Drew of ProPhotohome offers a new tool for white reference.

[Doug Kerr]

Hi, Drew,

Thanks so much for getting back with this data.

Hi Doug,

Here are the density measurements from the expodisc and the color parrot (v2.0).

I suspect you don't really mean "density" (which is the negative logarithm of the transmissivity). Perhaps these are just CMYK values on a scale of 0-1.

Are those linear CMYK values? Ordinarily, CMYK values are expressed on a scale of 0-255, which are just the complements of the complementary RGB values, which are not linear (so the CMYK values are not linear).

Your CMYK values seem to be on a scale of 0-1. My guess is that they are are "linearized". (Your instrument or its manual should tell which.) If not, then the "0.82" value would correspond to a transmittance of only 0.026 (2.6%), not likely.

ExpoDisc Density Readings in CMYK
Shows around the 18% stated in the cyan and the black (subtract to get transmission numbers)

Color Parrot Density Readings for center target area in CMYK
Shows around 60% light transmission

I trust all these were taken with a transmissive color densitometer, yes?

The Color Parrot version 2.0 transmits around 3 times the amount of light that the ExpoDisc does.

Version 1.0 (the one you have) shows light transmission of about 40% through the center. This would be about twice the light transmission of the ExpoDisc.

Wow! As measured in place (that is, based on the camera exposure meter report of measured Bv with the "partial" metering pattern in place), I show my Color Parrot to be about 0.5-0.75 stops "lighter" than the my ExpoDisc (a ratio of about 1.4:1 to 1.7:1). I guess telephone engineers should not fool around with densitometry.

If those values for the ExpoDisc that you measured were in fact linear CMYK values, that would correspond to a chromaticity that departed from neutral by about du'=0.009, dv'=0.016, du'v'=0.021. That's almost 13 times "worse" than intimated by the QC report numbers sent from the factory with my ExpoDisc.

Through the diffuser one would see a very deep blue.

Well, thanks again for the data.

Best regards,

Doug

 

[Doug Kerr]

Hi, Drew,

You can also use it exactly like the WhiBal, just hold it out and take a reflective reading from the surface, or have the Bride hold it for you while she talks to all of her guests and does the first dance.

Are you speaking of using the Color Parrot as a reflective target? Mon dieu! Perhaps even zut alors!

Well, at least we have measurements of its reflective neutrality, if not of its transmissive neutrality. From those, it looks as if might be pretty good in that mode.

Which side would one use? (We have determined that it is durable enough to sustain a "flip" for that, though.)

Thanks for the great idea.

Best regards,

Doug

 

[Doug Kerr]

A little bit about CMYK and CMY/K

Some of you may be a little baffled about the "CMYK" readings in Drew's latest posts.

Of course, the CMYK color space is intended to conceptually parallel the realization of colors in four-color printing.

Those of you who are familiar with this color space will recognize that the combinations of values on Drew's bar charts are impossible in the CMYK space. They would, if converted to RGB form, have large negative R, G, and B values.

More likely, those readings are a combination of a CMY (not CMYK) representation of the color involved, plus (separately) an assessment of its overall "density" (as if converted to gray scale, essentially the complement of luminance), expressed with the symbol K. (It seems likely that the first three are actually the linear subtractive color coordinates, c, m, and y, and that the last is similarly on a linear scale.)

For example, in the Adobe gray scale color spaces, there is of course only a single (nonlinear) color coordinate, which is represented by the symbol "K". I assume this was chosen to draw the parallel to (but not equivalence with) the "K" ("black") coordinate in the CYMK color space.

Best regards,

Doug

 

[Asher Kelman]

A Breather: Clarifying Optical Density: simple technical reminders!

This is a helpful discussion. Let's pause to refresh as to terms. That way no one will be lost!

Drew, it would be nice to have the vertical axis marked.


With a piece of rough glass, the light is reflected and refracted in all directions by the fact that the light arrives at the glass surface at all odd angles and ends up all over the place in the glass which then becomes illuminated and a source of light to the camera. So loss of light is 3 fold:

1. Light is reflected away from the disc
   
2. Light is reflected and refracted in the disc and also towards the camera.
   
3. Light is lost absorbed by the disc material itself as it contains molecules which are not glass and therefore absorb light at particular wavelngths preferentially.

For those who would like a brief reminder.

