Maker Pro
Maker Pro

Driving many led's (400+)

J

Jon Slaughter

I have an application where I need to drive 400+(~150 rgb's) individually.

I was thinking about using the

TLC5947---TI---24CH 12bit PWM LED Constant Current Driver

as this was the first one I looked at where each LED had it's own channel.
That way I had complete control. It whould require 20+ of these and each one
costs about 4$. It seems like a nice package and the only problem I have is
setting the same q current for all of the IC's(so they all have the same
brightness at the same level).

Also, there is the

ST---STP24DP05---24ch Constant Current Sink Driver.pdf

Which is similar but doesn't have PWM but has some other features. (although
I can PWM it myself)

But then I stumbled on

http://datasheets.maxim-ic.com/en/ds/MAX6952.pdf

which is a matrix driver. I was simiarly thinking about driving the LED's in
a matrix but I couldn't figure out how to do it. (Maybe was being dense)

Would there be any way to reduce the number of IC's (TI or ST) by somehow
configuring in a matrix but still get "individual" control?

I assume that the matrix stuff for the MAX6962 only works through
"refreshing" because it is impossible to turn on arbitrary LED's in the
matrix at the same time that are in different row's and columns? (seems to
be the only way it can work but maybe I'm missing something?)

What I'm thinking:

I can configure 2 TI's ran where there are 24 colums and 24 rows. I would
then have just have just have to refresh/scan each row at at time.

Then the problem is to get the PWM to work with the refreshing or abandon
the IC's PWM and do it myself.

Point being, I only need 2 drivers? If I had 1 led per channel and each led
used 30mA then thats about 15A that I need. If I run them as a matrix then
it's about 1000x more efficient?

In that case does anyone know which driver IC is better? Or see any
problems?

Thanks,
Jon
 
F

Frank Buss

Jon said:
I can configure 2 TI's ran where there are 24 colums and 24 rows. I would
then have just have just have to refresh/scan each row at at time.

I don't think this will work. But maybe you can use one TI for feeding into
the rows and some MOSFET arrays for connecting the columns to ground.
Then the problem is to get the PWM to work with the refreshing or abandon
the IC's PWM and do it myself.

Point being, I only need 2 drivers? If I had 1 led per channel and each led
used 30mA then thats about 15A that I need. If I run them as a matrix then
it's about 1000x more efficient?

And 1000x darker :) If you run it as a matrix, you have to use higher
currents, because each column is ON for 1/24 of the time. I don't know, if
the TI chip allows such high currents for all 24 outputs and if the LEDs
allows it. Maybe group it in 3x8 columns. Maybe something like this:

http://www.freescale.com/files/analog/doc/data_sheet/MC33291L.pdf

which can even help reducing the number of pins for your control logic,
because you can select the switches with SPI.
 
S

Spehro Pefhany

I have an application where I need to drive 400+(~150 rgb's) individually.

I was thinking about using the

TLC5947---TI---24CH 12bit PWM LED Constant Current Driver

as this was the first one I looked at where each LED had it's own channel.
That way I had complete control. It whould require 20+ of these and each one
costs about 4$. It seems like a nice package and the only problem I have is
setting the same q current for all of the IC's(so they all have the same
brightness at the same level).

Also, there is the

ST---STP24DP05---24ch Constant Current Sink Driver.pdf

Which is similar but doesn't have PWM but has some other features. (although
I can PWM it myself)

But then I stumbled on

http://datasheets.maxim-ic.com/en/ds/MAX6952.pdf

which is a matrix driver. I was simiarly thinking about driving the LED's in
a matrix but I couldn't figure out how to do it. (Maybe was being dense)

Would there be any way to reduce the number of IC's (TI or ST) by somehow
configuring in a matrix but still get "individual" control?

I assume that the matrix stuff for the MAX6962 only works through
"refreshing" because it is impossible to turn on arbitrary LED's in the
matrix at the same time that are in different row's and columns? (seems to
be the only way it can work but maybe I'm missing something?)

What I'm thinking:

I can configure 2 TI's ran where there are 24 colums and 24 rows. I would
then have just have just have to refresh/scan each row at at time.

Then the problem is to get the PWM to work with the refreshing or abandon
the IC's PWM and do it myself.

Point being, I only need 2 drivers? If I had 1 led per channel and each led
used 30mA then thats about 15A that I need. If I run them as a matrix then
it's about 1000x more efficient?

