Dear Asadi,

Now I think I got what do you want to achieve.

You want to measure the TC of the solar panel or any PV array.

You defined the TC = FFT(I)/FFT(V)

So, you want to get the admittance spectroscopy of the two terminal network under certain load and illumination condition.

It is clear to me that you do not yet develop such admittance spectroscopy y(f)

So, as I was a pioneer for the admittance spectroscopy of the solar cells i want to stress some facts:

y(f) depends on the operating point of the array. As the operating voltage increases the admittance will be larger and vice verse. So, the admittance curve is not the same at all operating conditions of the solar panel.

As the panel is nonlinear such that I versus V is not linear, then one has to speak about small signal admittance.

Then one can conclude in order to use this method in diagnosing the panel, one has to keep constant the operating conditions of the panel such as the load and or the terminal voltage and the illumination intensity.

Now we come to the important question how we measure y(f)?

We apply a small signal ac voltage on the panel and measure the the drawn AC current in magnitude and phase.

The ac voltage source is coupled to the loaded and illuminated solar panel by a coupling capacitor. The ac current is sensed by either a sense resistance or by a hall probe. The use of sensing resistance is the most common method.

I made such measurements for the small signal characterization of large area solar cells. I developed a method to measure the ac admittance.

Really you can use them again for diagnosis of solar panels.

I would like that you follow the paper in the link: Article A distributed SPICE-model of a solar cell

You can use the measuring method brought in the paper or use a more advanced method for admittance measurement by following the paper:Article Four Voltmeter Vector Impedance Meter Based on Virtual Instrumentation

Best wishes

Hi Meysam Asadi ,

at 10 MHz, you should have no problems with most common resistors. But I guess that "suddenly" means that the switching generates harmonics well above 10 MHz.

In the datasheets of most resistors, the behavior at RF and microwave frequencies is not documented. But there are resistors especially developed for high frequency applications. For example, the resistors described in the first attached file are well-behaved up to 6 GHz.

The second file describes the high frequency properties of another resistor family, suitable for lower power dissipation. As you can see, at high frequencies the impedance (capacitive component vs inductive component) depends strongly on the nominal value of the resistor. In most cases, there is an optimal value, but that's certainly above the 5 Ohm in your application.

Generally, it's possible to compensate for the deviation of the impedance by adding capacitors and/or inductors. However, because these additional elements deviate from the ideal behavior, too, it's preferable in most cases to choose suitable resistors.

Of course, the rest of the circuit (switching transistors, PCB tracks, etc.) must be suitable for high speed swiching, too.

Thanks,@ Joerg Fricke I have another question about resistance.

By changing the resistance from 5 ohms to 10 ohms, my voltage and current will also change, and I sent you photos of them below.

Then I take the measured voltage and current to the frequency domain and calculate their FFT.

Then I calculate their Transfer function according to the following formula:

TF=Abs (FFT(I)/FFT(V))

I also sent you the image of the Transfer function of my voltage and current.

My problem is that every time I measure the voltage and current by changing the resistance and calculating their Transfer function, in each test their Transfer function does not overlap.

While the Transfer function must always be the same.I will also send you a picture of the electrical circuit by which I suddenly change the resistors.(last picture)

please help me.

Meysam Asadi

While you've given away your resistance values, you did not give us the power rating required :(

When it comes to "high frequency" (OK - 10 MHz isn't considered high frequency these days any more), low-inductance resistors are uses. Which precludes wire-wound resistors. Best would be "carbon composition resistors" (basically a piece of carbon mixed with some binder), but these are rarely available and might exhibit tolerances in the 5 .. 10 %. Beyond that, carbon film or metal (oxide) film resistors are available with tighter tolerances and still reasonable power ratings. These may have somewhat higher parasitic inductance and capacitance values but might still apply.

I don't understand the voltage diagram.

What kind of voltage source do you use?

Could you please provide circuit schematics wherein the green and yellow wires are represented, and the points for voltage measurement are marked?

