Erik's web

  • Increase font size
  • Default font size
  • Decrease font size

Sony Alpha 900 vs. 67 analogue, round 2

E-mail Print PDF
Article Index
Sony Alpha 900 vs. 67 analogue, round 2
Dynamic range
The issues with the red flowers
Good comments from Dominique Ventzke
Downloads
All Pages

Note: Planned updates (waiting for films from lab)

- The DR test was reshot

- New measurement of DR

- A tabletop arrangement shot on both Ektar, Velvia and digital

 

 

Note: This is an article in progress and some findings are preliminary.

All original files can be downloaded here

  • A small amount of camera shake was discovered on the "flower and castle" digital image due image stabilization not being deactivated.
  • When I get the negatives back from Germany I'll rescan the Ektar image to ensure that same image was use for both scans.
  • Another image will be added for resolutions analysis (that image unfortantely only scanned at 3200 PPI on CCD)

In a previous article I compared my Sony Alpha 900 to my Pentax 67 using Velvia 50. The conclusion was essentially that the Sony Alpha was in general preferable. Image quality was better than my Velvia scans, and even if the Velvia may have resolved some more fine high contrast detail the Sony Alpha was much more efficient and had better DR (Dynamic Range).

Two questions my former test did not answer was:

  • Would negative film work better than the Velvia?
  • Would higher resolution scans on drum scanner give better results?
I finally get around to prepare a set of new test shots using Kodak Ektar and my Pentax 67 equipment.
While doing my assessment of the images there was a comparison on Luminous Landscapes comparing the new Phase One IQ180 back on a high end technical camera with 8x10" on film. This test I found interesting. On the other hand, the test raised a lot of questions as it was widely seen that the 8x10 images were probably "under par". One of the issues was that to low resolution was used when sampling the 8x10" film. For that reason I looked a bit at my new test shots and posted some of my own scans of 6x7 images.
Dominique Ventzke, who owns "High End Scans.de" invited me to test his scanning services at reduced cost. We arranged a 6096 PPI scan of two of my images from 6x7 film. One was one from the recent test shot and the other one was an old Velvia slide that I believed to be both sharp and challenging.
This article describes the experience I made doing these tests. We are going to discuss the following issues:
  • Dynamic range
  • Resolution
  • High and low contrast detail
  • The advantages/disadvantages of high PPI scans
  • We also are going to discuss a bit about oddities of "bayer filtered images".


What is dynamic range?

Dynamic range is the range of luminance levels that can be handled without excessive noise and acceptable tonal separation. There is a technical definition of dynamic range, that is log(2) of the maximum signal divided by the signal yielding a signal to noise ratio of one. In digital this translates to what is known as "full well capacity" (the number of electrons each pixel can hold) divided by read noise (also in electrons). So DR = log(2) FWC/read noise. There are some issues with this definition of DR for photography, one is that it does not say much about the character of the noise.
It is widely accepted that film has a wide dynamic range. This does absolutely not hold for transparency film, but has a lot of validity for negative film. For film much of the available DR is achieved by compression of shadows and highlights. We can illustrate this by shooting a desnity wedge. The "Stouffer wedge" is one of the more common ones. I have acquired a 41 step wedge for this test, and shot it under identical conditions on three cameras, Pentax 67 with Ektar 100, Sony Alphas 900 and also the Sony Alpha 55 SLT. The film image was scanned on my Minolta Dimage Scan Multi Pro. The images were loaded in the DR module of Imatest, giving the following results.

Th data indicates that both Ektar and Alpha 900 have a similar DR, with the Ektar having about one step wider dynamic. For high quality, the Alpha has wider DR, 6.6 stops instead of 5.3. One very interesting aspect of the curves is that highlights on the Ektar are compressed. Compare the red boxes. This means that highlight contrast is reduced.

 

Note: There are some pitfalls in this measurement, one is being that the scanned data may be affected by stray light in scanning. I plan to do a better test shot and scan both on my CCD-scanner and Dominique's high end drum device.

 

Now, this could also be related to scanner limitations. The pictures below shows a small crop of a 250 MP image. The structure of the wall is here clearly visible on the Sony Alpha image, while the scanned images show little detail. The detail that is visible is much sharper on the "High End" scan than on the other images.

 

Sony Alpha 900 (upsized) Ektar 100, High End Scan at 6096PPI
Ektar 100, Dimage Scan Multi Pro at 3200, upsized

Conclusion: DR on Negative film is similar to a decent full frame CMOS sensor. Film has compressed highlights and may therefore handle highlight smoother at the cost of loosing contrast.

