LED 'afterglow' time?

OBW0549

Joined Mar 2, 2015
3,566
I don't have any data on this, but my understanding is that these parts exhibit significant afterglow, lasting on the order of many microseconds up to several milliseconds. IIRC, I've read they have a complex afterglow "tail" since the white color is obtained by a mix of phosphors, each having its own decay characteristics. (Caveat: this is all secondhand info, unverified and cited from my somewhat dim memory.)
"Somewhat" dim memory was evidently an understatement; turns out I was wrong.

I took a PIN photodiode, put it in series with a 1000Ω load resistor and gave it 10V of reverse bias to reduce its capacitance, then illuminated it with various red, green and white LEDs driven by my pulse generator while observing the voltage across the load resistor with my oscilloscope.

The red and green LEDs gave output rise and fall times of around 100 ns. The white LED (not a high-power illumination-type LED, just an ordinary 5 mm indicator type) showed rise and fall times of about 200-250 ns. But there was no observable afterglow from the white LED on any timescale-- nanoseconds, microseconds or milliseconds.

Santa is bringing me some 3 watt white LEDs to play with, and once we get past Christmas I'll rig up a high-power driver for them and give them a try. I'll also put together a fast transimpedance amplifier for the PIN photodiode which will hopefully take any detector speed limitations (like stray capacitance across that 1000Ω resistor) out of the picture.

But for now, the bottom line is it appears I was wrong about that "tail."
 

Thread Starter

Hypatia's Protege

Joined Mar 1, 2015
3,226
"Somewhat" dim memory was evidently an understatement; turns out I was wrong.

I took a PIN photodiode, put it in series with a 1000Ω load resistor and gave it 10V of reverse bias to reduce its capacitance, then illuminated it with various red, green and white LEDs driven by my pulse generator while observing the voltage across the load resistor with my oscilloscope.

The red and green LEDs gave output rise and fall times of around 100 ns. The white LED (not a high-power illumination-type LED, just an ordinary 5 mm indicator type) showed rise and fall times of about 200-250 ns. But there was no observable afterglow from the white LED on any timescale-- nanoseconds, microseconds or milliseconds.

Santa is bringing me some 3 watt white LEDs to play with, and once we get past Christmas I'll rig up a high-power driver for them and give them a try. I'll also put together a fast transimpedance amplifier for the PIN photodiode which will hopefully take any detector speed limitations (like stray capacitance across that 1000Ω resistor) out of the picture.

But for now, the bottom line is it appears I was wrong about that "tail."
Many, many thanks for your effort and independent research into this matter!:):):)

Very best regards
HP
 

RichardO

Joined May 4, 2013
2,271
"Somewhat" dim memory was evidently an understatement; turns out I was wrong.

But there was no observable afterglow from the white LED on any timescale-- nanoseconds, microseconds or milliseconds.
I was wrong as well... Thanks for setting us both straight. :)

As Tektronix says: "Don't just observe, measure."
 

OBW0549

Joined Mar 2, 2015
3,566
I put together a fast transimpedance amplifier for the PIN photodiode using an LM6171 100 MHz opamp with a 1.0KΩ feedback resistor in parallel with a 10pF capacitor for stabilization, and repeated my measurements.

Results:

Red LED: 40 ns rise time, 60 ns fall time
Green LED: 100 ns rise time, 50 ns fall time
White LED: 300 ns rise time, 200 ns fall time

Santa is bringing me some 3 watt LEDs, and once I've worked up a suitably fast power driver for them I'll see what they do and report on my findings.
 

cmartinez

Joined Jan 17, 2007
7,313
I put together a fast transimpedance amplifier for the PIN photodiode using an LM6171 100 MHz opamp with a 1.0KΩ feedback resistor in parallel with a 10pF capacitor for stabilization, and repeated my measurements.

Results:

Red LED: 40 ns rise time, 60 ns fall time
Green LED: 100 ns rise time, 50 ns fall time
White LED: 300 ns rise time, 200 ns fall time

Santa is bringing me some 3 watt LEDs, and once I've worked up a suitably fast power driver for them I'll see what they do and report on my findings.
Nice... maybe you could post your results (circuit and all) in the completed projects forum. For someone like me at least, it would be excellent to have it readily available for future reference.
 

