Measuring SRM

Homebrew Talk - Beer, Wine, Mead, & Cider Brewing Discussion Forum

Help Support Homebrew Talk - Beer, Wine, Mead, & Cider Brewing Discussion Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

swankyswede

Well-Known Member
Joined
Mar 21, 2010
Messages
91
Reaction score
2
Location
Somerville, MA
So I realize not many people have access to the necessary equipment to directly measure the color of a beer, but fortunately, I work in a research lab and have all this stuff sitting around, waiting to be used by curious brewers such as myself. For those that do have access to a UV/Vis spectrophotometer, centrifuge, and syringe filters, I'll note my procedure below.

Wikipedia (which was the only source I could which gave the spectroscopic determination of beer color) defines the SRM of a beer as the absorbance at 430 nanometers times a constant and a dilution factor.

SRM = 12.7 * D * A, where D is the dilution factor (for an undiluted sample, D = 1, for a sample diluted 1:1 with deionized water, D = 2, etc.) and A is the absorbance at 430 nm.

When I ran my ESB through, I got an absorbance of 1.36, which resulted in an SRM value of 17.3 -- what? An SRM of 17.3 is something expected from a dark lager, not a golden ESB. However, an EBC value of 17.3 looks right on, and conversion to SRM using the formula SRM = EBC * 0.508 results in an SRM of 8.8, which is close to the color calculated by my brewing software, 9.5.

Is Wikipedia wrong, or are my measurements incorrect? Does anyone have a reference for these measurements?

For those interested, here is my procedure:

1. Dispensed 35mL of beer into a centrifuge tube
2. Centrifuged at 10,000 RPM for 10 minutes
3. Passed supernate through a 0.3 micron syringe filter
4. Decanted filtrate to a small vial
5. Pipetted 3 mL into a plastic cuvette
6. Measured the absorbance at 700 nm (This measurement will tell you if your sample is clear of turbidity - a value under 0.039 times the absorbance at 430 nm means your sample is acceptably turbidity free)
7. Measured the absorbance at 430 nm

Results:
A(700) = 0.005784
A(430) = 1.36414
 
I recently corresponded with a guy in the lab at Alaskan Brewing because their listed SRMs are all about twice what their beers look like. Basically, he convinced me they really are measuring and reporting their SRMs correctly using equipment to measure light transmittance. I don't know what it means when their pale ale looks normal and is a 14, their red looks red and is 35, and their 1.065 stout is a seemingly impossible 119. Most breweries don't list their SRMs so I don't know how often this happens, but it certainly doesn't line up with style guidelines and color guides on paper as to what SRM looks like.
 
SRM = 12.7 * D * A, where D is the dilution factor (for an undiluted sample, D = 1, for a sample diluted 1:1 with deionized water, D = 2, etc.) and A is the absorbance at 430 nm.

Yes, that is the correct procedure (for EBC multiply by 25)

When I ran my ESB through, I got an absorbance of 1.36, which resulted in an SRM value of 17.3 -- what? An SRM of 17.3 is something expected from a dark lager, not a golden ESB. However, an EBC value of 17.3 looks right on, and conversion to SRM using the formula SRM = EBC * 0.508 results in an SRM of 8.8, which is close to the color calculated by my brewing software, 9.5.

That's not atypical for an ESB. Here's a list of a few beers of about the same SRM. Note that Fullers is at 17.4 (the less significant figures are useless)

Mad Fox Molotov Hoptail 14.1732
Bitter 14.4145
Samuel Smith Winter Welcome 14.7828
Vienna 15.1638
O'Fest 15.1977
Lindemanns Kriek 15.2654
Anchor Steam 16.002
FdlrsElbow 16.3068
Boom Framb 16.637
Boom Kriek 17.399
Fuller's ESB 17.4498
Oktoberfest 18.0721
Speckled Hen 18.2118
Hercules Double IPA 18.3896
Negro Modello 18.796
Lindeman's Framboise 18.796


Is Wikipedia wrong,
Don't think so, but then again I wrote that article.

...or are my measurements incorrect?

Don't think so again. They seem reasonable to me.


Does anyone have a reference for these measurements?

Yes. I have several. 1 (ASBC) and 3 (MEBAK) in the Wikipedia article are the accepted standard practices.


For those interested, here is my procedure...

This all looks good to me assuming, of course, that the instrument has been checked for wavelength accuracy, linearity, stray light... Spectral bandwidth doesn't really have much to do with it, despite what the ASBC MOA says, because of the low pass nature of beer spectra. But out of curiosity, what is your instrument's bandwidth?


I recently corresponded with a guy in the lab at Alaskan Brewing because their listed SRMs are all about twice what their beers look like.

I am most intrigued by the comments made by both swankyswede and Orthobrewsky to the effect that they know what a given SRM looks like. I too felt that I could guess the SRM based on appearance but thought this strange as what you actually see when you look at a beer depends greatly on the path. A 5 SRM beer in a 10 cm wide glass has the same color (approximately) as a 10 SRM beer in a 5 cm wide glass. I found myself way off when I tried to guess. I had attributed the ability to tell to some sort of adaptation (for example the way we adapt to the different white points associated with different light sources). I think my opinion on this is changing.


