Control of a light source

tucker wrote:


Your message raises many questions.

What does "follow (only) the luminance variations" mean?

What is an offset condition?

How do you propose to control luminance and temperature separately?

How can you determine color temperature with a single diode?

Are the relations of color temperature and luminance to temperature
really linear, as your message suggests?

Have you modeled the resistance of the filament as a function of current?

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Jerry Avins wrote:

  ...


  ...

Make that
"Are the relations of color temperature and luminance to current
really linear, as your message suggests?"

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Jerry Avins wrote:


I mean that, for the color temperature, a correction of the supplyed
current is necessary only if the error is > 10K


When the lamp is supplyed with its nominal current of 6.8 A, the
relation between current and luminance is L = A + B*I for small
variations of luminance. I can vary the current only with step of 4.5
mA that correspond to a variation of 0.5% of luminance. If the
luminance error is in 0.1% - 0.5% I cannot vary the current, I have to
change the term A (L = A + B*I) in order to have an error  < 0.1%.



I cannot control luminance and color temperature separately because a
change of the current supplyed causes a modification of both the
luminance and the temperature. It is not guaranteed that both the
specifications on luminance and color temperature can be achieved.



Doing two measurements on the light source using first a red filter and
then a blue filter.



They are not linear, that relations were estimated by a physicist that
worked on the project.



No, all the information I have on the plant come from a doctoral thesis.

tucker wrote:

Are you saying you can vary the A term?  Surely not.

Even more frightening.

Depending on the hardware that you're using to implement your controller
you may be able to get better resolution on the luminance that 0.5% --
see my article about sigma-delta converters on
http://www.wescottdesign.com/articles/sigmadelta.html .

Your lamp is going to have some time-dependent behavior, and you can
never be sure of estimated gains (the 'B' term in your equation).
Unless you can afford to have the system respond very slowly you should
measure this behavior -- I advocate using swept-sine measurements, but
there are plenty of people who feel you should get a step response and
do an ARMA system identification procedure.  Then you should use this
measured behavior to determine the feasibility of increasing your
apparent DAC resolution through dithering, and to design a solid
controller with some sort of stability guarantees.

Were it me, I would first determine if I could make an
ultra-conservative controller.  If I could I'd just make a PI
controller, possibly with a sigma-delta output stage to increase the
current resolution, possibly with some deadband to prevent a limit cycle
around the current step.  I'd design and tune the PI controller per the
methods outlined in http://www.wescottdesign.com/articles/pidwophd.html .

If I needed better time response than what an ultra-conservative
controller would give me I'd do a swept-sine measurement and design a
PID controller using frequency domain techniques.  I'd measure the
response per
http://www.wescottdesign.com/articles/FreqMeas/freq_meas.html , and
design it per chapter 6 in my book "Applied Control Theory for Embedded
Systems" (http://www.wescottdesign.com/actfes/actfes.html ).

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google?  See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html

Tim Wescott ha scritto:


Many thanks for your suggestions.

I read the photodiode voltage using a digital voltmeter and the current
is supplyed using a programmable DC power supply, both connected to a
pc through IEEE488. I'd like to do a swept-sine measurement but I have
not at my disposal the necessary instrumentations, I can try to get a
step response instead. The system hasn't any requirement on the time
response so a PI controller should works fine.

Tucker

tucker wrote:

I think the first thing you need to do is clarify you specifications.
Your first item sounds OK, your second bullet sounds like design rather
than specifications, and your third item sounds like an alarm, not
control, which contradicts your opening paragraph.

If you just need to hold the luminance steady and sound an alarm if the
color temperature goes off too far then an LTI controller for luminance
and a threshold -- possibly with a filter -- around the color
temperature measurement should be all that you need.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google?  See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html

Sounds like a great place for an LM3524 if they still make them..I've
built this circuit...does a great job for less than $10.

http://www.hep.fsu.edu/~wahl/phy3802/expinfo/electronics/AN-292.pdf