The Holding Tank

Home Page Photography

Photography RSS 2.0 Feed

| Photos & Themes RSS | Hardware | Technical | Links | Books and Magazines  | Downloads |

Technical

| Flash Tests | RAW Converter Tests | Depth of Field, Image Quality and Sensor Size |

| MTF | Resolution | Cropping and Printing | Image Stabilization and Magnification |

| Lens Image Simulation | Anti-Alias Filter and Sharpening | | Tilt and Shift Lenses |

Lens Distortion Correction and Image Detail | Digital Darkroom Workflow | Optical Coatings |

RAW Converter Tests

Table of Content

Tonal Range Tests

Colour Tests

Over Exposure Tests

Noise Tests

Detail Tests

DPP and ACR Feature and Performance Comparison

Sharpening Effects 

This page covers some test with the Canon EOS 20D using in camera and supplied RAW converters. Unless otherwise stated the software versions are as in the table below.

Tonal Range Tests

In these tests three targets of black, average gray and white have been photographed under indirect daylight. The camera and all processing was on auto white balance and all processing carried out with "as shot" settings. 

Manual exposure was used exposing for the average gray. In addition a Gossen Lunasix F exposure meter with spot meter attachment was used to measure the light reflected from each target. The exposure settings are summarized in the table below for ISO 200 at a shutter speed of 1/8.

The actual exposure used for the shot was 1/8 at f7.1.

The shot was captured with Raw + LJPG, the RAW file was then subsequently processed in EOSV and DPP to create 16bit TIFF, 8bit TIFF and JPG files for analysis. This is summarized in the table below. For DPP sharpness 0 and quality 6 (jpg only). For EOSV the jpg is quality 2. Tone Levels are given in Photoshop.

If you wish to inspect the files download the RAW file and in-camera JPG and convert using the default settings.

This suggests that the basic tonal range, at least in the main section of the tone curve, are effectively identical for all RAW conversion outputs and in-camera JPG. This only covers basic reflective subjects so there may be more subtle differences for highlights due to light sources or specula reflection and deep shadows. 

Back to TOC

Colour Tests

This test was carried out in indirect daylight with an incident level of ISO 200 1/30 @ f6.7 measured with the Gossen LunasixF and a manual exposure of ISO 200 1/30 @ f6.7 including some colour test prints from digitaldog.

Here is the RAW file and in-camera JPG, the conversions in DPP and EOSV are all done the same way with the "As Shot Defaults" unless otherwise stated. The below values are measured using Photoshops dropper tool and are listed as R, G, B. 

Note that the original target has been printed in order for it to be photographed so variations from the original are mostly due to the printer.

Within experimental uncertainty there does not seem to be large differences between the in-camera jpg and the various RAW outputs. 

It seems DPP may be slightly less saturated for reds but perhaps a little more saturated for blues but these variations are small enough to just be due to measurement.

There is a significant difference in tone with DPP between as shot and faithful settings, nether are exactly like the EOSV output.

The reddish circles in the top right of the frame (numbered 1 & 2) are not quite so well differentiated with DPP. Using DPP's "faithful" setting the two colours become muddied and less well separated.

Eyeballing the printed target whilst looking at the test shots on the screen I would judge that  EOSV has done a better job of colour rendition than either DPP mode.

Back to TOC

Over Exposure Tests

In these test three RAW shots were taken with manual exposure on the 20D.

Metered of the wood of the table, +1 stop, +2 Stop.

Inspecting the LCD: For the +1 stop a small area on the left of the frame was flashing for clipping, with the +2 stop almost all the white paper was flashing for clipping.

Each frame has subsequently been processed to 8 bit TIFF by the RAW converters and in some cases 1 or 2 stops of digital exposure correction applied as appropriate. All settings "as shot" unless otherwise stated. All TIFFs were then down sampled and converted to JPG using Photoshop's "Save For Web" feature.

For DPP two sets of results are presented with the colour rendition of either "as shot" or "faithful". The software versions are as per the table below:

It seems that DPP has coped much better with extreme over exposure of 2 stops than EOSV or ACR, particularly for yellows and light blues which were rendered almost completely desaturated by EOSV. ACR and DPP are much closer in recovering 1 stop of over exposure although ACR seems to introduce some colour shoft in the white paper. This suggests that light tones of these colours will also be rendered more accurately with DPP.

Back to TOC

Noise Tests

This test compares DPP on its standard colour setting with zero sharpening with ACR with no sharpening, luminance or chrominance noise filtering. In both cases digital exposure adjustment was set to zero and the colour temperature set to 4100K. The test frame is an interior dim available light shot at 3200 ISO.

The raw file and side car file are downloadable.

We can see ACR tends to a warmer and brighter interpretation.

Now looking at some detailed areas at 100%.

Some colour artifacts are visible in both the DPP detail section and the ACR.

The ACR shadow areas  seem to have more chrominance and luminance noise than DPP although this is difficult to judge to the fundamental exposure/contrast/colour interpretation differences.

This may be due to some inherent noise limiting in DPP, although the similarity of detail noise colour artifacts suggests this may not be the case.

Back to TOC

Detail Tests

In the following tests DPP and ACR shots are compared with zero or moderate sharpening applied.

