Questions on LED PWM

I need to drive some RGB LEDs to generate a bit of eye candy and I'm
wondering if anybody has some experience on a few questions.  First,
I'm thinking of using 5 bits of intensity per color since the entire
RGB value would fit nicely in 16 bits.  Anybody have an opinion
whether 32 intensity levels for an LED is "smooth enough" or
"disconcertingly jumpy"?  The LEDs are going to be pretty busy so I
think (hope) I might be able to get away with fewer steps than
otherwise.

2nd question, how would the eye perceive the linearity of 32 equal
intensity steps (that is, ON times of e.g. 100us, 200us, 300us ...
3100us out of 3100us)?  I'm wondering if I may need to go to unequal
intensity steps to get a more apparently linear intensity ramp.
Anybody have experience with this?

I'll be wiring this up and playing with it before too long, but in the
meantime I thought I'd find out if anybody else has dealt with these
questions.  Many thanks!

Mike

Mike Silva wrote:

> 2nd question, how would the eye perceive the linearity of 32 equal
> intensity steps (that is, ON times of e.g. 100us, 200us, 300us ...
> 3100us out of 3100us)?  I'm wondering if I may need to go to unequal
> intensity steps to get a more apparently linear intensity ramp.
> Anybody have experience with this?

Some articles about it:

http://www.telescope-optics.net/eye_spectral_response.htm
http://en.wikipedia.org/wiki/Weber-Fechner_law

Looks like it is logarithmic. Another example is the magnitude value used
for measuring star brigthness:

http://en.wikipedia.org/wiki/Apparent_magnitude

This means the perceived intensity p=2.5*log(r), with r=real intensity =>
r=10^(p/2.5). Maybe there are some additional non-linearities if your PWM
rate is near 0% or near 100%, but for good looking results use a high
resolution PWM and use a logarithmic lookup table.

Another interesting article about intensity correction:

http://en.wikipedia.org/wiki/Gamma_correction

-- 
Frank Buss, fb@frank-buss.de
http://www.frank-buss.de, http://www.it4-systems.de

Use 6 bits for _green_, since human eyes is more sensitive to green lights.

Mike Silva <snarflemike@yahoo.com> writes:

> I need to drive some RGB LEDs to generate a bit of eye candy and I'm
> wondering if anybody has some experience on a few questions.  First,
> I'm thinking of using 5 bits of intensity per color since the entire
> RGB value would fit nicely in 16 bits.  Anybody have an opinion
> whether 32 intensity levels for an LED is "smooth enough" or
> "disconcertingly jumpy"?  The LEDs are going to be pretty busy so I
> think (hope) I might be able to get away with fewer steps than
> otherwise.
>
> 2nd question, how would the eye perceive the linearity of 32 equal
> intensity steps (that is, ON times of e.g. 100us, 200us, 300us ...
> 3100us out of 3100us)?  I'm wondering if I may need to go to unequal
> intensity steps to get a more apparently linear intensity ramp.
> Anybody have experience with this?
>
> I'll be wiring this up and playing with it before too long, but in the
> meantime I thought I'd find out if anybody else has dealt with these
> questions.  Many thanks!
>
> Mike

Mike Silva wrote:
> I need to drive some RGB LEDs to generate a bit of eye candy and I'm
> wondering if anybody has some experience on a few questions.  First,
> I'm thinking of using 5 bits of intensity per color since the entire
> RGB value would fit nicely in 16 bits.  Anybody have an opinion
> whether 32 intensity levels for an LED is "smooth enough" or
> "disconcertingly jumpy"?  The LEDs are going to be pretty busy so I
> think (hope) I might be able to get away with fewer steps than
> otherwise.
> 
> 2nd question, how would the eye perceive the linearity of 32 equal
> intensity steps (that is, ON times of e.g. 100us, 200us, 300us ...
> 3100us out of 3100us)?  I'm wondering if I may need to go to unequal
> intensity steps to get a more apparently linear intensity ramp.
> Anybody have experience with this?
> 
> I'll be wiring this up and playing with it before too long, but in the
> meantime I thought I'd find out if anybody else has dealt with these
> questions.  Many thanks!
> 
> Mike

Look at some of the LED driver devices, with PWM.
That will give you an idea of ranges - often they have
a RGB section, and a separate brightness control, so
that you are not relying on the Colour bits
for wide dynamic range.
Depends if you want this to operate over a daylight/nightime
dynamic range, or just a narrower range.

