Brewing reference using refractive index instead of extraction yield

Coffee preparation techniques besides espresso like pourover.
dukeja

#1: Post by dukeja »

Recently I realized the availability of cheap refractometer option such as this one:
https://www.amazon.com/gp/product/B07V5K9HXY/
(I made a mistake of purchasing a 0~40% Brix one, which does not have enough angular resolution. Then I got the above one, which has 0~10% Brix and it will come in a few days)
And I start to have fun using it as the reference for my brewing.

I am an optical researcher and have been working on all kinds of sensors for refractive index (RI)-based optical sensing for two decades, so I am comfortable playing with this very simple gadget to see if I can get most out of it. There are all kinds of sensing techniques and temperature compensation schemes that one can use for better accuracy or repeatability. But that is not the point of this post, so let's not get into that direction. (But you do need to carefully control your sample temperature, as it is shown in this post https://coffeeadastra.com/2019/09/21/me ... precision/)

What I am interested is that I realized extraction yield (E.Y.) that was often talked about may not be a good "brewing reference" for me. I am not trying to link E.Y. or TDS (R.I.) to the taste in the cup, which is a lot more complicated than what an single R.I measurement can tell. However, I do be able to roughly using R.I. to gauge the variation of brewing methods such as V60, Kalita, Aeropress, etc. and brewing profile (bean/water ratio, temperature, grind, water, pouring profile, etc) and maybe even roasting profile, provided we use the same batch of beans. This is what I am interested.

E.Y= (total amount of extracted coffee) / (wt. of bean) * TDS
or in my case E.Y. = (total amount of extracted coffee) / (wt. of bean) * (0.85*Brix)

Looking at the equation and we all know that E.Y. is a poor indicator of "strength" of the coffee, since one can use more water to extract and raise the E.Y. and yet get a over-extracted and even diluted coffee. Not to mention Kalita often withholds small amount of coffee between filter and bottom of the cup, or coffee still inside of the ground in V60. If you accidentally spilled a few drops on the table, the "total amount of coffee" can be quite unreliable.

That is why I think the TDS or Brix or R.I. (which was what the refractometry raw data) is much better and normalized "brewing reference". We certainly can use a carefully measured E.Y. as a global sanity check. However, if I do a batch of several brewing and try to figure out the correlation between "best taste cup" and a measurable brewing reference. R.I. (or TDS or Brix) is a lot easier indicator, since the water/bean ratio is not in the equation.

I found this brain exercise beneficial to me. Now I understand why barista will do so-called "by-pass" in their pouring profile and the its effect is clearly shown in R.I measurement but ambiguous in the E.Y. measurement!

Enjoy!

jpender

#2: Post by jpender »

dukeja wrote:Looking at the equation and we all know that E.Y. is a poor indicator of "strength" of the coffee, since one can use more water to extract and raise the E.Y. and yet get a over-extracted and even diluted coffee. Not to mention Kalita often withholds small amount of coffee between filter and bottom of the cup, or coffee still inside of the ground in V60. If you accidentally spilled a few drops on the table, the "total amount of coffee" can be quite unreliable.

Extraction yield isn't a poor indicator of strength; it's orthogonal to strength. In the context of a specific recipe one could use strength instead of E.Y. but in general strength doesn't by itself tell you anything about how well the coffee is extracted. Strength does affect taste perception and E.Y. isn't a reliable indicator of taste but at least there is a correlation.

I'm curious about the refractometer you linked. Its resolution is poor for use with brewed coffee (<2% TDS) and doesn't have enough range for espresso which is frequently greater than 10°Bx. I would imagine being frustrated with it.

Thanks for linking that article on Gagné's blog. I've wondered about the effect of both sample temperature and evaporation. It's nice to see it laid out so clearly.

dukeja

#3: Post by dukeja »

jpender wrote: ... in general strength doesn't by itself tell you anything about how well the coffee is extracted. Strength does affect taste perception and E.Y. isn't a reliable indicator of taste but at least there is a correlation.
That is exactly what I said. Neither of the measurements can tell the quality of the coffee. It is the correlation that you established in the series of trials that give "interpolation" (repeatability within the constrain), meaning, provide the guidance for future roasting and brewing of the similar beans.

