C41 film in CD3 (RA4 developer)

I have read about developing C41 film in CD3 developer from the RA4 process before in a number of forums but very few samples were posted and lots of arguments among the authorities on development and some enthusiastic free thinkers willing to experiment claiming success.
I stumbled into this DIY experiment after acquiring some RA4 as a hack to my C41 chemicals to process ECN-2 film. By mistake I loaded the wrong roll on to the reel. When I opened my changing bag I saw the canister of a Fuji NPZ instead of the Fuji Eterna that I intended to process. As I was out of C41 developer, I stood in front of the decision to either reattach the film to the canister and load the Fuji Eterna, or just go for it. I did the latter.

Processing

The processing was done in a rotary processor at 38degC.

Step 1: 3min in RA4

Step 2: 6:30min in C41 Bleach

Step 3: 3:30min wash

Step 4: 6:30min in C41 Fix

Step 5: 3min wash

Step 6: 1:20min stabilizer

Apart from the CD3 from RA4 the other steps used C41 chemicals from a Fuji Hunt C41 kit, where the developer (CD4) had gone bad.

Scanning

So how what was the result? Well, ocular inspection of the film looked normal.

The film was shot at EI400, which is how I normally rate my expired Fuji NPZ.

I occasionally shoot calibration shots, unfortunately I had not done it this time as I did not plan for this first C41 film in RA4.  After this initial trial, I did shoot three more rolls of Fuji Pro400H and NPH in RA4 developer with calibration shots which I used to create a profile for this more systematic test (see Second Round below). I was pretty certain that with film calibration the results would be good, but what about standard C41 color conversion? Below I have taken the same framed scanned/converted it with Silverfast and Vuescan with different settings.

As you can see, the result varies. I think while none of the is impossible to adjust to satisfaction in Lightroom. I would select (C) or (D) as starting point. The skin tones look natural, there is some small cast that needs to be address but in general quite OK. (F) might look good at a first glance but is over saturated, even if saturation is brought back the skin tone does not look right, and  it also suffers some loss of shadow details.  (E) also suffers loss of shadow detail.

Interesting is that correcting (A) (the one with heavy magenta cast) is down to two corrections.  Just change the white point for blue and green in Lightroom and the cast is gone. A slight hint of magenta cast still can be spotted but many Fuji films have this characteristic.

 

                

It seems strange that Silverfast gets this wrong. According to the histogram there seems to be almost no samples in the top 15% of the range for the blue and green component. Maybe CD3 uses 15% smaller space for green and blue and Silverfast profiles are operating with some pre known knowledge in which range the color samples can be found?? (note I did not touch the "auto tolerance" which does improve the result somewhat)




There is a contradiction the sample material if you are looking for what CD3 (RA developer) instead of CD4 (C41 developer) does to the negative.
With Vuescan the most color accurate one (F) was converted with a C41 NPZ ICC film profile that I created in Vuescan based on a IT8 target in a Fuji NPZ frame from a previous roll developed in C41. It would lead one to believe that CD4 and CD3 produces the same result. However the worst color accuracy is (A) based on using the Fuji NPZ profile in Silverfast (w/o CCR). The result becomes better with CCR turned on, which is probably closer to the Vuescan autolevel feature does to the processing. My guess is that CD3 (RA4 developer) changes the dye components density, confusing the Software which part of the scan range for each color (RGB) to apply the curve to. When the start and end points are correctly located for each color the response (curve) remains similar between the two developers. What this experiment does not tell is if that start to end piece of the range is larger or smaller just that when the range is located the response curve is correct applied to that range. I have no experience trying to produce a print with photo paper in RA4 process so I cant tell if this matters for this.

decode key
SF = Silverfast
VS = Vuescan
SF NPZ C41 profile  = Standard Silverfast Fuji NPZ profile
VS NPZ C41 profile = Custom calibration profile for Fuji NPZ developed in C41 created in Vuescan based on a IT8 target. You will not have it unless you create it yourself.
Pro400H RA4 profile = Custom calibration profile for Fuji pro400H developed in CD3 (RA4 developer as this blog entry describes) created in Silverfast AI based on a QP203 target.
You will not have it in SF unless you create it yourself.
CCR = Color Cast Removal in Silverfast language
autolevel = one of the methods in Vuescan to manage white balance 
Generic = Generic profile for C41 film not brand to type related

Going indoors

In the mixed lighting, the CCR in Silverfast needs to be turned on to reduce the cast.
In my mind (I) is the best base to start from. 

Second round

Encouraged by the partial success I decided to tryout with some expired Fuji Pro400H that I exposed at EI200. Processing was done as described above. With some more attention to avoiding water marks this time and also adding calibration shots.

As a the template for the calibration I used the Fuji NPH profile in Silverfast AI and calibrated the grey scale patches on a QP203 card. I also added a IT8 target in the calibration shot to create a calibrated ICC profile in Vuescan.

This time there is no doubt (A) is the best starting point for further processing.

Final thoughts

Is it possible to develop film in CD3 (developer for RA4)? The answer is yes.
Will scanning software Vuescan/Silverfast be able to produce usable scans with the built-in profiles? The answer is yes, but autolevel/ccr or manual tweaking is needs to be applied.
For best results I have to admit custom calibrated profile's are needed.

Should I continue to develop C41 film in CD3 (RA4 developer)? This is not a clear cut. While it is possible to produce good end results, I am hesitating as it is sort of proprietary and require calibration to achieve high quality color results. What if one try to rescan w/o having the tools for calibration at a later stage? I am less concerned over the often claimed issue of going through a paper printing process due to different color response on the intermediate negative as I (or someone else) unlikely will ever do that with my shots.

PS

Working with all setting in Silverfast and Vuescan can produce better starting positions in Lightroom but I seldom do as a part of my work flow. Only with tricky negatives and slides I work more inside the scan tools, all other work I prefer doing in Lightroom as it is easier to experiment and change the mind.

Techie notes
All frames were scanned with a Microtek M1 using Silverfast AI Studio 6.6.
Vuescan converted images was first scanned with Silverfast in raw mode "48 bit HDR Color" as Lasersoft calls it. Output from Silverfast was set to Adobe RGB TIF files(all negfix profiles as based on Adobe RGB in Silverfast)

In Silverfast I applied each profile, I wanted to test, to the scan and used image auto adjust. No other changes in Silverfast. The Silverfast images are prefixed with SF.

In Vuescan I loaded the frame and used the generic film profile and did variations using autolevel/fluorescent (where applicable) and selected no film ICC vs IT8 calibrated film ICC profiles. Output profile was set to Adone RGB in TIF files.

I did manual exposure adjustments in Lightroom to make them look similar bright. 

The "contact sheets" were produced in Lightroom through the print to file function with SRGB output.