OBW0549

Joined Mar 2, 2015
3,566
Santa is bringing me some 3 watt LEDs, and once I've worked up a suitably fast power driver for them I'll see what they do and report on my findings.
Santa arrived right on time bearing my precious cargo, and I'm done testing. The attached .pdf file includes schematics for the LED driver and the photo detector, as well as a scope screen shot showing the LED light output (upper trace) when driven by a 900 mA peak, 1.0 μs pulse signal.

I was quite surprised at the speed of this 3 watt LED, and the absence of any "tail" (at least, any tail large enough to register on my DSO).

From what I observed, it seems to me the biggest impediment to using these things for stroboscopic work is that they don't produce very much light; driving the LED at its rated current with 1.0 μs pulses at ≈1% duty cycle (9.4 kHz), the light output was visually quite dim.
 

Attachments

cmartinez

Joined Jan 17, 2007
7,313
Santa arrived right on time bearing my precious cargo, and I'm done testing. The attached .pdf file includes schematics for the LED driver and the photo detector, as well as a scope screen shot showing the LED light output (upper trace) when driven by a 900 mA peak, 1.0 μs pulse signal.

I was quite surprised at the speed of this 3 watt LED, and the absence of any "tail" (at least, any tail large enough to register on my DSO).

From what I observed, it seems to me the biggest impediment to using these things for stroboscopic work is that they don't produce very much light; driving the LED at its rated current with 1.0 μs pulses at ≈1% duty cycle (9.4 kHz), the light output was visually quite dim.
Wow... I didn't know you could use a mosfet driver to drive a LED. Are you using it because of its capability of driving a highly capacitive load? What about current, is the 6.8Ω resistor the only component that's limiting it? Or is the MCP1406 also playing a part in that?
 

OBW0549

Joined Mar 2, 2015
3,566
Wow... I didn't know you could use a mosfet driver to drive a LED.
Oh, yeah! The MVP1406 will put out 6 amps, according to its datasheet. Good for all kinds of things!

Are you using it because of its capability of driving a highly capacitive load?
Mostly because it gave me a minimum-parts-count way of driving the LED, and it's extremely fast.

What about current, is the 6.8Ω resistor the only component that's limiting it? Or is the MCP1406 also playing a part in that?
The datasheet says the output resistance of the MCP1406 in the output high state is typically 2.1Ω, so I expect it was mostly the 6.8Ω resistor that was limiting the current.

The question on my mind right now is, how much can one of these 3 watt LEDs be overdriven (in short pulses at low duty cycle) to increase its light output, without suffering damage? Can they take 1.0 μs pulses at 10 amps? 100 amps? More? Inquiring minds want to know...[/QUOTE]
 

Thread Starter

Hypatia's Protege

Joined Mar 1, 2015
3,226
it seems to me the biggest impediment to using these things for stroboscopic work is that they don't produce very much light; driving the LED at its rated current with 1.0 μs pulses at ≈1% duty cycle (9.4 kHz), the light output was visually quite dim.
IMO such is more than compensated by the precision with which they may be switched!:):):)
Wow... I didn't know you could use a mosfet driver to drive a LED.
Oh, yeah! The MVP1406 will put out 6 amps, according to its datasheet. Good for all kinds of things!
Yes, indeed! Gate drivers are often the ideal solution where high speed/low EMF pulse generation is required:)
Good Lord... that's more than $10 bucks a pop! You may want to reconsider using the see-how-much-current-it-can-take method
'Tis but a pittance - In the name of science!;):D

@OBW0549
Again -- many sincere thanks for your efforts on and interest in this matter!!!:):):)

Best regards
HP:)
 

OBW0549

Joined Mar 2, 2015
3,566
IMO such is more than compensated by the precision with which they may be switched!:):):)
One can always hope.

But as I said, running at its rated maximum current and 1% duty cycle, this LED was very dim; I doubt it would be usable as a strobe light for observing, say, rotating machinery. Also, for that kind of work you'd want quite a bit less than 1% duty cycle, probably more like 0.1%, to avoid blurring the subject; and that would make it VERY dim indeed.

I'm thinking the solution would have to involve one or more of the 100 watt LEDs, run well above their rated power. Can they take kilowatt pulses of 10 μs duration, repeated at a 60 Hz (3600 RPM) rate? I'd like to find out.
 

cmartinez

Joined Jan 17, 2007
7,313
One can always hope.