The current diagram shows ringing, probably caused by the wires. Could you please describe your overall goal, e. g. is this just a project for testing the properties of the resistors, or is it a part of a larger project?

@ U. Dreher

Thanks.

I sent you the power specifications

Meysam Asadi

OK - that's funny: You're stating that Voc = 60 V and Isc = 0.99 A. Which gives an output resistance of about 60 Ohms for your (whatever) device.

Considering this, it does hardly matter whether you're attaching 10 or 20 Ohms: both are next to a short-circuit, which is the reason why the currents measured do not really differ. It's just the residual voltage that differs.

If you want to measure anything substantial, your load resistances should be significantly larger - maybe in the 30 Ohms (absolute minimum) range...

P.S.: The spikes (and ringing) you see might come from the manual switch bouncing. And/or from some other parasitic effects - your wiring, some capacitance(s) on the driving device etc.

Joerg Fricke

I use dc source

I have already written articles about this project.

https://ieeexplore.ieee.org/document/9007139/

https://ieeexplore.ieee.org/document/8893609

This project has already been done in the form of simulation and is now being done in the laboratory to collect experimental data.

U. Dreher

Thankful

In the last answer, I fully explained my project and showed my circuit schematically. For some reason we have to use 5 and 10 ohm resistors. Given these circumstances, do you have another suggestion for us?

Meysam Asadi Since I had to pay $ 14,95 each for the IEEE articles, could you please send me copies attached to a personal message? I promise not to sell copies.

Meysam Asadi

Oh - a PV application ?!

You' may well ignore the transients as the PV panels have "their own life" (aka "parasitics"). Exempt if you want to characterize exactly these parasitics.

(Their step response is somewhat slow and they incorporate a lot of capacitance. Plus the inductance from the long traces/wiring. So these oscillations are intrinsic.)

But loading these panels with 5 and 10 Ohms when their Isc is just 1 Amp doesn't make sense at all. You'd better just use a fraction of the panels.

(If the photo is correct, you currently have 21 panels, making it around 3 Voc per panel. So even 2 panels in series would be at or beyond overload when loaded with the 5 Ohm resistance.)

U. Dreher

The panels were connected in series (3) and in parallel (7).

U. Dreher

series (3) and in parallel (7).

Joerg Fricke

Yes, Sure

I sent you the articles

OK - so these are 20 Voc/0.14 A Isc each. Making a source resistance of 143 Ohms per panel.

Paralleling 14 of these would bring the source resistance down to 10 Ohms, maybe rendering the configuration useful.

(You didn't tell us what exactly you want to achieve/measure.)

You got answer from two respected colleagues.

You have to tell us what do you want to test and then we may tell you how do you measure it. There are static characteristics of the panels and there is the dynamic performance.

It is so that you must operate the solar cells at their maximum power point.

So the load required for such condition is RL= Vm/Im

The maximum power point is given in the datasheet.

Then if you want to study the effect of the switching load suddenly then you have to consider the following:

- Using high frequency resistance such as metal resistance. You can also use carbon resistance. They are more noisy than the metal resistance but in your applications it is irrelevant.

- Use fast switches to change the load. much faster than the dynamics of the panels.

So, you can use power mosfets with strong drivers for high switching speed.

Best wishes

Abdelhalim abdelnaby Zekry

Thankful

We want to calculate the Transfer function of solar panels.

TF=Abs (FFT(I)/FFT(V))

The best solution for calculating the Transfer function is to use a sweep frequency response analyser.

Unfortunately, we do not have access to this device. Therefore, we must use transient state techniques. By changing the resistance, we create a surge voltage in the system and measure the current. In a healthy state, the Transfer function must be the same when there are no faults in the system. By creating a fault in solar PV, the Transfer function of the system also changes, which can be used to detect the fault.

Our problem is that in a healthy state, our Transfer functions are different and do not overlap with each measurement.

I review some results I saw in my experiments:

By changing the solar panels and using a larger, more integrated solar panel (one solar panel 100 W), the new Transfer functions measured did not overlap in each test.

