• Ultraviolet Photography
  •  

Lens and camera tests with sparticle and monochromator

60 replies to this topic

#1 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 27 July 2018 - 17:25

Have some time now to get back to some earlier explorations that I didn't have chance to post before.

First my work with a monochromator looking at camera response and then modelling predicted false colours. Then some lens tests with sparticle.

METHODS
The camera used was a Nikon d3200 full spectrum conversion (UVIRCameras). Response was measured using a 900 W xenon arc lamp and monochromator [Applied Photophysics clinical photoirradiator system] whose output was connected directly to the camera via a water filled light guide. Images were taken with a Schott BG40 2mm visual pass / IR block filter in place using a pinhole to remove any lens effects and then with a Nikkor EL 80mm f/5.6 (old metal style) enlarger lens. The bandwidth was set at 10 nm and images taken between 300-500 nm in 10 nm steps and 500-700nm in 50 nm steps. RAW format images were imported in Rawdigger software and the average RGB (Raw values prior to white balance) values noted over a 1500 pixel square centred over the light guide image. Exposure was adjusted to give high counts whilst avoiding overflow in any channel. Image RGB channel counts were normalised for camera exposure values and incident irradiance. The irradiance was measured using a wavelength insensitive thermopile meter (dexter 2M) which add previously been calibrated with reference to a standard thermopile traceable to the UK NPL. The wavelength calibration of the monochromator had previously been checked by measurement using a double diffraction grating spectroradiometer [Bentham], in turn calibrated with reference to the characteristic lines of a mercury vapour lamp.

The spectral response was modified to model the effect of two commonly used filters, Schott UG1 2mm and UG5 1.5 mm based on transmission values provided via Schott’s filter tool

The effect of imaging in sunlight was modelled by multiplying the response curve by a solar spectrum model for the UK obtained from (https://www2.pvlight....au/calculators)

The effect of “white balancing” was modelled by scaling the RGB channel responses so that they each had the same total response summed over all wavelengths.

False colours were generated base on this model by summing RGB values individually over 10 nm wavebands from 340 to 410 NM and displaying then using the cell fill style functionin Excel. This was saved as an image and imported into Lightroom for further adjustment.

RESULTS

Raw camera response with modelled sunligh and filter is shown below.

Modelling sunlight illumination and a UG1 2mm filter indicated that radiation less than 340 and 350 nm accounts for about than 1% and 5% respectively of the total camera response. Modelling a source with uniform output between 300-400 nm increases the cumulative sensitivity only slightly at 350 nm to 7%.

Attached Image: raw response pinhole.jpg

The modelled response after white balancing and using EL Nikkor 80 mm lens (Old metal style) is shown below:

Attached Image: EL nikkor response after WB.jpg

Using this response model the following false colours are predicted:

Attached Image: Capture modelled false colours 2.JPG

Which appear as below after auto tone adjustment in Lightroom:

Attached Image: false colour predcited nikkor el sun WB UG1 2 after LR auto tone.jpg

DISCUSSION

1. It appears that below 340 nm the overall system response is dominated by the Bayer transmission. Therefore there is little to be gained from using lenses with significant transmission below this. This might not be the case for certain specific applications.

2. The model false colours are in keeping broadly with what is seen in UV phtography. In broad terms this system produces images responding to radiation in the range 340-410nm, showing as false blue for the longer wavelengths and false yellow (equal red and green) for the shorter wavelengths. In the cross over between this two bands a false grey is produced where response in all channels is roughly equal. To test this more precisely some sparticle images would be helpful (imgages of sunlight transmission through a series of narrow band pass filters).

I have also used the model in conjunction with published flower reflectance data to successfully predict false colours from several common flowers

I am working on relating the model response to colour responses in bees and birds

Edited by Jim Lloyd, 27 July 2018 - 17:27.