But as I said, running at its rated maximum current and 1% duty cycle, this LED was very dim; I doubt it would be usable as a strobe light for observing, say, rotating machinery. Also, for that kind of work you'd want quite a bit less than 1% duty cycle, probably more like 0.1%, to avoid blurring the subject; and that would make it VERY dim indeed.

I'm thinking the solution would have to involve one or more of the 100 watt LEDs, run well above their rated power. Can they take kilowatt pulses of 10 μs duration, repeated at a 60 Hz (3600 RPM) rate? I'd like to find out.
This rises the question, why does a LED normally fail in the first place? Is it because of overvoltage, overcurrent, overheating? Or is it a combination of the previously mentioned factors?
 

hp1729

Joined Nov 23, 2015
2,304
Kind friends:

It occurs to me that stroboscopic tachometery represents a highly satisfactory solution to determination of rotor bearing condition (Re: Coolidge RA types) --- Moreover, owing to enhanced switching stability and favorable (i.e. compact) geometry, LED technology would seem a better choice than traditional ‘flash-tube’ systems…

Hence my questions:

1) Do ‘ultra bright’ white LEDs exhibit significant ‘after glow’? If yes, what is the antcipatable range in characteristics (i.e. luminance vs time following interruption) –- failing precise data - how do such LEDs compare to Xe 'photoflash tubes' in this regard?

2) Are such LEDs 'tolerant' of rapid switching?

Many advance thanks for any and all assistance!:)

Best regards
HP
how is the project progressing? Have you found the white LEDs to ne slow? Would red LEDs work? Multiple LEDs?
 

GopherT

Joined Nov 23, 2012
8,012
This rises the question, why does a LED normally fail in the first place? Is it because of overvoltage, overcurrent, overheating? Or is it a combination of the previously mentioned factors?
In rapid high current pulses, lead-wires typically fail and either melt or become de soldered from the die. If the current is lower for a prolonged period, their are various physical/material damages (various dopants start diffusing on the substrate and fail, Gallium nitride layer delaminates from the silicon or sapphire substrate, ...)
 

cmartinez

Joined Jan 17, 2007
7,313
In rapid high current pulses, lead-wires typically fail and either melt or become de soldered from the die. If the current is lower for a prolonged period, their are various physical/material damages (various dopants start diffusing on the substrate and fail, Gallium nitride layer delaminates from the silicon or sapphire substrate, ...)
So it's mainly overcurrent that makes them fail, and not actually overheating? Or maybe it is overheating, but it happens right at the LED's core, and there's no practical way of effectively removing that heat?
 

kubeek

Joined Sep 20, 2005
5,751
So it's mainly overcurrent that makes them fail, and not actually overheating? Or maybe it is overheating, but it happens right at the LED's core, and there's no practical way of effectively removing that heat?
It is mainly temperature of the die, if you will run a led in a 150°C environment at tiny current, it would still fail very soon.
Only in the situation of short but grossly overcurrent pulses (which on average could have the same power as a normally powered led) come into play the bond wires and die thermal capacity and thermal impedance and resultant heat spots or short term overheating.
EDIT: that is why in the osram datasheet I posted earlier the allowable peak current doesn´t rise unbounded with lower duty cycle, but is capped at 2.5A for however short bursts.
 
Last edited:

Thread Starter

Hypatia's Protege

Joined Mar 1, 2015
3,226
One can always hope.

But as I said, running at its rated maximum current and 1% duty cycle, this LED was very dim; I doubt it would be usable as a strobe light for observing, say, rotating machinery. Also, for that kind of work you'd want quite a bit less than 1% duty cycle, probably more like 0.1%, to avoid blurring the subject; and that would make it VERY dim indeed.

I'm thinking the solution would have to involve one or more of the 100 watt LEDs, run well above their rated power. Can they take kilowatt pulses of 10 μs duration, repeated at a 60 Hz (3600 RPM) rate? I'd like to find out.
I've two conditions working in my favour:)

1) While the maximum working anode AV is, in fact, 3600 RPM -- 'sub multiple' flash rates are acceptable inasmuch as calculation of true AV based upon observation at different strobe frequencies is not burdensome for periodic inspection...

2) Inasmuch as the light must be directed through the (~30 mm) beam port aperture, 'concentration' via optics may increase the intensity -- additionally, the interior of the housing is dark sans application of cathode current...

Linked image of a typical tube/housing assembly...


Again many, many thanks!
HP:)
 
Last edited:
Top