Max current for multiplexed LEDs

[pimpom]

I want to multiplex a 5-digit, 7-segment display, each segment
made up of an array of low-power red LEDs. The few LED datasheets
I've seen rate them at around 20mA continuous, while peak ratings
range from 50 to 70mA even at low duty cycles (10% or less).

Does this mean that low thermal inertia and/or the danger of
local hotspots do not permit operating the LEDs at higher peak
currents to get a bright display? (I'm referring, not so much to
the effects of persistence of vision as to maintaining the
*average* current near 20mA).

 

[Jon Kirwan]

I want to multiplex a 5-digit, 7-segment display, each segment
made up of an array of low-power red LEDs. The few LED datasheets
I've seen rate them at around 20mA continuous, while peak ratings
range from 50 to 70mA even at low duty cycles (10% or less).

Does this mean that low thermal inertia and/or the danger of
local hotspots do not permit operating the LEDs at higher peak
currents to get a bright display? (I'm referring, not so much to
the effects of persistence of vision as to maintaining the
*average* current near 20mA).

I am using an LED in a product, right now, that is spec'd at 20mA
continuous and 100mA at 10% duty. That's a current factor of 5, not
the 10 you might expect from the duty cycle/average calculation. It's
not unusual.

I haven't spent any more time thinking about it than just now, but
part of the reason may be due to the higher voltage required when
pulsing larger currents and the associated dissipation that resutls
from the difference. For example, the old 1980's style red LED had a
simplified (linearized) model that looks about like Vf= 1.55V + 21*I.
Power is then Vf*I, so P=1.55*I+21*I^2. At 20mA, this is about 40mW.
At 100mA, this is 365mW... and divided back down by 10, this isn't far
from the 40mW continuous figure. Anyway, that's about how I see it.

There are some 2mA per segment 7-seg displays to be had. You might
see if their specs may give you a little more margin.

Jon

 

[pimpom]

I am using an LED in a product, right now, that is spec'd at
20mA
continuous and 100mA at 10% duty. That's a current factor of
5, not
the 10 you might expect from the duty cycle/average
calculation. It's
not unusual.

I haven't spent any more time thinking about it than just now,
but
part of the reason may be due to the higher voltage required
when
pulsing larger currents and the associated dissipation that
resutls
from the difference. For example, the old 1980's style red LED
had a
simplified (linearized) model that looks about like Vf= 1.55V +
21*I.
Power is then Vf*I, so P=1.55*I+21*I^2. At 20mA, this is about
40mW.
At 100mA, this is 365mW... and divided back down by 10, this
isn't far
from the 40mW continuous figure. Anyway, that's about how I
see it.

That seems to explain it. Better to stay within specs then.
Thanks for the reply.

There are some 2mA per segment 7-seg displays to be had. You
might
see if their specs may give you a little more margin.

I'm not working with small standard displays using a single LED
for each segment. I need a large display 1-2 ft high and since
ready-made ones in that size are not easily available from where
I live, I'm going to have to build them - two 5-digit units.

I need a bright display because it's to be used outdoors where it
won't be possible to keep it in the shade all the time. Pity
about not being able to use the continuous rating as a
multiplexed average. It's probably wise to reduce the peak
current even further because the black mounting board will absorb
heat from sunlight.

 

[Jon Kirwan]

That seems to explain it. Better to stay within specs then.
Thanks for the reply.

No problem.

I'm not working with small standard displays using a single LED
for each segment. I need a large display 1-2 ft high and since
ready-made ones in that size are not easily available from where
I live, I'm going to have to build them - two 5-digit units.

I need a bright display because it's to be used outdoors where it
won't be possible to keep it in the shade all the time. Pity
about not being able to use the continuous rating as a
multiplexed average. It's probably wise to reduce the peak
current even further because the black mounting board will absorb
heat from sunlight.

Since you are dealing with the outside environment, contrast
becomes a whole new ballpark. You really need to carefully
think the optical side through. You can waste a LOT of power
and expense where it may not be needed, ignoring optical
contrast issues along with choosing color well. The whole
field is called "contrast enhancement" and it's worth looking
into, I suspect.

You want to maximize the contrast between ON and OFF
conditions. You can help this by reducing the reflected
ambient (I gather that's the reasoning for your choice of a
black background) and by concentrating on maximizing the
light reaching the intended eye.

