Any practical limit to multiplexing LED minimum duty cycle?

[Klaus Kragelund]

Hi

I am working on a LED driver

Standard LED with average 2mA current. I am drinking it at 10mA, so duty cycle is 20%

Just wondering, is there any lower limit for the eye for minimum duty cycle?

Cheers

Klaus

[DemonicTubes]

That would probably be frequency dependent.

[Klaus Kragelund]

Let's say we use a frequency so the eye cannot detect flicker, 50Hz

And then drive the LED within the specs at 20mA, 10% duty.

The figure of merit of best efficiency point may be non optimal, but I do not consider that right now

[Klaus Kragelund]

Data sheet:

http://www.i-lcd.com/PDFs/SMT-LED/19-213-G6C-AP1Q2-3T_datasheet.pdf

[rickman]

If you drive with a fixed 10% duty cycle it will appear 10% as bright. What
is your question?

I have used duty cycle to reduce intensity before, but the blink rate was
*much* higher. At 50 Hz you can see the blinking when you move your eyes.
I find this effect very disturbing.

--

Rick C

Viewed the eclipse at Wintercrest Farms,
on the centerline of totality since 1998

[klaus.k...@gmail.com]

On Friday, 25 August 2017 17:35:52 UTC+2, rickman wrote:
> Klaus Kragelund wrote on 8/25/2017 11:30 AM:
> > Let's say we use a frequency so the eye cannot detect flicker, 50Hz
> >
> > And then drive the LED within the specs at 20mA, 10% duty.
> >
> > The figure of merit of best efficiency point may be non optimal, but I do not consider that right now
>
> If you drive with a fixed 10% duty cycle it will appear 10% as bright. What
> is your question?
>
> I have used duty cycle to reduce intensity before, but the blink rate was
> *much* higher. At 50 Hz you can see the blinking when you move your eyes.
> I find this effect very disturbing.
>
> --

LED driven with drop resistor with the following current and duty cycles:

40mA, 5% = 2mA avg
20mA, 10% = 2mA avg
10mA, 20% = 2mA avg
5mA, 40% = 2mA avg

Is that clear enough?

Cheers

Klaus

[DemonicTubes]

I worry that at 50Hz flicker would be detectable and annoying at duty cycles under 50%, especially since there is no phosphor to help out between pulses. It would be more noticeable in your peripheral vision and/or when moving your eyes.

[rickman]

Yes, this is clear. What is your question? A lower limit to the duty
cycle? I believe tests have shown there is no practical limit. I think you
reach the max current through the LED first.

[Phil Hobbs]

IIRC the eye does get less sensitive for very short pulses, but I forget
how short is short.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

[Tim Williams]

"Klaus Kragelund" <klau...@hotmail.com> wrote in message
news:9639bcdf-8aa5-469c...@googlegroups.com...

> Let's say we use a frequency so the eye cannot detect flicker, 50Hz
>

FYI: please use a high frequency, even just for displays. The flicker of a
low duty cycle at 50Hz is irritating to many people; at 100Hz, it may be
passable, but is immediately obvious as soon as one moves their eyes.

It can be a safety issue. Every time I'm driving at night, I see cars with
pulsed tail lights. Every time I scan my surroundings, bip-bip-bip-bip-,
there's a line of dots across my vision and in that instant, I don't know
where that car is.

My own experience has shown that over 1kHz is quite acceptable: it's hard to
even see the pulsing, and when it can be seen, it's more like a dotted line
than a series of overlapping frames.

I mean, it's not like it's hard to pulse LEDs at any speed. Unless they're
being turned on and off from the CAN bus. *Shudder*. (You can even pulse
them at >100kHz, add an inductor, and get higher LED efficiency! w00t!)