Instead of panels, an ideal DC source was used, the results being the same as before, and its Transfer functions did not overlap. (That is, we tested the resistance change on a DC source).

Therefore, we found that by changing the system (small solar panel, large solar panel and DC source), the Transfer functions do not overlap in any case, so the problem is the resistance change circuit.

To create the surge voltage in the system, we have used one circuit, which consists of resistors and a Thumb knob. I guess the problem is either with the resistors or the Thumb key.

PS: Initially, I used brick resistors instead of carbon resistors, but found that as the brick resistors warmed up, their Transfer functions lost their overlap. I sent you the results.

That's why I guessed that the problem must be the resistance. Because we divide the current by the voltage and consider a constant function, the switch should not have much effect on our results.

Figure 1. We prevented the brick resistors from heating up

Figure 2. The brick resistors became very hot.

In this experiment, we used the DC source instead of the solar panels and changed the resistances to an ideal DC source.

Dear Asadi,

Now I think I got what do you want to achieve.

You want to measure the TC of the solar panel or any PV array.

You defined the TC = FFT(I)/FFT(V)

So, you want to get the admittance spectroscopy of the two terminal network under certain load and illumination condition.

It is clear to me that you do not yet develop such admittance spectroscopy y(f)

So, as I was a pioneer for the admittance spectroscopy of the solar cells i want to stress some facts:

y(f) depends on the operating point of the array. As the operating voltage increases the admittance will be larger and vice verse. So, the admittance curve is not the same at all operating conditions of the solar panel.

As the panel is nonlinear such that I versus V is not linear, then one has to speak about small signal admittance.

Then one can conclude in order to use this method in diagnosing the panel, one has to keep constant the operating conditions of the panel such as the load and or the terminal voltage and the illumination intensity.

Now we come to the important question how we measure y(f)?

We apply a small signal ac voltage on the panel and measure the the drawn AC current in magnitude and phase.

The ac voltage source is coupled to the loaded and illuminated solar panel by a coupling capacitor. The ac current is sensed by either a sense resistance or by a hall probe. The use of sensing resistance is the most common method.

I made such measurements for the small signal characterization of large area solar cells. I developed a method to measure the ac admittance.

Really you can use them again for diagnosis of solar panels.

I would like that you follow the paper in the link: Article A distributed SPICE-model of a solar cell

You can use the measuring method brought in the paper or use a more advanced method for admittance measurement by following the paper:Article Four Voltmeter Vector Impedance Meter Based on Virtual Instrumentation

Best wishes

Meysam Asadi

From your articles, I gather that one of your basic ideas is that a defect is more easily detected in a spectrum (0 .. 15 MHz) than in the results of applying a single frequency. Using a (simulated) sweep generator yielded useful results. (The amplitude of the generator signal would be interesting but is not given in the articles.) BTW, according to the simulation diagram in Fig. 2, the solar panels were in open circuit mode during this simulation.

Now - if I understand correctly - you are trying to replace the simulated frequency sweep by the generation of harmonics which should cover at least the range 0 .. 15 MHz by applying load steps (non-ideal, of course). Currently, you achieve this by manually switching a part of the load on or off.

I think manually switching is not an option in this case: FFT presupposes a periodical signal. Such a signal cannot be generated manually, of course. So, you have to mask your signals by a window function which generates artefacts; especially, any jitter of the switching event relative to the window strongly affects the result.

In order to pursue your approach further, I would recommend switching a part of the load by a transistor (power MOSFET) with rise time and fall time at or below 20 ns, controlled by a stable square wave generator. Such a generator is easily implemented using a microcontroller, clocked by a quartz stabilized oscillator. (E. g. ATmega328P (Arduino Nano), clocked by 16 MHz resp. a suitable frequency, please see below, will do.)