Still I would argue that R.I. (or TDS) is a normalized intensity (just like density), and the E.Y. is the total measure (just like total mass). "Density" is a lot better indication on the strength (not quality) of the coffee.
jpender wrote: I'm curious about the refractometer you linked. Its resolution is poor for use with brewed coffee (<2% TDS) and doesn't have enough range for espresso which is frequently greater than 10°Bx. I would imagine being frustrated with it.

Thanks for linking that article on Gagné's blog. I've wondered about the effect of both sample temperature and evaporation. It's nice to see it laid out so clearly.
It could be pretty frustrating. Fortunately, for engineer like me that are used to control lab variable to reach 1E-6 R.I measurement or better, it is a breeze. :) There is some technique needed to enhance the reading in addition to the temperature, such as illumination angle control, viewing angle, or if you could have some camera doing the machine vision for you. It is not too bad. And as I said, I have another one coming with 1/4 resolution, which will make it even easier. I'll report back.

dukeja

#4: Post by dukeja »

Thanks for chiming in for quality discussion. Sometimes it is pretty quiet here. :)
dukeja wrote:That is exactly what I said. Neither of the measurements can tell the quality of the coffee. It is the correlation that you established in the series of trials that give "interpolation" (repeatability within the constrain), meaning, provide the guidance for future roasting and brewing of the similar beans.

Still I would argue that R.I. (or TDS) is a normalized intensity (just like density), and the E.Y. is the total measure (just like total mass). "Density" is a lot better indication on the strength (not quality) of the coffee.



It could be pretty frustrating. Fortunately, for engineer like me that are used to control lab variable to reach 1E-6 R.I measurement or better, it is a breeze. :) There is some technique needed to enhance the reading in addition to the temperature, such as illumination angle control, viewing angle, or if you could have some camera doing the machine vision for you. It is not too bad. And as I said, I have another one coming with 1/4 resolution, which will make it even easier. I'll report back.

DamianWarS

#5: Post by DamianWarS »

dukeja wrote:Recently I realized the availability of cheap refractometer option such as this one:
video
(I made a mistake of purchasing a 0~40% Brix one, which does not have enough angular resolution. Then I got the above one, which has 0~10% Brix and it will come in a few days)
And I start to have fun using it as the reference for my brewing.

I am an optical researcher and have been working on all kinds of sensors for refractive index (RI)-based optical sensing for two decades, so I am comfortable playing with this very simple gadget to see if I can get most out of it. There are all kinds of sensing techniques and temperature compensation schemes that one can use for better accuracy or repeatability. But that is not the point of this post, so let's not get into that direction. (But you do need to carefully control your sample temperature, as it is shown in this post video)

What I am interested is that I realized extraction yield (E.Y.) that was often talked about may not be a good "brewing reference" for me. I am not trying to link E.Y. or TDS (R.I.) to the taste in the cup, which is a lot more complicated than what an single R.I measurement can tell. However, I do be able to roughly using R.I. to gauge the variation of brewing methods such as V60, Kalita, Aeropress, etc. and brewing profile (bean/water ratio, temperature, grind, water, pouring profile, etc) and maybe even roasting profile, provided we use the same batch of beans. This is what I am interested.

E.Y= (total amount of extracted coffee) / (wt. of bean) * TDS
or in my case E.Y. = (total amount of extracted coffee) / (wt. of bean) * (0.85*Brix)

Looking at the equation and we all know that E.Y. is a poor indicator of "strength" of the coffee, since one can use more water to extract and raise the E.Y. and yet get a over-extracted and even diluted coffee. Not to mention Kalita often withholds small amount of coffee between filter and bottom of the cup, or coffee still inside of the ground in V60. If you accidentally spilled a few drops on the table, the "total amount of coffee" can be quite unreliable.

That is why I think the TDS or Brix or R.I. (which was what the refractometry raw data) is much better and normalized "brewing reference". We certainly can use a carefully measured E.Y. as a global sanity check. However, if I do a batch of several brewing and try to figure out the correlation between "best taste cup" and a measurable brewing reference. R.I. (or TDS or Brix) is a lot easier indicator, since the water/bean ratio is not in the equation.