No. To a first order you need the same amount of average current to
get the same visual brightness regardless of the duty cycle (provided
the frequency is high enough that the individual flashes "fuse"
together (no visible flicker).
In that case does anyone know which driver IC is better? Or see any
problems?

Thanks,
Jon

NXP (nee Philips) had some. Dunno what's best for you. Often something
like 8 source drivers and 50 sink drivers makes sense, but you'd need
quite high speeds to control them all with, say, 7 bits of brightness
resolution. Eg. scan at 100Hz with 100 brightness settings per LED,
you'd need to calculate and spit out data at 50 * 100 * 100 = 500kHz.

There are drivers wot do more stuff internally if you need to do
something like RGB video.
 
J

Jon Slaughter

So basically I need to drive the rows with C*I where I is the current per
LED. This gives the peak current through an LED while it is on as C*I but on
average, since there will be C cycles, of just I. The driver needs to be
able to supply C*I current too(per row).

Unfortunately the peak current per LED is about 100 which is just 4 times
the average current. This means I can only have 4 rows ;/ but reduces the
number of drivers I needed originally by a factor of 4 which is better. (I
have 18 rows and 24 columns)

The TI IC can only deliever 30mA so this is out of the question as I would
need about 100mA or so. The ST has 80mA maximum. I might have to go with
this as there are no too many 24-ch drivers. Of course thats absolute max ;/

Maybe I would just have to go with 24 discrete mosfets for the column's and
18 discrete for the rows and then use some mosfet driver chips + uP to
drive? This of course is more work but solves the power capability issues.
Alternatively maybe I can find some LED drivers that have a larger capacity
at 16 ch.

Of course I'm looking at the worse case of all the LED's being on but since
it's not an impossibility...
 
J

John Devereux

Jon Slaughter said:
So basically I need to drive the rows with C*I where I is the current per
LED. This gives the peak current through an LED while it is on as C*I but on
average, since there will be C cycles, of just I. The driver needs to be
able to supply C*I current too(per row).

Unfortunately the peak current per LED is about 100 which is just 4 times
the average current. This means I can only have 4 rows ;/ but reduces the
number of drivers I needed originally by a factor of 4 which is better. (I
have 18 rows and 24 columns)

The TI IC can only deliever 30mA so this is out of the question as I would
need about 100mA or so. The ST has 80mA maximum. I might have to go with
this as there are no too many 24-ch drivers. Of course thats absolute max ;/

Maybe I would just have to go with 24 discrete mosfets for the column's and
18 discrete for the rows and then use some mosfet driver chips + uP to
drive? This of course is more work but solves the power capability issues.
Alternatively maybe I can find some LED drivers that have a larger capacity
at 16 ch.

Of course I'm looking at the worse case of all the LED's being on but since
it's not an impossibility...

Are you trying for individual proportional control of each led, or
just on-off control maybe with overall dimming (of all together)?
 
J

Jon Slaughter

John Devereux said:
Are you trying for individual proportional control of each led, or
just on-off control maybe with overall dimming (of all together)?

No, I need individual control because each LED is an RGB LED and I want to
mix for different colors.
 
J

Jon Slaughter

Paul Hovnanian P.E. said:
Is that actually a requirement?

? I have 100+ rgb LED's where each rgb LED will be used for multi-color.
Hence every individual LED needs brightness control(hence the PWM).
If you are going to be controlling things individually (i.e. in a matrix
display), equal brightness isn't as big an issue. With some LEDs on and
some off, and the display changing, nobody will notice some
non-uniformity.

Depends. This isn't a "display" and the LED's are scattered about(not random
but they are not packed tightly). They may or may not change but they will
not change often(maybe at most a few times a second).

Slight fluxuations in brightness won't be a huge deal and they calibrated
out if necessary but I'd rather not have to do that.

My idea right now is to basically use C mosfets, 1 for each column, and R
mosfets, 1 per row. This gives me the matrix and current capacity. The rows
will be driven by a ring counter and the columns by a mosfet driver that is
controlled by a uP.

I believe that will work and the only problem is coordinating the row's with
the individual PWM(which becomes sorta modulated by the refresh rate). I
think by choosing the refresh rate to be large would be ok but might cause
other problems.
If the point is to produce a uniform brightness across a panel, you are
going to have problems with LED tolerances. Particularly if you can't
get 1000 of the same P/N, from the same manufacturing batch.