 

Other views on DR

To begin we can look at Kodaks data sheet:

This curve does only show the straing part of the characteristic curve. So it says little about what the shoulder looks like, if we use the pulished data to estimate DR we start at exposure -2.2 end the curces end around 1.2, so that gives a DR of (1.2 + 2.2)  / 0.3 -> 11.3 stops. As said the curves say little about what happens past 20 lux-seconds.

 

Roger Clark has an excellent article comparing discussing film and digital: Clark's article

 

Clark estimates that DR for Velvia to be five steps, Kodacolor Gold seven steps and digital camera to cover 11.7 steps.

My interpretation of Clark's findings is that film gets very noisy at low exposures.

Observations from Dominique Ventzke

What you measured in Imatest was the DR of your particular workflow that includes the signal quality of your scanner. With a different scanner the results may be different.
Tim Gray posted some film DR tests on flickr: http://www.flickr.com/photos/tgray1/collections/72157623656649261/
It’s a pretty good indication of what happens with negative film when over- or underexposed. Note in the 400NC-2 test there is still an image when overexposed 12 stops though the definition and color information is very low at this stage. One of the reasons for the high DR that went into the design of these films was the use in point and shoot film cameras where you have one f-stop and shutter speed, and exposure corrections have to be handled at the scanning/printing stage from the film’s latitude.
You are correct that the increased tonal compression in the shadow and highlight range increases the noise levels within the DR of negative film . I’m not sure how Imatest interprets this as this is actually how we process many images in order to maintain good midtown contrast, and thus the noise gets compressed in the shadow and highlight roll-off. In the highlight part of the curve of color neg film you have reduced levels of detail (resolution) but it still allows the differentiation of extreme brightness levels.
For digital sensors there is no roll-off. You can expose to the right to maximize tonality. But how can you get this soft grading of specular highlights or within light sources like film does? If you want some tonality within those areas you have to set the exposure that nothing clips at all, but that would ruin your shadow detail from the massive underexposure necessary in some situations like high contrast night scenes. Digital technology will probably get there at some point. I found the footage of the Arri Alexa very impressive.
A small remark on the DR of transparency films: These films are designed to look great on a light table or in projection, and thus you get an image with a lot of proccessing already built in which also includes roll-offs in the shadows and highlights. While there is a lot less detail in the highlights and shadows compared to color negs I’ve seen excellent results from slides in high contrast situations. Some reversal films like Fuji Astia 100F or Kodak E100G have a wider DR than others.”

 

 

Resolution and fine detail contrast

Note: I have discovered that the Sony Alpha had some slight camera shake, due to image stabilisation not being disabled. So it would perform a bit better if it was perfectly used. Most findings here are valid. But absolute sharpness cannot be judged. This is of course not much to do about, except to retest, which I will do in due course.

One of the reasons I bought my Pentax 67 was that I wanted to make larger prints with good sharpness and no grain. With digital I started out with APS-C but later upgraded to full frame DSLR. No doubt, I will upgrade when the new 36 MPixel DSLRs arrive, especially after this test. To make large prints we need both resolution and fine detail contrast. The image below demonstrates this quite clearly. The top left image has good contrast (it's blown uop from 24 MP full frame digital), the top right image has good resolution but little contrast, due to highlight compression.

 

 

 

Sony Alpha 900 (upsized) Ektar 100, High End Scan at 6096PPI
Ektar 100, Dimage Scan Multi Pro at 3200, upsized

 

Now, look at the image below, it is another crop from the same set of 250 MPixel images. This part of the image is mid tone. The "High End" scan is definitively the cleanest image.

Sony Alpha 900 (upsized) Ektar 100, High End Scan at 6096PPI
Ektar 100, Dimage Scan Multi Pro at 3200, upsized