I found this brain exercise beneficial to me. Now I understand why barista will do so-called "by-pass" in their pouring profile and the its effect is clearly shown in R.I measurement but ambiguous in the E.Y. measurement!

Enjoy!
a common equation is Extraction Yield (E) * Dose (D) = TDS (T) * Beverage Mass (B) or ED = TB. with algebra you can isolate one value to suit your needs:

ED = TB
E = TB/D
D = TB/E
T = ED/B
B = ED/T

of course, EY is only useful when we're talking about consistent beverage mass or at the very least consistent with a similar brewing methods like espresso, brew, Moka pot etc... So when you have a 25% EY in a pool size of water it's only going to taste like water. EY doesn't care how much water you use it's only a measurement of how much coffee stuff you got out of the coffee. If you keep pouring water over the coffee bed eventially your going to extract everything you can out of it and achieve the highest extraction but at the cost of over diluting it . TDS is based on a small sample size that is consistent if it's more diluted/concentrated your sample size is diluted/concentrated with it.

the most known coffee refractometer is the VST by a guy named Vince Fedele (more properly his company) where he had to
find the relationship between TDS and RI to get a meaningful measurement. before refractometers, we had to dry out the coffee and weigh the dry contents to know how much of the coffee got in the cup which was tedious and expensive as it required special lab dehydrating ovens to do it properly. Vince Fedele took thousands of these samples and developed some sort of algorithm to determine the relationship between TDS and RI and this algorithm is his IP and really the meat of the VST. Other coffee refractometers like the Atago do the same thing but it's just based on their algorithm, not VST. Proper dehydration is still the most accurate way but I think with the introduction of refractometers to the world of coffee the refractometer is the defacto norm even if it's not the most accurate.

SCA says 18-22% is our aim but this is based on research conducted in the 50s by a guy named Earl E. Lockhart and it is still used today (I just took an SCA course and it taught exactly this) but I think this can be challenged or at least new models for determining good extraction. To start Lockhart's research doesn't include the interstitial fluid (like your example about using the Kalita). All coffee methods have trapped fluid in the coffee bed that we do not factor in when determining the EY which comes back to your point that TDS is a better way.

another thing this magic number of 18-22% or any EY percentage is that it actually doesn't tell us if the coffee is good or not. We can brew better in such a way that we raise the extraction ceiling to allow a better EY at 25%, but poor techniques can do the opposite and lower the extraction ceiling so even at 18% the coffee is still astringent. We may hit the target but because of being sloppy with all kinds of other factors, it's bad coffee so I think EY or even TDS in that matter should be measured after we taste the coffee so we don't approach the coffee with some sort of confirmation bias because we already know where these numbers fall into and have predetermined what is good and what is bad. Let's determine it's good first by the most important senses (our own taste perception) and then look at the RI.

dukeja

#6: Post by dukeja »

Thank you for the reply and some interesting background about TDS and RI. And we all agree that none of these measurements can be the absolute indicator of the taste, but use it as a one-dimensional measuring stick then it may guide us better than walking in the dark. Ultimately our taste is the judge, but the funning thing is that our taste also can be altered by our health condition and even the food/drink before we taste, so even taste is pretty unreliable. That is why a 1D measuring stick is still better than nothing, as a brewing reference.

Ultimately, if we can identify a few critical chemicals, we can do some absorption spectroscopy either UV/VIS/IR, which will provide the multi-dimension indicator corresponding to the absolute taste better.

So is RI measurement still useful? It certain is for me. I can use the exactly same bean, same grind, same pouring condition (water chemical, temperature, drain down time, etc) but achieve quite dramatic RI difference (and taste quite different) with different V60 pouring mechanics (pouring height, location, and timing). And it is extremely useful to determine the "by-pass" water, which is the last critical stoke of the brewing process. A brewing reference, whatever it is for you, is beneficial to perfect the art!

jpender

#7: Post by jpender »

dukeja wrote:So is RI measurement still useful? It certain is for me. I can use the exactly same bean, same grind, same pouring condition...
Sure. In that case it's essentially equivalent to E.Y., differing only by a constant.

