tinkering with: colour sensor

[@AnotherMaker]

You got way further along with this thing than I did. I wanted to use it in an escape room and never got it to work reliably. Very nice work.

[@ronm6585]

Colours were close by my old eyes. Thanks for sharing.

[@DubiousEngineering]

Mooooooo brew... straight out of the cow :-). I tried the colour sensor on the robot lawnmower and struggled to sense green grass... big hugs man hope all is well in Canada!

[@tinygriffy]

Very interesting ! Thanks :) datasheet says this IC delivers 16 bit values for each colour channel and has a dynamic range of 1:10000000 for both colour temp and ambient lightning, it also says that it is best operated behind dark glass (an OLED) .. I guess it could do much more, if the library would allow for it (maybe it does..)

[@Taran72]

This is an interesting sensor! You covered pretty much everything, thanks. DroneBot's codes are the only ones that work 100% all the time.....this from my personal experience. If DroneBot has it I don't go around looking for other sources.

[@adam850]

The colors were pretty close. There was some lightening of the colors, which may have been glare from the LEDs. Definitely want to see those RGB values piped to a LED panel.

[@markfergerson2145]

Agree on the glare- a short piece of (something) around the sensor to block direct light from the white LEDs but permit bounce light from the, uh, reflective surface under test would prevent washout. A RGB LED display wouldn't help the color-blind a whole lot though... maybe some stepper motors driving a pointer over a CIE diagram? ;>)

[@azariahjosiah2123]

i realize I am quite off topic but does anybody know a good website to stream newly released series online ?

[@cyrusthatcher5936]

@Azariah Josiah I would suggest Flixzone. You can find it on google :)

[@TheUnofficialMaker]

would be interesting to use a fiber optic cable with one end to the sensor and the other to an RGB led to see if you can detect the colors. Project in mind: detect color of led on alarm panel to feed another circuit without electronic connection to panel. May have to pick one up.

[@GnuReligion]

An interesting programming exercise would be to make an RGB LED reproduce the color just recorded by the sensor, using feedback. I like to use wax paper over a conical reflector for homogeneous appearance. Were you telling us you are color blind?

[@onecircuit-as]

Tassie beer for the win! Interesting sensor and a nice exploration of its capabilities. 👍🍺

[@stevetobias4890]

Dronebotworkshop is an excellent resource for micro processor resources. This would be a good module for anyone into art or colour mixing. For the rest of us it's next to useless. 😂

[@johncoops6897]

It'd be totally useless art or colour mixing too, due to inaccuracies in colour sensing!

[@frankowalker4662]

The colours were pretty close. I too would like to see it hooked up to an RGB LED.

[@sirgoodenough65]

Another beer might make the numbers even better but better than I expected!

[@TheEmbeddedHobbyist]

So now we can have more numeric data on how black the beer is, great. :)

[@markclark787]

I have great color vision, this was interesting. If you could get a hold of true color test samples I think your results would be better.

[@johncoops6897]

"true color test samples" ??? ROFL, what the hell are you babbling about?

[@kirkb4989]

So can it measure the color of beer??

[@pileofstuff]

I'll have to work on a modified pint glass...

[@ElmerFuddGun]

You need some code to convert the RGB colour values to an actual colour (hue). The RGB values will change with brightness even though the colour is not changing. Just looking at these RGB values doesn't tell us (humans) much except for the when one or two colour values are way different than the others. ie. is a RGB of 0,200,200 different from 0,100,100? Also, the small blue colour sample might have been letting the white board reflect back to the sensor.

[@johncoops6897]

RGB values certainly change with brightness - the numbers represent the amount of Red, Green and Blue. So, a RGB of 0,100,100 is the same colour as 0,200,20 but not as bright. RGB cannot really translate properly between paint and a screen because they work in completely different ways.

[@jyvben1520]

3d print a ring, or make the leds only illuminate upwards

[@johncoops6897]

FFS, you don't need to use a 3D Printer to create a ring. He already made one... didn't you watch the video? BTW - the LEDs already illuminate upwards.

[@jyvben1520]

@@johncoops6897 and sideways, an inner ring like on the other model he showed might help

[@colinwinterburn6136]

I have a q!uestion. Do you live in Winterpeg?

[@pileofstuff]

Yup.