Transmissive measurements: Optical Density as used scientifically to measure film or quantitate dissolved molecules when the O.D. linearly increases proportionately to the concentration of the measured light-absorbing molecule at that wavelength of light. Less and less light survives tranmission through the material. However, as you see below, the numbers are altered to give a simple positve scale. O.D. 0.0000 means no light is lost as it would be by a massive block of absolutely pure glass. An OD of 2.5 is already perceived as black. However 4.0 is even blacker. Few instruments measure that high as one is actually detecting very few photons arriving at the sensor.

Scanning Film: We see O.D. of scanners marketed at 3.9, 4.0, 4.1 meaning that the blacks are very very black. Very little light incident light comes back to the detector. (In fact anything under $5,000 is unlikely to be better than O.D. of 2.6-3.7.).

O.D. is defined as:
O.D. = log10( 1/ Tr) = - log10(Tr). The reason for using this "log10" scale
is to make the numbers positive and to make their range smaller since the
O.D. in the power of 10 in the exponent. Despite the use of the same term
"density" O.D. has nothing directly to do with the mass density.source.

Density as in mass/volume is one meaning and that is the Mass Density and that does increase the Refractive indext and bending power of glass lenses. However in the term O.D. or Optical Density we mean an answer to the question how effective is the material in absrobing light passing through a medium. In that case it's proportional to the concentration of the absorbing molecules for the particular specified wavelength.

Here, however, when we measure the transmitted light that remains, it has been reduced by more than just the actual absorbtion by the molecules material it passes through.But in fact when we measure the optical density of, for example, a filter, the "loss" due to reflection at the front, and due to backscattering, are included - anything that doesn't get out doesn't count.

Perhaps a better definition here would speak of how much the device "reduces" the light passed through, not (just) how much it absorbs (not real precise, but perhaps suitable for this mission). As we pointed out above, light is reduced by reflection from front surface, refraction and scatter within the material so Doug suggest we talk not of the O.D. of the transmissive material in scattering devices color balancing devices added to the lens. Yes we do use a device that is also used for measuring pure optical density, but since in this case, the loss of light is more complex a new broader definition of Optical Density for the partially transmissive Color Parrot™ or ExpoDisk™ would refer to reduction of light transmitted for all of the complex reason mentioned above. So we might still use 0.D. units, however we are measuring the sum of light flux loss for any reason.

Asher


Bart/Doug: feel free to add or edit

 

[Drew Strickland]

Hi Doug,

I can't find specific information about how the density numbers are derived in the manual.

Please feel free to have a look. The Density Scratchpad information is on page 91.

http://www.prophotohome.com/ColorShop X User Guide.pdf

Yes, please ignore the color numbers. This is a result of the pitted surface of the expodisc. This is only to show the amount of light reflected. This is why I measured each piece separately.

 

[Doug Kerr]

Hi, Asher,

This is a helpful discussion. Let's pause to refresh as to terms. That way no one will be lost!

It is unlikely that the ExpoDisc has a transmission of about 0.81. (The manufacturer says mine has a transmission of about 0.179.)

In any case, recall that Drew has said earlier that his determinations are made with a reflective spectrophotometer. (If you're new to this fray, you may find that "startling".)

So perhaps these results suggest that the ExpoDisc has a diffuse reflectance of about 0.81. I'm not sure from which side (see discussion below).

I can believe that. I've had no reason to measure it before, since it doesn't have anything to do with anything. I'll check mine this afternoon. (I can only make the determination on the rear face - the front face has a lenticular acceptance lens, and the definition of reflectance there would have to be very peculiar.) I will do it on a "specular reflectance excluded" basis.

Well, the White Rabbit and I are going back through the looking glass to have lunch.

Best regards,

Doug

 

[Drew Strickland]

It is unlikely that the ExpoDisc has a transmission of about 0.81. (The manufacturer says mine has a transmission of about 0.179.)

Hi Doug,

No one said, or otherwise suggested, that the transmission was .81. This measurement is the density number reported by an xrite dtp22 in ColorShop X. I have not used the density measurments for anything in terms of the design of the unit, or for comparisons to the ExpoDisc. Obviously, the color numbers are skewed for the expodisc due to the "funky" nature of the pitted surface.

You asked for some numbers regarding density, so I pulled this for you. My best guess is that the density numbers are showing the density in a "reflective" sense. In other words, 81% of the light is being reflected. Leaving the other 19% to either pass through (the vast majority of it would), or be absorbed by the device. This is consistent with their claim of 18% light transmission. For my analysis, all that really matters is that you gain a few stops with the new unit.