#2 JMC

    Member

  • Members(+)
  • 717 posts
  • Location: London, UK

Posted 27 July 2018 - 17:46

Thanks for sharing Jim. Can I just check something though - in the graph it says 'g total' for the green. Are you adding the two green channels together (which is what i read when I see 'g total'), or are you taking an average of them?

#3 Andy Perrin

    Member

  • Members(+)
  • 2,258 posts
  • Location: United States

Posted 27 July 2018 - 17:54

Nice! It would be interestiing if you and Jonathan could repeat exactly the same test with your respective cameras (including same filter stack). One thing I've always wondered about is differences between camera models.

Edited by Andy Perrin, 27 July 2018 - 17:55.


#4 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 27 July 2018 - 18:04

The next part of this work is about the sparticles. This is very much work in progres. Jonathan very kindly ordered two set of filters for his sparticle so my set is identical to his (I think he measured both to confirm) See his post here for details of the filter transmissions.

EDIT - forgot some crucial info! - D3200 full spectrum modified + UG1 2mm + BG40 2mm

This was something of a lashup and I would do it differently next time. Anyway the filters are arranged in two row - top row (Right to left ) Teflon tape, 405,404,396,382 - bottom row (right to left) 364,355,341,321,303

I found some problems with reflections from the back side of the filters (non-shiny side towards the camera) This was mostly cured by putting the camera the sparticle board and camera at either end of a cardboard tube (a very crude version of Jonathan's beautiful construction).

I pointed this towards the sky, avoiding direct sun. The tests below were done in two sessions and there was some variable cloud so it is difficult to compare absolute values directly. Also for all lenses I had the camera at the same distance and images were cropped afterwards to roughly the same size. WB was done in Lightroom, first using a calibration file which gives a result in the right ball park and then using the WB tool on the teflon. (

I wasn't sure the best way to normalise images to compare. I started by experimenting with a range of settings so that I just started to overflow in the brightest part of the image. ISO=800, t=1 sec, f/4 (except for a few lenses f/5.6). Then in processing I made adjustments to the exposure so that the 396 filter was just clipping.

This is intended as a broad first pass. Then I plan to improve the sparticle mounting and focus on a smaller number of lenses. Also I noted that better colours are obtained if I use Photninja, but I don't have a full licence for that and am not familiar with that - so for now I will stick with Lr, but in future will probably use Pn

Here are the results (these are all my lenses):

Attached Image: Sparticle lens tests.jpg

Comments:

I don't think there are any major surprises here.
  • EL Nikkor 80 mm old is best - the newer version not bad, but I have problems with light leak so might sell that one
  • Nikon E 50 mm is better than I thought it would be, although I think it is known good
  • Nikon E 100mm not suitable
  • All the 35mm f/3.5 I have are good - I need to rationalize these (don't need so many versions of the same thing)
  • The M42 - F adapter with lens reduces transmission, but not too bad - something to explore more
  • The 28mm lenses are not very good - maybe need to look at this again as I though the Soligor was OK from flower photos - but not sure it offers much over 35mm
  • Samuron lens 135mm no good for UV, and I haven't used very much, but few test images are nice. Can find very little about this lens on internet.
  • Nikkor 24 mm and kit lens not good - as expected

Edited by Jim Lloyd, 27 July 2018 - 18:16.


#5 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 27 July 2018 - 18:13

Jonathan - yes I added the two G channels in that graph. I figured it didn't really matter at that stage as that was before WB. Once WB is done it doesn't make any difference if you started with the G total or average.

Thanks Andy - We have had a few discussions - certainly that would be a good plan.

I have done some searching to see if this kind of result has been published already. I didn't find much, but came across this website and this paper - this suggests that he has measured camera, lens and filter response for a wide range of systems. I found this rather suprising as I got the impression from here that the Bayer response are not that well known in the UV.

