Looking for a specific DC CC driver for LEDs

Yes. Normally, a DC Vf is a function of DC current, just as DC current is a function of DC Vf.

The question is whether the PWM generation of the current source will illicit a different Vf response. For example, assume a 36Vf LED: if the LED runs at only 32V DC and 1.5A DC current. Would pulsing 1.5A DC drive the LED at 32V only, or would it be 36V?

If the response were linear, it wouldn't be much of an issue. The fact that the response is highly non-linear, especially at low currents, is potentially a large source of error.
 
karimwassef said:
Yes. Normally, a DC Vf is a function of DC current, just as DC current is a function of DC Vf.

The question is whether the PWM generation of the current source will illicit a different Vf response. For example, assume a 36Vf LED: if the LED runs at only 32V DC and 1.5A DC current. Would pulsing 1.5A DC drive the LED at 32V only, or would it be 36V?

If the response were linear, it wouldn't be much of an issue. The fact that the response is highly non-linear, especially at low currents, is potentially a large source of error.
Click to expand...



If you don't mind losing the full range PWM dim capability, output capacitance is the solution :)

(There is also the theory that pulsing LEDs is not ideal for biological processes, which I haven't seen scientific data myself on but is an interesting question none the less).
 
Since you're the one with the beefy power supply... can you run the test to see if there's a difference in IV response with a pulsed current source vs. a DC voltage source?

At this point, the only way I can tell if a chip is degraded is to capture it's "NEW" state and determine how much the voltage increases for the same current.

How would capacitance reduce the full range PWM dim capability? I'm assuming you mean the input capacitance so that it slows down the ramp of the voltage?

As far as pulsed LED vs. DC LED having an effect... my opinion is that it doesn't matter. Biological tissue is a great integrator unless you're in the limits of physical destruction. Most biological processes also operate at a much slower rate than you can influence with kHz range pulsing...

Also, if it were a significant effect, it would be published for hydroponics where growth rate = $$$ ... Since LEDs are now being used to grow vegetables in space, I would think NASA would have spent a little time studying the effects.

I think the myth comes from LASERs that are based on diodes, like LEDs. High power LASERS are pulsed to burn away material without overheating. So, if you're in the range of instantaneously destructive energy, then I think it matters.
 
karimwassef said:
Since you're the one with the beefy power supply... can you run the test to see if there's a difference in IV response with a pulsed current source vs. a DC voltage source?
Click to expand...

Yes. I do expect a difference, but not much.

At this point, the only way I can tell if a chip is degraded is to capture it's "NEW" state and determine how much the voltage increases for the same current.

How would capacitance reduce the full range PWM dim capability? I'm assuming you mean the input capacitance so that it slows down the ramp of the voltage?
Click to expand...

Output capacitance. Its going to limit the full dynamic range of dimming, as the hold-up is going to be more than one cycle of the inductor current.

As far as pulsed LED vs. DC LED having an effect... my opinion is that it doesn't matter. Biological tissue is a great integrator unless you're in the limits of physical destruction. Most biological processes also operate at a much slower rate than you can influence with kHz range pulsing...

Also, if it were a significant effect, it would be published for hydroponics where growth rate = $$$ ... Since LEDs are now being used to grow vegetables in space, I would think NASA would have spent a little time studying the effects.

I think the myth comes from LASERs that are based on diodes, like LEDs. High power LASERS are pulsed to burn away material without overheating. So, if you're in the range of instantaneously destructive energy, then I think it matters.
Click to expand...

My CO2 laser cutter agrees on the power of pulses ;)

Its pretty easy to end up with near-0 ripple current on LED drivers with enough bypass capacitance, and control any intensity via current limiting. The mass use of PWM dimming in horticultural at least would indicate several kHz of PWM frequency don't matter.
 
Ok. Finally. Results.

So, Here is the model that I ended up using.

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/0_zpsgfjeovps.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/0_zpsgfjeovps.png" border="0" alt=" photo 0_zpsgfjeovps.png"/></a>
 
This allows me to separate the light output in W and heat output in W as a function of measured PAR, I, and V.

Including the exponential (or equivalent natural log) diode model and the saturation characteristic of the light output allows for a very high fit across all data.

First, here's the calculation for converting PAR to Light Power (W):

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/8_zpse4fjixnz.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/8_zpse4fjixnz.png" border="0" alt=" photo 8_zpse4fjixnz.png"/></a>

This represents the power density for a cone of projected light based on the spread from the LEDs used.
 
Oh, my, geek speak! this thread has gone way, way over my head..... :lolspin: can't wait to see what results in terms of a controller that can monitor LED health.....
 
