Jump to content
UltravioletPhotography

A Sunflower shoot-out in UV


nfoto

Recommended Posts

Inspired by various discussions of late, I collected a lot of older and newer UV-capable cameras today and set up a small experiment to better understand what our cameras can deliver. It actually is a sequel to the earlier report (http://www.ultraviol...or-uv-captures/) on a sunflower (Helianthus annus) and in fact the very same flower specimen was used as a test subject. I had procrastinated to the very end of the sunflower's life so today was the day the job had to be conducted.

 

I used my Coastal Optics 60 mm f/4 APO lens and the Baader U2" (Venus) filter for all pictures acquired with the following cameras,

  • Nikon D1[R] modified with a naked sensor (no cover glass in front of the sensor, AA-filter pack removed). As I got the camera for free many years ago, I modified it for UV and IR usage. Resolution is 2.7 MPix on DX format. No Live View.
  • Nikon D1H. Stock camera thus no internal modification is conducted. Resolution is 2.7 MPix on DX format. No Live View.
  • Fujifilm Finepix S3 Pro UV/IR Limited Edition. Factory modified for broad-band use (range of 350-1000 nm specified). 12 MPix resolution nominal, effective resolution varies. Sensor has 6 M S-type and 6 M R-type pixels on DX format and not all converters can address (literally) these properly. Very primitive Live View functionality.
  • Nikon D40X. Modified as broad-band with internal cover glass by LifePixel. Resolution 10.2 Mpix on DX format. (I had another of these with internal Baader filter, but this is now an IR unit). No Live View.
  • (Nikon D200 broad-band camera was not available at the time of testing)
  • Panasonic GH-2. Modified as broad-band with internal cover glass by LifePixel. Resolution 16 MPix on m4/3 format. LiveView and EVF finder. Full HD video.
  • Nikon D3200. Modified with internal Baader U2" filter (in fact, originating from my earlier D40X). 24 MPix on DX format, LiveView and Full HD video.
  • Nikon D600. Modified as broad-band with internal cover glass by LifePixel. Resolution 24 MPix on FX format. LiveView and Full HD video.

For lighting I deployed two Broncolor Minicom 80 studio flashes with uncoated Xenon tube. These units were run at full output for the entire test sequence which makes them hum happily along.

 

I used the last day in the life of my Sunflower specimen to secure the UV captures. As UV w/b balance targets I used 3 PTFE discs (2*white, 1 grey) and a Passport Colorchecker. In order to get the cleanest possible RAW file I set the cameras to their respective base ISO and only increased that setting if f/11 on the lens did not achieve a good exposure.

 

Image magnification was kept at 1:3.3 to stay away from any hot-spot issues and in addition I deployed the specific lens shade described earlier (http://www.ultraviol...al-60-apo-lens/). No hot spots were seen. The selected magnification of detail resulted in the sunflower head being slightly cropped by the DX frame, even more so with m4/3 format.

 

RAW files were processed with PhotoNinja 1.10 using the neutral targets to set a proper UV "white" balance. PhotoNinja makes a mess out of the Fuji RAF files, making them rotated 45 degrees and the factory specified 12 MPix ends up as a small 6 MPix file instead. Probably PhotoNinja just silently discards all those small pixels embedded between the main pixel matrix, a feature that Fuji claims give their "Super-CCD" additional headroom.

 

So, how similar, or different, are the UV records this broad array of cameras can capture? After all, the models span a decade of hectic development in the digital camera sector and sheer pixel resolution has increased ten-fold.

Link to comment

Here is the D1[R], a spin-off of the camera that threw Nikon into the front of the race for better and better digital cameras. At its time it was a tremendous breakthough in digital technology. Now, some ten years later, most people will just laugh. Well, it still can do nicely in UV.

 

D1_T1308257668.jpg

 

Next out is Nikon D1H, the updated version of the D1 and in many ways a much more polished product. Still using old battery technology so is not practical for extensive field use, but here I hooked it up to the mains.

 

D1H_DSC_5499_v1.jpg

 

Being the first "invisible spectrum" camera offered to the public, the S3 Pro UV/IR LE (Limited Edition) was a surprise from Fuji around 2004. It is surely the quaintest of all my present cameras.

 

Fuji_T1308259097.jpg

 

The humble Nikon D40X was a big sales success in the low-end segment in the mid 2000s and the volume sale no doubt made a lot of later pro models possible.

 

D40X_DSC1104_v2.jpg

 

The Panasonic Lumix range earned itself a good reputation as capable m43-format cameras. The GH-2 was top of the range 2009-2010.

 

GH2_T1308250097.jpg

 

The DX-format D3200 was introduced in 2012 and initiated the move towards higher megapixel models in the economic camera segment.

 

D3200_T1308250811.jpg

 

Finally, the D600 also from 2012 has become popular as an entry camera into the FX systems.