#6 Andy Perrin

    Member

  • Members(+)
  • 2,258 posts
  • Location: United States

Posted 27 July 2018 - 18:18

Quote

Anyway the filters are arranged in two row - top row (Right to left ) Teflon tape, 405,404,396,382 - bottom row (right to left) 364,355,341,321,303

Jim, do you have the direction of the filters mislabeled? The teflon tape must be the white, but it's on the left...

#7 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 27 July 2018 - 18:39

Hi Andy, no the teflon comes out grey and is on the right. Its actually 2 or 3 layers of tape to make it a bit darker as sometime its difficult to white balance if its too bright. The white circle on the left is the 382 filter. The whiteness is somewhat exaggerated in the way I have processed these and the WB in lightroom.

Just to demonstrate I have WB this example in Photoninja and then reduced the exposure and lightness and increased saturation:

Attached Image: Capture PN sparticle.JPG

The circle on the left looks more blue now.

However, If you look at my monochromator results (and model) you will see that there is a wavelength around 380 nm where the RGB responses are equal after WB. So I think that a grey spectral response is predicted. Of course we also see a metameric grey (folliage for e.g) in our images due to a broad spectrum response. But I really don't want to get too much into all of that!

Edited by Jim Lloyd, 27 July 2018 - 18:40.


#8 Andy Perrin

    Member

  • Members(+)
  • 2,258 posts
  • Location: United States

Posted 27 July 2018 - 19:21

I guess if you have a blue response on one side and a yellow on the other than it has to be gray in the middle somewhere (if the curves are continuous).

#9 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 28 July 2018 - 12:12

Yes, although I haven’t seen grey on many sparticle images on here

Ideally of course one wouldn’t have the red and green channels overlapping as much ...

#10 Andy Perrin

    Member

  • Members(+)
  • 2,258 posts
  • Location: United States

Posted 28 July 2018 - 14:03

This is why it would be interesting to see if Jonathan gets the same results with the same processing. It could be a camera effect.

#11 JMC

    Member

  • Members(+)
  • 717 posts
  • Location: London, UK

Posted 28 July 2018 - 14:23

The plan is for us to get together at some point and do some comparison tests. At the moment, the issue is I don't have the filters or the lens that Jim has. So, too many variables.

I did do an experiment a while back, comparing a Canon and Nikon camera, both with the same ACS UV filter and lens, using my monochromator setup. Looking at the responses from the red, green and blue channels, there was a slight difference between the 2 cameras, so I certainly expect a difference in how they render colour in the UV, and how they respond to white balancing. I have yet to see a grey response at 380nm though.

I'll see if I can get some white balanced sparticle shots with a couple of different UV filters and see if I can replicate the effect.

#12 JMC

    Member

  • Members(+)
  • 717 posts
  • Location: London, UK

Posted 28 July 2018 - 15:16

Bit of a write and dash this one, as I am heading out in a minute. If I forget anything or it isn't clear, let me know and I'll sort it.

Experiment setup. Sunny day but with clouds moving passed. Sparticle photos (aimed at a patch of blue sky), Canon EOS 5DSR multispectral mod, Rayfact 105mm UV lens, ISO400, 2", f5.6. 3 filters; Baader U, SEU MkII, LaLaU. White balance image for each filter was captured using a PTFE tile. Images then white balanced in DarkTable for each filter individually. Sparticle images saved from Darktable after white balance, then cropped in photoshop with no further processing.

The 382nm filter, where Jim would predict grey, is on the right hand end of the middle row.

This is what they look like.

Baader U
Attached Image: 1R4A8838 Baader U WB cropped.jpg

SEU mkII
Attached Image: 1R4A8841 SEU II WB cropped.jpg

LaLa U
Attached Image: 1R4A8845 LaLaU WB cropped.jpg

I'm not seeing what I would think of as grey at 380nm (actually 382nm peak transmission measured through the filter).