Basically, he convinced me they really are measuring and reporting their SRMs correctly using equipment to measure light transmittance.
If they are following ASBC procedure then they are doing it correctly.


I don't know what it means when their pale ale looks normal and is a 14, their red looks red and is 35, and their 1.065 stout is a seemingly impossible 119.

You could hardly be expected to. Most people are not aware that the SRM actually conveys a lot of color information.

SRM 14 means that, in a 5 cm wide glass viewed under illuminant D65 the beer would have a luminance of about 37, a hue of about 59 ° and chroma ~ 74

35 means that, in a 5 cm wide glass viewed under illuminant D65 the beer would have a luma of about 9, a hue of about 26 ° and chroma ~ 35

Thus in going from 14 to 35 the beer gets darker, and redder.

119 implies, under the same conditions, that a luminance of 0.04, a hue of 10 ° (almost on the red axis) and chroma of 0.24 IOW, black. But if you put a very bright light behind it it would look pure red.


Most breweries don't list their SRMs so I don't know how often this happens, but it certainly doesn't line up with style guidelines and color guides on paper as to what SRM looks like.

The color guides on paper or film are an impossibility. The first problem,of course, is that color depends on SRM depth i.e. SRM*thickness_of_glass. The second problem is that you don't need many SRM-cm before you run out of gamut for computer displays, film dyes and CMYK inks. for example, even your SRM 14 beer in a 5 cm glass (a Nonix is 8 cm at the top), has a color which cannot be displayed on your computer monitor or printed on paper.

For example the color chart in the Wikipedia article at http://en.wikipedia.org/wiki/Standard_Reference_Method is way off. No mention is made of illuminant, observer, or path, all of which have an effect on color and with path it is a profound one.

For a beer at 2 SRM to have the brightness (luminance) of the 2 SRM patch in that article it would have to be viewed in a 1.8 cm wide glass (D65) and would have a* of -2 and b* of 20. The patch in the article has a* of -10 and b* of 78.

The SRM 6 patch in the chart has L* = 76.4, a*=1.7 and b* = 76
An average SRM 6 beer would have to be viewed (D65) in a 2.8 cm wide glass to have L* = 76 and would exhibit a* = 11 and b* = 63.

I have no idea where the colors in that chart came from but they aren't beer colors. Do any of us drink beer in 2.8 cm wide glasses (a Kölsch stange is 5 cm)?. I didn't take it out when I re-wrote the article because I knew they'd just put it back in. It is therefore little wonder, I suppose, that most people don't have a feel for the relationship between SRM and color.
 
Yes, they do, but it appears that these may be leading people astray, for the reasons given in #3, more than they help. Other than a photometer (and I still convinced that I ought to be able to build one for under $100), however, there isn't really another option. I suppose home brewers could go back to the iodine scale. It should be more accurate than the cards/transparencies.
 
I'm glad to see someone knowledgeable about this issue stepped in here because I'm very curious about this. Before I talked to the guy at Alaskan, I knew that SRM was measured with actual instruments and there would be some standard depth through which a standard intensity light passes. Of course it is true that beers look darker in thick containers (like a carboy).

However, for example, I've poured a SN Torpedo into a variety of glasses at various levels of daylight and I never think it looks terribly different. There are naturally color expectations in different styles and BJCP stipulates this in terms of SRM. How did BJCP set this up? Didn't they at some point take some beers typical of the expected color range and measure their SRM?

35 SRM is toward the upper end for an Irish stout and it is what BeerSmith computes for the standard Guinness clone recipe. Alaskan Brewery has a red ale and it looks normal in their picture, but has an SRM of 35. I don't know how they lit that pint glass, but it's hard for me to believe it really looks like Guinness. If Guinness really has an SRM of 70, why do both BJCP and BeerSmith think it has 35 SRM? Why are almost all of Alaskan's beers outside style guidelines on measured SRM? Their beers look normal, not just in their own website's pictures, but also in a variety of consumer photos on a Google image search. Also, they've won lots of awards.

Is it the case that two beers with the same SRM can look vastly different in pint glasses sitting on the same table? If so, I wonder what it is about those beers which cause them to look so different. From what I've read, the SRM machines use a very specific wavelength of light and my eyes are using whitish light. Is this the issue?
 
However, for example, I've poured a SN Torpedo into a variety of glasses at various levels of daylight and I never think it looks terribly different.
That's because your eye adapts not only to the level of the light but it's color quality. Our color vision evolved such that we could tell a red (ripe) apple from a green (unripe) over wide extremes of illumination (fire-light, noon day sun, sunrise and sunset) as long as there is enough illumination for photopic (cone) vision. L*a*b* colors, such as the ones I mentioned in the previous post are calculated normalized to the 'color' of an object that reflects or transmits all wavelengths equally i.e. something white. If your eye can't find something white it averages over the whole scene and calls that neutral. You can find lots of examples of how this works is various places. Typically you will be shown a bunch of small squares of different colors inside larger squares of different colors and then, when you read the material, be told that all the small squares are actually the same color but appear different because the surrounds are different colors.