The means aren't entirely about making a black background,
though. For example, circular polarizing filters (linear
polarizer + quarter-wave) can be placed in front to extingish
reflections from a specular surface. Doesn't do so much for
non-specular ones, since that kind of surface just messes up
the electric field angle all over again. Also, at steep
angles, all polarizations of sunlight reflect but at shallow
angles horizontally polarized light reflects better, so you
might consider using an appropriately aligned linear
polarizer (to extingish those horizontal polarizations)
depending on likely angles. There are anti-reflection
coatings; louvered filters (like the window coverings, in a
fashion); cross hatch versions of the louvered filters;
wavelength filters can help, too, but some of that depends on
what wavelength you are considering and the kind of days
there are (clouds filter out a lot of the red, whereas broad
daylight is pretty much "flat".)

Can you talk a little about your planning on this aspect? I
don't know a lot, but I might be able to suggest some things
to try out or consider. My guess is that you will do a LOT
more for yourself worrying about this part of the problem
than worrying about just how many milliamps you can pump into
the segments. That part is pretty much set for you by the
better devices you decide to use. So consider that a 'done
deal' and worry a lot more now about the optics side and do
some experiments before committing to a final design.

Jon

 

[krw]

I am using an LED in a product, right now, that is spec'd at 20mA
continuous and 100mA at 10% duty. That's a current factor of 5, not
the 10 you might expect from the duty cycle/average calculation. It's
not unusual.

It's not unusual for any electronics. The time component in the SOA
is an important concept. Localized heating causes all sorts of grief.

 

[Jon Kirwan]

It's not unusual for any electronics. The time component in the SOA
is an important concept. Localized heating causes all sorts of grief.

I held off discussing the more general issues with ICs, since
this was an LED. I didn't imagine an LED was complicated by
different trace widths, local densities of power dissipation,
differential ion migration from heating. But yes, in
general.

Jon

 

[krw]

I held off discussing the more general issues with ICs, since
this was an LED. I didn't imagine an LED was complicated by
different trace widths, local densities of power dissipation,
differential ion migration from heating. But yes, in
general.

Why would LEDs be any different? All packages have a thermal time
constant so one would expect some limit to PWM here too.

 

[Jon Kirwan]

Why would LEDs be any different? All packages have a thermal time
constant so one would expect some limit to PWM here too.

Complex ICs cannot be readily designed to equally distribute
watts across the surface, I'd imagine. There will be a
number of "weak" points and the weakest link determines it.
Much like the longest combinatorial pathway determines the
shortest cycle time for a particular micro design.

LEDs aren't entirely uniform blobs, granted. But the
calculations I gave seem to work well enough, just using a
linear approach. So I don't think a lot of extra worry about
nth order effects achieves much more.

As far as PWM goes, it seems to me that if the junction to
ambient is truly a constant (it may not be) independent of
the junction temperature, then it's merely about staying
under some peak temperature. I think the expoxies have a
glassification temperature point where it would be best to
stay underneath. (I think that point occurs well below any
worry about metal or dopant migrations.) This would explain
some of the derating (in my calculations, earlier, I showed a
40mW average and with a 10% duty a 36.5mW average -- which I
think comes from such a derating. If the PWM frequency is
very fast compared to the thermal times, then one could
probably hold closer to that 40mW average. If the PWM
frequency is low enough, the peak is worrisome and derating
is needed. I believe specs usually include a frequency at
which they have rated their 10% figures, which makes sense
here.

Jon

 

[Jon Kirwan]

I think you will have a hard time running a large board at a decent
refresh rate. You don't want to run into overvoltage conditions due to
a large L*di/dt. I'm inclined to say you should avoid multiplexing. I
am assuming the segment will be a string of LEDs, so you will have
high voltage (relatively speaking) driver circuits to deal with.
Remember, you need dead time when muxing to avoid ghosting. I think
it's going to be a lot of work, and certainly will be generating a lof
of RFI.

Many of these LED traffic lights are RFI sources since they flash the
LEDs a bit in the attempt to get higher perceived brightness. The
reality of brightness perception has been argued back and forth on
this newsgroup, never to a decent conclusion in my opinion, but I
think the pro-DC crowed argued better than the muxed and boosted
crowd.

It's not hard to test. I've some software available up on
the MSP430 web site that can be used with a $10 device from
TI to test things. Everything is already in place for those
wanting to play.

The first-order approximation is that humans perceive average
intensity when the pulse rate exceeds the critical fusion
frequency (CFF.) This is the Talbot-Plateau law. For 100%
modulation (on/off control), which is the usual case folks
talk about here, and assuming we are talking about photopic
vision and fairly bright illumination then somewhere between
45Hz and 70Hz is where things seem to fuse. Since the
contrast won't be perfect so as to make it appear to be 100%
modulation, that CFF may be lower in this case.