Back to TOC

DPP and ACR Feature and Performance Comparison

This section is a feature comparison between DPP and ACR identifying what edits can be performed without conversion from RAW. Missing features that are a problem for me are noted in red.

[1] DPP set for quality priority, all times seconds

[2] Time to open into CS + time to save as TIFF, all other timings for ACR from bridge

[3] DPP only saves changes to RAW setting when navigating away from the folder. Thumbnail update on change of RAW settings is instantaneous. 

[4] On Bridge the copy RAW settings and save RAW settings occurs simultaneously. However the thumbnail update then follows and uses lots of CPU resources and takes quite a while.

[5] All tests conducted on the following PC: Windows XP Pro SP2; Pentium 4 3.06GHz; 1.5Gb RAM.

The two programmes seem to operate with thumbnails quite differently. DPP seems to derive thumbnails from the JPG embedded in the CR2 raw file. CS2 generates thumbnails from the CR2 and sidecar files. The CS2 method is very CPU expensive and lengthy on any directory with a reasonable number of files but does generate very accurate thumbnails. The DPP method is fine as long as very large exposure corrections are not required. When this occurs significant thumbnail colour casts can arise although this does not effect the editor preview which is derived from the RAW file.

The following section is my experience using ACR 3.1 with files from a 20D having used DPP up to now.

The first thing to do with ACR is turn off the automatics, all of them. The automatics produced high key images with blown out highlights.

The second was to ignore the preset white balance settings, these are miles away from the correct values for the camera in bright sunlight. I found for daylight I needed to set the slider to 4700-4800K. This is 600-700K out from the 5500K one world expect but this quick fix gets me close the WB I would get with the daylight setting on DPP. 

The strange thing is ACR was quite close with "cloudy" 6500K or "shade" 7500K for a frosty early morning shot in shade matching life and DPP with its own "shade" setting. Daylight on both converters resulting in a blue interpretation.

Also flash was OK which gives close to the same WB as DPP. Some shots in daylight gave good colour on the "daylight" preset as long as tint was zeroed. It is worth noting that tint gets set by selecting on of the preset colours. 

All in all, white balance seems to be much more unpredictable on ACR that on DPP, with the quality of the balance varying from shot to shot. The camera AWB is the best choice unless a dominant colour has biased this.

Clipping of strong yellows and blues was a problem with ACR. This can be reduced by setting the output space to a wider gamut such as ProPhotoRGB, however this only creates problems at the stage of conversion to the output colour space. Further tests suggest this may be more of a problem with ACR than DPP because ACR has a higher default level of colour saturation than DPP.

Back to TOC

Sharpening Effects

EOS Viwer

Digital Photo Professional

Photoshop Unsharp Mask

In this section I have tried to evaluate the sharpening provided by the various RAW converters using the spectral analysis method shown in the anti-alias filter evaluation. Note DPP version here is 1.6.1.0.

The following shot (down sampled) was taken using the EF 200mm f2.8 ISO 400 f3.5 1/400 sec with the camera mounted on a sturdy tripod. 

A small central section of this image was then used to provide a random section of detail.

The spectral analysis routine was then used to evaluate the various RAW converter outputs with differing levels of sharpening. (download RAW file here).

EOSV

This section covers the EOS Viewer utility shipped with the 20D.

The boost with the EOSV sharpening peaks at about 52 cycles per mm and then rolls off.

The image sections are in the table below.

DPP

This section covers the Digital Photo Professional software shipped with the 20D and also supplied with the mark II pro Canon bodies.

The DPP sharpening shows no roll-off and probably slightly more sharpening intensity. Clearly the minimum sharpening in DPP is more than in the EOSV off case. However, Canon’s DPP uses edge sharpening, rather than Unsharp Mask, which has the effect of creating more defined edges in detail while maintaining a low-noise image [Ref].

Photoshop Unsharp Mask (USM)

These plots evaluate the Canon recommended sharpening (Range = 0.3 for EOS 1Dx cameras Ref “GETTING THE MOST FROM YOUR EOS-1 CLASS DIGITAL SLR” ) following the minimum sharpening case with EOSV off.

The next plot shows the relative frequency response of the a sharpening level from the EOSV converter and the DPP converter relative to the recommended Photoshop sharpening.

This suggests that compared to the Photoshop USM algorithm the EOSV algorithm may include some noise filtering, causing the frequency response to fall off above 50 cycles per mm. However the sharpening is greater than that recommended by Canon.

The DPP algorithm, even at quite low levels, provides a similar level of sharpening to the EOSV but does not so the roll-off above 50 cycles per mm. 

Assuming the EOSV off condition is representative of no sharpening then both viewers seem to provide more than the recommended sharpening.

Back to TOC

Last Updated 05/06/2008

All Content © 2005-09 Lester Wareham All Rights Reserved     

All material is supplied  as is and without warranty, use at your own risk.  

All opinions stated are the authors own.

All quoted information remains the copyright of the respective authors.

Contact: (anti spam remove underscores)

Note this email gets spammed a lot so if you want to get your message through please prefix "Holding Tank" in the title

f_ w_ a_ r_ e_ 9_ 0_ @_l_i_n_e_o_n_e_._n_e_t