As far as revisiting the reflective vs. transmissive spectrophotometer issue for determining spectral neutrality. Xrite has recognized this approach as a functional method. That is good enough for me. I guess it isn't for Doug. As far as I know, he doesn't own a reflective spectrophotometer, or a transmissive one. I love you anyway, man.

 

[Doug Kerr]

Simplistic measurements made here this afternoon suggest that the back of our ExpoDisc diffuser exhibits a diffuse reflectance of about 0.75 (specular reflection excluded).

This was done with an arbitrary incident source and an observation angle of roughly 20° from the normal.

This of course has absolutely nothing to do with anything insofar as the use or operation of the diffuser is concerned (I do not advocate its use as a neutral reflective target owing to its specular reflection), but might be of interest in trying to divine just what property of an ExpoDisc was measured in data recently supplied by Drew Strickland.

The measurements were made using the exposure meter of my Canon EOS 20D, using the partial metering pattern. The results were taken from the measured Bv reading of the metering system as reported in the makernote section of the file metadata by ExifTool. The value is mislabeled "MeasuredEV" (evidently per the Canon documentation). Note that this is not the bastardized form of EV often used by Canon (which is Bv+5) but rather the real Bv.

The reading frorm the back of the ExpoDisc was then compared with the WhiBal gray card, for which we have credible reflectance information from the manufacturer.

Again, let me emphasize that there is essentially no significance to the diffuse reflectance of either face of a measurement diffuser. The measurement was only made in support of forensic inquiry.

Best regards,

Doug

 

[Doug Kerr]

Hi, Drew,

Hi Doug,

No one said, or otherwise suggested, that the transmission was .81

Actually somebody did, but it was edited out upon second thought but after I had made my comment!

This measurement is the density number reported by an xrite dtp22 in ColorShop X. I have not used the density measurments for anything in terms of the design of the unit, or for comparisons to the ExpoDisc.

Other than in your post today here.

Obviously, the color numbers are skewed for the expodisc due to the "funky" nature of the pitted surface.

Ah, you measured the front face. Sure.

You asked for some numbers regarding density, so I pulled this for you. My best guess is that the density numbers are showing the density in a "reflective" sense. In other words, 81% of the light is being reflected. . . .

Strange. Normally, in a reflective test, a larger value for K (or for C, M, or Y) means less light being reflected. A surface with larger values is "darker".

But maybe the conventions used in your analytical software are the other way up. I'll have to look at my manual.

. . . Leaving the other 19% to either pass through (the vast majority of it would), or be absorbed by the device. This is consistent with their claim of 18% light transmission. For my analysis, all that really matters is that you gain a few stops with the new unit.

Mine's not that good. 0.5 to 0.75 stop.

As far as revisiting the reflective vs. transmissive spectrophotometer issue for determining spectral neutrality. Xrite has recognized this approach as a functional method. That is good enough for me. I guess it isn't for Doug.

This is true.

As far as I know, he doesn't own a reflective spectrophotometer, or a transmissive one.

This is true.

I love you anyway, man.

We love you too, Drew.

Thanks for responding to my request for more information on the Color Parrot. I understand more every day.

Best regards,

Doug

 

[Bart_van_der_Wolf]

Would you mind uploading that spectral curve and density readout. I would love to see it.

Hi Drew,

Here is the remission spectrum of both a WhiBal and a BabelColor White target:

That translates to a reflectance value (illumination=D50, observer=2 degrees) of:
BabelColor: L*=99.7 , a*=-0.2 , b*=0.4 (diffuse optical density= 0.0034 or 99.2% reflection)
WhiBal: L*=70.8 , a*=-0.4 , b*=1.4 (diffuse optical density= 0.3778 or 41.9% reflection).

Data was collected with an I1 Photo spectrophotometer, conversions from L* a* b* to other units were done with Bruce Lindbloom's calculators.

Bart

 

[Drew Strickland]

Thanks, Doug and Bart.

The BabelColor target certainly looks nice and smooth.

Off-topic. I just found out why we can't get enough of this topic. According to this funny story on the wsj we're all "infovores."

What is it about a Web site that might make it literally irresistible? Clues are offered by research conducted by Irving Biederman, a neuroscientist at the University of Southern California, who is interested in the evolutionary and biological basis of the human need for information.

Dr. Biederman first showed a collection of photographs to volunteer test subjects, and found they said they preferred certain kinds of pictures (monkeys in a tree or a group of houses along a river) over others (an empty parking lot or a pile of old paint cans).