[@andymouse]

I figured the sketch would light up an LED according to what colour was in front of the sensor, so 3 LED's on the breadboard connected to 3 pins on the Arduino, stick a red patch in front of sensor, red LED lights up, I think the numbers would need some 'math' though, anyhow interesting vid ...cheers.

[@TheWareek]

as we always say in maintenance if you got it done right its a fluke

[@AndyHullMcPenguin]

A fun little sensor and the results were pretty close, but not quite correct. This may be a limitation caused by your setup, or by the sensor itself, however there is also the colour spectrum of those LEDs to factor in. You could try the same thing with some improvised gel filters (transparent coloured sweet wrappers perhaps) using daylight. You would then have a better set of daylight spectrum reference colours to calibrate with.

[@pileofstuff]

I actually have some real gels around here somewhere. But they're probably not labelled, so it would be hard to know if they were being detected correctly.

[@TerryLawrence001]

I would grab some paint chips from the paint department at the Home DIY store and build a library of colours. Orange and yellows are hard to reproduce on an rgb lcd so they look wrong but are actually pretty close if you discount the poor lcd rendering.

[@pileofstuff]

In a normal time, I would have gone to the paint store - just doesn't feel right going anywhere I don't have to these days.

[@fredflintstone1]

Odd little video, not sure of what use the item would be though:-) now if you could get it to read resistor colour codes then:-)

[@pileofstuff]

That would be useful. I suspect it would take quite a lot of messing around with optics to make that happen reliably.

[@wooferhound7571]

Now you can automatically sort out your Skittles and M&M's Or Definitively rate your beer: Red, Brown, Amber, Stout . . .

[@strayling1]

That'll come in handy for when rock stars visit.

[@johncoops6897]

The limitation here will be the dreadful colour rendering of those shitty little 5mm white LEDs, and that'll be the biggest challenge to getting any form of logical outputs from the module. No, even having a high "CRI" won't help, because the CRI system that LED suppliers use is totally flawed. A simple white/black calibration won't go close to correcting the inaccuracies. As for the colour samples, the problem was that the monitor gamma was set too high, and/or there was too much brightness.

[@pileofstuff]

Sure it's not an "instrumentation grade" device, but it's also only $6. It was fun to play with it, so I already got my money's worth out of it.

[@johncoops6897]

@@pileofstuff - Ahhhh, I didn't realise it was so cheap! For $6 the performance is perfectly fine. It would be a lot easier to calibrate if the illumination LEDs were changed to something with better spectral performance. The huge challenge is to find such a thing, especially if buying online..

[@nexpro6985]

Or I can just use my Mimolta color meter 😁

[@pileofstuff]

Yeah, but where's the fun in that.

[@RottnRobbie]

I know you're not a programmer, and that you in essence copied Dronebot's code, so I'm fine with letting you off the hook. But those "delay to stabilize sensor" [ delay(200) statements ] bothered me. I noticed was that the readings are started and finished in the subroutine, so delaying after that shouldn't affect the reading at all. So I dug into it a bit by going to the tutorials you linked to, and while all of them have a delay statement after reading the frequency, none of them give a rational for it (other than Bill's comment statement about stabilizing the sensor). In every case, the code sets the colour to sense and immediately reads the frequency with a pulsein. Any delay after that won't have any effect on the reading, so it seems it's completely unnecessary! I tracked down the manufacturer's product documentation (ams.com/tcs3200#tab/documents) and checked the datasheet for the stabilization time, and here's the quote: "The response time to an input programming change or to an irradiance step change is one period of new frequency plus 1 microsecond". So the delay should be *between selecting the colour to read and reading the value*... in other words, a delay in each subroutine before the pulsein() call. How long a delay is needed is less clear, because the minimum frequency is harder to figure out. Looks to be different for different colours, and then adjusted by the set scale factor, but it does say that the "Dark Frequency" is typically 2 Hz - so up to 500 ms between pulses. That means that in the worst case, the delay should be 500 (not 200) plus it's in the wrong place. Like I said - you get a pass, but It'd be interesting to know what was going through the mind of the other tutorial creators.

[@pileofstuff]

I noted that all 3 of the tutorials I linked to used almost identical code, so it is possible that they all cribbed it from the same place, and didn't go too deep into that delay. it's also possible it's one of those "It won't work without it - don't know why" kind of things.

[@RottnRobbie]

@@pileofstuff Agree on all points. The why still niggles at me, but I'll get over it eventually...