Struggling to get accurate color from narrowband RGB scans. Always punchier/more contrasty than white light

[ToddSch1979]

Setup:

Camera: Nikon D800 (tethered via gphoto2)
Light: Big Scanlight, RGB narrowband mode
Software: NegPy v0.52.0
Narrowband frame (Frame012_R.nef): Capture set to Linear RAW and Narrowband, both enabled
White-light frame (Frame013.nef): Capture set to Linear RAW and Narrowband, both disabled

What I've tried so far:

• Ran the Camera Scanning rebate-patch calibration (clear film base metering) to set exposure/per-channel light levels — this fixed my exposure/clipping issues (Sh/Hi now consistently under 1%).
• Trichrome Calibration Profile is set to "None" on both frames . Linear RAW is enabled on the narrowband frame,
• Tried various Crosstalk Matrix options (Generic C41, Fujicolor Superia X-TRA 400, Fujicolor Natura 1600, Fujicolor 400) at various Strengths (currently 0.48) — helps somewhat but doesn't fully resolve it.
• Tried Hue Trim — didn't meaningfully help for this symptom.

The problem: Compared to a white-light scan of the same setup/negative, my narrowband RGB scans consistently come out more saturated and contrasty. Not a hue-rotation or crosstalk-magenta issue anymore (that's resolved) — just an overall punchier, less naturalistic look than white light gives me on the same film.

Attached: two frames from the same roll for direct comparison — Frame012_R.nef (narrowband RGB capture, Crosstalk Matrix "Fujicolor 400 (approx)" at Strength 0.48) vs Frame013.nef (single white-light capture, no crosstalk correction needed/applied). Same roll, same light source, same processing defaults otherwise. The narrowband frame is visibly more saturated and contrasty than the white-light frame.

Both frames were shot back-to-back on the same physical setup (same D800, same Big Scanlight, same aperture/focus) — only the capture mode differs, so this should isolate narrowband capture itself as the variable.

Question: Is this expected/inherent to narrowband capture (as scanlight's own docs suggest, richer channel separation reads as more saturated), or is there a calibration step I'm missing? Specifically should I expect the Trichrome Calibration Profile (built from 3 bare-light R/G/B exposures) to meaningfully help here on top of the Crosstalk Matrix, or is Crosstalk the primary lever and Profile mainly targets a different kind of error?

Any guidance appreciated. This is otherwise a fantastic tool, just losing my mind!

[seanharding]

My experience with my Nikon + Big Scanlight scanning setup (albeit with a quite different sensor than the D800 — I'm using a D3300) is that a trichrome calibration profile is crucial for getting the colors to look right with narrowband scans. In fact, that feature came out of my issue #561 about a similar topic. Once I made a profile and started using it, I got much better results. That is the primary lever here, I think. At least, worth trying. It only takes a couple of minutes to make.

[ToddSch1979]

OK, I’ll definitely give that another go. I must have done something wrong because I was struggling with the process.

I used ScanLight Control to “dial in” the Red channel so the histogram was just below clipping, then did the same for G and B. I then shot individual frames with no film carrier in front of the light at those individual settings with just the light source and kept the same exposure for each.

My settings were 255 for Red, 225 for Green, and 28 for Blue at f/11, ISO 100, and 1/160 shutter.

I then loaded those files into the calibration tool, but honestly, when I select the profile it created, I don’t really see a difference in the file.

Is this the same process you used, and does it sound like I’m doing this correctly?

For reference, I actually have a print on some Kodak Endure.

reference Print
White Light

RGB

[seanharding]

Interesting. I'd be curious to see what RGB triplets look like. In my setup, the narrowband + profile is almost indistinguishable from RGB triplets on the photos I've tested it on.

When I made my profile, I actually didn't mess with the scanlight control at all. I left all colors on 255, and exposed so they didn't clip. Though to be honest, I have experimented with different exposures, including using the same exposure across all of them and letting blue clip, and for my rig it didn't make a huge difference. My impression was that the exposure doesn't have to be perfect to get a result from this process, but your mileage may vary.

[seanharding]

I don't think it would make a difference here, but I should also mention that I'm not doing tethered scanning, because my cheap little camera doesn't support it. I'm just shooting on the camera and then importing via SD card. Again, I have no idea why that would make a difference, but it is one thing that differs between our processes.

[ToddSch1979]

Thanks. I built a profile the way you described: didn't fuss over matching per-channel light levels, just made sure each channel wasn't clipping. Wanted to share the full process and what I'm seeing, in case any of it's useful.

Calibration captures (bare light, no film, R/G/B one at a time):

Each channel individually looks clean and well exposed. No obvious clipping or noise issues in isolation.

One thing that stood out vs. your approach: my calibration ended up with fairly different shutter speeds across the three exposures (R 1/125, G 1/200 and B 1/1600). Wondering if that's expected/fine, or if a big spread in per-channel shutter speed could introduce inconsistency into the resulting matrix in a way that doesn't show up when looking at each channel individually.

Results applying the new Profile to real frames:

• Brick alley/street scene: visible difference with Profile on — subtle but real, looked closer to the white-light reference.
• Motorcycles (mostly red/chrome/dark tones): no visible difference at all with Profile on vs. off.

previous

and lastly, C-Print

For reference, here are the individual R/G/B triplet channels for the alley frame that did show improvement — each channel has good detail and a reasonable tonal range on its own, so the underlying capture for that frame seems clean:

Curious whether the shutter-speed spread during calibration rings a bell for you, or if the frame-to-frame inconsistency (helps here, invisible there) matches anything you saw before you landed on a working profile.