Slight variances are not a big issue. And Again, I'm sure they can be
calibrated out if I have control of individual brightness. (i.e., say one
LED is too bright, I can just reduce the current to it... I'll have a lookup
table for all the LED's for changes)

I don't think it will be a huge problem though. I have to keep a table
anyways because I don't think brightness is proportional to the PWM
frequency?
 
G

Greegor

? I have 100+ rgb LED's where each rgb LED will be used for multi-color.
Hence every individual LED needs brightness control(hence the PWM).



Depends. This isn't a "display" and the LED's are scattered about(not random
but they are not packed tightly). They may or may not change but they will
not change often(maybe at most a few times a second).

Slight fluxuations in brightness won't be a huge deal and they calibrated
out if necessary but I'd rather not have to do that.

My idea right now is to basically use C mosfets, 1 for each column, and R
mosfets, 1 per row. This gives me the matrix and current capacity. The rows
will be driven by a ring counter and the columns by a mosfet driver that is
controlled by a uP.

I believe that will work and the only problem is coordinating the row's with
the individual PWM(which becomes sorta modulated by the refresh rate). I
think by choosing the refresh rate to be large would be ok but might cause
other problems.


Slight variances are not a big issue. And Again, I'm sure they can be
calibrated out if I have control of individual brightness. (i.e., say one
LED is too bright, I can just reduce the current to it... I'll have a lookup
table for all the LED's for changes)

I don't think it will be a huge problem though. I have to keep a table
anyways because I don't think brightness is proportional to the PWM
frequency?

Are you trying for individual proportional control of each led, or
just on-off control maybe with overall dimming (of all together)?



Do these people source their subsystems out to others?

How long have RGB LEDs been around?
Are they economical yet?


http://www.lsi-industries.com/contentindex.asp?ID=6

http://www.smartvision.com/eng/products/indoor.asp?id=30

Silverstar YG-LED 109 Matrix Panel
http://www.nordiskmusik.se/default.asp?id=17992

http://www.ledsignsupply.com/?gclid=CM3IlMrZm5gCFRwwawod0joUnA

http://www.ecvv.com/product/vp1345904/indoor-full-color-led-display.html



http://www.panstadia.com/vol5/54-028.htm

http://www.ledsmagazine.com/news/3/6/25



Old, but I thought some of the description might help,
like the part about "shouldering" for example.
But if you're using RGB LEDs that shouldn't be trouble, right?

http://digitalcontentproducer.com/mag/avinstall_led_displays_light/

LED DISPLAYS: light up the big screen
Jan 1, 2000 12:00 PM, Peter H. Putman

Previous columns covering large-screen displays mainly focused on
projectors and monitors using emissive (CRT and plasma) technology for
direct-view monitors, and we have concentrated on transmissive (LCD)
and reflective (LCD and DLP) imaging for high-brightness large screen
displays. In many of these applications, the ambient lighting
environment is controlled to maximize image contrast and legibility.
That is not always possible when designing and installing electronic
displays in arenas and stadiums. Outdoor stadiums in particular
present a lighting environment that can be characterized as hostile -
extremely bright sunlight, pronounced shadows, variable color
temperature and plenty of glare.

The new breed of arena builders are not satisfied with simple black-
and-white dot matrix displays for player stats, animated graphics and
advertising/promotions. The typical sports fan (or concert attendee)
is used to a steady diet of video and fancy visual effects, not to
mention instant replay. For them, full-color, full-motion video is the
only way to go.

How do they get there? Front projection systems are certainly bright
enough to light up 40 foot (12 m) wide screens with long projection
throws but cannot produce enough contrast under daylight levels that
can exceed 20,000 lux. Videowalls can crank up the brightness, but
they do not provide wide enough viewing angles, and their rear-screen
surfaces are easily washed out in direct sunlight.

The answer is to construct an emissive display that delivers a point
source of light bright enough to be seen more than 200 yards (183 m)
away under any kind of lighting. This ideal display would be visible
from viewing angles up to 160 degrees, provide full-color imaging (at
least 8 bit processing per color channel) and have a refresh rate fast
enough to show video. Perhaps this emissive display could also be
constructed in a modular fashion for easy assembly. While we are at
it, let's try to keep the weight down and the footprint small.

A pipe dream? Not really. Using nothing more than the garden-variety
light emitting diode (LED), a switching matrix and a lot of wiring,
SACO Smartvision of Montreal has been assembling some pretty
impressive direct-view displays for indoor and outdoor venues, among
them the Oakland-Alameda County Coliseum, the MCI Center in
Washington, D.C., and PSI Net Stadium in Baltimore, home of the
Baltimore Ravens football team.

