Full brightness multiplexing.

All my projects till now have used latched segments of 7 segment displays. ( or occaisionally 5 x 7 text character latched displays )

I run the LEDs at 18mA DC for maximum brightness when in direct sunlight, and maximum LED reliability.

I now need to make a display with 16 letters and 8 numbers, so I have decided I might as well go for 8x5 dot characters, but if I latch them my usually way, I will need 120 TPIC6B595s !

So now I am thinking multiplexing ( which I have avoided to avoid RF interference, which doesnt apply now as I dim the displays with pwm anyway )

I have done some searches and found there are some conflicting ideas out there about duty cycle and brightness.

I know that if I use 50% duty cycle drive with my latched system, the brightness appears half as bright.

Some say pulsing the LED at 50mA for 10% will give the same apparent output as a steady 25mA, the only way I can see this happening is if the light output is 5 times brighter at 50 mA, but the graphs dont show this.

Can anyone suggest the best way to get maximum brightness from the LEDs in a matrix format ?

I dont expect you will get any answers (any useful ones at least), have you done any testing? Maybe you could make a test rig, with several LEDs, and run them each with different duty cycles and visually compare them.

Even if you had a bunch of fancy equipment to measure everything, what will/wont work will depend on how it looks, so setting stuff up and testing it seems the most appropriate.

Two things and they contradict each other:

  1. after a certain point, high current barely increases brightness: most small leds light up at 5ma, almost as much as they at at 20ma.
  2. the only other thing you can do to increase brightness in a multiplexing is to increase current. And led rated at 20ma will typically take multiples of that current, happily.

I have experimented quite a bit with pwm dimming, and with current through the LEDs.

I generally limit the current to 18 mA, even though the LEDs I use are rated at 25mA, I find theres not much difference in brightness, and I have had less than 10 LEDs fail in the last 10,000 LEDs I have used.

I have decided to use latched TPIC6B595s to drive th LEDs, as I want them as bright as possible in full sunlight. I have designed a single sided board today, which was not as complicated as I feared.

Can anyone suggest the best way to get maximum brightness from the LEDs in a matrix format ?

I think what you're looking for is "peak forward current". Most of the time you can pump a lot more current through an LED if your pulse is short enough (typically 0.1ms), this can counteract the dimming you see.

You'll have to look at the LED spec sheet to find the maximum peak forward current because it varies; for example you can drive these at up to 175 mA:

http://www.us.kingbright.com/images/catalog/SPEC/APTR3216SYCK.pdf

but these you can only push up to 60 mA:

http://www.ebay.com/itm/3000-yellow-green-super-bright-1206-smd-led-light-bulbs-car-SMT-Brightness-High-/370691404486?pt=US_Car_Lighting&hash=item564eedf2c6

In any event, as dhenry states you should be able to get close to what you want by giving the LEDs more current.

Good luck!

I use quite a lot of peak current with IR LEDs, you can pulse some at 1 amp, in some TV remotes they dont even have a resistor in series, but use the internal resiastance of the battery to limit the current.

But if I can only drive my LEDs at 150Ma, and its only for 1/8th of the time, it would be about the same " power" being supplied to the LED as driving it at 18mA DC.

So what I was trying to find out ( as I havn't actually looked into multiplexing before ) is that if I have all 800 LEDs on at once, and I want them the same brightness as if they were all on at 18mA, would I need 8 rows and 100 columns with 100 LEDs on at once ?

It depends on your duty cycle (how you multiplex them) and your frequency.

Assuming sufficiently high frequency (100hz for example), you are then at the mercy of your multiplexing arrangement. If you scan 1 led at a time, your dc for each led is 1/800; and your need 800*18ma for each led to reach an average brightness of a 18ma constantly-on led;

If you scan 1 column at a time, your dc for each column is 1/100 and you need 100*18ma on current;

If you scan 1 row at a time, your dc is 1/8 and you need 8*18ma on current.

etc.

I think I will stick to my latched 18mA DC system, I don't really want to be pulsing 1.8 amps around the display screen . ( I have a wireless link controlling it and dont want to compromise reception )

I will play around with multiplexing for my next indoor displays, when I dont need the grunt to fight the sun.

Thanks all for comments.

But if I can only drive my LEDs at 150Ma, and its only for 1/8th of the time, it would be about the same " power" being supplied to the LED as driving it at 18mA DC.

Sort of yes, sort of no.

Human eyes respond to light logarithmically and not linearly...pumping twice as much power through an LED may or may not make it emit twice as many photons, but your eye definitely won't see twice as much brightness.