There are naturally color expectations in different styles and BJCP stipulates this in terms of SRM. How did BJCP set this up? Didn't they at some point take some beers typical of the expected color range and measure their SRM?

I really don't know how BJCP came up with their numbers. There are various places one can find lists of beers and their SRM ratings.


35 SRM is toward the upper end for an Irish stout and it is what BeerSmith computes for the standard Guinness clone recipe.
That seems quite low for an Irish stout. Most of mine come in at about 80 SRM. I have measured various Guinness products from 50 to 71.

Alaskan Brewery has a red ale and it looks normal in their picture, but has an SRM of 35. I don't know how they lit that pint glass, but it's hard for me to believe it really looks like Guinness.

A beer of 35 SRM doesn't transmit much light. In a 5 cm glass the luminance is going to be around 9-10 and in an 8 cm glass around 2 (you can think of those numbers as percentages - white object has L* = 100). As I mentioned in the previous post, the color of an SRM 35 beer is too pure to be represented photographically or on a TV or computer screen but not by much. There are various ways to 'clip' colors such as printing the highest chroma color available from the primaries (display) or inks (printed) that has the same hue. The fact that the beer looks normal in the photo is a testament to the photographers lighting and Photoshop skills.

If Guinness really has an SRM of 70, why do both BJCP and BeerSmith think it has 35 SRM? Why are almost all of Alaskan's beers outside style guidelines on measured SRM? Their beers look normal, not just in their own website's pictures, but also in a variety of consumer photos on a Google image search. Also, they've won lots of awards.
It's a simple matter for me to go to their website and measure the color depicted. It is miles away from what a 35 SRM beer is going to look like in a 8 cm (the width of a Nonix at the top) glass. The fact that I can see it on my computer screen says it is within the gamut of the Rec. 709 (ITU) primaries and it is - well within at x = 0.53, y = 0.34 implying that z = .13 which in turn says that this picture contains blue. It also contains green. The color of a 35 SRM beer in an 8 cm path is pure red of wavelength 625 nm with x = 0.74, y = 0.26, z = 0. That picture does not show the color of a beer. But you see the picture as normal. I can only think that this is because by experience you know what beer looks like. Put some of this beer in a glass and hold it next to the monitor and see if you think it looks normal then.

Another interesting aspect of this is that if I ask someone looking at a glass of beer like the one in the picture whether all the beer in it is the same color they say "Of course!" when inspection of the photo with your computers pixel color meter shows that the color is appreciably different between the top and bottom of the glass. Again I think it is the mind playing a trick on you. You know full well that it is the same beer throughout the glass and it must, therefore, be the same color. Now when you tell a person to take a careful look and then tell me whether the color in the thick and thinner parts of the glass are the same, they clearly see the difference.

Is it the case that two beers with the same SRM can look vastly different in pint glasses sitting on the same table? If so, I wonder what it is about those beers which cause them to look so different. From what I've read, the SRM machines use a very specific wavelength of light and my eyes are using whitish light. Is this the issue?

In a sense, yes. SRM is based on a single measurement of absorption at 430 nm. The thing that causes beers with the same SRM to look different is that the color you see depends not only on the beer's absorption at 430 nm but on the absorption at all the other wavelengths between 380 and 780 nm. To compute visible color a complete set of spectrum data is needed. Two beers with the same absorption at 430 nm but different spectrum shapes will appear to be different colors. An extreme example of this would be a lambic and an all malt beer. For this reason I don't record the colors of my beers by just the SRM but rather by what I call the 'Augmented SRM' which I write as, for example, SRM 12.38 [ -0.516, 0.104,-0.008]. The three numbers in brackets are measures of the extent to which the spectral shape for the beer deviates from the spectral shape for an average beer and with them I can reconstruct the spectrum of the beer and from that compute actual color.
 
35 SRM is toward the upper end for an Irish stout and it is what BeerSmith computes for the standard Guinness clone recipe. Alaskan Brewery has a red ale and it looks normal in their picture, but has an SRM of 35. I don't know how they lit that pint glass, but it's hard for me to believe it really looks like Guinness. If Guinness really has an SRM of 70, why do both BJCP and BeerSmith think it has 35 SRM? Why are almost all of Alaskan's beers outside style guidelines on measured SRM? Their beers look normal, not just in their own website's pictures, but also in a variety of consumer photos on a Google image search. Also, they've won lots of awards.

Beer recipe software will generally derive its SRM values from the summation of dilution equations specifying Malt Color Units (MCU) and converted to SRM using some non-linear relationship equation such as Morey, Mosher or Daniels. What you have when you do this, is Morey SRM, Mosher SRM or Daniels SRM but you don't have an instrument measured SRM nor do you have a value that can be labeled Lovibond.

