SPD2Matrix - Correlation Files to Matrix Conversion

As said, without a decent spectro to verify, you have no idea what is correct...
All you can really do is decide which result you prefer.
That is the same with any Correlation File application.
You just have no idea on the actual accuracy.

Steve

That is what I thought, Steve. No way to tie it down with either method. Both the display factory tune and the meter are variable enough to cause errors.

Will try with the closer 2026 G6 .spd and see how it looks in HDR with 1D and 3D LUTs. Will see what Wonder Woman looks like with that combination.

LeRoy

With SPD2Matrix you are eliminating the probe errors, but are left with any potential display mismatch.
But that mismatch is likely to be there, albeit differently, when using the Correlation File.

Steve

Steve,

Thanks for the recent updates. The added state assessment and probable original peak-Y information make the intended workflow much easier to follow.
I have been working through several trials using a TCL and the correction file. SPD2Matrix identified the source as native/uncalibrated, with a very cool white, native wide gamut and an original peak around 295 nits.

From my testing, the closet available state was:

reset the display
selected native/wide gamut
disabled the normal correction file and Probe Matching
disabled local dimming
set luminance near the inferred source level
tested the untouched factory temperature presets and selected the one that produced the closest overall RGBW result. This ended up being the coolest color temp available.

This was the closest native state available:

Patch Local Active Y/x/y BPD Reference Y/x/y
R 56.08 / .6883 / .3111 72.78 / .6761 / .3060
G 198.57 / .2289 / .6900 202.48 / .2176 / .6688
B 32.52 / .1541 / .0708 26.39 / .1553 / .0506
W 288.19 / .2813 / .3089 288.19 / .2803 / .2872

I then generated the BPD using the exact local white Y, loaded the local RGBW as Active and the generated BPD as Reference, and enabled the simple matrix match.

For a quick practical check, I used only the 2-point controls to align the display to x .3114 / y .3283. This is not D65, but it is the perceptually matched white used by my primary mode, so it allows a direct visual comparison with the same mode calibrated using the normal correction file.

Without changing the display controls, the finished white measured:

Measurement method x y
Raw i1D3 .3064 .3493
Normal correction file .3079 .3503
SPD2Matrix Probe Match .3117 .3283
Target .3114 .3283

The Probe Match reports essentially the intended target, while the raw and normal-correction readings remain close to one another.

I understand that two separate panel samples will never produce identical RGBW values, even when placed into the same nominal underlying state. The practical question is therefore: how much residual difference is acceptable before the matrix is correcting panel variation rather than individual probe variation?

I also found that changing the factory temperature preset could improve the white point match while worsening primary xy or relative RGB luminance. When selecting the closest source state, should white chromaticity, primary chromaticity or relative RGB luminance take priority?

Unless you use the exact same probe as the Correlation File was generated with, and ideally on the exact same displays, you will never know what is/is not the true readings based on the original display settings.
So you are basically asking an impossible question...

And this is basically the same when using Correlation files, with the added unknow on the underlying variable of each individual i1D3...

And all this really does explain why direct Probe Matching to a decent spectro on your actual display is always the better approach.

Steve

I have a LG C1 and a single i1d3 probe, no spectro.

The correction file pack has CSV file created with C1 and JETI 1501 4.5nm. SPD2Matrix states that CSV White measument coordinates are x=0.309089 and y=0.318940. LG OLED can have those coordinates for white only after precalibrating white point. If I precalibrate white point into those previous coordinates with only i1d3 and correction file enabled or disabled, there is no way to say without spectro, what is i1d3 deviation of x and y errors on white compared to JETI measument.

During regular probe match this would roughly equals situation, where bpd file would have created RGB measuments from spectro and W measument from tristimulus. Clearly a method leading into inaccurate results compared to normal spectro use.

So many consumer with only a single i1d3 would still prefer the use of correction file directly and there are two main use cases for SPD2Matrix:
1) Eliminate calibration deviation between different probes when owning multiple tristimulus probes.
2) To apply correction files other probes than i1d3s.

Am I on right track on my totally uneducated guess?

EDIT: Based on the readme, the 2026-07-24 version of SPD2Matrix.exe is much easier to use by novice HTL/ZRO user. Some display models / csv correction files might be more suitable for this method than other display models and exe can now do this assessment. Thx.

When doing a standard probe match with a spectro and a colourimeter you measure RGBW for both - hence the inherent accuracy.

When using an i1D3 with a Correlation File you do not know what the underlying inaccuracy with your i1D3 is.
(Nothing to do with having multiple i1D3 probes, and as shown in the i1D3 User Guide.)
When using SPD2Matrix you cannot be totally sure what state the display was in when the Correlation File was generated.

