High Quality Spectral Response Data Incoming

the setup = responsecurves

Link: https://github.com/ilia3101/SpectralMeasurements

11 Likes

Sounds right up @ggbutcher 's alley.

7 Likes

very exciting, looking forward to more data in your repo!

Interesting! Very curious about your monochromator, and its construction if DIY. That’s the gold standard for this work, but commercial offerings as you probably know are pretty expensive.

For constructing profiles, consider dcamprof. It makes ICCs or DCPs, will start with target shots or SSF data, and make matrix or LUT profiles. But, what it does of interest to making comparisons is to produce a boatload of reports of various dE related to the training reference. If you use dcamprof and the ColorChecker spectral reference to make your profiles, we can compare dE against other methods, including my single-shot spectroscope data.

Of note is that you need to adjust your data against the spectral luminance of the light source. I’ve been winging it using someone else’s measurement of a tungsten-halogen light, but that’s really not accurate. I was going to get a XRite i1Studio, but I’ve found a place to procure a Hamamatsu C12880MA MEMS micro-spectrometer breakout board for almost half the price:

https://www.tindie.com/products/groupgets/c12880ma-mems-u-spectrometer-and-breakout-board-2/

15nm resolution between 340 and 850nm should be good enough for spectral power distribution data.

6 Likes

@ggbutcher I’ve been working on this for almost two years, watching ebay constantly, that’s how I managed to obtain a proper monochromator and integrating sphere. The monochromator is a “.2 meter monochromator” by Verity Instruments (seen in quite a few youtube videos). I removed the monochromator’s built in detector and now use that hole for light input. I also addedd a closed loop stepper motor, and calibrated the wavelength with a helium neon laser (as it had gained a significant zero error thanks to me unscrewing the micrometer and screwing it back in while trying to attach the stepper motor).

I use a photodiode (TSL235r) with a known response to measure the light in the integrating sphere at the same time as the camera. In processing, the diode’s readings get divided by it’s response at any given wavelength, and then the entire result is divided by that, which cancels out both the lamp and the monochromator’s influence automatically, as well as that of any filtration.

Here’s what happens if I didn’t divide by the diode reading, very spiky as I’m using a xenon lamp:

The drop in red is the long pass filter for red wavelengths getting engaged. Maybe a little too early, considering how strong that drop is. But most imporatantly, it doesn’t affect the shape of the final result. I can swap out the light source or change whatever I want without worrying.

4 Likes

ah, you’ve got it going better than I…

The prices for used Verity devices are a lot better than what I’d been finding just six months ago, $355US is moving to within my budget.

I started a github repo to collect data that I measured and also found during my research. Right now, have data I haven’t posted because I haven’t been able to clarify the license, but I’m probably just going to make profiles from that, derived work is okay. The repo is here:

https://github.com/butcherg/ssf-data

You might find it useful for validation.

I wrote a set of threads here to document my investigation back in 2020, here are links to each in sequence:

The Quest for Good Color - 1. Spectral Sensitivity Functions (SSFs) and Camera Profiles

The Quest for Good Color - 2. Spectral Profiles “On The Cheap”

The Quest for Good Color - 3. How Close Can IT8 Come to SSF?

The Quest for Good Color - 4. The Diffraction Grating Shootout

I started a github repo to collect data that I measured and also found during my research. Right now, have data I haven’t posted because I haven’t been able to clarify the license, but I’m probably just going to make profiles from that, derived work is okay.

I thought scientific data couldn’t be copyrighted?

By the way, feel free to inlcude my data. Should I do anything to make the license clear? Should I add something to the repo?

The prices for used Verity devices are a lot better than what I’d been finding just six months ago, $355US is moving to within my budget.

Are you in North America? If you are, you have much more choice than me! Have you set a search alert on ebay? Those can help you catch a well priced listing.

I found my monochromator through a UK listing (extremely rare occurance), around mid 2020. It was around £250.

I wrote a set of threads here to document my investigation back in 2020, here are links to each in sequence:

Wow I hadn’t seen before, thank you for sharing.

Also have you seen the spectral data from here - Results from an IDT evaluation - Tech/Engineering - Community - ACESCentral?

That report is what originally got me interested in this topic :slight_smile:

Although that report containts little of value, it inspired me go down this rabbit hole, so I’m happy with that.

1 Like

Hi Ilia, brilliant work!
The above quote is the only potential Issue I can find in your approach. I’m not 100% sure wether that HeNe-one-point calibration is enough. If you can, find another wavelength to check your calibration, for example green laserpointers usually use an output at 532nm. Red laserpointer diodes IIRC can have several wavelengths in the red, same with blue laser diodes (which makes them harder to use as a calibration tool if you do not know if it’s 475nm or 480nm or 460nm). Now, I know that grating-angle and diffraction angle should “behave” and the multipoint calibration should not be critical. But if by whatever coincidence the blue or red end calibration of the diffraction spectrum are off by 5nm, that would be painful.
It’s the only reason why I would suggest checking the calibration.

Looking forward to everything to come out of this and @ggbutcher work. And of course the OpenDRT-like thing you’re working on.

Cheers

I’d love to check the calibration. But I don’t know enough about lasers. Are those 532nm lasers reliably 532nm?

I know there’s also reliable 543nm HeNe lasers, but they’re super expensive on ebay.

I could do a wavelength sweep of a fluorescent lamp, and we can see how well the peaks match up to the known values. I think I have some old CFL bulbs.

