I am performing an imaging experiment in which LED stability is critical. I use 2-3 red LEDs to illuminate a surface while imaging that surface. In each LED circuit, I use a constant-voltage laboratory power supply (typically 11 V / 0.05-0.10 A) and a buckpuck LED driver to achieve constant current. The buckpuck's output current is adjustable using a 5k potentiometer between the REF and CTRL pins (higher resistance = higher current output = brighter LED). The circuit diagram is illustrated below.

The LED itself is attached to a heat sink to avoid overheating. Despite this, I am still having issues with LED instability. Does anybody have any suggestions to make the LED brightness more stable?

Here are the specific components I am using:

  • Power Supply: TENMA 72-7245 Dual Output Power Supply
  • LED Driver: LuxDrive 1000mA Buckpuck 3021-D-E-1000
  • LED: Osram LRW5SN Platinum Dragon Red 625nm LED
  • Potentiometer: Bourns 3590 Precision Potentiometer, 5k 2W 10-turn linear wirewound

Any help is much appreciated! Thank you.

LED Circuit Diagram

EDIT: Thank you for the feedback so far. Ideally, I would like to avoid large changes to the circuit design if at all possible. Given that, would the following things help reduce high-frequency noise at all?

  1. Grounding. I'm not really sure how to properly ground this circuit, or if it needs it.
  2. Decoupling capacitors. Could a capacitor between Vin +/-, LED +/-, or CTRL/REF reduce high frequency fluctuations?
  • 3
    \$\begingroup\$ I've written about my experiences many times on this. I spent time trying to create "standard candles" out of LEDs. Used expensive, NIST-traceable current sources accurate to 0.05% and maintained LED die temperatures at about 75 C while monitoring their output for 48 hours. Only about 1% of LEDs from the same wafer were irradiance-stable. The rest "flickered" the entire time and we just threw them away. Even with that, they weren't the same as each other. So they still need to have different currents calibrated for them to get the same output. You have all my sympathies. \$\endgroup\$
    – jonk
    May 17, 2019 at 19:35
  • \$\begingroup\$ You could try caps at the input to the buckpuck -- but I think the suggestions about using an LDO or the lab supply should be heeded. twisting your power leads together may help, particularly if it's an area that has high radiated emissions from something. \$\endgroup\$
    – TimWescott
    May 17, 2019 at 21:54
  • \$\begingroup\$ Thanks Jonk, and thanks Tim! I'm looking into an LDO long-term if the simpler solutions don't pan out. I will try the capacitor between supply leads, and I will also try twisting my wires. Thanks again! \$\endgroup\$
    – nckcard
    May 17, 2019 at 21:58
  • 1
    \$\begingroup\$ @nckcard We experimented to see if the idea was viable. A hope was to achieve a long-term drift uncertainty of 1% for some "useful life," once calibrated, before another calibration was required. (I haven't mentioned some other practical problems because they haven't come up here, yet.) 48 hours appeared to be a sufficient bake-in period to separate those that were promising from those that were hopeless (in the sense that perhaps they might stabilize in a week or two, but we couldn't afford the time/costs to find out if more would settle down.) I wish you better success with modern LEDs. \$\endgroup\$
    – jonk
    May 18, 2019 at 2:37
  • \$\begingroup\$ To avoid "high frequency fluctuations" you should use power source with low ripples (low frequency fluctuations), and potentially a linear regulator instead of various light fixtures. \$\endgroup\$ May 18, 2019 at 5:49

3 Answers 3


Ditch the buckpuck and run the LED off of the constant current mode of the LAB supply. If it is more steady then it's probably the buckpuck.

Another thing to do would be to check the voltage supplied to the buckpuck, some of these cheap lab supplies dont have a very steady output.

If you really need stability an LDO is probably the way to go in constant current mode (except the load would be the LED, and you need to deal with thermal issues of the LDO):

enter image description here
High Side Constant Current Source

Here are some other cool ideas for stabilty:

  • \$\begingroup\$ Thanks, I will try ditching the buckpuck. We tried this once before and saw even more instability, but I am beginning to think that grounding may be the true culprit in both scenarios. Frankly, I am at a bit of a loss for how to properly ground the circuit. Also - do you think adding any capacitors would help weed out high frequency noise? \$\endgroup\$
    – nckcard
    May 17, 2019 at 20:30
  • \$\begingroup\$ Caps may affect stability worse or better. \$\endgroup\$ May 17, 2019 at 20:47
  • \$\begingroup\$ @nckcard Yeah, but it could be the switching that is affecting the stability also. A cap gives another current pathway that could be dissipating more or less power. I would think that not switching at all would be better than switching. If you are going to use a constant current swticher and you need stability, you might want to find one that has noise an stability requirements you need. Linear power products generally have better stability, because most of their products operate from 125C to -40C \$\endgroup\$
    – Voltage Spike
    May 17, 2019 at 20:59

You may need a pulsed current burn-in period to burn off electrode interface impurities and then optical intensity feedback such as used in lasers. Stability is defined over different time intervals. What are you specs? What were your results?

I assume you have checked ripple current if this is your problem, use an LDO for a CC source, allowing for LED thermal stability.

Even a fixed power series resistor with CV on your power supply would be better. to drop 1V. That will stabilize in a minute with temperature.

Efficiency also depends on temperature.


The LEDs you are using are about 10 years old.

At 1000 mA your LED's forward voltage is 2.7 V and generating 2.7 watts. What is the temperature of the LED's thermal pad?
Can you hold your finger on it for an extended period of time?

Being connected to a heatsink may not be sufficient.
You need substantial thermal management at 1 A.
Red LEDs luminosity is very sensitive to heat.
You can lose up to 70% flux due to heat.

enter image description here

In 2008, when these LEDs were developed, the LEDs manufactured on the same wafer had a large differences in their performance characteristics.
Different luminous output.
Different relative Luminous Flux (both current and temperature.
So two, especially red, LEDs with same current will very likely emit amounts of flux.

I'm not sure by what you refer to as "stable". I really need a better explanation than just "instability".
The human eye cannot easily detect a difference in luminous intensity very well so I doubt that is the issue.
If do you mean the flux of a single LED fluctuates or if the difference in flux between two LEDs varies too much.

I suspect the temperature is too high. You are pushing the LEDs to their limits if you run at 1000 mA. The lab supply may not have enough power.

I would suggest you run the buckpuck where you can hold your finger on the LED without getting burned.

If you need the luminous intensities to match, then you may need a separate current source for each LED.

looking into an LDO long-term if the simpler solutions don't pan out. I will try the capacitor between supply leads, and I will also try twisting my wires

I doubt caps (or twisting wires) will help a constant current source.
The Buckpuck is a fairly decent (not cheap) CC supply and doubt it is causing the instability. If find the lab supply specs a little troubling: 11 V / 0.05-0.10 A

Try powering the buckpuck with a battery.

The LDO is a good approach. An adjustable LDO, one for each LED.
If you set the LDO's max output at 2.8 V, then you do not need a resistor.
You can adjust each LDO's output for the desired luminous intensity.

Or you could use a 3.3 V LDO for all the LEDs and use a different resistor for each LED.

  • \$\begingroup\$ Hi Misunderstood, thank you for your input! I am looking into LDO options now. One good option seems to be illustrated here with the LT3085: allaboutcircuits.com/technical-articles/… Does this solution seem reasonable? I am of course a novice when it comes to LDO's, though it appears that I could set the output voltage to ~2.8V here based on the R1 SET value, and then adjust the constant current source based on the R2 OUT value. Thanks! \$\endgroup\$
    – nckcard
    Jun 6, 2019 at 16:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.