As for frequency, duty and current: AFAIK, at frequencies above the
persistence of vision, the eye is a perfect integrator. At least up to
pulse widths where transient vaporization occurs (TW, fs pulses?). :)

Tim

--
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com

[Pimpom]

On 8/25/2017 10:09 PM, Phil Hobbs wrote:
> On 08/25/2017 11:10 AM, Klaus Kragelund wrote:
>> Hi
>>
>> I am working on a LED driver
>>
>> Standard LED with average 2mA current. I am drinking it at 10mA, so duty cycle is 20%
>>
>> Just wondering, is there any lower limit for the eye for minimum duty cycle?
>>
>> Cheers
>>
>> Klaus
>>
>
> IIRC the eye does get less sensitive for very short pulses, but I forget
> how short is short.
>
> Cheers
>
> Phil Hobbs
>

Like Phil H, I'm not sure about the relation between pulse
duration, duty cycle and apparent brightness. One thing worth
noting though is that there's an SOAR-type limit to how far you
can reduce the duty cycle to get the same average current.

To take an extreme example for illustration, you cannot safely
run a common 5mm LED (T-1 3/4) at 2A peak at 0.1% duty cycle to
get 2mA average even if it has a continuous rating of 20-30mA.
The maximum peak current is probably in the region of 100mA. It's
best to check the manufacturer's datasheet if you're thinking
along those lines.

[upsid...@downunder.com]

At least on the CREE web site, there was a white paper about running
LEDs in pulsed mode.

[rickman]

Tim Williams wrote on 8/25/2017 1:46 PM:
> "Klaus Kragelund" <klau...@hotmail.com> wrote in message
> news:9639bcdf-8aa5-469c...@googlegroups.com...
>> Let's say we use a frequency so the eye cannot detect flicker, 50Hz
>>
>
> FYI: please use a high frequency, even just for displays. The flicker of a
> low duty cycle at 50Hz is irritating to many people; at 100Hz, it may be
> passable, but is immediately obvious as soon as one moves their eyes.
>
> It can be a safety issue. Every time I'm driving at night, I see cars with
> pulsed tail lights. Every time I scan my surroundings, bip-bip-bip-bip-,
> there's a line of dots across my vision and in that instant, I don't know
> where that car is.

YES! The first time I saw this I was merging onto a road that can be busy
and it appeared there was only a single car as I prepared to merge. On
moving my eyes back to forward it suddenly looked like a dozen taillights
were beside me and my instinct was to twitch the wheel away. Fortunately I
didn't as there was no place to go. That was a Cadillac and since then I
have seen this on many cars. I can't believe the auto makers haven't seen
that this is a problem and upped the frequency.

[Joerg]

If you stay below safe pulse power levels for the light output I believe
there is no limit in the percentages you are looking at. PWM-dimming
(which is duty-cycle dimming) is done down to factors of 0.001 and less.

Assuming you want to keep a reasonably high brightness with fairly harsh
multiplex duty cycles: For the LED itself you'll have to request the
peak forward current rating from the manufacturer. It is often stated in
the datasheet with infra-red LEDs but less so with LEDs for the visible
range.

A typical rating would be ten times the continuous max and if you go
above the rated pulse limit the lifetime begins to deteriorate. I
believe the mechanism of wear is degradation via local electromigration.
When you push the current to duty cycle ratio to extremes you could also
smoke a bond wire.

--
Regards, Joerg

http://www.analogconsultants.com/

[Klaus Kragelund]

Yes, I wrote that in the original post. I wont drive them higher than the 25mA rated current, 60mA peak. Mind you, that is a thermal limit the first one, the second a pulse current limit

Cheers

Klaus

[Klaus Kragelund]

Good info

I will contact the manufacturer for more information, just to be sure

Cheers

Klaus

[John Larkin]

Given the responses that you are getting here, maybe you should try
it.


--

John Larkin Highland Technology, Inc trk

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

[Joerg]

Can be risky depending on what the product is. For example, it might all
work fine on the bench and even for a few weeks. Then in the field LEDs
start to dim or fail way too early. We've something similar that at a
client where a contract design & assembly house had not protected an RF
input well enough. They claimed it was fine and that everything was
bench-tested. Problem was, after a number of thunderstorms rolled
through more and more transceivers became desensitized to the point
where end customers considered them deaf. Ranges went from hundreds of
feet to tens of feet. The customers wanted new units on warranty, lots
of them, and every one of them required a truck roll to install it.

I've got some more stories like that.