There is a Broca-Sulzer effect that isn't entirely understood
(at least, not before when I last read about it) which holds
that there is a modest increase in the appearances of how
bright a source looks when the ON pulse width is on the order
of 75ms! But that pretty much leaves out the usual PWM case.
You'd have to keep the ON pulse at about 75ms and vary the
OFF time and I don't think that works well for displays. (Not
at all well.) So it's not an effect that can be much used to
advantage.

If anyone is interested, it's not hard to set up a test
station to play with Broca-Sulzer and, as I mentioned, I've
already developed an educational set that includes docs and
all the necessary software to test out Talbot-Plateau for $10
US. (TI gets the money, not me.)

Bottom line is -- forget it. If using PWM, there's few free
lunches to be had. Plan on average and realize that the eye
sees things on a log scale so a 50% duty cycle is NOT the
same as half-brightness.

Oh, almost forgot. The CFF also varies a bit depending on
the size/area of the emitter. With LEDs on signs seen a ways
out, this probably isn't an issue. But for significant
areas, it turns out that the extrafoveal retina does a better
job at recognizing flicker (probably because of the need for
all mammals to see something coming from the periphery of
their vision) than the foveal retina does. So wider areas
require faster flicker rates.

Jon

 

[_]

That seems to explain it. Better to stay within specs then.
Thanks for the reply.


I'm not working with small standard displays using a single LED
for each segment. I need a large display 1-2 ft high and since
ready-made ones in that size are not easily available from where
I live, I'm going to have to build them - two 5-digit units.

I need a bright display because it's to be used outdoors where it
won't be possible to keep it in the shade all the time. Pity
about not being able to use the continuous rating as a
multiplexed average. It's probably wise to reduce the peak
current even further because the black mounting board will absorb
heat from sunlight.

Why not use those flip-segment displays - they have a little coil that you
pulse for each segment; painted in a flourescent green, on edge they almost
disappear.

 

[krw]

Why would LEDs be any different?  All packages have a thermal time
constant so one would expect some limit to PWM here too.

For electromigration, generally 1ms on-time is the limit where you can
play duty cycle games.[1ms on, 3ms off , 4x current, etc.) Obviously
you would need to contact the factory to get their opinion. The number
can be relaxed a bit. Anyway, that implies the strobing has to be
neighborhood of a few hundred Hz on the low end.

The easy way to check for an abused LED is to measure the reverse
leakage.

Ok, but that's a long-term effect. I'm talking about limits for
short-term use, which are thermal.

 

[krw]

Complex ICs cannot be readily designed to equally distribute
watts across the surface, I'd imagine. There will be a
number of "weak" points and the weakest link determines it.
Much like the longest combinatorial pathway determines the
shortest cycle time for a particular micro design.

Right, but it's not just complex ICs.

LEDs aren't entirely uniform blobs, granted. But the
calculations I gave seem to work well enough, just using a
linear approach. So I don't think a lot of extra worry about
nth order effects achieves much more.

You lose.

Thirty years ago, or so, we had a problem with a semi-custom contactor
driver. The short story was that they were getting clobbered by the
inductive kick back, not of the contactor, but the 20' of wire from
the control board to the contactor (the free-wheeling diodes were at
the contactors). TI's devices were crap, Motorola's not so much, and
Sprague never had a failure. TI basically took a logic device and
added a oversized 200mA transistor on the output, where Sprague took a
power transistor and added a little logic to the front end. No wonder
the Sprague was a better power device (and they did have good power
devices). The problem was secondary breakdown (localized heating).

As far as PWM goes, it seems to me that if the junction to
ambient is truly a constant (it may not be) independent of
the junction temperature, then it's merely about staying
under some peak temperature.

But at some frequency it won't be. There will *always* be a
temperature gradient.

I think the expoxies have a
glassification temperature point where it would be best to
stay underneath. (I think that point occurs well below any
worry about metal or dopant migrations.) This would explain
some of the derating (in my calculations, earlier, I showed a
40mW average and with a 10% duty a 36.5mW average -- which I
think comes from such a derating. If the PWM frequency is
very fast compared to the thermal times, then one could
probably hold closer to that 40mW average.

