More questions on color reader...

[Charlie E.]

Hi All,
You have given me good advice in the past, and I am now really close
to shipping this thing, but still running into some of the same old
problems.

Basically, when I program a unit, it works great here on the bench,
and around the house, but when I go out into the real world, all heck
breaks loose!

My present problems seem to revolve around dark colors. Browns shift
to dark red, or green, blacks suddenly become dark greens, dark denims
become black, dark green, or even dark blue-green.

Trying to determine the cause is difficult, because the problems never
happen in the lab when I am in debug, and can get full data on what is
going on internally. My present guesses all point to shifts in the
strengths of the LEDs and other electronics, perhaps with temperature,
or maybe with differences in background lighting leaking into the
unit.

So, can anyone offer any suggestions? You can find a schematic and a
photo of the unit at
http://edmondsonengineering.com/RainbowColorReader.aspx

Thanks in advance!

Charlie

 

[Joerg]

Hi All,
You have given me good advice in the past, and I am now really close
to shipping this thing, but still running into some of the same old
problems.

Basically, when I program a unit, it works great here on the bench,
and around the house, but when I go out into the real world, all heck
breaks loose!

My present problems seem to revolve around dark colors. Browns shift
to dark red, or green, blacks suddenly become dark greens, dark denims
become black, dark green, or even dark blue-green.

Trying to determine the cause is difficult, because the problems never
happen in the lab when I am in debug, and can get full data on what is
going on internally. My present guesses all point to shifts in the
strengths of the LEDs and other electronics, perhaps with temperature,
or maybe with differences in background lighting leaking into the
unit.

So, can anyone offer any suggestions? You can find a schematic and a
photo of the unit at
http://edmondsonengineering.com/RainbowColorReader.aspx

As Jim hinted you might have background light get in. Since there is
only one phototransistor and not three with suitable filters the unit
can't predetermined the ambient light energies per band. At night you
may also see errors due to mercury vapor street lighting and so on,
worst case also some PWM modulated stuff.

This will be worse with darker colors because the signal to noise ratio
is lower.

BTW, "rainbow" can mean something entirely different in places like San
Francisco

 

[Charlie E.]

Lab has controlled background ambient? Real world does not.

PIC in an analog circuit

...Jim Thompson

It IS a PIc in an analog circuit! ;-)

Charlie

 

[[email protected]]

As Jim hinted you might have background light get in. Since there is
only one phototransistor and not three with suitable filters the unit
can't predetermined the ambient light energies per band. At night you
may also see errors due to mercury vapor street lighting and so on,
worst case also some PWM modulated stuff.

This will be worse with darker colors because the signal to noise ratio
is lower.

BTW, "rainbow" can mean something entirely different in places like San
Francisco

You mean like

"Red, Orange, Yellow, Green, Blue-Green, Blue, Purple, Pink"
"Plus: Whitey, African-American, Gray Davis, Jerry Brown, ..."

?

 

[Charlie E.]

Why is VCC connected to both power pins of U2?

The + input of U2C has no DC path.

What's going on with U8/U9/U10, especially U10?

John

Hi John,
The upper pin is connected to ground. The VCC is on top of the bypass
cap for the chip...

U2C hasn't given me any problems, to this point. ;-)

U8/U9/U10 are all complementary MOSFET pairs, doing the switching of
the LED power to the LEDs.

A note on background leakage. The LED and PT are both oriented to the
front of the case. I have black felt glued to the top and bottom of
the case (about 1" wide) to block reflections and reduce ambient. My
typical 'CLEAR' reading (no LEDs active) is in the <0.5% range. My lab
has a large east facing window, and I have done testing both in the
morning, afternoon, and evening without a lot of difference. Direct
sunlight is another matter...

Charlie

 

[Joerg]

You mean like

"Red, Orange, Yellow, Green, Blue-Green, Blue, Purple, Pink"
"Plus: Whitey, African-American, Gray Davis, Jerry Brown, ..."

No, that sort of stuff (this one from back east):

http://www.wheretraveler.com/classi...where_archive/Where_Magazine/2007_07/0025.jpg

 

[Joerg]

With nice linear phototransistors you could modulate the LED, and
subtract out the "dark" current. That'd make it more robust to ambient
light.

I think Charlie does modulate the LED already. But no matter how linear
the phototransistor, there comes point where it can't cope with ambient.
Especially if things are modulated.

 

[Joerg]

Hi John,
The upper pin is connected to ground. The VCC is on top of the bypass
cap for the chip...

U2C hasn't given me any problems, to this point. ;-)

U8/U9/U10 are all complementary MOSFET pairs, doing the switching of
the LED power to the LEDs.