The preferred pictures had certain common features, including a good vantage on a landscape and an element of mystery. In one way or another, said Dr. Biederman, they all presented new information that somehow needed to be interpreted.

...

When you find new information, you get an opioid hit, and we are junkies for those. You might call us 'infovores.' "

 

[Doug Kerr]

Hi, Drew,

Thanks, Doug and Bart.

The BabelColor target certainly looks nice and smooth.

Off-topic. I just found out why we can't get enough of this topic. According to this funny story on the wsj we're all infovores!

An excellent outlook!

Thanks.

Best regards,

Doug

 

[Doug Kerr]

Hi, Drew,

Hi Doug,

I can't find specific information about how the density numbers are derived in the manual.

Please feel free to have a look. The Density Scratchpad information is on page 91.

http://www.prophotohome.com/ColorShop X User Guide.pdf

Thank you so much. That should be very useful. I'll let you know my thoughts after I've had a chance to look it over.

Best regards,

Doug

 

[Doug Kerr]

Remission spectra

Hi, Bart,

Very interesting.

Thanks for doing this.

Best regards,

Doug

 

[Bart_van_der_Wolf]

Perhaps a better definition here would speak of how much the device "reduces" the light passed through, not (just) how much it absorbs (not real precise, but perhaps suitable for this mission).

When we talk about a semi-transparent material, it is perfectly okay to talk about it's transmission. It is usually expressed as a percentage of output versus input (within a given spectral band of interest). The input is usually assumed to be a diffuse source of illumination of a given magnitude, although one might also want to determine a certain scattering property of the medium for which a ratio between collimated and diffuse transmission can be used.

As we pointed out above, light is reduced by reflection from front surface, refraction and scatter within the material so Doug suggest we talk not of the O.D. of the transmissive material in scattering devices color balancing devices added to the lens.

The OD only describes the overall integrated opacity (opacity=1/transmission) along the entire spectral pass-band, which is fine from a 'filter factor' perspective, but it doesn't help us to understand it's color filtering characteristics.

Besides the total integrated transmission over the entire spectral pass-band, we can also determine the transmission for smaller spectral pass-bands. When the transmission for these sub-bands varies along the entire spectral range we want to investigate, a color filtering will occur (the spectral composition of the illuminant, as seen by the sensor array, is changed). This can usually be expressed as a shift in color temperature when one compares the unfiltered illumination spectrum with the filtered one (which requires a spectrophotometer if one wants to avoid ambiguity).

A low cost alternative for a spectrophotometer could be a measurement of an observed change in color (e.g. expressed as a Lab value offset) by using the filter itself. The Lab color space allows to separate the Luminance from the Chromaticity components of color. I encourage the use of a diffuse light transmission source (such as omnidirectional ambient light diffused by e.g. a sandwich of 2 layers of opaline glass), or even better (although a bit more expensive) the use of a reference diffuse reflection object such as the BabelColor White Target.

Our methodology choices are governed by exactly what we want to determine.

Bart

 

[Doug Kerr]

Hi, Bart,

The Lab color space allows to separate the Luminance from the Chromaticity components of color.

It certainly does. But as I'm sure you know, a* and b* do not directly denote a chromaticity (it is actually a chrominance plane).

The two factors that define chromaticity (when working with the L*a*b* coordinate system) are actually a*/L* and b*/L*.

That is, these two colors have the same chromaticity:

L*=70, a*=+6, b*=-4

L*=35, a*=+3, b*=-2

Just a teeny point, lest someone get confused.

Best regards,

Doug

 

[Bart_van_der_Wolf]

It certainly does. But as I'm sure you know, a* and b* do not directly denote a chromaticity (it is actually a chrominance plane).

Thanks for the correction. The point I was trying to get across is that we should avoid CMY or RGB colorspaces if we want to describe color shifts.

The two factors that define chromaticity (when working with the L*a*b* coordinate system) are actually a*/L* and b*/L*.

That is, these two colors have the same chromaticity:

L*=70, a*=+6, b*=-4

L*=35, a*=+3, b*=-2

I don't think that's exactly correct. Try the CIE Color Calculator to verify that they result in slightly different coordinates in e.g. xyY chromaticity coordinates. It would also result in a small shift of the correlated color temperature of -24.8K (not much, but still) if used with a D50 reference white.

By using 'Lab' and the above calculator, we can also get a feeling for the shift in (correlated) color temperature (which is one of the topics we're investigating).

Bart

 

[Doug Kerr]

Hi, Bart

Thanks for the correction. The point I was trying to get across is that we should avoid CMY or RGB colorspaces if we want to describe color shifts.