[Joerg]

The graph "Luminous intensity vs Forward Current" in the middle of page
5 on your datasheet goes to 60mA at 10% duty cycle which is encouraging.
I guess the dashed part means "only pulsed" but Everlight should be able
to put something in writing for you.

http://www.i-lcd.com/PDFs/SMT-LED/19-213-G6C-AP1Q2-3T_datasheet.pdf

Mind the temperature derating curve on page 5. For that to start
dropping already at 25C is a bit disappointing.

[John Larkin]

On Fri, 25 Aug 2017 16:47:15 -0700, Joerg <ne...@analogconsultants.com>
wrote:

Klaus said he wouldn't exceed the LED max current spec, so it should
be safe to play with duty cycles and see how it looks.

[k...@notreal.com]

On Fri, 25 Aug 2017 08:46:49 -0700 (PDT), klaus.k...@gmail.com
wrote:

The brightest and most efficient will be 2mA, 100% duty cycle.

[Phil Hobbs]

Luminous intensity has only a vague correlation with physiological
reality when short pulses are in view.

Cheers

Phil "down with photometrists" Hobbs

[Jasen Betts]

well, yeah, but

At 0.1% duty cycle the mean current would be 2mA, not 20mA

And at 1% the RMS current would be 200mA, enough to cause accelerated
aging through resistive heating

--
This email has not been checked by half-arsed antivirus software

[rickman]

For what range of values for "short"?

[Spehro Pefhany]

I don't know about the eye, there does not seem to be a strong effect,
but from reliability considerations I stay above 10% duty cycle for
"normal" brightness. Proabably less is acceptable in dim settings with
modern led dice, and more would be better if you are pushing the
apparent brightness towards limits.

--sp


--
Best regards,
Spehro Pefhany

[Robert Baer]

If you do not care about background lighting (eye does not seem to
care much if indirect), most bright LEDs are visible in the sub-microamp
(DC) region.
Eye may "see" things differently with PWM drive but same average
power/current.

[Robert Baer]

Phil Hobbs wrote:
> On 08/25/2017 11:10 AM, Klaus Kragelund wrote:
>> Hi
>>
>> I am working on a LED driver
>>
>> Standard LED with average 2mA current. I am drinking it at 10mA, so duty cycle is 20%
>>
>> Just wondering, is there any lower limit for the eye for minimum duty cycle?
>>
>> Cheers
>>
>> Klaus
>>
>
> IIRC the eye does get less sensitive for very short pulses, but I forget
> how short is short.
>
> Cheers
>
> Phil Hobbs
>

How about dark-adapted, seeing a single ten nanosecond light pulse
from a Huggins lamp at minimum voltage (about 30V if i remember correctly)?

[Robert Baer]

Beats the heck out of me what a crude measure of luminous intensity
for that 10nSec 32v Huggins lamp pulse is, but it seems bright.
100V drive seems to not increase brightness, tho.
The 32V is the ionization potential which makes testing at lower
drives not too useful to say the least.

[Pimpom]

That's what I said.

> not 20mA,

then I went on to say "*even if* it has a continuous rating of
20-30mA."

> And at 1% the RMS current would be 200mA, enough to cause accelerated
> aging through resistive heating
>

I didn't say anything about 1% at 2A.

[Klaus Kragelund]

I do not think that is true

If you draw up the efficiency curve normalized to current, then it's most efficient at 13mA. So my choice would be to drive the LEDs at 13mA, and fit the dutycycle to what light output is desired

Cheers

Klaus

[Klaus Kragelund]

We have a current design that I need to be able to match, but if I can do it smarter and better, I would like to do it :-)

Cheers

Klaus

[Klaus Kragelund]

The peak current of the LED is 65mA, so that is probably the point at which lifetime effects come into play

Cheers

Klaus

[Piotr Wyderski]

Klaus Kragelund wrote:

> Just wondering, is there any lower limit for the eye for minimum duty cycle?

My experiments show there is none (at least practical). Even 1e-3 is
pretty bright in the darkness. The eye has a shocking dynamic range.

Best regards, Piotr

[Joerg]

60mA is exceeding the abs max limit significantly but it may be ok. The
datasheet has dashed lines continuing to 60mA but only in graphs which
aren't guaranteed data. I'd ask the mfg.