Here is the data applied to the 5 tests that include all datasets:

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/1_zpsdyulw4tr.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/1_zpsdyulw4tr.png" border="0" alt=" photo 1_zpsdyulw4tr.png"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/2_zpscquld3si.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/2_zpscquld3si.png" border="0" alt=" photo 2_zpscquld3si.png"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/3_zpsyeddpota.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/3_zpsyeddpota.png" border="0" alt=" photo 3_zpsyeddpota.png"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/4_zpsxa4az8c0.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/4_zpsxa4az8c0.png" border="0" alt=" photo 4_zpsxa4az8c0.png"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/5_zpsputhtlqo.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/5_zpsputhtlqo.png" border="0" alt=" photo 5_zpsputhtlqo.png"/></a>

So the fit is very representative and aligns with all the data, including the compromised LEDs that have suffered stress damage.
 
The exponential coefficients "k" capture the diode response. Without this factor, the low and high current data fits wouldn't align. I originally thought this was temperature driven, but adding this coefficient drove very tight alignment.

For most cases, the coefficients are set to 10^15 where the fit response was k> 10^6. The error from this artificial constraint was very small, but it allows the data across all LEDs to be compared.

The exception is for the stress damages devices. Those exhibited a small coefficient that moves the response much closer to the diode range / or alternatively, we can say that the diode response moves higher up in current with a softer transition to a linear IV characteristic.

The easiest way to see this is to compare the compromised diodes 4,5 to the good diodes 1,3. The curvature of the IV curve that is usually constrained to very low currents begins to show at higher currents.
 
Here's the comparisons of the 5 test (note that LED 3 was run twice, once with a lens and one without)

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/6_zpszvkgfesr.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/6_zpszvkgfesr.png" border="0" alt=" photo 6_zpszvkgfesr.png"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/7_zpswwsjeiub.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/7_zpswwsjeiub.png" border="0" alt=" photo 7_zpswwsjeiub.png"/></a>
 
LED 4 (blue LED that started at a very high voltage) is clearly off.

LED 5 (white LED that experienced thermal stress due to running temporarily without a fan) is also different, but I can't immediately tell from the data capture.

I think this is because I have no other white LED data yet to create a baseline for white LEDs. I expected PAR for white to be higher than blue (based on the definition of PAR), but even a compromised white LED seems to put out a lot more than a comparable good blue.

To be fair, LED 5 didn't suffer through the same stress as LED 4 - so maybe I'm just being a little picky.

I would say that the most direct indicator of a compromised chip is the low current IV characteristic (represented by the small k coefficient). That doesn't mean that the LED is not putting out enough light or dissipating more heat. It just means that its life is shortened.

A bad LED (one that dissipates a lot of heat and produces little light) with poor wall efficiency can be identified by an increased voltage at the same reference current. This means that a baseline LED measurement can be used to capture true light degradation.

I couldn't tell that the LED was failing just by visually looking at the light output... but the IV readings were very clear. The PAR just confirmed this.
 
Last edited:
zachts said:
Oh, my, geek speak! this thread has gone way, way over my head..... :lolspin: can't wait to see what results in terms of a controller that can monitor LED health.....
Click to expand...

NOOOOO... It's not over your head. I'll answer any questions to make sure we're all on the same page.

The first step was to convert a PAR reading to power in Watts. That allows me to compare the power consumption of the LED to the light output.

All that math was just to walk through the conversion formula. Now that it's done, anyone can apply it. Just divide PAR by (4.6 x 1550) and multiply that by the area being covered in square inches... that'll give you the power in Watts. (That's if you believe my formula).

The next step was to get the data to match a function that makes sense. Then, I look for a pattern in the terms that represent resistance or radiation, etc...

The big conclusion is that there are a couple of markers that can be used to tell a good LED, from a weakened LED, to a failing LED...
 
Another result of this is that the wall plug efficiency is actually quite low, even for good LEDs (even if it is 3x that of MH :) )
 
This should make it clearer

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/9_zps3edupau5.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/9_zps3edupau5.png" border="0" alt=" photo 9_zps3edupau5.png"/></a>

4.4 is the failing LED
5.2 is the reduced life LED

The other 2 LEDs (three tests) line up with the expected output.
 
I have 4 DC-DC converters (2 boosts and 2 bucks), multiple current shunts and an arduino mega...

I also have many more LEDs to test and collect a bigger dataset.

And finally, multiple lenses to evaluate the distribution. This will probably go in its own thread... I'm very curious about the high end street light lenses... so here's the teaser.