Yes, a point well made.

I don't think that's exactly correct.

Well, Duh, I see what you say. (By the way, I use Lindbloom's calculator and other material all the time. it is wonderful!)

I had been aware for a long time that the a*b* plane was more of a chrominance plane than a chromaticity plane - that is, a given pair of a*,b* would only imply a chromaticity if interpreted in connection with a value of L*. (This is of course the same situation as for YIQ, YUV, etc.)

I had not assumed (earlier) a relationship as simplistic as that I asserted here.

Then, following some chain of conversion equations, I came upon a pair that seemed to imply exactly the very simple relationship I then adopted. But I can't find those now! Maybe it came to me in a dream!

But here is the reality (based on the actual equations for L*a*b* on Lindbloom's site) site:

(Note that this may not work if we are in the "linear tail" region for of small values - I haven't checked my derivation there yet.)

The factors that remain constant for a given chromaticity are:

a*/(Y* + 16)

and

b*/(Y* + 16)

(Well, I was "close"!)

Thus, these colors have the same chromaticity:

L*=54, a* = +6, b* = -4

and

L* = 19, a* = +3, b* = -2

Indeed, for any chromaticity (above the tail), the ratio a*/b* is constant.

In any case, thanks for disabusing me of my error!

And thanks again for your help in this and many other matters.

Best regards,

Doug

 

[Doug Kerr]

Measurement technique

Hi, Drew,

Perhaps the reason I am not comfortable with the way you determine the transmissive spectral uniformity (or transmissive chromatic neutrality) of a diffuser with a reflective spectrophotometer is that I don't understand just what you do. You could help by reminding me.

Just to kick it off, let me offer some possibilities on a multiple choice basis:

1. You use the reflective spectrophotometer to measure the diffusers from their front face, with the device in midair, or against a black background, and:

1a. You take that data as indicative of their transmissive properties.

or

1b. You take the "inverse" of that data as indicative of their transmissive properties.

or

2. You use the reflective spectrophotometer to measure the diffusers from their front face, with the device against a nominally "white", neutral background, and:

2a. You take that data as indicative of their transmissive properties.

or

2b. You take the "inverse" of that data as indicative of their transmissive properties.

or

3. You take the data as in 2 above, subtract from it the data as in 1 above, measure the reflective data for the background itself, subtract that from the previous result, take the square root of the residual, and use that as indicative of their transmissive properties.

Of course, it may be something else again. I'm certainly no expert in the use of spectrophotometers (I don't have any, as you have noted), so I may have missed the best answer.

Thanks for your help.

Best regards,

Doug

 

[Doug Kerr]

Readin' the data

I have deciphered the "four bars" data used by Drew to report his "measurements" of the transmission of a Color Parrot and an ExpoDisc.

It took downloading the same software he used with his instrument (fortunately a "14-day free trial" package) and a bit of reverse engineering.

Here is the concise story:

1. The three bars are variables the manufacturer labels "C", "M", Y", and "K".

2. They often call this kind of "K" (more about that in a moment) "V", including on the dialog box in which one can manually enter color descriptions into the "measurement log". When it is called "V", the convention is to use the order V,C,M,Y. (So we enter V,C,M,Y and get C,M,Y,K, where K=V. There's probably some historical reason for that. Maybe.)

3. In the CMYK color space we regularly encounter in image editing, the K parameter is (with C, M, and Y) part of a recipe for the color (not a unique one - the value of "B" is rather arbitrary, but the other values have to suit).

In this notation, K is not (with C, M, and Y) a part of the description of the reflective color of the surface being measured. That is fully described by C. M, and Y alone. Rather it is a supplementary descriptor, the "monochrome density" of the surface (as if we had measured it with a "non-color" densitometer). It is apparently reported as a "bonus" by their color densitometers.

4. All four values are in terms of density (that is, the negative log of the reflectance). In the case of K, the significance of that is obvious. For example, here are some density values and the corresponding (monochrome) reflectance:

K _____R

0.1____0.79
0.2____0.63
0.3____0.50 (That's "-3 dB" to the electrical engineer!)
0.4____0.40
0.5____0.32
0.6____0.25 ("-6 dB")
0.7____0.20
0.8____0.16
0.9____0.12
1.0____0.10
1.2____0.063
1.5____0.031
2.0____0.01
2.5____0.0032
3.0____0.0010

5. In the case of the other values, the "backward" nature of a subtractive color model (CMY) makes it a little hard to visualize what they do.