Contrast that to a measured value of SRM using the absorbance at 430nm with a certain path, which has a relationship to Lovibond through EBC. ( I believe the previous poster discusses some of that approach. )

The loose use of color units and their context leaves something to be desired. Compound this with the fact that most software is not using ASTM E-308 much less computing the transmission sprectrum, but rather using a simple SRM->RGB lookup table.
 
Beer recipe software will generally derive its SRM values from the summation of dilution equations specifying Malt Color Units (MCU) and converted to SRM using some non-linear relationship equation such as Morey, Mosher or Daniels. What you have when you do this, is Morey SRM, Mosher SRM or Daniels SRM but you don't have an instrument measured SRM nor do you have a value that can be labeled Lovibond.

SRM is linearly dependent on the concentration of coloring materials in the brew because it is proportional to absorption. Beer-Lambert says

SRM = 12.7*A(430) = 12.7*1*sum_over_i(Ci*ei(430))

where Ci is the molar concentration of colorant i and ei(430) the molar extinction of i at 430 nm. The problem with programs that attempt to model color is they don't know Ci and ei. Ci depends not only on how the malt has been treated but on how the beer is brewed. Heat results in the production of colorants, yeast absorb some coloring material etc. Color estimation schemes can at best come up with a WAG.

Contrast that to a measured value of SRM using the absorbance at 430nm with a certain path, which has a relationship to Lovibond through EBC. ( I believe the previous poster discusses some of that approach. )

SRM = 12.7*A(430); EBC = 25*A(430) with A(430) measured in 1 cm. The 12.7 factor was tweaked so that SRM numbers would agree with Lovibond numbers in the low part of the range.

The loose use of color units and their context leaves something to be desired. Compound this with the fact that most software is not using ASTM E-308 much less computing the transmission sprectrum, but rather using a simple SRM->RGB lookup table.

E-308 is irrelevant to discussion of SRM and EBC color as they derive from one measurement at a single wavelength (430 nm). E-308 is the prescription for the computation of tristimulus (visible) colors in the CIE system and requires absorption data measured over the entire range of visible wavelengths (380 - 780 nm). ASBC does have an MOA (Beer-10c). Some breweries are using it but it is impractical if your spectrometer is not capable of automatic collection of the data set and easy transmission of it to a computer (as, of course, most modern instruments are). Some instruments will do the computation of at least X,Y and Z internally. These can be plugged into the formulas for L*, a* and b* (or L*, u' and v' or whatever you want).

SRM to RGB lookup is impossible without specifying path, illuminant and observer and, as I mentioned in a previous post, anything over about 40 SRM-cm (8 SRM in a 5 cm glass etc.) is too saturated to be displayed by the R, G and B primaries used in computer and TV displays.
 
E-308 is irrelevant to discussion of SRM and EBC color as they derive from one measurement at a single wavelength (430 nm). E-308 is the prescription for the computation of tristimulus (visible) colors in the CIE system and requires absorption data measured over the entire range of visible wavelengths (380 - 780 nm). ASBC does have an MOA (Beer-10c). Some breweries are using it but it is impractical if your spectrometer is not capable of automatic collection of the data set and easy transmission of it to a computer (as, of course, most modern instruments are). Some instruments will do the computation of at least X,Y and Z internally. These can be plugged into the formulas for L*, a* and b* (or L*, u' and v' or whatever you want).

SRM to RGB lookup is impossible without specifying path, illuminant and observer and, as I mentioned in a previous post, anything over about 40 SRM-cm (8 SRM in a 5 cm glass etc.) is too saturated to be displayed by the R, G and B primaries used in computer and TV displays.

Right. I would guess most software isn't generating the transmission spectrum -> tristimulus values -> whatever color space.

You have to think non-scientific when it comes to most programs. A simple lookup table could be made by taking a chart from the internet, reading the RGB values with a paint program and associating them with the SRM value, perhaps interpolating between values.

1 -> 255, 255, 255
2 -> 255, 250, 230
etc...

The programmer just has to use the MCU -> SRM calculations and voila they have a crude "lookup" table.
 
Right. I would guess most software isn't generating the transmission spectrum -> tristimulus values -> whatever color space.
How would one do that without knowing the effects of the mashing and fermentation processes on the concentrations of chromophores and the effects of wort chemistry on their extinction coefficients? For example, lowering pH deepens color.

You have to think non-scientific when it comes to most programs. A simple lookup table could be made by taking a chart from the internet, reading the RGB values with a paint program and associating them with the SRM value, perhaps interpolating between values.

1 -> 255, 255, 255
2 -> 255, 250, 230
etc...

Lots of people have tried to do this and failed. There are no good charts on the internet for the reason I have mentioned several times here: one cannot depict the actual colors of beer within the sRGB gamut. And the other aspect I keep mentioning is that color changes dramatically with path. The same beer viewed in a Kölsch stange is a very different color from what it is viewed in a Maß.

