Not a member yet?
Click here to register.

Forgotten your password?
Request a new one here.
You must login to post a message.

03/27/2023 6:44 AM
ivitamin welcome cup

03/23/2023 10:31 AM
Welcome, bob22

03/22/2023 8:52 AM
Thanks skpacman! all help with our site is much appreciated! ThumbsUp

03/20/2023 8:57 PM
Added some code to combat against bots. If anyone has issues with it, let me know. ❤

03/19/2023 4:55 AM
frankvanhaarlem, fenix and Andyleighmyers welcome to forum

In Memory Of Ginny

Latest Donations
dmccallum - 10.00
JackH - 25.00
snwcmpr - 10.00
Anonymous - 2.00
Anonymous - 5.00
Users Online
Guests Online: 7

Members Online: 0

Total Members: 7,831
Newest Member: ivitamin

View Thread

Who is here? 1 guest(s)
 Print Thread
Roast Color Nerdy Stuff
Is the color really relevant? ... not enough to read the darkness only ?
How many hues could have a "coffee brown" to change the judgment, compared to a monochrome image ?
In colorimetry, brown is an mix of red and green. The coffee brown seems to be an equal balance of the both, I am unable to see any red or green hue in a coffee bean roasted of any degree. Many decades photographer eye...

The guy from Roast Vision sell for 300 his device made using a $5 pulse oximetry sensor, able to distinguish only 36 levels of gray... not even fifty Grin
And SCA quote him as innovative... Shock
Edited by renatoa on 02/20/2023 6:29 AM


renatoa wrote:

If would be mine, according to graph, this would be medium roast.

The Color chart I have, I have no idea how accurate it is but putting the beans on the card they are Full City to Dark.
The Chew taste test dark!

Now the artisan log shows my temps and I think they are low, using my guess at temps Medium?

Using the meter? Because they are positive numbers.

L= lightness
a= (red-green) or redness
b= (yellow-blue) or yellowness

So using the meter reads of the chart I have, the only one I have.

L- 27.38 (How light or dark) puts it at Full City
a=5.82 (How Red) puts it between Full City and Full City+
b=7.62 (How Yellow) puts it between Full City and Full City+

How relevant is the color? I don't know but visually using the chart I put the beans in between Full City and Full city + and I think the meter reading backs that up. Based on eating a bean (Blindfolded) I would say it was a Dark roast.

I'm sure some beans will present darker or lighter then others and the Sumatra is a darker earthier tasting bean?

Do you think I went wrong with comparing the bean to my created color chart?
I’ve been keeping an eye on progress with this thread. Very interesting indeed. Within the visible spectrum Renatoa is spot on with the red green thing, it’s often used in agriculture for detecting ripeness of fruit and veg although not so good for brown.

From a little research it seems the best data might be found in the NIR range so I’ve ordered some bits to make a cheap colour spectrometer. There’s a chap on YouTube with instructions and code on GITHUB.

There are also some nice cheap, visible and NIR spectrum sensors on Sparkfun - plenty to ‘waste time’ on.


HarryDog wrote:

Do you think I went wrong with comparing the bean to my created color chart?

No idea, especially on the chew test...
For my roast I am using mainly the weight loss to judge their roast level, regardless of any color and taste.
But weight loss judgment is valid only if related to greens humidity, else...
If you want an evaluation during roast, then color seems the way to go. For me is development time.
With my current roaster it's not so easy to check the beans during the roast so just checking after as one more data point.

Weight loss is 16.98%, so the sweet maria chart puts this at Burnt?
Other sites put this at high end of Medium.

What are some trusted numbers for roast levels based on temp and or weight loss? A site I should consider correct values to use as guidelines?
17% loss is definitely FULL ! Shock You are good at chewing beans Grin
My roasts are mostly around 14%

Trusted number depends on your green moisture, is it in the 10-11% ballpark, at best ? Or 8-12%, at the worst...
My actual greens is 9.8%, measured in the shop at the moment of buy.
Depending on the storage conditions, you can experience severe loss, even 1% per week during summer and open bag.


ACEMBL wrote:

Within the visible spectrum Renatoa is spot on with the red green thing.

That's only part of it but yes. Info from the wiki.

The CIELAB space is three-dimensional and covers the entire gamut (range) of human color perception.
L* is the lightness value
a* axis is relative to the green–red opponent colors
b* axis represents the blue–yellow opponents

This link is a color representation of my result on this roast.
For those non familiar with the figures above...
CMYK is the color mode intended for printing with ink, i.e. reflected light, i.e. subtractive color model.
RGB is the color mode intended for screen displays, i.e generated light, i.e. additive color model.

CMYK should be used when evaluating real beans color (in the trier for ex.), under a neutral white light.
RGB should be used when looking on a display, a picture of those beans, taken under ideal conditions of exposure and white balance.


HarryDog wrote:


ACEMBL wrote:

Within the visible spectrum Renatoa is spot on with the red green thing.

That's only part of it but yes. Info from the wiki.

The CIELAB space is three-dimensional and covers the entire gamut (range) of human color perception.
L* is the lightness value
a* axis is relative to the green–red opponent colors
b* axis represents the blue–yellow opponents

This link is a color representation of my result on this roast.