For example, the CMY density triple 0,5,5 represents (almost) pure red (RGB=255,0,0).

Remember, the CMY model is a metaphor for colored inks, where the presence of "ink" of a certain color actually subtracts a certain part of the incident light spectrum. No ink at all: "white" light. Full dose of all inks: "black" (not really, of course, but that's another story).

A density of 5 is a large density (surviving light of 1/100,000 what there would be otherwise). With the 0,5,5, specification,, we can think of this as meaning that there would be overwhelming activity by the magenta and yellow "inks" in suppressing their parts of the spectrum, leaving just the part that comes out looking like "red".

The triple 0,1,1 would also mean "almost" pure red, but not as pure as 0, 5,5 - RGB=255,25,25, in fact. Here, the magenta and yellow inks have only suppressed 90% of the light they are responsible for controlling, the "anti green" and "anti blue" light respectively.


So that's how it works.

The bottom line is that what Drew measured for his Color Parrot (presumably the reflectivity of whatever surface of it he measured), giving a "K" (or "V") of 0.40, means a reflectance of about 39.8% * - and the reflectance is pretty neutral.

* 10^-0.40=0.398

That's great for its newly-touted use, as a reflective color target. It of course tells us nothing about its transmittance or transmissive chromatic neutrality.

Best regards,

Doug

 

[Doug Kerr]

Measurement

Drew has recently called attention here to the fact that I do not feel that measuring the transmissive properties of a transmissive optical measurement device with a reflective instrument is appropriate.

He says:

As far as revisiting the reflective vs. transmissive spectrophotometer issue for determining spectral neutrality. Xrite has recognized this approach as a functional method. That is good enough for me.

What is he referring to? Well, we don't know. Maybe it is this:

The following was quoted by Andrew Rodney shortly after the emergence of the Color Parrot. He had evidently asked X-Rite about the use of a reflective instrument for the purpose. This is their response (as Andrew cites it):

If you are using it to create white balance then I assume it is used in transmission (essentially a diffuser to evaluate the light to make it “white”). If this is not how it is used the following is irrelevant.

As such you it would be preferred to evaluate it’s neutrality (or flatness of the spectral curve) in transmission mode. Assuming it is translucent you could measure it in reflection by measuring over a neutral white substrate and essentially subtracting the readings from the readings of the substrate to see the difference (there are a number of problems with this approach as it ignores all sorts of issues of latteral diffusion, internal reflection etc. but it would be okay for a quick and dirty test). . . .

Best regards,

Doug

 

[Bart_van_der_Wolf]

Assuming it is translucent you could measure it in reflection by measuring over a neutral white substrate and essentially subtracting the readings from the readings of the substrate to see the difference (there are a number of problems with this approach as it ignores all sorts of issues of latteral diffusion, internal reflection etc. but it would be okay for a quick and dirty test). . . .

One also has to keep in mind that light emitted by the reflection densitometer passes the translucent layer twice, once through the layer and again after reflection by the neutral white substrate. Therefore all inaccuracies (as noted in Andrew's quote) and absorptions/refractions/scattering of the illuminant's spectrum introduced by the translucent layer are doubled.

Bart

 

[Asher Kelman]

Now, when do we get measures of the transmission comapred to other devices or did I miss that?

We need criteria by which we would find such a device useful for particular types of work. I can just put a cut of a milk bottle as a diffuser. So what are the specific requirements, what is the best methodology to test with?

Asher

 

[Doug Kerr]

Hi, Asher,

Now, when do we get measures of the transmission compared to other devices or did I miss that?

I have reported on the difference in overall transmission of the Color Parrot (v1.0) compared to my ExpoDisc (2006 model). Note that the total transmission will depend on the in incident lighting distribution (there are several standard configurations from which one typically chooses when using a densitometer, none of which can I directly emulate).

For the test situations I used, the Color Parrot was 0.5-0.75 stops "hotter" (1.4-1.7x) than the ExpoDisc. That would be a range of about 25%-31%.

Drew has cited in this thread that the transmission of the 1.0 Color Parrot is about 2 times that of the ExpoDisc (as divined from surface reflectance measurements). He says it will be "2-3 times" for v 2.0.

We need criteria by which we would find such a device useful for particular types of work. I can just put a cut of a milk bottle as a diffuser. So what are the specific requirements, what is the best methodology to test with?