The programmer just has to use the MCU -> SRM calculations and voila they have a crude "lookup" table.

You could start with the beer mentioned by Orthobrewsky which is pictured at http://www.alaskanbeer.com/our-brew/limited-edition/pilot-series/alaskan-imperial-red-ale.html. It clearly states that it is a 35 SRM beer. So, using the color meter in my computer I get

35 -> 189,76,1
35 -> 182,59,27
35 -> 117,47,49

Which would I pick, especially as I know there is no blue in a 35 SRM beer in any reasonable path?
 
How would one do that without knowing the effects of the mashing and fermentation processes on the concentrations of chromophores and the effects of wort chemistry on their extinction coefficients?

Given only an SRM you could take the antilog of the average spectral characteristics to generate a transmission spectrum -> tristimulus -> random color space.

How are you getting the initial SRM? Probably through MCU and one of the SRM estimate equations (Morey, Daniels, Mosher). Though I don't know that the Lovibond -> SRM equivalence is valid unless you're basing your calculations on EBC prior to 1991, for which Lovibond = SRM is close but still not precise.

Lots of people have tried to do this and failed. There are no good charts on the internet for the reason I have mentioned several times here: one cannot depict the actual colors of beer within the sRGB gamut. And the other aspect I keep mentioning is that color changes dramatically with path. The same beer viewed in a Kölsch stange is a very different color from what it is viewed in a Maß.

My point exactly.

You could start with the beer mentioned by Orthobrewsky which is pictured at http://www.alaskanbeer.com/our-brew/limited-edition/pilot-series/alaskan-imperial-red-ale.html. It clearly states that it is a 35 SRM beer. So, using the color meter in my computer I get

35 -> 189,76,1
35 -> 182,59,27
35 -> 117,47,49

Which would I pick, especially as I know there is no blue in a 35 SRM beer in any reasonable path?

Roll a dice?
 
Some interesting things have been brought out here. Apparently, two beers can look very similar to me while one absorbs a lot more blue light than the other. I've found some breweries in which their SRMs correspond to what I'd expect and others like Alaskan. I'm still curious about what specific properties of the beer from different breweries is responsible for a greater or less bias against blue light.

By the way, I realized I haven't posted the actual reply from Alaskan here, though I did on an Alaskan Amber Clone thread. Here it is:

Hi Ed,



My name is Brian Ross and I work in the lab at Alaskan Brewing Company and studied brewing at Oregon State University. I understand your point on the stated and perceived color of our beers, though I assure you that we are measuring color correctly. It's a measurement that we have made routinely for many years.
In response to your question about why our beer pint pictures don't appear to match the stated SRM level, there are several possible reasons for this. The SRM and EBC scales are based on light transmittance and the color of the transmitting liquid is simply implied based upon % transmittance. Two beers with the same measured SRM can appear to be very different colors. Our Red ale color as compared to an Irish stout is a perfect example of this situation. This is one of the downfalls of the SRM system and is part of the reason that much of the rest of the food industry uses a tri-stimulus color scale which is more 3 dimensional, as compared to the two dimensional SRM scale.

Use of the tri-stimulus scale hasn't become the standard for communication of beer color yet because the equipment for measuring the color of a liquid in tri-stimulus is very expensive. Light path length is also a factor in the perception of color as you probably realize. When we measure SRM, that path length is standardized. In real world perception, the path length and lighting can vary greatly which affects your perception of a beer's color. I hope this helps explain the visual discrepancy.

Cheers,

Brian Ross
 
Given only an SRM you could take the antilog of the average spectral characteristics to generate a transmission spectrum -> tristimulus -> random color space.

That is, in fact, how I calculated all the colors in came up with in #3. But I thought we were talking about programs that try to estimate the beer color from the grain bill. WAGing SRM should be possible based on Beer-Lambert but it isn't. WAGing L*,a*and b* is even less likely to succeed.

How are you getting the initial SRM? Probably through MCU and one of the SRM estimate equations (Morey, Daniels, Mosher).
I measure it.

Though I don't know that the Lovibond -> SRM equivalence is valid unless you're basing your calculations on EBC prior to 1991, for which Lovibond = SRM is close but still not precise.

There is no 'equivalence'. The Lovibond system was based on a series of numbered glasses manufactured by the Tintometer Company. The brewer put the beer into a cell, held it up to the light and then dialed in the slide that looked most like the beer. From early on it was recognized that the colors of the standard glasses were not the colors of beer and adjustments were proposed to render them closer. Tintometer took up the recommendations but as not all beers of a given SRM or EBC have the same color the slides still did not match the beers. When the spectrophotometer measure was developed people were asked to determine the Lovibond color of those beers which were then compared to the proto-SRM measurements and the SRM scale adjusted for minimum mean square difference. That's as close to equivalent as the scale gets.

