More questions on color reader...

I did offer a thought, earlier. No response to it.

Have you read and do you fully understand the CIE 1931 and

1964 color standards? (If you are really into this, I'd also recommend Edwin Land's work papers from the late 1970's to early 1980's -- you may not need to, but it is just very interesting to study and it addresses directly some points that relate squarely on various lighting situations as you are encountering.)

Which is exactly the area that Edwin Land highlights in his research reports...

In other words, how much do you understand about human color perception?

By the way, what are you using as your comparison "standard?" Your eye? Or?

Jon

No need to get snotty. The question on the floor is _electronic_, lab versus _field_. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
      The only thing bipartisan in this country is hypocrisy

Which _is_ a question at hand, I suspect.

Jon

close

old

bench,

heck

shift

what=20

the

and a

device.

colors,

ways

emulate.

every

dawn.

FCC-

Declaration

Probably

require

wholesale

not

The

add

info):

7-vol1-part15.pdf

with=20

see=20

a=20

goes=20

Doing so would go a long ways in demonstrating due diligence. Test as many units as you can afford and document them all.

While i did not read Jon's post as snotty, i can see how it can be read that way. It struck me more as a "did you know of this(?), which seems like it would be helpful".

Thanks. It was offered in that vein.

The funding of accurate human color perception _measurement_ dates back at least to the time when colored house paints began to be sold (by Sears, for example) as something to 'spruce up' homes. (Turn of 19th to 20th century, roughly.)

Customers would buy a 'brown' for an addition, hoping to get the same 'brown' they ordered two years before, and getting something that 'any idiot' could see wasn't even close. The control of paint chemicals and dyes wasn't sufficient by itself at the time and there was a strong need for some "feedback" to help adjust the dyes, as appropriate.

This commercial desire played into a university research desire regarding color blindness and provided a substantial funding source for this research to proceed.

A method was needed so that a 'brown' paint bought today and used on the shady part of a home would look the same as the same 'brown' paint bought next year and used on a well-lit side. Side by side, the two paint jobs should "look the same" to a viewer -- VERY WIDELY varying lighting conditions.

The CIE color system has remarkable fidelity for that purpose. It also can deal with lighting conditions within some limits.

What it does not do is cope with the surroundings, which in human perception is very important. That is one reason why I pointed him at Edwin Land and one reason why I wondered how Charlie was determining when the color reading was wrong or right.

If you take a canvas and place it on a painting easel and put swatches of colors nearby each other on that canvas and light it with a bright 100W tungsten bulb, people will see the colors a certain way. Then, as you dim the bulb down, the Planck radiation curve emitted will shift dramatically in wavelength -- all of us here are well aware of this fact -- causing reflections and the resulting distribution of wavelengths from the surfaces to be markedly different than before. Yet even operating at 10W or even 1W of output, with the distribution almost totally different than before, a human will still "see" the same "colors."

Do the same experiment, now instead with a cover sheet that blocks out all but one selected color swatch and the human will NO LONGER see the same colors when the lighting is changed. This proved that the human vision system uses the reflections from nearby areas to help adduce any nearby color. As the reflections shift dramatically for one color swatch, so it also shifts for others nearby. Lose that additional information source and the brain can't maintain the perception. Keep it, and it can.

Edwin Land spent years studying this phenomenon well after both CIE standards groups put out their results and it turns out that humans also use nearby colors to "calibrate" their perceptions.

In any case, it is VERY illuminating to study this material (the CIE _and_ Dr. Land research papers) and it _may_ bear on this "issue." It made sense to me to at least ask if the OP is fully familiar with the existing research here.

I spent some years working with OSRAM on these issues. I worried that Charlie may be fighting an issue that might be more easily 'understood' from a different domain. He is in a better position to generally know one way or another, but a small pointer might help if he wasn't already aware.

Jon

Hi Jon, Basically, yes, I am using my own eyes as a calibration source, sort of...

Since I have chosen my illumination source, and RGB LED run one color at a time, I am not dealing with different illumination sources. I am also not trying to do high accuracy color determination, just red, green, blue, orange, etc. My frustration is that two units, apparently calibrated identically, will still see certain colors differently. Trying to identify the source of the differences led me to consult with my friends here to see if there was some electronic design error on my part. I suspect that, such little things as different alignments of the LED and PT, different arrangements of my blocking black felt, and even temperature of the unit may be enough to skew the results so that certain 'border' colors change nations, i.e beige becomes pink, gold becomes orange, and blue becomes purple, or blue green.

Fortunately, in my market, it appears that this is a conundrum that hasn't really been solved yet! Some makers require a calibration before every read. I think I may just ship what I have, and see if it is useful enough at my price point, to make sales.