Typical important performance measures are:

1. Directivity pattern. Again, this must be defined based on certain assumptions about how the light from the rear of the diffuser is "accepted" by the actual measuring element, which in our case is some lens, with some lens settings, on some camera, which regards some part of the frame in making chromaticity determinations.

These latter complications essentially wash out if the diffuser exhibits a uniform luminance across its entire rear surface, and that luminance is constant with angle of observation.

In any case, there are those who feel that a "cosine" directivity pattern is ideal. It is hard to say what departure from that should be considered acceptable, since even that as an ideal is controversial.

Some feel that a rather "narrow" directivity pattern is beneficial (although I have not yet head any understandable theory as to why).

2. The transmissive spectral response should be flat within some stated criterion. As an alternative, the chromatic response should be neutral within some set limit. We might for example want it within four MacAdam steps of true neutrality (that is a way to describe chromaticity differences in terms of human perception of those differences).

3. We might want to establish a limit on overall transmission being not less than a certain amount (probably something like 10%). Some believe tthat higher values are more desirable. I have heard no supportable theory as to why.

4. Then of course we may wish to establish "mechanical" criteria, such as:

a. What diameter is it?

b. How is it held to the lens?

c. What kind of drop to what kind of floor surface will it withstand without damage (and how do we define "without damage".

d. What kind of lens cleaning solvents should it resist?

Best regards,

Doug

 

[Bart_van_der_Wolf]

Now, when do we get measures of the transmission comapred to other devices or did I miss that?

We need criteria by which we would find such a device useful for particular types of work. I can just put a cut of a milk bottle as a diffuser. So what are the specific requirements, what is the best methodology to test with?

My take on it is as follows:
1. One requires a diffuse uniform (both spatially as well as spectrally) source of (preferably incandescent or full spectrum) light. A good quality slide viewbox, or a white LCD screen, would suffice.

2. One 'measures' that lightsource; once without, and once with the 'CWB diffuser'. The CWB diffusor is placed flush to the front of the lens, such that no reflected light can influence the measurement.
The preferred measurement method is by using a spectrophotometer set for measurement of (ambient) light emission, while pointing in the direction of the lightsource (the probe is preferably shielded for other light).
Alternatively, the measurement can be done by shooting images with a camera, but a lot of variables could enter the equation. To avoid vignetting, one should use an aperture of f/5.6 to f/8. Smaller apertures would increase the risk of sensor dirt showing up. Linear gamma files, or Raws, are best for quantification.

3. Both measurements/images are subtracted from eachother, and the result is optionally normalized.

This will allow to determine the net color shift introduced by the CWB diffuser, and it's transmission differences over the measured angle of integration.

By shooting an image, one approximates the net result of all specific variables. The variation in spatial uniformity can be plotted accurately with e.g. Imatest's Light Falloff method. I wouldn't mind doing that if supplied with appropriate Raw files, and someone to host the resulting figures.

4. Finally it should be established how the CWB diffuser's color and its spatial non-uniformity influences the various camera CWB algorithms, although Doug has already done some of that. This should allow to test the scenario of measuring the ambient light color from the camera's position, pointing towards the subject (which is an alternative scenario for when we cannot measure from the subject position pointing towards the camera, the latter scenario being more accurate).

Bart

 

[Drew Strickland]

I have sent this information to Doug regarding information in this thread and thought it would make sense to post here as well.

I appreciate everyone's candidness, here. I really do. And, I hope you will continue to help.

This whole density thing is a very minor issue. In no way do I need to use this tool to demonstrate approximately how many stops difference there is between the two devices. ExpoDisc is the company that is attempting to get by with simple density numbers when referencing the "neutrality" of their product, rather than showing the full true spectral curve.

I continue to believe that the ExpoDisc is not neutral. On that point we disagree. Could I be wrong? Quite possibly. I could have a faulty instrument, etc. But, my practical research backs up this claim. The ExpoDisc is too warm, plain and simple.

They claim +-2% Neutrality in all images. What kind of a claim is that? Is that in mixed lighting when you are using the reflective method, or using the incident method with one predominant light source? Perhaps we should be investigating that. What exactly does that mean? +-2% of what? Was the ExpoDisc always within +-2% of a truly neutral gray card?

We simply state the facts, that the components used in the manufacture of the Color Parrot are more spectrally neutral than the components used in the manufacture of the ExpoDisc. This is a simple fact. If you, or others, choose to believe that somehow magically adding three layers of wrong together (the 3 expodisc layers) will make a right, then well, what more can I say? We also, much more accurately, simply state "Great Color" in 15 seconds or less.

