Nature of the problem
Recently, while upgrading my solar garden shed PV installation a problem popped up with DC loads connected to the new Renogy MPPT charge controller. The original system had a single 10W 120 VAC LED light fixture in the center of the shed .I added two more fixtures at either end of the shed and switched the wiring over to 12 VDC supply. All three lights are wired in parallel and controlled by a standard light switch. The Renogy MPPT controller can supply up to 20 A at 12 VDC through its Load terminals. The two new light fixtures are pull string types that can be turned on as needed. Each new fixture was tested in isolation. The Tento 850 lumen bulbs drew about 880 mA each. A problem arose when I tried to turn on all three lights at the same time. The lights did not illuminate. Powering three 10 W LED lights wired in parallel tripped an error condition in the charge controller which disconnected power to the load terminals. The load can be reconnected by pressing a button on the charge controller, but I'd rather not have to do that every time I turn on the interior lights. Three bulbs in parallel only draw about 2.4 A, which is well below the 20 A limit on the load terminals, so I surmised that maybe I had some form of in-rush current problem. A single bulb being turned on didn't trip the load output, but two, or three in parallel did.
My plan to solve the problem was to install In-rush current limiter (ICL) devices in series with the two extra LED lights. Specifically I planned to install Negative Temperature Coefficient (NTC) thermistors. The resistance of an NTC thermistor decreases as its temperature increases. As current flows through the thermistor, heat is generated. This heat decreases the devices resistance, allowing more current to flow. The idea is that at ambient temperatures the NTC ICL device will present enough resistance in series with the bulb that when it it is energized the initial current will be kept below the trip threshold in the charge controller. As current flows, the NTC ICL device rapidly heats up and its resistance decreases, eventually reaching an equilibrium at a much lower resistance, I decided to go with Amphenol CL-120 NTC ICL thermistors. These devices have a nominal resistance of 10 Ohms at 25 C and can handle steady state currents of 1.7A at 25 C. Current handling capability decreases as temperature increases according to the following equation provided in the Amphenol data sheet:
I estimated that on the hottest summer day the temperature in the rafters where the light fixtures are located might reach +40C. The derated current calculates to be about 1.25A, so I felt the device would work on the hottest days. At the other end of the year when ambient temperatures might hit -30 C the problem will be too much initial resistance. The effect I expect in extreme cold will be low brightness when the lights are turned on. They should then brighten rapidly as the ICL devices heat up. I installed the ICL devices in series with the two new fixtures and the problem disappeared. All three lights can now be turned on simultaneously without tripping the cut off circuit in the charge controller.
The measurement challenge
I wanted to know what the current flow looked like with and without the ICL device in series with the supply and I wanted to know the steady state temperature of the ICL device when the lights were in use. There isn't a strong technical reason to perform these measurements because installing the ICL devices seems to have solved the problem, but I am in the middle of reviewing the Keysight . Measuring in-rush current seemed like a cool application to try, so here we are. Now, the has a built-in In-rush current measurement feature available through the Analyze button. However, when I investigated this feature it turned out to be designed to measure in-rush currents going into power supplies. It also required the use of a differential probe and a current probe, neither of which do I own. The set up screen for the in-rush test is shown below.
I would have to figure out my own test set up in this situation. I decided to go with a small sampling resistor in series with the power connection to the LED bulb. The lowest value resistor I had available was a 0.01 Ohm 1W current sense resistor. This would be great because its low value would barely impact the circuit, however the voltage drop across a 10 mOhm resistor at the expected current levels is very small. The problem with very small amplitudes is noise. I gave the 10 mOhm resistor a quick try and found the noise was at least double the size of the signal I was interested in. The next smallest value I had available was a 1 ohm resistor in a resistance decade box. Not ideal because 1 Ohm will likely impact overall circuit behavior, but still small enough to get useful information. My test set up using the decade box is illustrated below.
The power supply is a Keysight E36313A. The ICL device is mounted to the breadboard. The orange instrument laying beside the breadboard is a Keysight Infrared thermal imaging camera situated to monitor the temperature of the ICL device. The decade box, set to provide 1 Ohm of resistance, is below the breadboard with a scope probe measuring voltage on either end of the resistor (relative to the supply negative terminal). The LED lamp is glowing brightly on the left side. To display and compare the two in-rush currents I used the Math and Ref features on the . The math menu allowed me to set up a differential calculation that would display Ch2 - Ch1. This signal is the voltage appearing across the 1 Ohm sensing resistor. Because I = V/R, this signal will provide a proxy measure of the current flowing through the circuit.
My first measurement with 1 Ohm sampling was with no ICL device in the circuit. The result was disappointing. Even thought the sensed voltage was 100 X the voltage that appeared across the 0.01 Ohm resistor there was still a great deal of noise in the captured waveform. See the screen capture below.
At slower sweep speeds (in this case 10 ms/div), the High Resolution acquisition mode provided by Keysight averages in more samples to smooth out noise and increase effective vertical resolution to 12 bits. The same signal captured in High resolution mode looks like this:
Much better. Now, this waveform will be saved to Ref 1 and a new acquisition will be obtained with the ICL device in the circuit. The resulting traces are shown below.
The ICL device makes a significant difference in the initial current draw. I also notice that the current waveforms don't look like typical in-rush current waveforms seen on power supplies. Normally in-rush current spikes then falls to a lower steady state value. In this case there is a spike, then a dip, then an exponential rise to steady state. The insertion of an ICL device makes the in rush current waveform look more typical. Though it should be noted that as the ICL device heats up, current slowly rises to a final steady state value. These measurements focus on the initial turn on behavior as that behavior seemed to be the cause of the breaker trip in the charge controller.
The other thing I notice is that the initial slew rates are different between ICL and no-ICL device. The just happens to have a built-in Slew Rate measurement feature. Let's give that a try. After adding a slew rate measurement for the Reference waveform (no ICL device) and the math waveform (with ICL device) I did a "wait, what?" when I saw the displayed values. See below.
Visually it is clear that the no-ICL device current has a steeper (i.e. higher) initial slew rate than the current with the ICL device, yet the oscilloscope measurements indicate the opposite. The no-ICL device current is measured with a slew rate of 30.764 V/s whereas the current with the ICL device is measured with a slew rate of 58.255 V/s. Turns out this is another case of checking assumptions before assigning blame. The cursors indicate the points used by the 'scope to take values for the slew rate calculation on the no-ICL waveform. The cursors were automatically placed at 10% and 90% points on the overall trace based on steady state conditions. Given this, then yes, the slew rate to steady state is slower on the no-ICL trace. So, how do I make the 'scope measure the slew rates on the initial slopes? The solution is as simple as turning on the zoomed time base and selecting a span that only covers the duration of the initial turn on slopes. As long as the Measurement Window setting is on Auto Select, the 'scope will remeasure the slew rates on the zoomed window. Now the numbers look like this:
Under these conditions the no-ICL device slew rate is 117.55 V/s and the slew rate with an ICL device is 49.025 V/s. I wonder if the reduced slew rate is responsible for preventing the load output from tripping?
One more thing
The ICL device, as mentioned, heats up as its resistance decreases. How hot does it get under these conditions? The Keysight IR camera pointing at the ICL device captured the image below 200 s after current started flowing through the ICL device.
The relevant reading in this image is the Max temperature detected near the center of the ICL device. +104.6 C. That is pretty warm. I will have to investigate methods of dissipating that heat. Later.
Once again the has delivered the measurements I needed. There were a few missteps on my part as I learned how to operate various features, but in the end I was able to get the insights I needed.