Comments from Dominique Wentzke

“I’m not sure if the lack of contrast in the fine wall textures has to do with highlight compression. In my experience this sort of softness or reduced definition of low contrast fine detail is also in the midtones though it is stronger in the compressed highlights. You can also see it in the green leaf crop. The veins of that leaf are more clearly defined in the Alpha image than in the Ektar drum scan.
The tonality in film comes from different sizes of grains and their amount finally building up density. The number of grain sizes in an emulsion layer is limited. Thus it takes a larger area to represent tonal values that only vary very slighty in brightness and hue. Or it takes a threshold contrast to be detected on a small area of film. If the contrast is too low the detail will be lost in the grain structure. Near this threshold film tends to create a textural feel for the detail in a very vague form of rendering. It’s when only a few grains in an area respond to these slight tonal variations. It’s the analog nature that there is something between all or nothing which makes it hard to pinpoint the resolution of a film to a number.
In my experience reversal film is better at the rendition of low contrast fine detail because the tonal range is stored over a larger density range so each recorded f-stop is stretched over much more density than in color negs, and the tonality is not compressed like in color negs. Digital sensors measure the brightness directly and quite accurately wich makes them very efficient at recording very low contrast details as long as it isn’t in the deep shadow end of the signal. I would agree for images where this level of low contrast detail is critical digital is probably better unless you’re using a much larger area of film.”

 


The issues with the red flowers

Note: This issue was pointed out to me by Dominique, who also referred to this article

Another crop from the same image shows a red flower. This flower has some issues both in Analogue and digital. The digital image is heavily pixelated. The main reason for this is probably that the sensor uses a so called "Bayer matrix". The Bayer matrix has red, green and blue pixels laid out this way:

 

Only one fourth of the pixels are red, so for the read color the 24 MP Alpha 900 turns into a 6 MPixel camera. Essentially all DSLRs and MFDBs use the "bayer matrix" solution, with the "Foveon sensor" used in several Sigma cameras being the sole exception. Note within the red frame that the small red flower is resolved on both Ektar scans, but is essentially almost invisible on the Alpha 900 image.

Sony Alpha 900 (upsized) Ektar 100, High End Scan at 6096PPI
Ektar 100, Dimage Scan Multi Pro at 3200, upsized

It is possible to extract more "red flower detail" from the High End scan, Dominique Ventzke, who made the scan posted the image below made from the same scan (click on the image for actual pixels) :

 

Another image...

In the image with the red flowers I have some sensor shake caused by in camera image stabilization. I made another shot comparing the Alpha 900 and Ektar. The Ekatr image was in this case not scanned on a drum scanner but we can still compare with my MDSMP 3200 PPI sxan.

 

Based on the above samples I would draw the conclusions:

 

  • It seems that  a well executed Ektar 67 image scanned on a CCD film scanner will have similar detail to a well executed 24.5 MPixel picture by a full frame DSLR.
  • A drum scan will not necessarily extract more detail but the resulting image will be smoother.
  • In either case the resulting image can be varied a lot. With the scanner much depends on the operator.
  • It is not clearly evident that film has wider dynamic range than digital, but behaves very differently.
  • It is very clear that expert drum scanning at high resolution gives superior results to "amateur CCD" scans.
  • In the red flowers we see some limitations in handling very fine single color detail, due to the "Bayer matrix".

 

Finally, postprocessing matters a lot. The results depend much on sharpening, noise reduction and so on.

 

To me it seems that 67 film needs to be drum scanned to surpass 24 MP digital full frame.

 

 


Good comments from Dominique Ventzke

The Ektar neg was very interesting especially compared to the digital shot of your Alpha. In terms of resolution I think the 6x7 records much finer detail as long as there is enough contrast. But the digital is sharper when it comes to low contrast  textures and has higher micro contrast as can be seen in the wear and tear of the house walls. I was able to extract quite a bit of information from the highly saturated flowers. I don’t know how these flowers look like but to me the film image looks more real, three-dimensional and subtly detailed than the harsh pixelation of the Alpha image.


When you compare film and digital I think it’s not so much a matter of which is technically better in image quality. Both have their strengths and weaknesses. It’s not so black and white that you can say you need this many MP to match film in 6x7, 4x5” or 8x10”. To me the difference in the look of film based and sensor based photography is much more obvious. For example the open shadow detail and highlight rendition of color negative film corresponds very well with the way I see the world with my eyes. The grain however does not and yet it can look very pleasing. Film scales really well when you enlarge it. But the optical quality of the enlargement (enlarging lens or scanner) is a really important factor. You will start to see the grain already at moderate enlargements but it takes a really long stretch before the image falls apart. For example you can reduce the scans to 80 MP and still have a very good image quality on pixel level albeit at much lower microcontrast than a comparable MFD image.

 


Downloads

 

All original files can be downloaded here

 

Last Updated on Tuesday, 29 November 2011 00:33  

Statistics

OS : Linux u
PHP : 5.2.9
MySQL : 5.0.67
Time : 09:19
Caching : Enabled
GZIP : Disabled
Members : 1510
Content : 72
Web Links : 1
Content View Hits : 356705

Login

Banner1

J!Analytics