 

D600_cropped_DSC_5867_v1.jpg

(image is cropped so as to correspond to the DX class)

 

The focus for all examples was put on the top of the exserted style and recurved stigmas. Holding the ColorChecker may have upset the focus accuracy, however, so use these pictures as a visual clue to what the camera can do in UV. I'll present 100% crops for comparison later when I have collated all exposure data points.

Link to comment
So far, we have documented that the "UV white" treatment indeed makes the UV rendition similar for a huge range of cameras, old and new. Whilst the main features are in general identical, so conclusions of the UV signature of this specimen would remain unaltered were the camera used changed, there are slight discrepancies in colour details. However, I think this may be with in error range of any general-purpose RAW conversion software, in particular considering the fact several of these cameras are quite old and thus likely no longer are actively supported. In fact, I would turn the argument in the opposite direction and state that it simply is remarkable how well and consistently the UV traits of this sunflower are depicted whatever camera is applied for its capture in UV.
Link to comment

They will all look the same if you profile each camera's colour as a custom light in PN with the CC tool on a Vis shot made with the UVIR-cut. White balance step is not enough. This is in my write-up currently underway (in Stuff). The cam colour profile is selected under the colour correction tab but with a Manual setting for the WB preset.

This might be cryptic, but more detail in in the write-up.

 

This is a very cool set of examples for the different cameras. Looking forward to hearing about the pixel density thing.

Link to comment

Andrea: that constitutes a Stage 2, we're not there yet. The point is here to see what can be achieved with simpler means as we never can eliminate the weak point, which is the profiling done in the software. In an ideal situation white-balancing against a set of UV neutral Lambertian standards should suffice.

 

Next step is summarising the UV exposure required to give best result for each camera. Do note we should think in terms of relative differences here as the zero point is dictated by the subject itself and the lighting setup, both of which clearly can introduce their own alterations to the settings.

 

A reference frequently used by me when I set up such shooting is f/11 ISO 100. That usually provides a good starting point for further fine-tuning of the exposure for a given subject. So I'll give the values for each camera relative to this datum.

  • Nikon D1[R]: CCD. f/11 ISO 200=> -1 EV
  • Nikon D1H: CCD. f/8 ISO 400 => -3 EV (note this is a stock camera so a lower UV sensitivity is to be expected)
  • Fujifilm S3 PRO UV/IR LE: CCD. f/8 ISO 200 => -2 EV
  • Nikon D40X: CCD. f/16 ISO 100 => +1 EV
  • Panasonic GH-2: CMOS. f/11 ISO 160 => -1/3 EV
  • Nikon D3200: CMOS. f/16 ISO 200 to 400 => -1 to 0 EV
  • Nikon D600: CMOS. f/16 ISO 200 => 0 EV

Thus, the tested cameras do vary over a range of 4 EV in terms of the UV response (remember lighting and lens were unchanged factors). I intuitively knew the unmodied D1H would be at the bottom of the list, but wouldn't have thought the Fuji should fare almost as badly. That the D40X is a head above the pack as far as UV sensitivity goes was known in advance.

 

The compilation also indicates that UV responses (as optimum exposure to this test subject) varied less for the newer camera models, all of which deploy a CMOS sensor. The CCD models are at either end of the scale so one cannot say they are in any way "better" for UV.

 

A final comment is that UV captures do have a substantial headroom in terms of how they are exposed. As long as severe channel clipping is avoided you can do well even with suboptimal exposures albeit noise may start to creep in. This is a particular problem for the m43 cameras as their sensor headroom is lower than that of DX/FX models.

Link to comment

The comparison most people are eager to see, and which is tantalisingly difficult to make in a meaningful and objective manner, is how fine details the various cameras can record. Since I'm using formats ranging from m43 to FX and pixel numbers ranging tenfold this is not a trivial case to solve.

 

I'll continue tomorrow when I have found a practical solution for such a presentation. Until I'm ready to continue I lock down the thread (it'll open later of course). This is to avoid answering in advance questions that will be addressed by the coming treatment.

Link to comment

After pondering the issues for a while I decided to make comparisons by using a pixel density matrix for scaling. Thus I generated a density matrix from available data on sensor dimensions and pixel counts in order to find the scaling factor applicable to any specific camera combination. This approach is equivalent to saying "I want to print my file at a given end magnification - which file will look the sharper?". It does not equalise over field of view, but perspective and detail of magnification are directly comparable. It also confers the benefit of making possible any combination of pairs of cameras in an objective manner. Formats become irrelevant save for the question of coverage of field, but in the days of stiching this is less of an issue than it was earlier.

 

Here are the base data for the array of cameras. Please note that there is some uncertainty to the numbers because format specifications vary slightly. In particular, the pixel dimensions reported by conversion software almost always are slightly higher than stated by the maker. This results because there is a calibration band of pixels around the sensor and the programs sometimes can utilise all or part of this hidden area.