Absolutely backwards. If the PWM frequency is above the thermal time
constant there *will* be hot spots. The heat doesn't have a chance to
average out over the chip (and spreader). This has nothing to do with
dopant migrations. It's a very short term effect (milliseconds or
even microseconds)

If the PWM
frequency is low enough, the peak is worrisome and derating
is needed. I believe specs usually include a frequency at
which they have rated their 10% figures, which makes sense
here.

It depends on the "on" *period*. If that's less than the thermal time
constant you will have localized heating.

 

[Spehro Pefhany]

Right, but it's not just complex ICs.


You lose.

Thirty years ago, or so, we had a problem with a semi-custom contactor
driver. The short story was that they were getting clobbered by the
inductive kick back, not of the contactor, but the 20' of wire from
the control board to the contactor (the free-wheeling diodes were at
the contactors). TI's devices were crap, Motorola's not so much, and
Sprague never had a failure. TI basically took a logic device and
added a oversized 200mA transistor on the output, where Sprague took a
power transistor and added a little logic to the front end. No wonder
the Sprague was a better power device (and they did have good power
devices). The problem was secondary breakdown (localized heating).


But at some frequency it won't be. There will *always* be a
temperature gradient.


Absolutely backwards. If the PWM frequency is above the thermal time
constant there *will* be hot spots. The heat doesn't have a chance to
average out over the chip (and spreader). This has nothing to do with
dopant migrations. It's a very short term effect (milliseconds or
even microseconds)


It depends on the "on" *period*. If that's less than the thermal time
constant you will have localized heating.

There exist SPICE thermal models of some packages (more transistors
such as SOT-23 than LEDs). To model the actual effects there are
several time constants in series. Unless you're worried about how even
the current distribution is across the die (which would also be a
problem at DC in many cases, so perhaps it is built into the ratings),
one could step the current down to a relatively small fixed current in
the off times (maybe just parallel the switching device with a
resistor or an active current sink) and measure the actual (more or
less average) die temperature by looking at Vf. Easily characterized
in an environmental chamber at DC, or check it ast some fixed
temperatures such as 20C/100C and interpolate/extrapolate.



Best regards,
Spehro Pefhany

 

[Jon Kirwan]

Jon, what equation would you use to describe duty cycle versus perceived brightness?
I would like to include that in my LED controller.

Once you realize eye-response is log, the answer is obvious.
Use geometric progression. I wrote a series of tutorials
(most of which will be way below your interest, as they were
targeted largely for beginners) which culminate in using PWM
to drive an LED brightness according to an apparently linear
brightness curve. It includes some things you can toy with
and try out, as well. It's a series of 12 projects and a PDF
documentation file.

If you haven't already purchased one, buy one of these:
http://www.ti-estore.com/Merchant2/merchant.mvc?Screen=PROD&Product_Code=EZ430-F2013

They are $10 if you can find and use a discount code. Last
one I tried was this: MCU2009-03 but there is a change it
won't work, now. Yesterday it worked, though.

Then I can send you a zip file of materials I wrote or else
you can just download them from this folder:

http://tech.groups.yahoo.com/group/msp430/files/Member contributions/Simple EZ Lessons/

The direct file address for the zip file is given as:

http://f1.grp.yahoofs.com/v1/UGI-S1...ber contributions/Simple EZ Lessons/EZLED.ZIP

Which is __way__ long. Tiny URL gives me this from it:

http://tiny.cc/kV7pd

But I don't know if you need to be logged into Yahoo to get
it, or not. So that may be a problem.

You don't need to buy the device, obviously. If you read
Project 12 in the series, that may be enough for you.

Jon

 

[Jon Kirwan]

<snip>
It depends on the "on" *period*. If that's less than the thermal time
constant you will have localized heating.

I think that's what I was saying. The duration of the ON
time is key. Did I say something that disagreed with this? I
wouldn't have, had I realized it.

Jon

 

[Jon Kirwan]

I've looked around for 45-degree cross-hatch louvers -- they
seem better for outdoor viewing (if you can accept the
viewing angle limitations.) To avoid ghosting effects, some
translucency (10-15%) is helpful in the louvers, themselves.

#M has had something called "Light Control Film" for some
time. You might google that. It appears that it is
available now for computer monitors, ipods, and pretty much
anything that may get used outdoors.

There's this company to look at:

http://rpcphotonics.com/

I know nothing about them, but it appears they are in the
business area I'm talking about here, too. Namely,

http://rpcphotonics.com/news.asp?action=view&ID=9

Their president claims their films reduce energy consumption
in LED displays for outdoors, by 80%.