[Phil Hobbs]

Well, people are making pulses shorter than 1 fs these days, so there
are about 13 orders of magnitude to play with before the blinking
becomes visible. I don't recall at what time scale the nonlinearity
becomes important.

Cheers

Phil Hobbs

[Phil Hobbs]

I didn't say that we were blind to short pulses, just that the
sensitivity goes down. It's easy to see light from femtosecond lasers,
for instance.

[Phil Hobbs]

One good way to estimate when you're hitting the LED too hard is to
measure the 1/f noise. At high currents you start to get
electromigration due to impurity scattering, which gradually causes
dislocations to form.

You can watch that happening by measuring the 1/f noise.

[Martin Brown]

On 25/08/2017 17:39, Phil Hobbs wrote:
> On 08/25/2017 11:10 AM, Klaus Kragelund wrote:
>> Hi
>>
>> I am working on a LED driver
>>
>> Standard LED with average 2mA current. I am drinking it at 10mA, so duty cycle is 20%
>>
>> Just wondering, is there any lower limit for the eye for minimum duty cycle?

> IIRC the eye does get less sensitive for very short pulses, but I forget
> how short is short.

So do classical emulsion films but ISTR it has to be really very short
before there is a perceptible difference in brightness. I suspect
driving LED displays with 10mA for 2ms or 20mA for 1ms would not show
any detectable difference to the human eye (except perhaps moving in
peripheral vision where the length of the dash would be longer).

I find some car LED brake lights particularly annoying in this respect.


--
Regards,
Martin Brown

[Joe Chisolm]

Several years ago I did a project using some Lumex SML-LX1206 series red,yellow and green.
I was driving them about 15mA if I remember correctly. I did steady on, blink and a
fade mode where the led would ramp up from off to full on and back to off over the
course of a couple of seconds. I found that any on time less than about 1200us was really
off to our eyes in a "server room" type lighting.


--
Chisolm
Republic of Texas

[dagmarg...@yahoo.com]

On Saturday, August 26, 2017 at 11:45:46 AM UTC-4, Phil Hobbs wrote:
> On 08/26/2017 02:31 AM, Robert Baer wrote:
> > Phil Hobbs wrote:
> >> On 08/25/2017 11:10 AM, Klaus Kragelund wrote:
> >>> Hi
> >>>
> >>> I am working on a LED driver
> >>>
> >>> Standard LED with average 2mA current. I am drinking it at 10mA, so
> >>> duty cycle is 20%
> >>>
> >>> Just wondering, is there any lower limit for the eye for minimum duty
> >>> cycle?
> >>>
> >>> Cheers
> >>>
> >>> Klaus
> >>>
> >>
> >> IIRC the eye does get less sensitive for very short pulses, but I forget
> >> how short is short.
> >>
> >> Cheers
> >>
> >> Phil Hobbs
> >>
> > How about dark-adapted, seeing a single ten nanosecond light pulse
> > from a Huggins lamp at minimum voltage (about 30V if i remember correctly)?
> >
>
> I didn't say that we were blind to short pulses, just that the
> sensitivity goes down. It's easy to see light from femtosecond lasers,
> for instance.

I did experiments circa 2002(?) with an 'A/B' switch that would switch
an LED between a low-duty pulse, and the equivalent average as a d.c.
current.

The answer to Klaus' question is that the practical limit was set by the
increasing i^2r losses of running the LED at high peak current. But down
to very short pulses, there was no perceptible difference between a
pulsed LED's brightness and the same LED running on the equivalent d.c.
current.

I think I tested down to ~10uS, IIRC. I was trying to increase energy
efficiency of 'white' LEDs, and it quickly became clear that d.c. was best.

(Some older colored LEDs had a "threshold" current. Pulsing those can
increase efficiency, depending.)