So what I was trying to find out ( as I havn't actually looked into multiplexing before ) is that if I have all 800 LEDs on at once, and I want them the same brightness as if they were all on at 18mA, would I need 8 rows and 100 columns with 100 LEDs on at once

If you multiplex 8 rows you'll have a duty cycle of 12.5% which is higher than most LED's limits for peak forward current (10%). My guess is you could drive them to around 75% of peak forward current and be fine; worst case scenario you'd be shortening their useful life from 15 years to 12 years.

If you'd like a real-world scenario I'm currently working with a 24x16 matrix (24 LEDs controlled by 3 TLC5916s X 16 rows controlled by MOSFETs + a 74hc154) and at 10 Khz (0.1 ms strobe cycle) running them in the matrix at about 75 mA each seems to be about as bright as driving them at 20 mA directly.

I have done some searches and found there are some conflicting ideas out there about duty cycle and brightness.

I know that if I use 50% duty cycle drive with my latched system, the brightness appears half as bright.

Some say pulsing the LED at 50mA for 10% will give the same apparent output as a steady 25mA, the only way I can see this happening is if the light output is 5 times brighter at 50 mA, but the graphs dont show this.

Can anyone suggest the best way to get maximum brightness from the LEDs in a matrix format ?

I wished to avoid the whole problem of figuring out led brightness Vs scanning rate Vs duty cycle when I began to build a 5x5x5 led cube. So doing research on different designs I came a across a method that takes all the thinking out of it and also eliminated needing 25 current limiting resistors at the same time. There are shift registers that have constant current output pins (mine are active low sinking) where you add one 'programming resistor' for the chip that sets the constant current value amount to your desired wish for all the output bits. Mine used two 16 bit shift registers wired in series to drive the 25 vertical columns of the led cube, and then one of 5 pnp switching transistor were used to enabled to supply the source current for a whole level. Sized the two programming resistors for 20ma and the things runs perfect with nice even brightness no matter how many leds are commanded on for any specific level. Such registers cost a little more but the saving in series current limiting resistors and the board space they took up is well worth it in my opinion, plus it let me get on with developing the sketch software without having to consider if I would have to deal with 'variable or low brightness' led by having to play with scanning speed and duty cycle.

Lefty

I don't really want to be pulsing 1.8 amps around the display screen .

That's just for individual led. Multiply that figure by the numbers you have.

This is why large commercial displays do not utilize multiplexing.

Having said that, let me just say that it's nuts to wanting to run your led at 18ma average current.

You are creating a hurdle so high for yourself that you practically have to fail.

This is why large commercial displays do not utilize multiplexing

I was under the impression that all large displays, commercial or otherwise, use multiplexing like this one:

Using one driver per pixel seems like it would cost infinity money.

Having said that, let me just say that it's nuts to wanting to run your led at 18ma average current.

Using one driver per pixel seems like it would cost infinity money.

The display needs all of that and more to compete with the direct sunlight. Theres a pic of the display facing into the sun on the homepage

www.scorebauds.co.za

This particular display I am on now is for team names , for the actual scores and countdown timer I usually use 16 to 20 3,000 candela LEDs per segment, or up to 140 LEDs per digit, so a 7x5 array of similar LEDs will be a lot dimmer anyway.

Boffin1:
So what I was trying to find out ( as I havn't actually looked into multiplexing before ) is that if I have all 800 LEDs on at once, and I want them the same brightness as if they were all on at 18mA, would I need 8 rows and 100 columns with 100 LEDs on at once ?

Yes, and you would need to run each LED at about 144mA when it is on. So that's up to 14.4A per column. Possible using mosfets and good grounding.

thanks dc42,

I am sure its possible, but I have to have decide this weekend, so I have no time to experiment .

I think I would have needed a double sided board for the normal MAX chip type multiplexing ? and the MAX chips here are over 10 times the price of the normal TPIC6B595N

I managed to design a single sided board with 5 TPICs for a 5x7 module today, and I know it works and I can get the chips. ( double sided boards are a lot more expensive here, especially the setup charges for a prototype board like this )

When I get a chance I will try some multiplex boards

Why not use thirteen TPIC6B95s as the column drivers and eight P-channel mosfets as the row drivers?

That could be good, trouble is I cant find the max allowable current pulse for the chips I have, and I have designed a single 5x7 board that I can just stack side by side for any number of digits, or build them up into bigger panels later on ( there is no border around the pcb )

I will look at that idea when I get a break though, and also see if I can do it on a single sided board, thanks

The good news is I have an extended period to design this project ( the pcb manufacturer closes down this week til mid Jan and is not taking any new jobs before then )

So I think I will try the Mosfet and TPIC idea, or possibly the MAX 7219, I dont know how Tadya can sell them for $1.25 when Mouser is $11 or more, and I see Tadya have an online discount of 15% until the 12th!

Which library would be best to experiment with, ( I only want single colour )

I am now going to RTFM about multiplexing, thanks for the leads.

how Tadya can sell them for $1.25

Trust me, you do not want to know.