A note on background leakage. The LED and PT are both oriented to the
front of the case. I have black felt glued to the top and bottom of
the case (about 1" wide) to block reflections and reduce ambient. My
typical 'CLEAR' reading (no LEDs active) is in the <0.5% range. My lab
has a large east facing window, and I have done testing both in the
morning, afternoon, and evening without a lot of difference. Direct
sunlight is another matter...

Sunlight and thus outdoor ambient has a lot of IR content which goes
through just about anything. Indoors there will be very little, AFAIK CF
lamps emit very little in IR percentage and windows keep it out as well.

You might need some foil in there as well, between plastic and felt.

 

[Charlie E.]

Sunlight and thus outdoor ambient has a lot of IR content which goes
through just about anything. Indoors there will be very little, AFAIK CF
lamps emit very little in IR percentage and windows keep it out as well.

You might need some foil in there as well, between plastic and felt.

Maybe a possibiliity, but the phototransistor is an ambient light
detector, which has built in filters in the IR range to help shape the
response curve. Pam was more concerned with possible UV contamination
from the bright flourescents in the stores.

Charlie

 

[Joerg]

Maybe a possibiliity, but the phototransistor is an ambient light
detector, which has built in filters in the IR range to help shape the
response curve. Pam was more concerned with possible UV contamination
from the bright flourescents in the stores.

Well, it just says "photo npn" in your schematic. Which one is it? IR
filtered doesn't necessarily mean is has to be a very good filter.

 

[Charlie E.]

Well, it just says "photo npn" in your schematic. Which one is it? IR
filtered doesn't necessarily mean is has to be a very good filter.

It is a TEPT5600, a 5mm through hole part, which I bend to a 90 degree
angle to face the front. The LED is a Kingsbright WP154A4SUREPBGVGAW
RGB LED which I also bend to a 90. I have the LED at a 45 degree
angle to the face plate, and the PT and LED are on opposite sides of
the board, to reduce direct reflections between them.

Charlie

 

[[email protected]]

No, that sort of stuff (this one from back east):

http://www.wheretraveler.com/classi...where_archive/Where_Magazine/2007_07/0025.jpg

Are you saying that SF doesn't do rainbows?

 

[Joerg]

It is a TEPT5600, a 5mm through hole part, which I bend to a 90 degree
angle to face the front. The LED is a Kingsbright WP154A4SUREPBGVGAW
RGB LED which I also bend to a 90. I have the LED at a 45 degree
angle to the face plate, and the PT and LED are on opposite sides of
the board, to reduce direct reflections between them.

Figure 7 hints that this photodiode still lets in a ton of the near-IR
spectrum:

http://www.vishay.com/docs/84768/84768.pdf

Might also be good to put a snippet of metal tubing around it as a
shroud because it seems to be a plastic diode. The more you can reduce
the acceptance angle the better.

 

[Joerg]

Are you saying that SF doesn't do rainbows?

Oh, they do. There was a story in our paper but it makes no sense to
post links from there because they have obfuscated server access for
non-subscribers.

 

[Charlie E.]

Oh. OK.


It's probably banging up against its own ESD diodes in one direction
or another. C10 makes the power amp ignore the offset. You could
remove C4.

Interesting. It doesn't really do that much, does it... ;-)

Why the 100r and 0r resistors S-G on the Pfets? That wastes a lot of
Vled current... especially the 0r one.

Ok, that is an error in the schematic. Those are all three 10K
resistors. It was correct on the board... ;-)

Are you "auto-zeroing" out the background light? Are you checking for
sensor saturation?

John

I had built in a test for the 'CLEAR' component, where I measured the
ambient, and subtracted it from the measured values, built in testing
it was always so low I didn't feel the need to keep it in. Might need
to test that hypothesis now...

Thanks!!!!

Charlie

 

[Charlie E.]

Hey Charlie,
Can't help with your circuit, but it looks like a pretty cool device.
It's the sort of thing that someone probably would never think of
unless they were familiar with blindness.?
I wonder if it could have application in teaching kids their colors,
or for use with colorblind individual.
Maybe an advanced unit to detect world currencies...??

My only thought is that maybe light is bouncing around in weird ways
on the input.
Try setting up some lights on your bench and see if you can emulate.
Use different types of lights (flourescent, tungsten, basically, every
wavelength you can think of).
I would also experiment with different daylight times (outdoors, of
course), as the Sunlight temperture varies considerably dusk to dawn.
(You probably already know all this, right??!) Disclaimer: Amateur
photographer here.