The LED display concept is not new. LEDs have been used for crude, low-
resolution signboards for years, and they have also been put to work
in matrix configurations for numeric and character displays. The
reason they work well in large screens has to do with our perception
of image resolution. If you have a magnifying glass handy, hold it
over an advertisement or photograph in this magazine. You will see
thousands of colored dots or pixels, which make up the screen of the
photograph. At close range, these dots are quite noticeable, but at
normal viewing distances, you do not see the dots, just the image that
they form.

LED displays work exactly the same way. Viewed up close, they are a
jumble of red, green and blue dots. As you back away from the display,
the dots become less noticeable. Eventually, your eyes stop paying
attention to the display structure and concentrate instead on the
images being formed. Depending on how coarse or fine the matrix of
LEDs, this distance will vary from tens to hundreds of feet.

SACO's Smartvision screens are manufactured in a variety of screen
resolutions, primarily determined by the effective pixel pitch. In a
CRT display, that pitch is determined by the spot size of the electron
gun as it traces a raster and is typically less than 1 mm. Plasma
displays can have slightly large pixels close to 1 mm, while such
matrix displays as LCDs and DMDs are considerably smaller and are
measured in nanometers.

In contrast, the smallest pixel configuration in a Smartvision display
measures 4 mm, about 225infinity larger than a single pixel in a 0.7
inch (18 mm) diagonal SVGA LCD panel. The largest pixel matrix for an
indoor Smartvision screen is 15 mm, while the largest available for an
outdoor screen is 40 mm. Obviously, these screens were designed for
viewing at long distances. SACO's recommended minimum viewing distance
for a 4 mm pitch array is 10 feet (3 mm), while the 30 mm and 40 mm
arrays are specified for a 100 foot (30 m) viewing distance.

Using custom interfaces and driver boards, the individual pixel
matrixes are driven in a progressive-scan configuration. This allows
the display of both line-doubled video and computer graphics at a 60
Hz refresh rate.

How bright of an image will you see? SACO claims up to 2,500 candelas/
m superscript 2 for its indoor displays and between 5,000 candelas/m
superscript 2 and 6,000 candelas/m superscript 2 for the outdoor
versions. Ambient light levels indoors are more easily controlled, and
it is easy to obtain high contrast from such a display; there is
little light spilling on the screen surface that will reduce contrast.

Outdoors, however, things are different. As I write this article, it
is raining outside, and the sky is a dark, overcast gray. Even so, a
quick light reading shows that I still have more than 1,760 candelas/m
superscript 2 of daylight to contend with, which would reduce image
contrast to 2.8:1 using a 5,000 candelas/m superscript 2 Smartvision
LED array. So, another trick is used to kick up contrast - small
horizontal louvers that line the top of each Smartvision four-pixel
array. This louver can reduce stray light levels by a factor of eight
or more, boosting image contrast by a corresponding amount.

The secret to making these displays work was the discovery of a bright
blue LED, a process that has stumped engineers for many years. There
is something about the color blue that has also vexed scientists in
laser technology. In a typical Smartvision outdoor pixel matrix, four
individual five-LED arrays contain eight red LEDs, eight green LEDs
and four blue LEDs. The red and green LEDs surround the blue LEDs and
the effect looks like four dice rolled to show fives.

SACO claims 150,000 hours for each LED with a brightness fall-off of
only 15% after 100,000 hours of operation. It is possible to get even
more illumination from LEDs, but at an accelerated aging cycle.
Maximum viewing angles are specified at 170 degrees and 90 degrees
vertically.

There is a phenomenon to LED displays, known as shouldering, that will
affect normal viewing. Shouldering is caused by mutual obstruction
among adjacent LEDs, creating noticeable color shifts. For example, if
you are positioned at increasingly acute viewing angles from an
outdoor screen using the 4infinity1 matrix, you will notice more red
and green in the image and less blue as many of the blue LEDs are
partially blocked from view.

I traveled to two different arenas to check out a couple of
Smartvision screens recently, both installed by Professional Products
of Maryland. The first installation is part of a scoreboard in the
brand-new Sovereign Bank Arena in Trenton, NJ. This 8,000 seat venue
(10,000 for concerts) hosts both minor league hockey and basketball
games, and uses four 79infinity99 (2.1 m infinity 2.7 m) arrays of
Smartvision panels for video replays, advertisements, promotions and
special video graphics. The screens in this arena are the brightest
thing you see in the arena, even more so than the spotlights on the
ice.

