How to build your own LED TV-like color displays?

There are some companies manufacturing large LED based TV-like, color
display units.
http://www.monstavision.com/
http://www.polycomp.co.uk/led-display-video-&-graphic.htm
http://www.daktronics.com/video_prod/dak_video_products.cfm

http://www.gizmag.com.au/go/4880/
http://engadget.com/2005/10/08/daktronics-to-build-worlds-largest-hd-led-display/

Electronics is an amateur hobby for me and I would like to investigate (and
eventually build) the design issues of LED based TV-Like color display
units.
As a starting point I would see the following grouping of issues;
1.) Mechanical design and manufacture of the modular LED panels
2.) Automatic LED adressing protocol (this will initialize individual
LED panels as well as individual LED when panels are connected with each
other to form large 2D matrix)
3.) Selecting/sourcing bright and low-cost LED with almost identical
specs.
4.) Digitising video into NxM color pixels (N, and M are number of
pixels in Row and Column)
5.) Sending the digitized image to indivually addressed LED panels
6.) Each LED panel should be intelligent to receive part of the
digitized video data corresponding to its own spatial position
7.) Scanning LEDs according to received pixel data.

Do you have any idea, information, design, URL etc. about above mentioned
and/or any other related issue to share?

Thanks.

Nicky

 

Structural, comms, bandwidth and power hassles will excalate
at an x by y rate. (at least).

Consider reliabilty, repair, 'fail gracefully', self-test
and fault-diagnosis issues.

Add in variations in LED colour and luminance with age
and manufacturer and manufactured date.

Not a trivial or cheap project beyond around 150 x 40 pixles.

 

Definitely.

But wouldn't a 720 x 576 panel be one heck of a project!

Dave

 

These are large displays that draw a lot of power (some modules are specced
at 0.5KW and higher per module)!

That one is going to need special cameras and other gizmos to drive it
as well as possibly its own power station!

Well that all depends on what you are used to. The bits described above are
LARGE systems with teams of people working on different parts.

Not unsurmountable, if you use a scalable design methodology. But will
probably require lots of LED drivers to keep LEDs 'ON' until a new value
appears, CRT and LCD have a persistence that keeps the 'dot' bright for
longer than the sweep of 'addressing' each 'dot'.

Please also consider what resolution display you want to make and then
total up the power involved in doing this and how you are going to drive
the POWER.

If all LEDs permanently left ON for QVGA resolution and various LED
currents

Total LEDs 320 x 240 = 76800

@ 10mA per LED 0.01 * 76800 = 768A (yes AMPS)

@ 5mA per LED 0.05 * 76800 = 384A

One line of LEDs ON at a time = 320 * 0.01 = 3.2A

Sorting out what POWER distribution and LED control is a CRUCIAL part of
this sort of design. These figures are true even if they are ALL turned
on at the same time for a short period of time, just the same current
is seen for a shorter period of time repeatedly.

How to turn on LEDs, for how long and how many at the same time has effects
on power and how you see the image (too dim, pulsing, beat frequencies).
Line at a time can cause vertical beat frequency 'cycling' through the
picture, groups at a time on can cause bands or other distortions, where
as single LED at a time on

Being sure you can get colour balance across the 'screen' will be the main
issue there, where by you see the same level and mix of colours for the
same colour at all points on the screen (within a _small_ tolerance) as
the eye will pick out the differences.

Your real problems are not so much the digitising, but synchronising and
scaling/cropping when you have different number of LEDs (pixels) compared
to input resolution and how to handle it.
Not forgetting if you are taking PAL or NTSC video, you have the
de-interlacing to deal with.

The real crucial issue is anything beyond 1bit depth per colour (R, G and B)
is a complicated distribution

analog - switching and distribution without signal loss across
the dispaly modules

digital - transferring all the digital levels, between the modules
(lots of wires or lots of higher speed serial transmission)

Which requires some form of intelligent controller that parcels out the
data (or provides clocking signals to say module start/end).

Shift register controls (chips or PLD) are probably the easiest method
for scan control, and cascadeable between modules. See also above about
power and duty cycle of LEDs.

Look at some of the Disco Dance floor designs for how people have done
projects with modular LED panels if nothing else for driving tehniques.
You will need a higher density of LEDs.

Such as

<http://web.mit.edu/storborg/ddf/>
<http://betterthaneveryone.com/>

 

You are probably going to be overrun with the new OLED/PLED
Technology. Samsung has demonstrated a 40" OLED Display panel.
The resolution is 1280x800, which is just over 3 million LEDs.
Stacking smaller panels, in stead of individual LEDs probably makes
more sense for future products.

http://www.samsung.com/PressCenter/PressRelease/PressRelease.asp?seq=20050929_0000194163

Regards
Anton Erasmus

 

hmmm, that'll need around 4m x 3m of surface area if
you use 3mm leds.
Awkward for personal viewing, but if you don't close
the curtains then I'm sure that the neighbours will
appreciate it <g>

I'll let you pay for the 240 Amps (peak) to run it.

