Tanzania Iyela NA


Nice tips for soundtrack friends :slight_smile: but, I dont think its safe to put music like this under youtube videos (for a reasons discussed a lot on many youtube channels so I wont start ranting about it here :)) I try to add my sound so that a crack can be heard or other hints in the recorded sound, but have some problems with the way filmic pro records the sound that I need to solve.

Also - I have to say I am thinking hard if making the video is of as much more value as it is much more time spent on it, since I can just make the picture of the plot on the right side and not bother with 10 minutes of HD video at all :smiley: (most of the video is a static frame after all)

To the process and meaning - what I do is make a video with my iphone (lighted with its led “flash” light), on a manual exposure setting so that it does not drift and change in any way. I then make a small cropped part where I think is most representative of the colour (without the light reflected from glass etc) and blur the hell out of it to make average colour. I then make it very small rectangle and move it horizontally, leaving a trail of echoes of colours along the way - that way the coloured strip representing the colour changes is make. I then take it and extract only the luminosity of that strip (sort of black and white but not exactly) and use it to create the top curve that cuts the strip (after some median processing to remove noise and make it into nice curve) so the curve shape on top of the strip shows how the beans get darker (or brighter in the start when they dry to pale and yellow) … I then synchronize the length and size of it to the screenshot of profile and the recorded video … (blurred version representing the colour that was used to plot the colour changes is under the roasting video)
Uff :smiley:


for me the good stuff is in the graph, I do enough watching of my own roasts :slight_smile:


+1, i fully agree, the vid is fun for one time, but overal the graph is what you need.


Ok … that will help me plot those quite a bit faster. Still what I need to solve is some calibration or at least fixed settings so that I get every time same result with the same real colour.


the best way to do that is use a white balance sheet aka ‘grey card’ (or gray card, depending on your favourite anglo lingo). This I use for my photograpy, but I guess it can serve the same purpose for film. What software are you using to get the color histogram?


After effects cc at the moment, but I plan to remake the setup in Fusion which has a free version so anyone can use it.

Grey card is the most correct way, but I also,wanted to make it more accessible to anyone that would like to record this too, so I am thinking of using some prints from a laserjer … that might be enough for at least some calibration. But I will se how it works.


if you want easy acces calibration across multiple users who don’t have a gray card around, I’d suggest to use a single sheet of white paper. Its not as good as gray card but serves it purpose and everyone has easy access to a single sheet of white printing paper. You could even use the IKAWA casing (preferably in broad daylight) as a calibration point. The problem with prints is that you need the same machine making all the prints and then distribute them. If you send around color sheets and people print those out (even on the same type of machinie) its quite likely there will be shift. The ikawa casings however are very likely all exactly the same color. Granted, you’re calibrating ‘wrong’ in regard to an absolute standard, but since we’re comparing ikawa roasts anyway, its good enough to have a relative offset but the same all accross.


Sure … white paper is something accessible to everyone so thats also what I though of at first. But I need more than just a whitebalance (I think whitebalance is not as critical in fact) I need a way how to calibrate the luminosity levels, so to have a recorded “somewhat” white and “somewhat” black that can help with making the shape of the luminosity curve and overall brightness comparable.
My thinking is, that BW Laser printer uses the same material to make black, so if you print full black part is should be quite close in darkness. And also, I think it should be possible to print a chessboard like pattern that would have equal amounts of white and black in small squares (thats how raster in print works anyway) that should after bluring create 50% between those two for more precision.
But I have to find time to build a little rig for capturing the images and testing all those theories. and also I have some problems with setting whiteballance in the app the way I would like to (fixed setting I can recall every time) so thats also something to solve.


well I wouldn’t underestimate the importance of color temperature but yeah I see your point, luminosity is probably a more important factor in this. And interesting approach, interestingly enough I’ve been coaching a start-up company (as shareholder representative) that is working on inkless printing. Hence I’ve seen many many samples from them but also from the industry and although I don’t think its relevant to your approach there’s a LOT of difference in laser printed ‘black’. They all however try to meet an accepted industry standard which is why I think it doesn’t make it really to your approach.