A few people are focused on the expodisc because they assume that because they are a larger company, then they must be better. I put it in the original article simply to contrast and compare. I would much rather focus on the practical uses of the Color Parrot. I have not attempted to use the density numbers from this device in any way to design the product, or to report on it in a review or article. I made the mistake of pulling these numbers for everyone without really looking them over. When it comes to density, a reflective device is simply a mess. Fortunately for us, we make no use of these numbers whatsoever.

Drew

 

[Drew Strickland]

One also has to keep in mind that light emitted by the reflection densitometer passes the translucent layer twice, once through the layer and again after reflection by the neutral white substrate. Therefore all inaccuracies (as noted in Andrew's quote) and absorptions/refractions/scattering of the illuminant's spectrum introduced by the translucent layer are doubled.

This is true, but when the curve is essentially entirely flat, it is a moot point from a practical standpoint.

We may go ahead and pick up a device that can be used directly in a transmissive manner in the not too distant future. Just to cover all of the bases.

 

[Drew Strickland]

Drew has cited in this thread that the transmission of the 1.0 Color Parrot is about 2 times that of the ExpoDisc (as divined from surface reflectance measurements). He says it will be "2-3 times" for v 2.0.

Actually, it was divined by looking through my camera meter.

I quickly pulled those numbers for you because I know you like instruments. The numbers seemed to jibe with the readings I was getting from the camera, so I put them up without doing my due diligence. It is easier to show the chart than my camera meter.

But, I could do that as well.

 

[Asher Kelman]

Hi Drew,

The links to purchasing the Color Parrot within your website no longer work! Dead ends to B&H store. The B&H store itself does not have them either! I tried the link in your signature too. That's the same. Am I just not seeing the obvious? What's up?

Asher

 

[Doug Kerr]

Hi, Drew,

I have sent this information to Doug regarding information in this thread and thought it would make sense to post here as well.

Well, then I guess it would make sense to post here some of my responses to you.

I appreciate everyone's candidness, here. I really do. And, I hope you will continue to help.

This whole density thing is a very minor issue. In no way do I need to use this tool to demonstrate approximately how many stops difference there is between the two devices. ExpoDisc is the company that is attempting to get by with simple density numbers when referencing the "neutrality" of their product, rather than showing the full true spectral curve.

Although it takes the full spectral response to describe the whole story, it is almost impossible for one to readily compare the performance of two diffusers with regard to what we are interested in by looking at their spectral responses.

I continue to believe that the ExpoDisc is not neutral.

Nothing is neutral.

My new 2007 model ExpoDisc measures at a du'v' of about 0.0015.

On that point we disagree. Could I be wrong? Quite possibly. I could have a faulty instrument, etc.

Drew, you just have the wrong instrument.

But, my practical research backs up this claim. The ExpoDisc is too warm, plain and simple.

They claim +-2% Neutrality in all images. What kind of a claim is that?

I have no idea what that would even mean.

Is that in mixed lighting when you are using the reflective method, or using the incident method with one predominant light source?

ExpoImaging today absolutely (I would say, vehemently) does not endorse in any situation the method where the camera plus diffuser is at the camera, pointed at the subject (what is often, misleadingly, called the "reflective method"), so I doubt that their discussion is based on that method.

Perhaps we should be investigating that. What exactly does that mean? +-2% of what?

I'm mystified, too. Where do they say that? If I could see it in context, I might be able to figure out what that means.

Was the ExpoDisc always within +-2% of a truly neutral gray card?

I woiuld have no idea what that means, either.

We simply state the facts, that the components used in the manufacture of the Color Parrot are more spectrally neutral than the components used in the manufacture of the ExpoDisc.

I do not believe you have anything there that can determine the transmissive response of anything.

If you, or others, choose to believe that somehow magically adding three layers of wrong together (the 3 expodisc layers) will make a right, then well, what more can I say?

Of course one can pile up three layers with non-neutral response and have the overall thing be neutral. That is in fact what is done in the ExpoDisc.

The diffusing layers in the ExpoDisc (like those in the Color Parrot) are unavoidably a little more "yellow" than neutral, so they add a carefully chosen transparent "bluish" layer to restore the whole thing to neutrality. This is not a case of "three wrongs not making a right". Its a matter of "three non-zero numbers adding up to zero".

[/quote]We also, much more accurately, simply state "Great Color" in 15 seconds or less.[/quote]

What can I say. "Let's hear it for "Great Color".

Best regards,

Doug