-jg

Mike Silva wrote:
> 
> I need to drive some RGB LEDs to generate a bit of eye candy and
> I'm wondering if anybody has some experience on a few questions. 
> First, I'm thinking of using 5 bits of intensity per color since
> the entire RGB value would fit nicely in 16 bits.  Anybody have
> an opinion whether 32 intensity levels for an LED is "smooth
> enough" or "disconcertingly jumpy"?  The LEDs are going to be
> pretty busy so I think (hope) I might be able to get away with
> fewer steps than otherwise.
> 
> 2nd question, how would the eye perceive the linearity of 32
> equal intensity steps (that is, ON times of e.g. 100us, 200us,
> 300us ... 3100us out of 3100us)?  I'm wondering if I may need to
> go to unequal intensity steps to get a more apparently linear
> intensity ramp. Anybody have experience with this?

I suspect you will want to multiplex drive the display, just to
save wires, drivers, etc.  That means you will have a time pattern
to fire each one.  The only reasonable (IMO) control of intensity
will be the length of activation pulse in that pattern (which can
be delivered to only one end, if desired).  This lets you fool with
the intensities with a value to time converter somewhere.

Note that you can deliver the three colors in three different time
slots, provided the multiplex frequency is high enough.

-- 
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: <http://cbfalconer.home.att.net>
            Try the download section.

Op Mon, 06 Oct 2008 04:24:53 +0200 schreef Mike Silva  
<snarflemike@yahoo.com>:
> I need to drive some RGB LEDs to generate a bit of eye candy and I'm
> wondering if anybody has some experience on a few questions.  First,
> I'm thinking of using 5 bits of intensity per color since the entire
> RGB value would fit nicely in 16 bits.  Anybody have an opinion
> whether 32 intensity levels for an LED is "smooth enough" or
> "disconcertingly jumpy"?  The LEDs are going to be pretty busy so I
> think (hope) I might be able to get away with fewer steps than
> otherwise.

Obviously, this depends heavily on the output power and the transition  
time.  But if it is not enough, you could think about a simple routine  
that will transition between two 16-bit colours in several steps using  
24-bit precision.



-- 
Gemaakt met Opera's revolutionaire e-mailprogramma:  
http://www.opera.com/mail/

Mike Silva wrote:

>I need to drive some RGB LEDs to generate a bit of eye candy and I'm
>wondering if anybody has some experience on a few questions.  First,
>I'm thinking of using 5 bits of intensity per color since the entire
>RGB value would fit nicely in 16 bits.  Anybody have an opinion
>whether 32 intensity levels for an LED is "smooth enough" or
>"disconcertingly jumpy"?  The LEDs are going to be pretty busy so I
>think (hope) I might be able to get away with fewer steps than
>otherwise.
>
>2nd question, how would the eye perceive the linearity of 32 equal
>intensity steps (that is, ON times of e.g. 100us, 200us, 300us ...
>3100us out of 3100us)?  I'm wondering if I may need to go to unequal
>intensity steps to get a more apparently linear intensity ramp.
>Anybody have experience with this?
>
>I'll be wiring this up and playing with it before too long, but in the
>meantime I thought I'd find out if anybody else has dealt with these
>questions.  Many thanks!

Depending on what you are displaying, you might find that the 
difference between 00000/00000/00000(all off) and 00000/00000/00001 
is especially noticable, and that the color shifts at the lower
end of brightness -- not enough resolution to keep the ratios right.
An old trick is to use that extra bit to reduce the brightness of 
all three colors by a factor of 4, thus giving 7 bits of resolution
at the dim end of the range.      