The modern Tintometer is a spectrophotometer (or photometer). It reads in SRM or EBC or Lovibond. There are no colored glasses in it (AFAIK) and I assume, therefore, that it computes LAB color from full spectrum measurements and then compares that color to LAB colors it has defined as being the official Lovibond colors for the individual glasses. This is what I have done in 'measuring' Lovibond colors for beer and I've been able to do that because the spectral absortions of the glasses have been published. It's a simple matter to interpolate between two glass colors and find the point on linear interpolation curves that minimize the CIELAB difference between the interpolated color and the beer color. I reiterate that I am assuming that this is what they do now as it only makes sense to do it that way.





Roll a dice?
What's the point of that? It would be more meaingful to compute the colors of average beers using the hypothecated constant normalized spectrum and produce a chart of colored patches vs SRM-cm not just SRM. You would still be faced with the problem that not all beers at 30 SRM-cm have the same color (because their normalized spectra are not exactly equal to the hypothesized normalized spectrum) and that colors over ~ 50 SRM-cm cannot be reproduced using conventional means but at least you would have included the path which is just as important as the SRM in determining the visible color.
 
That is, in fact, how I calculated all the colors in came up with in #3. But I thought we were talking about programs that try to estimate the beer color from the grain bill. WAGing SRM should be possible based on Beer-Lambert but it isn't. WAGing L*,a*and b* is even less likely to succeed.

I measure it.



There is no 'equivalence'. The Lovibond system was based on a series of numbered glasses manufactured by the Tintometer Company. The brewer put the beer into a cell, held it up to the light and then dialed in the slide that looked most like the beer. From early on it was recognized that the colors of the standard glasses were not the colors of beer and adjustments were proposed to render them closer. Tintometer took up the recommendations but as not all beers of a given SRM or EBC have the same color the slides still did not match the beers. When the spectrophotometer measure was developed people were asked to determine the Lovibond color of those beers which were then compared to the proto-SRM measurements and the SRM scale adjusted for minimum mean square difference. That's as close to equivalent as the scale gets.






What's the point of that? It would be more meaingful to compute the colors of average beers using the hypothecated constant normalized spectrum and produce a chart of colored patches vs SRM-cm not just SRM. You would still be faced with the problem that not all beers at 30 SRM-cm have the same color (because their normalized spectra are not exactly equal to the hypothesized normalized spectrum) and that colors over ~ 50 SRM-cm cannot be reproduced using conventional means but at least you would have included the path which is just as important as the SRM in determining the visible color.

I think we're on the same page. ;)
 
Some interesting things have been brought out here. Apparently, two beers can look very similar to me while one absorbs a lot more blue light than the other.

The obvious question is, of course, 'Are you color blind?'. Almost 10% of men are. All beers absorb light at the blue end of the spectrum more than the red end. Only the paler beers transmit appreciable light at the blue end. The attached diagram shows the colors of beers in 1 cm (grey dots) and 5 cm (black dots). The blue (really x) component of their colors is measured along a line perpendicular to the diagonal straight line with numbers ranging from 535 to the 700[s along it. A 35 SRM beer in a 5 cm wide glass has color depth 175 SRM-cm. The curve of dots (beer color locations) leaves the triangle with vertices labeled 'Rec. 709 G', 'Rec. 709 R', and 'Rec. 709 B' at color depth of about 50 SRM-cm. That triangle is the boundary of colors which can be reproduced by the Rec 709 primaries (which are the same as the sRGB primaries) and thus cannot be displayed on your monitor because the blue 'gun' cannot be set to an intensity less than 0. Any beer outside that triangle can only be shown as a combination of the green and red primaries on your screen. That is why I say the beer in question has no blue though in fact it does contain some, but very little, in the real world. Your beer, in a 5 cm glass, would lie on or very close to the diagonal line with all the other black dots. That line is the line of 0 or close to 0 blue.

I've found some breweries in which their SRMs correspond to what I'd expect and others like Alaskan. I'm still curious about what specific properties of the beer from different breweries is responsible for a greater or less bias against blue light.

I think we've established that it isn't blue light differences that are at play here but rather the relative amounts of red and green. While all malt beers have normalized absorption spectra that are more or less the same, they aren't exactly the same and small changes can make quite a difference. I posted earlier that an 'average' beer of 35 SRM in a 5 cm glass would have LAB color
L*= 9, a*= 31, B* = 15. With a first spectral deviation component of -0.3 that would shift to L* = 15, a* = 35, b* = 27. Most beers have a first deviation component of magnitude less than 0.3 but it isn't that unusual. The CIELAB difference between these colors is sqrt( (9-15)^2 + (31 -35)^2 + (15 - 27)^2 ) = 14 which represents quite a difference (the least noticeable difference is about 1 and 3 or better is considered a good match). What the correlation may be between the spectral deviation coefficients and process or material is as yet unknown.

By the way, I realized I haven't posted the actual reply from Alaskan here, though I did on an Alaskan Amber Clone thread. Here it is:

Hi Ed,


Two beers with the same measured SRM can appear to be very different colors. Our Red ale color as compared to an Irish stout is a perfect example of this situation.


