Very cool. A couple comments:

1) Please contribute to this open source effort by uploading the 3D print files and other supporting info as a fork to the GitHub. Others may be interested in this idea.

2) How much coffee and surface area is needed to perform a reading with this case? One benefit of the current design is 4-5 beans are sufficient to do a reading.

Transparent Roaster wrote:As far as I can see, the color meters that are on the market typically take a top-down scan versus a bottom upwards scan. Devices are placed over a sample, and not vice versa.

I designed a new case with that in mind. With this approach, no coffee or residues will ever cloud the sensor window. Another detail is that it seemed the original roast vision calibration could vary depending on how much pressure was applied to the calibration material placed over the scan glass. That is less of an issue with this design, and the readings are stable on its calibration pad.

This design can be placed on a single sample and multiple readings taken for an easy average of the color... on the original, you'd have to dump the sample multiple times and measure again. Overall, the user experience is faster and cleaner. I use the meter on my espresso grind just before pulling the shot. I'm now taking a reading beforehand to better understand my roasts.
image

Brilliant, I was suggesting this a while ago but not having a 3D printer made it a mere thought.
Can you show the 'business end' underside too?

Transparent Roaster wrote:As far as I can see, the color meters that are on the market typically take a top-down scan versus a bottom upwards scan. Devices are placed over a sample, and not vice versa.

I designed a new case with that in mind. With this approach, no coffee or residues will ever cloud the sensor window. Another detail is that it seemed the original roast vision calibration could vary depending on how much pressure was applied to the calibration material placed over the scan glass. That is less of an issue with this design, and the readings are stable on its calibration pad.

This design can be placed on a single sample and multiple readings taken for an easy average of the color... on the original, you'd have to dump the sample multiple times and measure again. Overall, the user experience is faster and cleaner. I use the meter on my espresso grind just before pulling the shot. I'm now taking a reading beforehand to better understand my roasts.
image

I assume you are the person who posted about this on discord. You have done something very similar to what I've been working on. Or rather not working on as I haven't had any time for it recently. Perhaps I will abandon what I'm doing and copy yours -- provided you are willing to share the details of construction.

I'm curious how you control the distance between the sensor and the coffee surface. Very small differences affect the reading. I think this is at least part of the reason why pressing on the sample in the DIY meter can change the output.

The workflow improvement kind of depends on the context. For me, and perhaps you too, grinding 7-10 beans into Turkish powder is simply a waste of coffee and time. I'd much prefer to measure the espresso grind that I am about to pull a shot with. But clearly that's not true for everybody. Even if you have to grind three Turkish samples to get a reliable number that's still only a few grams of beans versus 15-20g for an espresso shot.

It was my hope that by making repeatability easy it would result in improved precision. Nobody wants to measure 10 samples in the DIY meter but it should be easy to do that in a portafilter basket full of coffee where you can just reposition the sensor in a second. But *is* there a big improvement? That's another aspect that I would hope you might share with us: data comparing coffee measured with your design versus the existing DIY meter design.

Another issue, perhaps not as important, is the accuracy. At present the DIY meter is calibrated with baking soda. That's all fine except that we're measuring coffee and its reflectivity is a function of the grind and surface preparation. So two people with identically calibrated DIY meters might get different readings depending on how they ground their coffee. So that's another question I have: how much of difference in reading do you get because you are grinding coarser than Turkish? And how much does your reading then depend on dialing a particular coffee in as opposed to the coffee actually being lighter or darker?

Thanks for the comments, questions and curiosity.

An 18 gram dose is my sample. It gets tamped for a nice, compact, flat surface then and measured before pulling a shot. As jpender, and most of you know, even slight variances of the sample height from the sensor will shift the reading a few numbers either way, not too badly, but definitely skewed. I tested calibrating the device with unbleached Hario filters(186-188), Aeropress GO white filters(209-210), and sodium bicarbonate(230). The current calibration is with Hario filters. As for variances in readings based on grind size, I haven't got there yet as I've been using the same roasted beans and the same grind setting for a few weeks now. The readings have been predictable and stable.

About my model, actually, I'm not on Discord or Github and haven't posted anything there. Similar to my various other posts here and on the HR site, the model and posted observations are an arrow to point others in a new direction to be able to design their own device. If someone would like me to build one for them, they can send me a personal message and we can work it out from there. I am still working on a few refinements though... there's always more R&D.

Transparent Roaster wrote:About my model, actually, I'm not on Discord or Github and haven't posted anything there. Similar to my various other posts here and on the HR site, the model and posted observations are an arrow to point others in a new direction to be able to design their own device.

I had you pegged as that guy on discord who recently posted about the same basic idea. The same idea I had and I guess Marcel too. Probably others as well.

It would have been fun to see what you had done exactly. Oh well. Cheers!

If distance to grounds is so critical, why not build this device as a tamper

Early on I did a variance analysis of multiple readings of a single sample and found maybe 1-2 Agtron difference.

Also, if the goal is averaging a roast's color, it's more valuable to take one reading from multiple samples than to take multiple readings of a single sample.

So I quite like the original coffee-on-meter design because it uses less coffee. And cleaning the glass after isn't a big deal.

renatoa wrote:If distance to grounds is so critical, why not build this device as a tamper :?

That's what I was thinking about. I believe the guy on discord has the same basic plan. The existing DIY meter is essentially the same in that the coffee is right up against the glass. But it's very sensitive. The distance between sample and sensor is about 3mm in the DIY meter. Just considering the 1/r² change in intensity you'd expect ~1% difference at the sensor with only a 15µm change in distance.

Brewzologist wrote:Early on I did a variance analysis of multiple readings of a single sample and found maybe 1-2 Agtron difference.

But don't you also see outliers? I found that I couldn't rely on a single reading because of that.

Brewzologist wrote:Also, if the goal is averaging a roast's color, it's more valuable to take one reading from multiple samples than to take multiple readings of a single sample.

I think the idea was that you'd essentially be taking readings of multiple samples by adjusting the sensor to look at different places on the surface. But I agree with you that the original design has merit.

the idea was indeed to use one larger sample to essentially provide multiple readings faster, you'll need to grind a certain amount anyway and that sample needs to be finely ground. I'd like to skip the glass covering since the sensor is protected by a glass cover already.
Playing around with the sensor directly on the sample showed very small variability, in my experiment. I was contemplating using some resin to end up with a solid block

Marcelnl wrote:I'd like to skip the glass covering since the sensor is protected by a glass cover already.

From the MAX30101 data sheet:

"The MAX30101 comes in an OLGA package that is not sealed from dust or liquid....
...The MAX30101 should not be cleaned with a liquid solution, baked, or coated with anything."


From an application note about the package type:

"We recommend that taller components are placed near the OLGA package to avoid damage to glass during handling....
...All OLGA parts have vent holes for better moisture sensitivity and reliability. Vent holes should not be blocked during or after SMT."


FWIW