Ah … interesting. I though the black should be for the most part a Carbon and some additives. So - if I get a print with a full coverage on the page (without a raster) …it should be nearly the same “blackness” (if the toner is opaque in the layer thickness that is applied to the paper). I guess it can have some colour to it that may change but I am interested more in how dark it is. What bothers me is that the prints are usually quite shiny so the reflection of the light source close to the lens (as is exactly the case here) would make it not so usefull (The same is true about the reflection on the glass … since I use now the led of the iphone, its simply reflecting on the glass a lot, and I think it changes the image - so I try to find some easy solution that anyone could make and use, but dont have it yet :slight_smile: )

What I would prefere much more, would be to have a ring light that could be very close to the glass to not reflect from it, and a small camera of hero5 session style but way cheaper, that could be setup to full manual and record at few fps, but high resolution and short shutterspeed. If the image of the beans could be sharp, the filtering could be tuned to ignore many areas that are not important, hotspots, chaff that is lighter, or bottom 10 - 20% in luminosity range that is mostly shadows between the beans - so it would give much more valid information about the bean colour. But using the camera and light of mobile device is so much more handy and almost everyone has access to it.


why don’t you use money (a dollar bill or pound or something) as the calibration? Use a light by the same company with x amount of lumens etc. and the flash off on the phone? y/n?


Money could work, but even with that there are several types of notes and lets face it, this is an international community. Using ‘the same light’ is also not gonna work as the recorded luminescence also depends on the distance between the source and the object and the environment AND the shutter-speed and sensitivity of the sensor of the recording device!. Making a good calibration is REALLY hard. Which is why I suggested using the exterior of the IKAWA for white-balance, but that probably will not give you enough accuracy in luminescence. As for light sources, I would suggest a heat resistant led on a wire which you place beneath the glass. Note that you don’t really require high shutter-speed (high shutter = less light on sensor) its the overall colour in which you’re interested, thus its not really that important that the image is a bit blurry.


Well, anything printed including money changes colours due UV light and other aging factors - so I do not think its any more stable. I still think the white paper and something black on it should be good enough, may it be a laser printed something, or say using a soft pencil to make a black shape there, or some other widely available thing, black marker of the same kind available worldwide etc.

The shutter speed is dependent on if we just blur it to get average colour, or if we want to do some filtering before to eliminate some parts of the image that we do not want to see in that average value.
If we just blur then image with motion blur is ok, if the processing prefered then no motion blur is better.


Hi @pavel, I just got possibly the same greens? or very similar from a local roaster. Imported by Nordic Approach.
It’s Tanzania Iyela (Kent AB), Songwe, Mbozi. Cup score 88. Washed. Roaster taste notes are Yuzu, hops and jasmine, 1800 to 1900m elevation. Think it might be lot #30.

My first roast (fast Pro mimic) is really nice. It is a very light bright coffee with a body like sweet fruit tea. Great fruit acidity with sweet citrus, (never had a Yuzu so I don’t know) but it tastes of sweet lime/grapefruit) apricot and a hoppy aroma. Really nice coffee.


Hello Stephen. Looks like a different farm from the farm name and taste profile. But probably also very nice from you description. So … tell us more :))


I thought the NA in your beans title was “Nordic Approach”? Maybe you meant AA for the grade?

What more can I tell you?


Yes NA is for Nordic Approach … but they have obviously more of the Tanzania Iyela versions, deiferent farms, lots … yours is not the same as the one I had … :slight_smile:


Yes. It must be a different lot. We are agreed though that is very nice coffee!


Yes indeed :slight_smile: could you share your profile for it?


Very probably the one I had is AAA … but I do not have the complete description so … just what I can see now.