-- 
Guy Macon
<http://www.GuyMacon.com/>

On Oct 6, 7:34=A0am, Guy Macon <http://www.GuyMacon.com/> wrote:
> Mike Silva wrote:
> >I need to drive some RGB LEDs to generate a bit of eye candy and I'm
> >wondering if anybody has some experience on a few questions. =A0First,
> >I'm thinking of using 5 bits of intensity per color since the entire
> >RGB value would fit nicely in 16 bits. =A0Anybody have an opinion
> >whether 32 intensity levels for an LED is "smooth enough" or
> >"disconcertingly jumpy"? =A0The LEDs are going to be pretty busy so I
> >think (hope) I might be able to get away with fewer steps than
> >otherwise.
>
> >2nd question, how would the eye perceive the linearity of 32 equal
> >intensity steps (that is, ON times of e.g. 100us, 200us, 300us ...
> >3100us out of 3100us)? =A0I'm wondering if I may need to go to unequal
> >intensity steps to get a more apparently linear intensity ramp.
> >Anybody have experience with this?
>
> >I'll be wiring this up and playing with it before too long, but in the
> >meantime I thought I'd find out if anybody else has dealt with these
> >questions. =A0Many thanks!
>
> Depending on what you are displaying, you might find that the
> difference between 00000/00000/00000(all off) and 00000/00000/00001
> is especially noticable, and that the color shifts at the lower
> end of brightness -- not enough resolution to keep the ratios right.
> An old trick is to use that extra bit to reduce the brightness of
> all three colors by a factor of 4, thus giving 7 bits of resolution
> at the dim end of the range. =A0 =A0 =A0

That's a very clever idea and I'll be sure to look at it.  Since I was
going to manipulate the colors as 24 bit data anyway (and only
transmit it as 16 bits) this sounds like a good fit.  I was definitely
thinking that that first level from all OFF would be too large a step,
and that the colors at low resolution would be off (but since these
will be very abstract patterns this latter may not matter at all, and
may even add to the effect).

And thanks to all who posted links and other info.

On Sun, 5 Oct 2008 19:24:53 -0700 (PDT), Mike Silva <snarflemike@yahoo.com> wrote:

>I need to drive some RGB LEDs to generate a bit of eye candy and I'm
>wondering if anybody has some experience on a few questions.  First,
>I'm thinking of using 5 bits of intensity per color since the entire
>RGB value would fit nicely in 16 bits.  Anybody have an opinion
>whether 32 intensity levels for an LED is "smooth enough" or
>"disconcertingly jumpy"?  The LEDs are going to be pretty busy so I
>think (hope) I might be able to get away with fewer steps than
>otherwise.

If you want a reasonably smooth fade, 256 levels per colour  would be an absolute minumum. 

On Mon, 06 Oct 2008 14:18:23 +0100, Mike Harrison
<mike@whitewing.co.uk> wrote:

>On Sun, 5 Oct 2008 19:24:53 -0700 (PDT), Mike Silva <snarflemike@yahoo.com> wrote:
>
>>I need to drive some RGB LEDs to generate a bit of eye candy and I'm
>>wondering if anybody has some experience on a few questions.  First,
>>I'm thinking of using 5 bits of intensity per color since the entire
>>RGB value would fit nicely in 16 bits.  Anybody have an opinion
>>whether 32 intensity levels for an LED is "smooth enough" or
>>"disconcertingly jumpy"?  The LEDs are going to be pretty busy so I
>>think (hope) I might be able to get away with fewer steps than
>>otherwise.
>
>If you want a reasonably smooth fade, 256 levels per colour  would be an absolute minumum. 

I agree (well, unless the changes are going to take place rather
quickly like a second or two). 

The eye is insensitive to small brightness differences, but quite
sensitive to small changes in brightness. 
Best regards, 
Spehro Pefhany
-- 
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com