#13 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 28 July 2018 - 15:50

I didn’t actually say that I would predict grey through the 382 band pass filter, but that there will be some single wavelength near here where Rgb responses are equal giving grey . And as I said before my images are close to being overexposed at this wavelength as I was using this as a reference to compare all the different lenses as a quick test to select suitable ones -based on relative transmission into uv only. So any colour is rather washed out

Not sure Jonathan why the individual filters in your images look tonally quite similar ? Did you adjust tones ?

Edited by Jim Lloyd, 28 July 2018 - 15:52.


#14 JMC

    Member

  • Members(+)
  • 717 posts
  • Location: London, UK

Posted 28 July 2018 - 16:02

Jim, oh right, I must have misunderstood then about the grey at 380.

No adjustment other than setting white balance and tint in Dark Table.



#15 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 28 July 2018 - 16:19

Jonathan - if you increase the exposure in processing so it’s just clipping does it look a bit more like mine ?

#16 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 28 July 2018 - 17:21

Looking back at the last time I reviewed my different lenses by photographing colt's foot and snow drops here (reproduced below), I am reassured that in broad terms I come to the same conclusions based on the sparticle tests (same camera and filters used on both occasions, D3200+UG1 2mm + BG40 2mm,:

Attached Image: Capture multi lens comp.JPG

#17 Cadmium

    Member

  • Members(+)
  • 2,606 posts

Posted 28 July 2018 - 19:42

Jonathan, Cool 3 x 3 layout.
This was using Canon 199A flash for backlight. Baader U on lens.
Attached Image: Sparticle_199A_Diffuser_Kuri_35_1020.jpg

Edited by Cadmium, 28 July 2018 - 19:43.


#18 Cadmium

    Member

  • Members(+)
  • 2,606 posts

Posted 29 July 2018 - 04:40

Jonathan, What are the BP filters you are using in your test? ###BP## of each?

#19 dabateman

    Da Bateman

  • Members(+)
  • 887 posts
  • Location: Maryland

Posted 29 July 2018 - 05:03

@ Cadmium,
Jonathan has the details here:
http://www.ultraviol...dpost__p__22555

Very cool white balance test and an interesting way to use the sparticle test.
This actually could be a good way to characterize different sensors. The slight peak changes between your Nikon and Canon is hard to argue. But a white balance shift with same setup, would be better to hammer down different dye used and relative sensitivity. There will also be differences between models as different dyes have been used for different Nikons that I know.
The 355 and 364 are very greenish. I am used to seeing stronger yellow here with Nikon and my Olympus.

#20 Jim Lloyd

    Member

  • Members(+)
  • 300 posts

Posted 29 July 2018 - 09:10

I would argue that the monochromator method would be best to compare camera responses and I hope that Jonathan and I can get together some time to do this on different cameras.

This is meant to be the main focus of this thread with the sparticles being used eventually as a way to corroborate the monochromator measurements and modeled response

This is a continuation of what started here

To recap on my overall objectives:

1. to produce continuous analytical curves of raw RGB channel relative responses versus wavelength - independent of source, lens, filter and white balancing
2. Obtain separately models for source output, lens and filter transmission
3. Combine source, lens and filter data with RGB responses to determine raw responses in any selected combination
4. White balance by normalising curves so that area under each is the same
5. As validation predict false colour in different sparticle wavebands by combining the sparticle transmission data with model - compare with actual images
6. Use validated model to predict false colours for different flowers
7. Use model to determine optimum match between camera/lens/filter response and animal visual sensor responses

The only problem with all this is that my enthusiasm is somewhat reduced as I think a lot of this has been done before here

What puzzles me is that I get the impression from discussion on here that this fundamental understanding of the sensor response in UV is not well understood (I mean not that individuals can't understand it, but that it isn't freely available in published literature) and yet the reference above appears to indicate that this information is easy to obtain.

Anyway, I will keep chipping away at this, so any ideas and feedback are of interest. It's difficult to keep fully motivated on this as I always want to move forward with the actual artistic image taking as well.