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/IMG_5217_zpslmtamx4l.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/IMG_5217_zpslmtamx4l.jpg" border="0" alt=" photo IMG_5217_zpslmtamx4l.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/9A335AF4-8B27-43D7-B36C-5FC636C41A5F_zpsk9pvx4vn.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/9A335AF4-8B27-43D7-B36C-5FC636C41A5F_zpsk9pvx4vn.jpg" border="0" alt=" photo 9A335AF4-8B27-43D7-B36C-5FC636C41A5F_zpsk9pvx4vn.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/4AFE6D3E-E5B1-4E9A-B47C-9F28CDA6DCD1_zpsvlkrcnbj.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/4AFE6D3E-E5B1-4E9A-B47C-9F28CDA6DCD1_zpsvlkrcnbj.jpg" border="0" alt=" photo 4AFE6D3E-E5B1-4E9A-B47C-9F28CDA6DCD1_zpsvlkrcnbj.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/CECDE1AD-CF7F-4D5C-A5FD-B877B53A2EC6_zpsxoctyjay.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/CECDE1AD-CF7F-4D5C-A5FD-B877B53A2EC6_zpsxoctyjay.jpg" border="0" alt=" photo CECDE1AD-CF7F-4D5C-A5FD-B877B53A2EC6_zpsxoctyjay.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/5833401D-6321-4B74-A9BF-9A7C02B1362E_zpskxdkn7kq.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/5833401D-6321-4B74-A9BF-9A7C02B1362E_zpskxdkn7kq.jpg" border="0" alt=" photo 5833401D-6321-4B74-A9BF-9A7C02B1362E_zpskxdkn7kq.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/7E66FDBB-4DFE-4275-9E46-AB3A37BB56DC_zpstevpjg5z.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/7E66FDBB-4DFE-4275-9E46-AB3A37BB56DC_zpstevpjg5z.jpg" border="0" alt=" photo 7E66FDBB-4DFE-4275-9E46-AB3A37BB56DC_zpstevpjg5z.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/4FB639FC-E9B1-4D33-85E5-BCDADDD1F0BC_zpsktcaftux.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/4FB639FC-E9B1-4D33-85E5-BCDADDD1F0BC_zpsktcaftux.jpg" border="0" alt=" photo 4FB639FC-E9B1-4D33-85E5-BCDADDD1F0BC_zpsktcaftux.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/E8F25031-9A62-4FDA-8F0C-BE7AEC2C180D_zpspn7ruepf.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/E8F25031-9A62-4FDA-8F0C-BE7AEC2C180D_zpspn7ruepf.jpg" border="0" alt=" photo E8F25031-9A62-4FDA-8F0C-BE7AEC2C180D_zpspn7ruepf.jpg"/></a>

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/37BD7ED9-4DF1-4EAE-A75F-CAC100657F9B_zpsn39sqnfx.jpg.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/37BD7ED9-4DF1-4EAE-A75F-CAC100657F9B_zpsn39sqnfx.jpg" border="0" alt=" photo 37BD7ED9-4DF1-4EAE-A75F-CAC100657F9B_zpsn39sqnfx.jpg"/></a>
 
Love the data. Going to have to digest it a bit.

A medium speed sampling of the power on V and I curves for the LEDs might show enough data to sort out the failing emitters.


Sent from my iPhone using Tapatalk
 
This is what I think we can use to capture the characteristics of the LEDs

<a href="http://s1062.photobucket.com/user/karimwassef/media/LED%20experiments/1_zpsirm9kinq.png.html" target="_blank"><img src="http://i1062.photobucket.com/albums/t496/karimwassef/LED%20experiments/1_zpsirm9kinq.png" border="0" alt=" photo 1_zpsirm9kinq.png"/></a>

The more data I have, the easier it'll be to spot the pattern for degradation. Right now, the first sign is the nonlinear diode effect creeping up to higher currents. That's the curved part on the left. The second is an increase in voltage relative to current. As the IV curve drifts up to higher voltage, this does not indicate more light - it indicates more heat and an accelerated degradation causing even less light.

What do you think?
 
That should be totally achievable, and can be added to the current LM3409 design. The only downside is current is sensed on the high side presently, which complicates some of the analog circuity (read: adds a dollar or two). Even if its running in switch mode, digital averaging is more than enough.

I'll noodle up some micro/sense circuit options - the MCU is going to be more expensive than the controller chip sadly.
 
You must log in or register to reply here.

Similar threads

MrineLfRlz
Ok to run a calcium reactor and dose kalk. for ph control?
Replies
4
Views
584
griss
griss
M
ESP32 - AQma WiFi LED Control - 8 x LED Channels, 3 x SSR. FOR FREE!
Replies
2
Views
2K
danielch2008
D
B
Budget LED options for 72" deep tank (no SPS plans)
2
Replies
30
Views
3K
wildman926
wildman926
D
Algae issue - and I feel like I've tried everything
2
Replies
21
Views
2K
JCOLE
JCOLE
E
I need a Return Pump Suggestion: COR20 Not Enough
Replies
0
Views
933
EuroDriver
E
Back
Top