So any use of a Correlation File, either directly or via SPD2Matrix has uncertainty.
So what to use is up to you...

However, the white point of the C1 Correlation File points to the TV being uncalibrated.

Steve

New SPD2Matrix update released.
Small change to better handle spaces in the input file name, and added info on the use of quotes to the ReadMe.

Steve

New SPD2Matrix update released.
Now with ability for the user to enter their RGBW xy(Y) measurements to better gauge a likely display match situation with the Correlation File.
Info in the ReadMe.

Steve

This is an extremely useful addition. I had already been experimenting with a crude manual version of the same RGBW comparison using an Excel sheet, but having it built directly into SPD2Matrix makes the workflow much clearer. It is also very helpful to see your cutoff values for excellent, good, marginal, and poor, as choosing a reasonable threshold was one of the questions I had been wrestling with. Thank you for putting the time into this update.

One point I would appreciate your view on is how you define an acceptable local "native state" for the comparison. On consumer TVs, different untouched picture modes and factory color-temperature presets can produce different RGBW balances, even with native gamut selected and all processing, LUTs, calibration controls, and CMS disabled.

Would you consider it valid to compare those naturally available states and select the closest one, provided no granular controls are used? For example, testing the available factory color-temperature presets, but not using 2-point, multipoint, Color/Tint, or CMS to force a match.

In practical terms, if the warmest preset does not align well with the source RGBW data, but a cooler factory preset brings the comparison within the acceptable range, would you consider that valid state selection, or too much manipulation and therefore circular?

If any display is in its true native/uncalibrated state it would be the same regardless of the Mode selected.
Unfortunately, for many TVs that is not the case, so the selected Mode is not in its native/uncalibrated state...
So even more unfortunately, you are asking an impossible question.

All you can do is select a Mode that you hope matches the state of the TV when the Correlation File was generated, and see if you like the result.
None of this, including using Correlation Files directly, provides any guarantee of calibration accuracy.

And again, this further explains why direct probe matching with a good spectro is so important for colour accuracy.
Even a cheap spectro will likely be more accurate that any Correlation File, especially with the unit to unit variations of i1D3 probes.

Steve

New SPD2Matrix update released.
Added output list for all lines of spectral data in the original file.
The means the original file can be edited to remove non-RGBW data to compare vs the original file.
Info in the ReadMe.

Note: due to the way Correlation Files work with i1D3 probes it is possible that non RGBW data may compromise accuracy

Steve

SPD2MATRIX conversion is not working for me. Here are the various WP readings. Obviously matching to my spectro is probably the most accurate but I am trying to get the process working so that I can measure other displays that my spectro won't work with. Any thoughts?

Meter Readings
NX7 Color Profile OFF - High Bright - Custom default gamma

Pro 2 Spectro .3040 .3541 93.6
i1d3 no correction .2974 .3551 95.6
i1d3 match pro 2 .3042 .3544 93.7
i1d3 corr nx9 .3011 .3526 95.7
i1d3 match nx9-spd2m .2999 .3241 108.8

SPD2MATRIX OUTPUT
input : 4 measurements x 401 bands -> 380-780 nm @ 1 nm
observer: built-in CIE 1931 2 deg
auto-detected (row = input row, 1-based):
RGBW rows: R=1 G=2 B=3 W=4 (keep these for an RGBW-only file)
target white Y: 95.6 cd/m2

[matrix]
x y Y cd/m2
R 0.675634 0.323426 20.0829
G 0.271982 0.700808 70.9359
B 0.145514 0.037181 4.8502
W 0.300740 0.325020 95.6000
R+G+B 95.8690 (+0.28% vs white)
display state (advisory, from R/G/B/W chromaticities):
white : xy 0.3007,0.3250 nearest D65 du'v'=0.0079 [off]
gamut : nearest Rec709 max primary du'v'=0.0526 [off]
peak : white integrates to 9.374e+04 cd/m2 [scaled ~x1000, likely mW->W]
original white Y probably ~ 93.7 cd/m2 (confirm with --white-y 94)
relative Y (W=100): R 21.0 G 74.2 B 5.1
=> NATIVE / uncalibrated; measure the new display in its native, uncalibrated state

I can't see an NX7 profile in the available downloads, so I can verify anything.

Steve

There is an NX 9 profile not an NX7. Optical block is the same. Just a slightly different lens.
My NX7 Spectro
Pro 2 Spectro .3040 .3541 93.6

Matching using as a correlation file works quite well.
i1d3 correlation nx9 .3011 .3526 95.7

Converting to matrix does not.
i1d3 match nx9-spd2matrix .2999 .3241 108.8