Apparently, they are working with Wahl Optoparts, which got
rebranded as Jenoptik Polymer and is now also working with
Lumenova:


http://ledke.spaces.live.com/blog/cns!833E0630C3F6F2D2!166.entry

But if you read about these things, you'll see why there is a
need to focus on the optics side to gain some real advantage
in the display. Just brute force alone will get you only so
far.

Jon

 

[Jon Kirwan]

wget --user-agent mozilla http://f1.grp.yahoofs.com/v1/UGI-S1...ber contributions/Simple EZ Lessons/EZLED.ZIP
--01:05:17-- http://f1.grp.yahoofs.com/v1/UGI-S1...ber contributions/Simple EZ Lessons/EZLED.ZIP
=> ZLED.ZIP'
Resolving f1.grp.yahoofs.com... 66.196.95.63
Connecting to f1.grp.yahoofs.com|66.196.95.63|:80... connected.
HTTP request sent, awaiting response... 404 Not Found
01:05:18 ERROR 404: Not Found.

Same with the short URL.

Yahoo groups may be a pain, here. I just went there, via my
login account, and it is definitely there and downloadable.

Forget the direct link. Probably has my session stuff in it.
Just go to the web link I provided before that. You'll
probably have to join or something, though. Want me to just
send the file to you?

Ah, hell. I'll load it to my own site. Here's the link:

http://www.infinitefactors.org/misc/EZLED.ZIP

Jon

 

[Don Klipstein]

I want to multiplex a 5-digit, 7-segment display, each segment
made up of an array of low-power red LEDs. The few LED datasheets
I've seen rate them at around 20mA continuous, while peak ratings
range from 50 to 70mA even at low duty cycles (10% or less).

Does this mean that low thermal inertia and/or the danger of
local hotspots do not permit operating the LEDs at higher peak
currents to get a bright display? (I'm referring, not so much to
the effects of persistence of vision as to maintaining the
*average* current near 20mA).

One thing I see in a lot of LED datasheets is pulse ratings amounting to
maximum RMS current not exceeding maximum continuous DC current. I forgot
how short the maximum pulse repetition period is for most of these, maybe
1 millisecond. Whatever that maximum pulse repetition period is, the RMS
current over that amount of time may have to not exceed the LED's maximum
DC current.

It appears to me that *somewhat* higher RMS current is OK (possibly
1.5-2 x the DC max. current rating? ) as long as the repetition period is
nice and short. With duty cycle in the 10% ballpark, the limiting factor
is usually bonding wires and maybe contacts to the die.

If duty cycle is short, this will put a crimp on average current.

I have pulsed LEDs with peak current as high as a majority of an amp -
and RMS maybe double the maximum DC rating of 30 mA. I had pulse
repetition rate of 5 KHz. The purpose was to achieve spectral change in
LEDs of a few types whose spectra change in interesting ways when
instantaneous current gets extreme. I do not guarantee that LEDs will
reliably survive this.

- Don Klipstein ([email protected])

 

[Jon Kirwan]

<snip>
I am looking at PRJ_12.CPP
<snip>

Yeah. That's the one. The idea is to just use geometric
progression. It's not tricky.

The comment says:
"The basic idea for recomputing
a new TACCR1 value is to multiply the on-time by 5/4ths if we are
increasing the brightness or else multiply the on-time by 3/4ths,
if we are decreasing the brightness."

It also points out, "This isn't an even-handed method, as the
dimming period will be shorter. But it is cheaper in terms
of code and doesn't harm things much."

Anyway, the only 'trick' in the code is that I switch to a
linear rise instead of an exponential one near the bright end
of the cycle to "visually blunt" the transition there. The
reason for doing it on the bright end is to "soften" the
transition when turning the direction around. To me, it's
more "pleasing" to see it that way. YMMV.

Jon

 

[pimpom]

I want to multiplex a 5-digit, 7-segment display, each segment
made up of an array of low-power red LEDs. The few LED
datasheets
I've seen rate them at around 20mA continuous, while peak
ratings
range from 50 to 70mA even at low duty cycles (10% or less).

Does this mean that low thermal inertia and/or the danger of
local hotspots do not permit operating the LEDs at higher peak
currents to get a bright display? (I'm referring, not so much
to
the effects of persistence of vision as to maintaining the
*average* current near 20mA).

Hi folks. Sorry about being silent for some time. My ISP was down
for more than 24 hours and I fell ill during that time. Nothing
serious, but I still don't feel up to going through your replies
in detail and answering them. Will come back ASAP. Thanks for
your interest.