The thing is, if I told you my coffee had an E.Y. of 13% you'd expect it to taste poorly. Whereas if I said the strength was 2% you'd have no idea without knowing more information.

Fortunately, for engineer like me that are used to control lab variable to reach 1E-6 R.I measurement or better, it is a breeze. :) There is some technique needed to enhance the reading in addition to the temperature, such as illumination angle control, viewing angle, or if you could have some camera doing the machine vision for you. It is not too bad.
I'm very curious about this. I've never wanted to spend hundreds on a refractometer so I have used my kitchen oven and a $20 milligram scale to determine strength. It's accurate to 0.03% TDS or better but it takes too long to be practical for most uses. I've wished for an alternative.

What precision are you able to achieve with your $30 refractometer? And what specifically do you do?

dukeja

#8: Post by dukeja »

jpender wrote:Sure. In that case it's essentially equivalent to E.Y., differing only by a constant.
Not trying to be tense here, but I respectfully disagree. Using light as an analogy, 60 W light bulb is nothing, but 60W laser will cut and kill. The difference is that laser spot is much smaller than your room in terms of "area" so the "normalized intensity" (called "irradiance") is much much higher in the laser case. E.Y. is like total power, but R.I. is the "normalized intensity" or "irradiance". E.Y. can be diluted by large volume, but R.I. will not, since the factor of volume (or area in the example) is "normalized" out.
jpender wrote: I'm very curious about this. I've never wanted to spend hundreds on a refractometer so I have used my kitchen oven and a $20 milligram scale to determine strength. It's accurate to 0.03% TDS or better but it takes too long to be practical for most uses. I've wished for an alternative.

What precision are you able to achieve with your $30 refractometer? And what specifically do you do?
I'll report it back when I get it on Monday. Even using the "wrong" one (0~40% Brix", I am already helped. :)

jpender

#9: Post by jpender »

dukeja wrote:Not trying to be tense here, but I respectfully disagree. Using light as an analogy, 60 W light bulb is nothing, but 60W laser will cut and kill. The difference is that laser spot is much smaller than your room in terms of "area" so the "normalized intensity" (called "irradiance") is much much higher in the laser case. E.Y. is like total power, but R.I. is the "normalized intensity" or "irradiance". E.Y. can be diluted by large volume, but R.I. will not, since the factor of volume (or area in the example) is "normalized" out.
No worries about disagreement, as long as you don't mind me disagreeing with you. :-)

For percolation brews, E.Y. = strength * (beverage size / dry grounds). So if you keep the recipe the same E.Y. = strength * constant. Sure, it's not going to be exactly constant but probably close enough that you won't be able to taste the difference.

Ignoring the fact that light intensity is different between lasers and light bulbs for reasons other than just beam diameter, I think that analogy is kind of like the difference between a shot of espresso and a long black. Same extraction, very different strength (e.g 10% vs 2%). Generally speaking, it's the quality of the extraction that correlates closest with how good it tastes, not the strength. You can't really fix a bad espresso by adding water. You can mute some of the bad tastes a bit but not make it great. On the other hand, a well extracted espresso makes a delicious long black.

dukeja wrote:I'll report it back when I get it on Monday. Even using the "wrong" one (0~40% Brix", I am already helped. :)
I must have misunderstood. I thought you were already doing something to improve precision beyond what the manufacturer claims.

dukeja

#10: Post by dukeja »

jpender wrote:No worries about disagreement, as long as you don't mind me disagreeing with you. :-)
...
I must have misunderstood. I thought you were already doing something to improve precision beyond what the manufacturer claims.
We are good here. In terms of coffee, whatever float the boat works. :)

I only have the 0~40% Brix reader at hand, and I have no problem reading it at 0.1% by eye. (even my eyes are pretty bad now) The 10% Brix reader will be even easier to read. Using a light moving around the flat top side to change the illumination angle will help to identify a very sharp edge, which is the boundary formed by the critical angle.

I have no need to use machine vision yet, but if there is enough interest I'll try my Raspberry Pi camera to make a demo. I'll also try to run some Zemax simulation that may come out some low-cost lens solution to replace the existing one. As I said, it is fun toy to play with.