Oh I agree CIElab is a good way of describing what’s seen. I was mainly pointing out the fact that the main change is observed in the red green axis. However, from what I’ve read (allowing for some companies not sharing their IP) most sampling equipment uses selective NIR wavelengths for the ‘agtron type’ scales as theres greater quantifiable difference across the roast range.
Yes from what I have read 880nm wavelength is used on some Agtron type units (Much more separation), I think I seen this on a spec sheet for an Agtron model as well. If I see that again will post the info, I just can't say for sure.

I was scanning that info to see what color space would be best to use on this cheap meter. So far samples of Hunterlab and CIElab were close. I just thought CIElab would be better at sampling more yellows? Then Hunterlab that expands in the blue area, but claims to be more uniform. Probably just a toss up.

Renatoa posted that paper from HunterLabs comparing 640nm to 880nm.
Edited by HarryDog on 02/20/2023 9:43 AM
In keeping with the title of this thread, here's some very nerdy stuff. It's a study that delves into time spent in the development phase and color or degree-of-roast, and how both influence the results in the cup.

More nerdy stuff on roast color: Understanding Roast Color
What a great, detailed, and professional presentation of the topic. A shout of thanks to Artisan for publishing these findings and research.
Technivorm Moccamaster, French Press, Aeropress GO, Ibrik, Moka pot, Breville Barista Express, Urbanic070s/static-free with zero retention mods, Breville SGP, Roaster908
I also read the article posted here by MaKoMo. Lot's of info and very nerdy indeed. Grin I recommend it highly for those who enjoy the deep-dive.

It's interesting to me that so much effort is put into accuracy and precision of reading color, which I understand, especially for high production commercial roasters to get something close to repeatability, but we are dealing with an agricultural product that on many levels, varies widely, not only in a single-year's harvest, but also geographically, by genetics and so on...

It would be interesting to see how much variation in the degree of color measured by a professional-level color reading device is actually detectable in cupping tests.
Just to stick in my $0.02: which bean do you select? I have probably several different levels of color in my last batch, not including the kind of green beans when I forgot to hit the start button on the new to me Behmore as it beeped for attention at the 70% roast point.
Typically, roast color evaluation is measured using a typical random sample comprised of enough beans for the sample, then ground and leveled before measuring. The odd bean in the sample should be measured with the rest if the others are left in and not removed.
OK, so a single spot sample will throw up all sorts of issues. Indeed, different lighting, variances in bean, any refraction … will all add to differing results. So multiple sampling controlled lighting and data regression seem to be necessary.

I’m having a little play to see whether I can identify the matching average colour to trigger an event (different method but similar result to ColourTrak). There are a couple of different methods I’m exploring so we’ll see whether either is suitable! My poor little Raspberry Pi is going to be busy! Just a shame my coding ability is almost non-existent.

Great little thread!
I now can see why bean temp is a criteria, that is something you can measure on the fly without so many variables.

Since the maillard reaction is a function of sugar/starch carmalization might co relating sugar content in the raw bean to roast color be more appropriate? Not sure how to determine starch quantity
Edited by Piotrkurak on 03/23/2023 10:50 AM
Thread on HB about a diy color meter.


github info seems to be a work in progress, but looks like it has potential to me.

To bad I don't have a 3d printer or I'd try it now.

If you don’t have a printer there are lots of project boxes out there. I’m doing something similar and am using a small electrical junction box. A bit of plastic putty, A sharp craft knife and you should be good to go!
Also, depending on where you are, there may be very inexpensive places to have the printing done. Where I live, Craigslist, Shapeways, 3dhubs, and others....
Still I am confused about the usefulness of such methods... if they aren't able to tell you degree of roast during development, and tell you with enough precision: Drop ! 3, 2, 1 ... NOW !

Assuming the grounds measurement is reproducible enough... discussions on HB shows they aren't... will ask those following here, generically, not only the thread author Grin :
How reproducible are our roasts ?
If we decide a roast give us the right Agtron, are we able to reproduce it within the same limits as the Agtron measurement ?
Jump to Forum:

Similar Threads

Thread Forum Replies Last Post
CafeMasy SCR-300 Roast Other Roasters 22 03/22/2023 7:51 AM
Fresh Roast SR540/800 Retrofit Logic Board: Comments Needed Fresh Roast 8 and NEW SR 500 13 03/12/2023 12:24 AM
New with Hottop roast profiles, flavors are unimpressive. Roasting Profiles 16 01/24/2023 7:15 PM
Good yet unremarkable Brazilian Roast. Roasting Coffee 11 12/19/2022 5:14 PM
I could use some help with my Burundi Roast. Roasting Coffee 19 10/18/2022 6:12 AM
Homeroasters Association Logo, and all Content, Images, and Icons © 2005-2016 Homeroasters Association - Logos are the property of their respective owners.
Powered by PHP-Fusion Copyright © 2023 PHP-Fusion Inc
Released as free software without warranties under GNU Affero GPL v3
Designed with by NetriX
Hosted by skpacman