Yeah, I’d do that at least once, just to gain confidence in the drive resolution. The single-image spectroscope approach requires this in order to interpolate the rest of the wavelengths against the spectrum image, so I use a household CFL bulb with its known peaks.

Oh, one more link; I wrote software to 1) pull data off the single-image spectrum TIFF and 2) to mangle data with respect to calibration and formatting:

https://github.com/butcherg/ssftool

Command line programs that do one thing at a time, but allow piping of the result into a subsequent invocation. With that an gphoto, I had a cute little scrip that automated the entire thing from capture to profile. I could do a profile in a little over a minute, which helped immensely determining the alignment sensitivity of my contraption…

2 Likes

Imho, for this application, yes.

HeNe’s are gas discharge laser cavities…the precision needed for the cavities makes them expensive.

Ah you can do the same with that xenon lamp! That doesn’t require buying a green laserpointer. Linewidths are probably a bit broader than from a laser-diode, but that sure would be a multipoint calibration/check.
With an additional CFL you could crosscheck the multipont calibration. Yeah, that should do for a good enough calibration.

One more thing: mechanical backlash in the grating positioning. If you can, sweep the wavelengths from one direction. The closed loop steppermotor is good, but the mechanical angular positioning might have some slack. Turning the grating just in one direction reduces errors nicely. (Or have a go-to-pos command for the steppermotor driver that always arrives at the position from one side)

Oh yeah! Great idea. This file will have the data: https://github.com/ilia3101/SpectralMeasurements/blob/main/Data/Camera/Canon/EOS%205D%20Mark%20III/001/diode.dat

2nm increments, and starting from 380nm (allegedly). It’s dark readings interleaved with actual readings, so you need to take every second value.

Already considered :slightly_smiling_face:

    /* Backlash correction - always finish on a positive rotation */
    int32_t backlash_steps = 50;
    if (steps_required < 0)
    {
        StepperStep(steps_required - backlash_steps);
        delay(20);
        StepperStep(+backlash_steps);
    }
    else
    {
        StepperStep(steps_required);
    }
2 Likes

Ah do you have an estimate for your spectral bandwidth? So, how narrow can you make your input slit and have what wavelength bandwidth? If you have actual one to two nanometer resolution that would be rather cool for spending 250quid!

Erratum: maybe you need to take the CFL! looking at a pure xenon lamp spectrum, hmm, those bumps at 480nm might be too broad for calibration. In my head those were sharp lines!

1 Like

I think the bandwidth is in the range of 1-2 nm, I got that impression while matching the HeNe laser. But I do not know for sure what the slit width is. Need to look at the datasheet of the monochromator and try to figure this out.

The highest peak is at reading #47 here:

380 + 47 * 2 = 474nm

From my quick googling, xenon lamps apparently have a strong peak around 475nm, which confirms that my calibration is likely very close.

Erratum: maybe you need to take the CFL! looking at a pure xenon lamp spectrum, hmm, those bumps at 480nm might be too broad for calibration. In my head those were sharp lines!

Maybe. Might become more clear with a 1nm step rather than 2nm.

I’ll definitely try the CFL (this will only be possible January)

2 Likes

Cool! those are not too broad it seems.

That would be cool if it is that good!

1 Like

I managed to get a i1pro (EFI ES-1000) for 36 pounds!!! WTF!

It was apparently dropped and untested, and the UV cut filter is smashed (good for me, actually).

However I can confirm the wavelength scale is still very accurately calibrated. Tested with fluorescent lamp and HeNe laser. So I got a steal.

So lens measurements are also incoming.

First tests:


Takumar lenses have interesting coatings, with a more flat transmission spectrum than modern lenses (I think), so the results always look magenta using standard matrices.

7 Likes

Thats interesting. What type is it? Super multi coated? Pentax/Takumar has lots of different versions and combinations of coatings but still label them smc. The new HD coating is even more magenta in my experience. But thats just the result of “better” coatings right magenta being a mix of both ends of the spectrum?

1 Like

WOW.

Oh those transmission curves look everything but color-(wavelength-)neutral. I would not have expected such wavelength dependence and variability between lenses.
Basically, you need a new matrix/LUT for every lens! Wow.
Oh and the Helios should appear visibly tinted yellow to the eye.

That would be my guess too.

1 Like

Yes, the better transmission at the ends of the spectrum, the more magenta (relative to your average greenish lens).

The lens is Super Multi Coated. The coatings are noticably green, so it all adds up.

I think most modern lenses lines fall somewhere between my Helios44 and Takumar plots, closer to the Helios 44 (I find it produces quite neutral results with Adobe matrices). I don’t think it makes much difference in most cases, but with the Helios 85mm it becomes significant. Yes: it’s very yellow, and has very blue coatings.

It basically transforms daylight in to incandescent.

1 Like

Ahh! Imagine having that sought after incandescent look in all images :smiley: /joke

Im surprised to hear you expect contemporary coatings not to be flat. I mean that takumar is a sixties lens right. Coatings must have improved since then. I did hear somewhere that Pentax patented the shits out of the processes though. Perhaps thats still a blocker.

Unless there are advantages to filtering off the edges? Chromatic aberrations?

Edit: I have the K mount, Pentax branded, version of the 3.5/135 by the way. Same optics, different barrel, possibly adjusted coatings. It can give very nice feathered or brushed bokeh in some circumstances.

1 Like