He seems to accept the notion of Irish stout as SRM as low as 35. This is news to me but maybe....?

This is one of the downfalls of the SRM system and is part of the reason that much of the rest of the food industry uses a tri-stimulus color scale which is more 3 dimensional, as compared to the two dimensional SRM scale.


The SRM scale is 1 dimensional. The second reading is only made in order to determine whether the beer is turbid. It is not used in the SRM calcuation. Note that clear beers will fail the turbidity test if they are deviant enough in which case the SRM measurement is still perfectly valid. EBC measures actual turbidity with a nephelometer.

Use of the tri-stimulus scale hasn't become the standard for communication of beer color yet because the equipment for measuring the color of a liquid in tri-stimulus is very expensive.


Not really. There is plenty of stuff in a commercial brewery that is orders of magnitude more expensive than a decent spec but I wouldn't go out and drop that kind of money on the spec first i.e. before I had picked up other things like an RO system, for example.

2DegCIE.jpg
 
If you're going to use a D65 Illuminant because of Rec. 709 B, then you may be varying from the spec. which is a C Illuminant w/ 10 Degree Observer. Of course if you're displaying on screen then you should convert from C - 10 Degree to your choice of Illuminant / Observer by using a Bradford Chromatic Adapation Matrix.
 
He seems to accept the notion of Irish stout as SRM as low as 35. This is news to me but maybe....?

You should look at the beers on this page for which measured SRMs are provided. They all line up well with SRM's given as BJCP style guidelines. Their "Coffee Oatmeal Stout" is 37.1 SRM.

http://www.goodpeoplebrewing.com/ales.php

This is an Alabama brewery and I have personally experienced some of their beers. They look, to me, very much as depicted in these photos. It seems that some brewers produce beers for which the true SRM matches the style guidelines. There seem to be others, such as Alaskan, in which all or most of the measured SRMs are far outside the style guidelines, but whose beers are judged by human eyes as within style guidelines.
 
If you're going to use a D65 Illuminant because of Rec. 709 B, then you may be varying from the spec. which is a C Illuminant w/ 10 Degree Observer.
Which spec? ASBC's?

Of course if you're displaying on screen then you should convert from C - 10 Degree to your choice of Illuminant / Observer by using a Bradford Chromatic Adapation Matrix.

I'm not displaying on a screen. I am just pointing out that trying to display on a screen is futile as many beer colors are out of gamut. I usually do computations based on D65 but have come to realize that there is really no intention to compute visible colors with the ASBC Tristimulus method but rather, as the bloke from Alaska pointed out, get a couple of more degrees of freedom into the color measurement process. One might as well use the equienergy illuminant in this case or better still, take away the pretense that color is being calculated and stick with the spectrum's principal components. That's actually the approach I advocate and how I measure beer 'color'. I also don't like the 10 ° observer. The chromaticity diagram folds back on itself by which I mean that as a monochromatic stimulus wavelength increases beyond a certain point the dominant wavelength decreases as the stimulant wavelength increases. What's that all about?

But look at the transformation matrices from C to D65: pretty close to unitary. This explains in part why D65 has replaced C as the 'standard' illuminant for most purposes.
 
There seem to be others, such as Alaskan, in which all or most of the measured SRMs are far outside the style guidelines, but whose beers are judged by human eyes as within style guidelines.
Remember that SRM has one degree of freedom and human vision 5. You just cannot measure SRM visually. That is why the Lovibond, Brand, Iodine and dicromate color scales are a thing of the past.
 
I just remembered the attached graphic which shows the colors of average beers over a range of path lengths. Run you computer's color meter over this in RGB mode to get a feel for the saturation and blue content story. I don't know what effects the JPEG process might have had on these colors nor any processing the site might impose on posted images.

As the 'average' beer here is based on an ensemble which included some lambics these colors are probably a bit more reddish than they would be had i excluded lambics from the ensemble.

ColorPatch.jpg
 
That diagram should be 4 Dimensional with Weight and Volume or at least 3 Dimensional with MCU.
 
The chart shows the color of beers as a function of SRM and path under the constant normalized spectrum hypothesis (beer color has one degree of freedom) which is approximately true. This means that beer color depends only on the product of the SRM and the path. How the beer was made, what the MCU's of its malts were, how big the batch was and its specific gravity have nothing to do with it. The chart (subject to gamut limitations) shows the approximate color of any beer. So actually the chart should be 1 dimensional as the color at 10 x 2 is the same as the color at 2 x 10.

Beers do deviate from the hypothesis and that's what the PC's are for. If I wanted to show their effects I would have to plot color against SRM, path, first principle component, 2nd PC, 3rd PC.... That is a) difficult to do because I don't know how to plot more than 3 dimensions b) worthless because of the gamut problem and c)assuming it possible, not very useful to anyone since they don't find out what the PC's are until the beer is made and analyzed properly which, as the guy from the Alaska brewery said, most breweries don't do because the gear is too expensive.