Each panel is made up of a 16 infinity 16 LED matrix containing 256
red, green and blue LEDs, and there are 99 panels per screen (11
horizontal rows and nine vertical rows). The effective pixel
resolution of each active display is 2,816 infinity 2,304 sets of red,
green and blue LEDs. SACO uses 10 bit 4:2:2 component signal
processing or 1,024 colors per red, green and blue channel. The
interface is all digital and conforms to SMPTE 259M and CCIR-601 using
single coaxial cables for signal distribution.

Laura Black, the technical services coordinator for Sovereign Bank
Arena, uses a variety of formats to feed video and graphics to the
hanging video board, including Sony DVCAM and VHS videotape playback.
A Media 100 workstation is used for editing and special graphics
effects, and a variety of Videotek DDRs and framestores are available
for replays and still shots. Up to nine cameras can be handled through
a Ross RVS210A switcher, and a Pinnacle Deko 500 system provides real-
time video SFX.

About 100 miles (161 km) to the south, the staff at the Baltimore
Ravens facility have configured a unique Smartvision screen into their
own proprietary game-day video system they call "Raven Image". The
pair of LED displays sit at opposite ends of PSINet Stadium and
measure 100 feet wide by 25 feet high (30 m infinity 7.6 m) with a
viewable area of 969infinity249 (29 m infinity 7.3 m). Unlike other
stadiums where electronic displays are mounted high about the
nosebleed seats, these two screens sit nicely between the first and
second levels of the stadium, providing a more natural sightline.

Both screens use the 30 mm outdoor 4infinity1 pixel array and are
actually made up of two complete 489infinity249 (14.6 m infinity 7.3
m) screens that are precisely aligned to provide a 4:1 panoramic
image. As a result, there are four remote controls setting up and
calibrating the two screens. All video and graphics originate in a
sophisticated production studio that takes 15 people to operate during
a game. All images are captured, edited and manipulated as 16:9 525-
line video in the studio then effectively cropped by the long, narrow
Smartvision screen.

According to producer/director Marcia Kapustin, no other professional
sports team uses such an unusual production and display format, which
she considers ideal for the perspective of a football game. What is
even more interesting is that Raven Image runs continuously during a
game, just like a network broadcast. In fact, the Raven Image crew
will sometimes take camera feeds from CBS NFL telecasts and mix them
with its own 16:9 widescreen coverage.

The control room at PSINet Stadium includes a full-bore Sony DVS-7250
digital switcher with a raft of Ikegami cameras set up with 16:9
monitors for acquisition and three Tektronix PDR200 Profile disc
recorders/players with 12 channels of video for instant replay and
video segments. Up to 10 cameras can be sourced, and several Type DeKo
and Pinnacle DVExtreme boxes are online for special effects. Four DPS
465 frame synchronizers feed each of the four individual screens, and
Panorama aspect ratio converters can be used to resize 4:3 material to
the 16:9 format.

During my visit, the sun moved in and out of cloud cover and
completely illuminated the west screen. Despite this much ambient
light, there was enough contrast (about 10:1) in the image to clearly
see the clips of an earlier game. No doubt, the louvers helped because
the full daylight levels would have been far in excess of 5,000
candelas/m superscript 2 (low haze, direct sunlight). Of course, both
screens are in the field of view of at least 75% of the spectators, so
there is always one screen that is fully legible.
 
J

Jon Kirwan

Is that actually a requirement?


If you are going to be controlling things individually (i.e. in a matrix
display), equal brightness isn't as big an issue. With some LEDs on and
some off, and the display changing, nobody will notice some
non-uniformity.

If the point is to produce a uniform brightness across a panel, you are
going to have problems with LED tolerances. Particularly if you can't
get 1000 of the same P/N, from the same manufacturing batch.

Yes, I had to develop instruments for binning displays based on LEDs
for exactly this reason -- in a matrix with human perceived intensity
along one axis and human perceived color along the other. (monochrome
devices, there.) I also developed instruments for calibrating RGB
arrays so that all the pixels could be driven for video with a uniform
"white" if all are driven that way as well as calibrated mixing of
colors consistent with the CIE models of human perception. There is
some fun work in there.

http://www.infinitefactors.org/misc/images/p1000046_640x480.jpg

I just glanced at Mike H's suggestion of Macroblock MBI5031. I'll
have to look at those a little more later on.

Jon
 
Top