I notice you're in the United States.
I thought I should point out a potential regulatory concern that
involves offering products for sale before they have either been FCC-
certified, or before the manufacturer (you!) has filed a Declaration
of Confirmity.
Honestly, I'm not sure which one your product would involve. Probably
a DoC, but I don't know if they consider the LED emissions to require
certification...?? Maybe someone here will know.
Your PIC certainly operates fast enough to qualify for testing. I
recall anything over 9kHz requires testing, though I could be wrong
about that lower bound.
8 to 32 MHz definitely qualifies, however.

I think the FCC might interpret your website as advertising of the
product, particularly in light of your web text suggesting wholesale
pricing. (I realize the product is not ready for sale, but that's not
really the point.)
If the FCC were to draw that conclusion, it could subject you to an
Official Citation or Notice of Apparent Liability. The former is a
slap on the wrist involving no out-of-poekct money. The latter can
lead to possible fines and in-rem forfeiture - (though all of these
outcomes are highly unlikely).

Either way, a quick review of the Rule may put your mind at ease. The
Rule is FCC 2.803
For convenience, here's the link:
http://www.gpo.gov/fdsys/pkg/CFR-2009-title47-vol1/pdf/CFR-2009-title47-vol1-part2.pdf

Not knowing your situation, my recommendation is to be safe just add
the following text to your web site (the page that has pricing info):

This device has not been authorized as required
by the rules of the Federal Communications
Commission. This device is not,
and may not be, offered for sale or lease, or
sold or leased, until authorization is obtained.

The above is the exact language specified by 47CFR2.803(c).

Enjoy!! And best of luck with the project/product.
-mpm

Hmmmm...
It is definitely not an intentional radiator, and no clock outputs
leave the chip. The two switchers are only going to a cap less than
10 mm from the chip. Would never have thought anything this simple
could need certification. Jeorge? Any thoughts?

Thanks,
Charlie

 

[Joerg]

[FCC regs]

Hmmmm...
It is definitely not an intentional radiator, and no clock outputs
leave the chip. The two switchers are only going to a cap less than
10 mm from the chip. Would never have thought anything this simple
could need certification. Jeorge? Any thoughts?

Well, there are paths of self-certification:

http://www.techintl.com/emcinusa.cfm

I run into this a lot. Is a change serious enough to warrant re-cert?
Most clients do it anyhow, send the stuff to an EMC lab. But it's
expensive, basically we rarely get out of there under $5k. If this is
more like a non-profit product and helps the visually-impaired you may
be able to convince a lab to do a "charity run".

But first I'd fix that "CLEAR" ambient subtraction routine that must
have fallen through the cracks ... whoops ... got to have that

 

[Charlie E.]

[FCC regs]

Hmmmm...
It is definitely not an intentional radiator, and no clock outputs
leave the chip. The two switchers are only going to a cap less than
10 mm from the chip. Would never have thought anything this simple
could need certification. Jeorge? Any thoughts?

Well, there are paths of self-certification:

http://www.techintl.com/emcinusa.cfm

I run into this a lot. Is a change serious enough to warrant re-cert?
Most clients do it anyhow, send the stuff to an EMC lab. But it's
expensive, basically we rarely get out of there under $5k. If this is
more like a non-profit product and helps the visually-impaired you may
be able to convince a lab to do a "charity run".

But first I'd fix that "CLEAR" ambient subtraction routine that must
have fallen through the cracks ... whoops ... got to have that

Ok, so if we market this device in the US, since it has a MCU in it,
then we have to get certified by the FCC that we don't radiate?

We may just have gone out of business!!!!!!

Charlie

 

[[email protected]]

Lots of places do rainbows... there are various samples here:
http://en.wikipedia.org/wiki/Gayborhood

That phallic-looking one in Boystown, Chicago seems quite fitting.

Looks like something NASA would do.

 

[[email protected]]

No. Kits are covered under the regs.

I agree that as a practical matter, nobody will complain unless /
until interference to some other licensed radio service occurs.

Or a competitor gets a burr under their saddle.

If / when that happens, the last thing you want is no DoC,
Verficiation or FCC ID on file for your gear. - particularly if
someone gets hurt.
That an invitation to get fined and sued into the poor house. (I'm
thinking about those motorized wheelchair controllers that went crazy
every time a taxicab radio keyed up in the vicinity!)

But before jumping ship, Charlie -- just call the folks at the FCC OET
in Washington, describe your product, and see if you can squeak by
with Verification.
If so, no money out of pocket -- assuming you know how to self-
certify, and have access to the right test equipment.

Even if they require a DoC, all is not lost. Shop around. It may be
cheaper than you think!

IME, EMI compliance is pretty cheap (though test labs vary quite a bit). It's
all the rest of the stuff that takes time and $$. If there is no AC power
(even a charger) most of that goes away.