The constant normalized spectrum assumption is a very powerful tool for teaching about beer color. You can gain a lot of insight from the picture if you understand what it is telling you even though it suffers from the gamut limitations of the display you are using to look at it.

I guess it would be possible to make something like the Munsell color tree with SRM-cm along the vertical axis and the first 2 PC's as coordinates in planes perpendicular to the SRM-cm axis and I would love to have such a thing on my desk but I couldn't even get the BJCP to print color patches to my specifications for their guide so all I see here is big bucks. Would you buy one?
 
I'd buy one only if it was a hologram hypercube with dynamic voice activated color lookup ;)

I would guess the production cost on something like that would be huge and you'd have a $300+ model. It'd definitely be interesting to see, maybe a virtual one could be made and manipulated in a computer program.
 
The Munsell tree sells for about $360 and I am quite sure their sales are appreciably larger than mine would ever be. I'm sure part of the cost is that many of the chips have to be 'spot colors' as no ink/pigment set can produce the more saturated ones.

I do have, somewhere, a program that allows the user to move sliders to see the effects of SRM, path and the first three PC's but it only runs under a program called IGOR and I'm not sure I could find it if I wanted it. And it has the gamut limitations.
 
It seems there's a difference between:

1.) Measuring SRM (Spectrophotometer, Measure Absorbance, Principal Components, Augmented SRM w/ different beer samples, etc...)
2.) Estimating SRM (Lovibond, EBC, Malt Color Units, SRM, Morey SRM, Daniels SRM, Mosher SRM, Calculated Tristimulus via Augmented SRM, etc...)

My question is, if you were to write a beer recipe software package, what is the best method of estimating SRM from a grain bill, since you can't feasibly expect a homebrewer to have the equipment necessary to measure the SRM.

(Apologies for hi-jacking the thread into a potentially misleading area.)
 
This would be a challenging problem for reasons I have gone into in other posts. The place to start would be with the maltsters' color data. A malt rated X Lovibond produces, in a Congress mash, a wort of color X °L. If you can figure out what SRM or EBC color that corresponds to you now have an estimate of concentration*extinction at 430 nm for that malt (and, assuming the constant normalized spectrum model, for other wavelengths as well). Using the effective concentrations of each malt you would have something like the current MCU model. This has been demonstrated to not work very well and that is doubtless because it does not account for color development in the kettle. The obvious modification is to try to account for that somehow. It appears the Morey approach is to plot realized SRM against MCU and fit the curve which minimizes the mean square error between the estimated and measured SRM values over some set of beers. If all the measured points cluster tightly around the best fitting curve it represents a good model. If they don't it doesn't and one must try to improve on the model by making allowances for boil time, for example.
 
The next step is obviously to convert from an SRM number to an RGB triplet to be displayed on a computer screen.

Two methods exist:

1.) Create an SRM->RGB lookup table by assigning RGB triplets to SRM numbers and interpolating between those that don't have an assigned RGB. Scanning an SRM chart with a colorimeter or spectrophotometer to produce the RGB values (in some color space).
2.) Given the SRM generate the spectrum via antilog of average spectral characteristics, then calculate the tristimulus values via ASTM E-308 and convert to some RGB color space.

What is your recommendation on the best method to get from SRM -> RGB triplet for display on a computer screen?
 
The next step is obviously to convert from an SRM number to an RGB triplet to be displayed on a computer screen.

Two methods exist:

1.) Create an SRM->RGB lookup table by assigning RGB triplets to SRM numbers and interpolating between those that don't have an assigned RGB. Scanning an SRM chart with a colorimeter or spectrophotometer to produce the RGB values (in some color space).

This won't work because there is no mapping between SRM and RGB. That mapping also depends on path, illuminant and observer. Thus there is no chart you can scan.


2.) Given the SRM generate the spectrum via antilog of average spectral characteristics, then calculate the tristimulus values via ASTM E-308 and convert to some RGB color space.

This you can do but you must scale for path. You can get an estimate of the CNS (constant normalized spectrum) from the spreadsheet at wetnewf.org (http://wetnewf.org/pdfs/Brewing_articles/MOAWorkbook.xls), scale it to the path you want and use the color matching functions and illuminant you want from E308 but as all this is built into the spreadsheet you can use it to calculate Rec. 709 RGB and Adobe RGB from a specified SRM. For the average spectrum set all the deviation coefficient (PC's) to 0.

What is your recommendation on the best method to get from SRM -> RGB triplet for display on a computer screen?

I'd say use method 2 (or the spreadsheet which implements it). The results, of course, depend on the CNS you use. The one in the spreadsheet includes lambics and so gives, for PC values of 0, colors that are redder than a CNS estimate estimated from all-malt beers. As I've noted in earlier posts, several times, you won't get very far in SRM-cm before you run into the limits of the Rec. 709/sRGB gamut as indicated by negative blue and green numbers or numbers the spreadsheet can't compute for these colors.

You still can't make a table of RGB vs SRM but you can make one for RGB vs SRM-cm based on the CNS.
 
Back
Top