Charlie

cut

Cut Your eye's may be quite insufficient. I compared some time ago with a college, we both had correct color vision(according to some standard tests). Then we used a little test device, with one halve circle true yellow and another half circle you could adjust red and green to get the same hue and intensity yellow. Well.... My yellow was pale green-yellow according to my college, and his yellow looked to me like orange. So when you fiddle around with an rgb led, your eyesight is a poor judge, and some hardware to test would be better.

[...]

Did you turn the black level subtraction back on? Also, I think just some felt inside a thin plastic enclosure isn't going to cut it, you need at least foil underneath and then make sure that nothing in the sensor area heats up too much.

Understandable but dangerous. At least I would compare it to other units that are in the market. Serious egg in the face never really wipes off clean in small markets, people remember. Best not to let that happen.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

=20

Of course some of that is in how it is marketed. If it is sold as right often enough to be useful that is one thing, if it is marketed as reliable and repeatable it is another.

[...]

Even in the first case it better be on par with simlarly priced other units or better. I don't know the market for Charlie's device but it is amazing how fast reputation is gained or lost with the Internet these days. All it takes is a major blog somewhere.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hi Jeorge, Yep, I turned on the black level subtraction (or, increased it to match temporal conditions...) and it has helped some. Also, sensor is in front of unit, power supplies are at least an inch away. I am more concerned with the opamps and the digital pot maybe self heating, especially when I have been testing for a while...

My real problem I think is engineer's disease... I keep thinking of improvements to the hardware and software!

Charlie

Exactly. Most of the products out there now in this market are in the 'right enough to be useful' category. What is interesting is some of the more expensive units, where you could buy a calibrated colorimeter for less...

Charlie

Opamps can be covered, either by using low offset versions or by old-fashioned clamping. Digital potmeters, different thing. One can never rely on the absolute value. The step-to-step accuracy would be listed in the datasheets and if that ain't good enough you'd have to develop a solution without those potmeters.

But I'd also be concerned about IR getting inside. Your photodiode isn't so stellar in rejecting near-IR and without metal between plastic and felt your box may let a lot of that pass through.

Don't we all :-)

Just like the typical SW/firmware guy won't stop until 95% plus of the available ROM space is filled.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hey, I am at 96%! Maybe I can ship now... ;-)

Charlie

On Mon, 26 Apr 2010 15:07:02 -0700, Charlie E. wrote: Well, it looks like it is time to call it a day. I spent two days this weekend testing, and calibrating four units. On the bench, they all worked great, and gave good results across my entire test samples.

This morning, I mounted them in their final cases, and hooked them up. Two failed immediately, basically decided everything I tested was white. Two appeared to function, but as soon as I started testing, failed on every 'corner' case in my test samples. Took one of those back to the bench, and the calibrations had shifted drastically. Funny thing was, the shift was to needing more gain, not less, which the 'all white' indications would have indicated.

Technically, I have been 'measuring' gain as the setting on the digital pot that gave an almost full indications on the ADC. This gave me values from 0 to 255. When I measured this unit on Saturday, it had gains of red 239, green 239 and blue 226. On Sunday, when I finallized the program, it read 231, 233, 214. This morning, after retesting, it calibrates at 245, 241, and 231. So, a shift of over 5% in just two days. There might have been temperature or background variations, but the background measuremnts have been stable at a reading of around 8 - 10 on a scale of 2048. I am totally baffled!

So, after a year and about $2000 in materials, looks we are going to forget this product, unless some of ya'll have any ideas.

Anyone out there know of anyone needing a good applications engineer?

Charlie

Nah, don't throw in the towel so fast :-)

Something is deteriorating. Assuming VCC is perfectly stable (check for dips with a DSO) this almost has to be the LEDs.

Question: How close to the max do you drive your LEDs? If in a healthy range hang a scope across RLED and check for fast spikes. I am not at all a fan of charge pump converters, who knows, maybe it's kicking out nasty ones.

Does your software turn LED_PWR_ON to off before changing position at the BSS8402 switches and then back on? It should, because the regulator will not be able to react in nanoseconds, it'll be more in the tens of microseconds.

Sorry, I don't. We were looking for an analog guy at a client but that's been filled by now.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Thanks, Jeorg, I am driving the LEDs at 20mA, which might be the problem if there are spikes that drive it above limits. I did think to turn the power off before switching. I even turned one LED on, then turned the other off for a while, to be sure there were not any no load conditions on the switcher. I don't have an o-scope at this time. It was on my list for purchases as soon as we made some sales... ;-)

I wish I was an analog guy, but this project has shown me how far I have to go to really consider myself to have the necessary experience to be one!

Charlie

Just for kicks, try hanging a couple of 10n caps from output to inverting input on those MCP op-amps. (between pins 1/2 and 7/6). It's possible they're oscillating, particularly the first one.