And 3x 414720 high luminance Leds at (say) $0.10 each
comes to around $124k.
oh, you don't want a visible-in-daylight display ?,
fine, we'll use ordinary cheap Leds at $0.01 each
to reduce the price of the visible-bits to around $12k.

 

Where on earth would you source 240 amps from? I don't have such a
circuit in my house.

Also, the time to fabricate, by hand, would be incalculable.

My question is can the LEDs be operated in various intensities, or is it
only full-off/full-on? If its it the latter, how do you produce gray? If
its the former, how to you duty-cycle modulate millions of LEDs?

 

240 amps at ~3V for the LEDs is only about 8 Amps at 110V. The power
supply shouldn't be that tough to build.

Not really---at least if you have a decent computer and Excel.

I suspect that it would be similar to the fashion in which you modulate
millions of pixels in an LCD display. Kodak is making OLED displays
with 521x218 resolution now---this should just be a step up in size!
;-)

http://www.kodak.com/eknec/PageQuerier.jhtml?pq-path=2/782/1473/1481/1486/1495&pq-locale=en_US

Mark Borgerson

 

I think that was the point.

Long, yes. Incalculable, probably not.

Usually the latter. You can operate them at variable intensity
by varying the amount of current, but that's difficult. To
acheive "gray" you usually use pulse-width-modulation to vary
the duty cycle.

Pulse width modulation.

Good question.

 

This current would be at around the 3 (or so) volt level, so
you'll only need about 750W (peak) from the mains.

Not if you get it hand-assembled in China !
I once spend half a day populating and soldering a 256 Led panel. yetch.

To create colour, you need different intensities of the primary colours,
and the usual technique is to use pulse-width-modulation (PWM) because
it reduces power consumption and heat.
You'll also need colour correction of some sort because the Leds
you buy now will have different characteristics from the
ones you buy in 6 months time. (even if it's the same part number).
You can reduce this to some extent by buying a particular 'bin'
of tested and sorted devices, but the price then goes up a lot.

As to how to control millions of LEDs.., that I'll leave as an
exercise for you <g>,
but bear in mind that control and communication techniques which
work ok for a small number of pixels tend not to scale-up very
well when the numbers get into the realm of thousands of pixels.

Most sign manufacturers organise pixels in groups that can be
easily manufactured, installed and repaired as one entity.
eg. vertical strips of 2 x 16 pixels or 8 x 8 pixel panels,
etc etc etc.

 

["Followup-To:" header set to aus.electronics.]

you could use a step-down transformer (etc), leds only need about 3V each,

or possibly run the LEDs in a number of series chains with transistors to
bypass individual leds (to make them dark(er)) and probably some sort of
switching regulator at the head of each chain..... and probably some sort of
custom IC doing all the switching.

if one IC can handle 10 leds 7680 ICs would be needed...
the IC could do PWM to modulate the brightness.... and only a little over
24A would be needed (at aroound 30V)

another way would be to matrix them with capacitors and scan at a
reasonable speed.

| //
+-->|--+-[R]->|--.
| | |
| === |
| | |
--|------+---------+-- ROW
|
COL

 

Where did you get that number from?
If he uses 20 mA LEDs, and turns on one row at a time, the peak
current would be 720*0.02=14.4 A. That is 43.2 W peak at 3 V. The duty
ratio per LED would be 1/576. I would prefer 172.8 W and 1/144. A
modern, high-brightness 3 mm LED produces a lot of light at duty
ratio=1. At 1/144, probably one single pixel still produces more W/m^2
at the eyes of an observer that is located at a distance such that
s/he sees the 2.16 x 1.73 m (min) screen covering the same angle as a,
say, 32" CRT display with only one pixel on.

Yes, price and self-"madeability" of this: horrible.

 

eh ? what's wrong with a 2 orders of magnitude error ? <g> <g>

I assumed a non-multiplexed display to keep the brightness up
for outdoor use.
So 720 x 576 x 3 colours = 1244160 leds, and in the worst case
scenario they are *all* driven on at 20mA = 24.8 kA !!

 

the worse display ive seen was one that used 4 leds per pixel 2red 1blue
1green. the best display ive seen used composite led's which were 5MM
led's with 4 di's per 'package' 2blue 1red 1green. 4 elements in a 5mm led.

if you wanted to you could bring the display down to VCD resolution
(350x288 or what ever the vcd res is), it wouldnt be HD or SD but still
perfectly usable.

 

test