 

Table 1. Base data to calculate pixel density

pixeldensity.jpg

 

The Fuji with its "Super-CC" design comprising large and small pixels in a 45 degree rotated matrix is handled in different manners by conversion software. Most programs simply ignore the small pixels and PhotoNinja belongs to this category meaning one only gets half the data off the sensor. So in case of the S3, I'll use numbers in the second row (6.1 MPix etc.). PhotoNinja also keeps the rotated aspect of the output so the processed file looks a bit "unusual".

 

We observe that there is a ten-fold variation in pixel density and about 3 times range of actual pixel size. Since density is a planar and size is a linear measure, this makes perfectly sense.

 

Now, in order to make a comparison, we need a scaling factor based upon these pixel density figures. The underlying assumption here is of course we have the same detail of magnification (image or reproduction scale) at the time of capture. Thus, broadly speaking the lens "sees" the same level of detail, the question now being what is actually recorded by each camera.

 

The scaling matrix S consists of scale factors defined as

 

Sij = Sqrt (wi/wj)

 

where S = scale factor, w is pixel density (from table 1) and i,j are camera labels such that i denotes the reference camera and j the model to compare against.

 

The entire S matrix is shown below as Table 2.

 

Table 2. The scaling matrix (values in %)

 

scaling.jpg

 

Working through an example will show how to use these factors. Say we want to compare a D1/D1H capture to that of D600.

 

From Table 2, S(D600, D1) = 199 %. So, we need to look at a D1 file at 199% enlargement to match it to a D600 file at 100%. Details will now match and can be compared directly whether on-screen or in a print. If on the other hand you wish to test against the D3200, the D1 file now must be scaled by 309%.

Link to comment

So, we need to look at a D1 file at 199% enlargement to match it to a D600 file at 100%.

 

But why don't we simply shrink the D600 file to the D1 size?

Enlarging a file often compromises apparent image quality.

Shrinking a file often improves apparent sharpness and noise.

Link to comment

You could do that as well. Just look in the table to find a D600 file needs to be downscaled 50%. It's all there.

 

Point is that it is easier to see differences between the outputs when they are viewed larger.

Link to comment

Well, perhaps. But if one image gets degraded, then that's not so great.

I would really like to see both kinds of resizing for fairness in any comparison survey.

 

But I digress. We have been stuck with this problem of how to compare images from different sized sensors and different pixel densities from the beginning of digital. Do we make "fill-the-frame" comparisons or do we make "equal distance" comparisons? Do we resize up or down for a comparison? Each type of comparison has its pros & cons.

 

From a practical point of view, I suppose that we simply try to put as many megapixels as possible over a subject. That's all we can really do with what we've got.

 

BTW, you mentioned:

I want to print my file at a given end magnification - which file will look the sharper?

This forces me to ask what you really mean here by sharpness. Are we talking about edge acutance & micro-contrasts? Depth of field? Capture of fine details? Or all of those? There are so many factors involved in the perception of "sharpness", that it might be helpful to clarify what you are looking for here.

Link to comment

You would surprised to see how "bad" a file can look on-screen yet make a nice print.

 

Since the resolution of the human eye is limited, it is important in these comparisons to make the comparison scale large. If we shrink everything sooner or later details start to disappear not because they are missing from the scaled file but because we cannot delineate them.

 

I'll work on a score matrix later today to enable a complete ranking of these cameras.

Link to comment

Of course printing does wipe out noise and so forth. Of this I am well aware.

Yes details disappear if we shrink too much. And blur intrudes if we enlarge too much.

It is a tricky problem.

Link to comment

This forces me to ask what you really mean here by sharpness. Are we talking about edge acutance & micro-contrasts? Depth of field? Capture of fine details? Or all of those? There are so many factors involved in the perception of "sharpness", that it might be helpful to clarify what you are looking for here.

 

All of these except for depth of field which really is irrelevant in the context I outlined (note: same final end magnification). The criteria you list are inter-related so capture of fine detail cannot exist without sufficient edge acutance and [micro]contrast. We usually can get by with an scaling of 300 % and perhaps even 400 % before empty magnification makes further upscaling meaningless (for fine detail appraised at normal viewing distances; will still work for murals and billboards though).

 

It is also food for thought that we almost never lit the subjects in a manner that produces the highest perceived sharpness. For this one would apply harsh lights at an angle so as to make hard shadows that help delineate and model fine detail. If you look at photographs from the '50s and '60s there are lots of such stuff published then. Today in the era of softboxes we simply would find this kind of lighting unacceptable. In doing so we willingly throw away a lot of the perceived sharpness. Some of it can be restored perceptually by judiciously applying high-pass filtration, the intensity of which is set by the final output size and medium characteristics.

Link to comment
Good point about the lighting being involved in the perception of sharpness. There is never much written about that. I've tried some strong raking light shots using one of the UV torches, but indeed the hard shadows such lighting produces are not pretty. But I did learn from those experiments to aim the SB14 at a bit of an angle to the main plane of the flower (when there is one) rather than flat on.
Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...