Cheers,
James Arthur

[mako...@yahoo.com]

>
> FYI: please use a high frequency, even just for displays. The flicker of a
> low duty cycle at 50Hz is irritating to many people; at 100Hz, it may be
> passable, but is immediately obvious as soon as one moves their eyes.
>

+1

yes at 50 or 60 Hz with any movement or viewing with peripheral vision, the flicker is very noticeable and annoying. It's not hard to use a higher rep rate.

m

[Klaus Kragelund]

That is good information, thanks :-)

Cheers

Klaus

[George Herold]

Huh, I did some testing on TH led's with lots of light output.
(well f'ing lumens on the DK site.)
~5-10 years ago. (I was looking for photon sources)
All at DC, the two or three I looked at carefully
all had a threshold. Below which light (current from a photodiode)
was less than linear with led current, (a 2/3rd's power law)
and above which was ~ linear..
until heating or resistance things took over at high current.
(I_threshold ~1mA, or so.) If you care about such things it's
not a difficult measurement. I've used the uA input on a DMM
with a PD as a meter. (fluke DMM) I'm not sure how deep a DMM
will go into the 2/3 region...

George H.
>
> Cheers,
> James Arthur

[dagmarg...@yahoo.com]

I also measured the photometric output vs. forward current for various
LEDs.

It's been too long for me to remember which LEDs had thresholds--I think
they were low-efficiency greens (GaP?) and or yellows. But the InGaN
blue, green, and white LEDs' outputs were impressively linear over the
three decade range I measured them carefully. And, separately, failure
to light at 1uA generally meant an ESD-damaged unit.

Lowering the forward current on the GaN LEDs /improved/ efficiency
substantially, down to fairly low currents, due to i^2r losses in
the LEDs' high ESRs. (I'd have to look up the data to give exact figures.)

> until heating or resistance things took over at high current.
> (I_threshold ~1mA, or so.) If you care about such things it's
> not a difficult measurement. I've used the uA input on a DMM
> with a PD as a meter. (fluke DMM) I'm not sure how deep a DMM
> will go into the 2/3 region...

For similar measurements I used a small solar cell as a PD, a small
test jig with an LT1014 TIA, into an old IBM analog interface card,
and logged by a BASIC program.

Since the pulse measurements were to determine perceived brightness,
for those I flipped the A/B switch and adjusted the currents or pulse
width for "no noticeable change" when flipping the switch.

Cheers,
James Arthur

[George Herold]

Right! Green and yellow.
https://www.dropbox.com/s/atyo4uvsb09fgd7/LED-PD.BMP?dl=0
(One of those part numbers is EOL)

>
> Lowering the forward current on the GaN LEDs /improved/ efficiency
> substantially, down to fairly low currents, due to i^2r losses in
> the LEDs' high ESRs. (I'd have to look up the data to give exact figures.)

Thanks, I'll have to look at GaN someday. In my case I wanted something
Orange or yellow..(green) a little above the band gap of the red LED that I
was using as a detector.

It'd be fun to look at all the different LED's.. it'd be an OK
high school science fair project.

George H.

[Phil Hobbs]

Nitride LEDs tune a lot with bias current, unlike phosphide ones. A bit
of a trap for young instrument-builders, that.

[rickman]

I've wondered why the auto makers use this type of circuit and I have to
assume they are trying to save those pennies from an ever so slightly more
complex current regulation circuit. Then I wonder why they don't just cut
off half the LEDs and I expect it is because they have designed the lenses
to work optimally with the minimum number of LEDs, so again, counting the
pennies. Damn those pennies!!!

[klaus.k...@gmail.com]

The circuit is used since it takes less connections (read smaller microcontrollers), less parts (shared resistors). So yes, pennies saved

Cheers

Klaus

[rickman]

No savings on MCU pins, just the $0.50 inductor required for the switched
current supply. I guess there is a diode as well.

[John Larkin]

Osram makes some really beautiful true-orange LEDs.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

[rickman]

I recall my customer's equipment having a bi-color LED that was orange and
purple, very pretty.

[George Herold]

Searches Digikey for orange osrams.. 15 found.

Most photons from this one (@50 mA!)

https://www.digikey.com/product-detail/en/osram-opto-semiconductors-inc/LO-E67B-U2AA-24-1-Z/475-1184-1-ND/810345

(same device shorter url)
https://www.digikey.com/products/en?keywords=475-1184-1-ND

George H.