard VIS light "narrow".
your converted camera has a very wide sensitivity... the exposure meter is limited
the only way to determine exposure is to look at the histogram after taking the photo.

[nfoto]

The white-balance enters here as well, as the histogram is from the embedded jpg, not the raw data directly.

[Andy Perrin]

Yeah, I usually exposure bracket and pick the best photo in UV. 

[dabateman]

Yes I will agree with Birna here.

When I have seen that its the whites balance thats off. Depending on your wavelength,  you are hitting a majority of blue color filter array pixels or not. The switch from yellow (red green) to blue is around 380nm.

So it can be a cliff when you hit hard into the blue zone from greater than 380nm to 400nm or not if below.

[Doug A]

@photonigood advice. I usually guesstimate the 1st exposure and adjust from there. Even adjusting from the histogram is iffy. @nfoto explains the histogram shortcomings well and points out how WB affects things. @dabateman adds extra info about WB and color channels. Lots of things I've never considered. Even though it isn't a perfect fix, I'm going to go back and review the color channel histograms. Just wish they were in Raw. @Andy Perrin has an excellent recommendation - exposure bracketing. At least digital makes bracketing affordable.

Thanks everyone for the excellent advice,

Doug A

 

[OlDoinyo]

UV scenes are often quite contrasty and the dynamic range of some cameras will  be challenged, especially if you are looking at sky and ground with vegetation at the same time. Metering can be an iffy proposition at times, especially if you are metering visible and guessing. Exposing for foliage can easily blow out the sky if you are not careful.

[ulf]

7 hours ago, nfoto said:

The white-balance enters here as well, as the histogram is from the embedded jpg, not the raw data directly.

To get a better understanding when analysing the captured RAW image you can check the histograms of the raw data.

Then you will see the source of information unaffected by any white balance.

 

RawDigger and FastRawVierwer are two good tools for that. I assume there are others that also show the RAW histograms, but am not sure about which, as I already have those two.

[Andrew Dayer]

6 hours ago, dabateman said:

Yes I will agree with Birna here.

When I have seen that its the whites balance thats off. Depending on your wavelength,  you are hitting a majority of blue color filter array pixels or not. The switch from yellow (red green) to blue is around 380nm.

So it can be a cliff when you hit hard into the blue zone from greater than 380nm to 400nm or not if below.

 

Is it established that UVR false colours correspond to discreet wavelengths? I had thought I'd seen something here saying the opposite?

 

Does the false response vary with sensor type or specific Bayer array 'models'?

 

I think I've seen different slightly colours in scene with different lens ie same subject / lighting / filters and 'same' white balance. But I don't think this is rigorous and I sometimes think that colour varies with exposure. Too subjective...

[photoni]

@Doug A For example, with the Sony A7 I have 3 fixed WB settings
one for UV ... ZWB2+TSN, one for VIS ... QB39 one for IR ... Hoya R72.
after the first shot looking at the histograms

(Birna is right, they are calculated from the Jpeg not from the RAW)

I have an idea whether it is underexposed or overexposed and I adjust the automatic exposure compensation times / ISO + or -

.

White balance changes with light, exposure, lens, etc
If the photo is overexposed even Teflon can fool you

[nfoto]

Setting the camera to b/w mode will often help in getting a good starting exposure for UV. Nikon user can apply the b/w Picture Control and other brands should have something similar.

 

Do note that the RAW file is wholly unaffected and will have its weird unprocessed colours. The b/w mode is just a helping tool as it were.

[Andrea B.]

Is it established that UVR false colours correspond to discrete wavelengths? I had thought I'd seen something here saying the opposite?

 

Some broad generalizations can be made but the correspondences are certainly not at all good enough to make any precise (or scientific) conclusions. Remember that different combinations of visible wavelengths can produce the same visible colour. So, similarly, different combinations of UV light can also produce the same false colours. The problem is compounded by the differences between sensors, filters, lighting and converter software, all of which all play a role in producing UV false colours.

---

 

Does the false [colour] response vary with sensor type or specific Bayer array 'models'? 

 

Yes, it certainly can. My converted Panasonic S1R does not produce quite the same false colours as my converted Nikon D610. And non-Bayer sensors differ from Bayer sensors in the UV response.

---

 

 

UV highlights often record non-linearly. I can take a shot and, upon review, the histogram might show exposure peaking near the middle. Add perhaps 1/3 to 2/3 steps more exposure and the histogram is close to or crashing into the right side.

 

The camera records light of any kind linearly in each sensor "well" or "pixel". What you are seeing is a non-linear, JPG version of the scene on your camera LCD. The raw data has had a "gamma curve" and a "histogram stretch" applied by the camera's JPG engine in order to produce a human-visible photograph.

---

 

 

 It might be interesting to work through an example of a photo made using BIrna's (and often, my) suggestion to shoot UV in B&W in order to make use of the JPG brightness histogram.

[Andrea B.]

I hope somebody read this. Did I get anything wrong? 

I suppose I'm feeling insecure today. 

[dabateman]

On 10/31/2023 at 1:59 PM, Andrea B. said:

I hope somebody read this. Did I get anything wrong? 

I suppose I'm feeling insecure today. 

All looks good and responded before I did.

I was going to say something similar. 

 

[Andrew Dayer]

On 10/30/2023 at 9:15 PM, Andrea B. said:

Is it established that UVR false colours correspond to discrete wavelengths? I had thought I'd seen something here saying the opposite?

 

Some broad generalizations can be made but the correspondences are certainly not at all good enough to make any precise (or scientific) conclusions. Remember that different combinations of visible wavelengths can produce the same visible colour. So, similarly, different combinations of UV light can also produce the same false colours. The problem is compounded by the differences between sensors, filters, lighting and converter software, all of which all play a role in producing UV false colours.

---

 

Does the false [colour] response vary with sensor type or specific Bayer array 'models'? 

 

Yes, it certainly can. My converted Panasonic S1R does not produce quite the same false colours as my converted Nikon D610. And non-Bayer sensors differ from Bayer sensors in the UV response.

---

 

 

 

UV highlights often record non-linearly. I can take a shot and, upon review, the histogram might show exposure peaking near the middle. Add perhaps 1/3 to 2/3 steps more exposure and the histogram is close to or crashing into the right side.

 

The camera records light of any kind linearly in each sensor "well" or "pixel". What you are seeing is a non-linear, JPG version of the scene on your camera LCD. The raw data has had a "gamma curve" and a "histogram stretch" applied by the camera's JPG engine in order to produce a human-visible photograph.

---

 

 

 It might be interesting to work through an example of a photo made using BIrna's (and often, my) suggestion to shoot UV in B&W in order to make use of the JPG brightness histogram.

 

Thanks @Andrea B.. Off @Doug A's original topic so I won't expand on your answers to my queries about colours and UV wavelengths beyond saying my path has now lead to Joylon Troscianko's work and to calibration, cone models tools and more here:  http://www.empiricalimaging.com/. Includes data for UV capable DSLR's.

Only previous mention I can find here is in this thread:

 

 

[Andrea B.]

I will try to look through the Troscianko material as time permits. Thanks for the link !!