The cutouts of the spectra in one of your articles seem to indicate that the sweep generator increased the frequency in 5 kHz steps. So, the fundamental frequency of the square wave generator could be 5 kHz (or 2.5 kHz since the even harmonics might be considerably weaker than the odd ones). 15 MHz / 5 kHz = 3000 (resp. 6000 for 2.5 kHz). Consequently, you'll need a 6000 (12000) point FFT. If the number of points should be a power of 2, 8192 (16384) points would be right. An important point is to apply the FFT to a period of 200 (400) microseconds (= the period of the square wave signal) as exactly as possible.

To put the dependencies the other way around, if your data aquisition system has a sample rate of 50 MS/s = 20 ns distance between successive samples, and if you apply a 16384 point FFT = 327.68 mys duration, then the frequency of the square wave should be 1 / (327.68 * 10-6) Hz = 3051.76 Hz. In this case, the period of the clock of the microcontroller should be 2n * 20 ns, for example n = 2 => T = 80 ns => f = 12.5 MHz; a 16-bit timer of the microcontroller has to divide this clock frequency by 4096.

One basic difference to the frequency sweep is that the amplitude (and power) of the harmonics depends on frequency while the amplitude of the sweeping signal is constant. If this reduces the recognizability of defects, you might apply digital high pass filtering before the FFT resp. some kind of frequency dependent zooming of the displayed spectrum.

In any case, you should use resistors with inductance as low as possible in order not to mask effects caused by a change of inductance inside the array of panels.

That is actually rather moderate frequency range ( 10MHz at 5..10 ohm), usual SMD resistor like 2512 or 1206 should be ok, however I would stay away from wirewound resistors (which many high power through resistor are at such low values) because of their high inductance.

One can see some low inductance current sense shunt techniques e.g. here: https://core.ac.uk/download/pdf/54204708.pdf

"Brick resistor" is no usual term, so I have to guess a bit: considering the power to be dissipated (some Watts), they might have used cemented resistors - - - which are wirewound. And the nice, gold-colored Dale types are wirewound as well.

When considering SMT resistors, those won't build small:

SMT resistors with 1 W and beyond start at size 2010 and end at size 2512 - not much choice. And don't ask for prices :)

Gold band resistors are the best for circuitry.

Use Mos as a resistor

metal film resistor is good for use in high-frequency applications because of its temperature stability.

Hello Meysam,

The issue on low impedance resistors for high frequency is that they get inductive and therefore increase their impedance. For a frequency up to 10MHz you should be fine with reversed shape version resistors like 0612, 1020 or 1225 e.g. https://www.we-online.de/katalog/en/WRIS-KWKB/.

If you need it very stable over a wie frequency range go for anti pulse resistors. The reason is related to its trimming. Standard resistors trimmed with meander, serpentine or plunge cuts. The best RF performance is reached with an edge or scan cut.

However, resistor manufacturers don’t tell you what kind of trimming they use (except on very expensive RF resistors). Only one manufacturer told me is using mainly scan-cuts which is Susumu.

A Vishay technician told me a another thing. All anti-pulse and anti-surge resistors facing the same issue. Any “inductive” trimming will kill the resistor on a surge pulse. Therefore, they are all trimmed low inductive meaning scan or edge cut.

That is why anti-pulse or anti-surge resistors (pulse resistant) are good for RF.

By the way a good tool to measure resistor impedance in your frequency range is Bode 100.

I hope this helps.

Guenther

One can find some reduced inductance, high precision shunt resistor at very low price here:

10Ω 1225 Thick Film High Power Surface Mount Fixed Resistor ±1% 1.33W - ERJA1AF100U

• RS Stock No. 572-873
• Mfr. Part No. ERJA1AF100U
• Manufacturer Panasonic
• https://ie.rs-online.com/web/p/surface-mount-fixed-resistors/0572873/

Hi Nikolay,

Thank you for the hint. Even as it looks like a resistor array it is a single resistor.

Interresting thing.

Guenther

The Carbon Composite Resistor is a low to medium type power resistor which has a low inductance making them ideal for high frequency applications but they can also suffer from noise and stability when hot.