# Learning About EMI From My DIY Lightning Detector

Monsoon season just ended here in southern Arizona. During this time of year, brief but intense storms roll through with lots of lightning — one of nature's most spectacular sources of electromagnetic interference (EMI).

A recent storm got me thinking about the EMI challenges we face with precision systems. I wondered if I could actually use EMI to my advantage — to detect lightning during a thunderstorm.

Making an antenna
I needed an antenna to convert the radiated EMI from the lightning bolt into a signal conducted to the input of my amplifier. I made a loop antenna on a PCB (4.65 inches square, eight turns) with an inductance of 20μH tuned to 10.27MHz using a 12pF capacitor.

A simulation from TINA-TI, TI's SPICE-based analog simulation program, shows the output of my antenna circuit. Here it's excited by a pulsed current source (IG1), representing the current induced in the loop from the magnetic field of a lightning bolt.

Figure 1

The response of the tuned antenna circuit (red, bottom) for an input
current pulse (green, top), representing a lightning bolt.

The antenna circuit produces a sinusoid at 10.27MHz. This will serve as an input EMI signal to an amplifier.

Designing the amplifier circuit
The EMI rejection ratio (EMIRR) measures how well an amplifier resists converting EMI into a DC offset. Designers normally select amplifiers with a high EMIRR to avoid EMI problems. In this case, I wanted the exact opposite. The OPA347 does not have an input EMI filter, and it exhibits low EMI rejection at 10MHz (around 13.5dB). With these stats, I knew it would make a great lightning detector.

The SBOA128 application report provides equations to predict the offset resulting from an input EMI signal. For example, we can calculate the change in the OPA347's input-referred offset for a 10mV-pk, 10MHz input signal.

Now, 211μV may not seem significant, but if the amplifier is configured for high gain, EMI might drastically affect the output voltage.

I configured the OPA347 for a gain of 100 using the topology shown in Figure 2. (The design equations for this topology can be found here.) Coupling capacitor C2 forms a high pass filter with the DC biasing resistors R1 and R2 to eliminate mains interference.

Figure 2

Complete lightning detector schematic.

Now for the fun
After receiving a prototype PCB for this circuit (Figure 3), I set up the detector in my backyard and monitored the amplifier output as a thunderstorm approached.

Figure 3

Lightning detector prototype. The PCB includes an oscilloscope trigger
and a one-shot circuit for camera shutter release.

Figure 4 shows the amplifier's output during intense storm activity. The amplifier's output dips as each lightning bolt strikes. The magnitude of this change is proportional to the EMI intensity; lightning discharges that are more intense and closer to the detector produce the biggest dips. A particularly strong bolt caused the output to shift by 728mV (at 0.189s) and was followed by smaller discharges.

Figure 4

Sample data taken during a July 27 storm. The sudden drops
in amplifier output indicate lightning discharges.

In the future, I'll use the board to trigger my camera for lightning photos. EMI rectification can be a major headache for engineers, but sometimes we can use it to our benefit.

Need EMIRR information?
In recent years, we started including an EMIRR curve in the datasheets for our new amplifiers. We're also providing reports for many amplifiers developed before this change. When available, these reports appear on the product page below the datasheet and have “EMI Immunity Performance” in the title.

Figure 5

A screen shot showing where to find the supplementary EMI immunity reports.

If the amplifier you're considering for your design has not been characterized, visit our Precision Amplifiers forum, and ask us to characterize it for you. Here's a curve I took for a customer who asked for help.

Figure 6

The EMIRR curve for the OPA347 was measured upon customer request.

For weekly tips, tricks, and techniques from TI precision analog experts, visit our companion blog, the TI Precision Designs Hub.

John Caldwell is an applications engineer in the Precision Linear group at Texas Instruments, where he supports operational amplifiers and industrial linear devices. He specializes in precision circuit design for sensors, low-noise design and measurement, and electromagnetic interference issues. He received his MSEE and BSEE from Virginia Tech with a research focus on biomedical electronics and instrumentation.

Related posts:

## 27 comments on “Learning About EMI From My DIY Lightning Detector”

1. RedDerek
September 15, 2013

I did a quick search and see there are several detection circuits out on the web; even ones for DSLR cameras. Curious as to why you selected 10.27 MHz for the loop antenna frequency and not go with a simple dipole antenna.

Nice article and I like the description of the operation. Not many lightning storms out where I live, but would build one just to snag the pictures.

2. jcaldwell
September 15, 2013

Thanks for your comment! You are correct, there are multiple lightning detector circuits online that serve the exact same purpose. I was just looking for an entertaining way to approach the topic of EMI Rejection Ratio in op amps, and using an amplifier with poor EMIRR to detect lightning seemed like alternative to the standard “textbook approach” to engineering articles.

I chose a loop antenna for a couple of reasons. First, it could be fabricated in the PCB itself and didn't require me to purchase additional parts. The PCB manufacturer I use has a flat rate up to a certain board area, so the larger board required to accomodate the loop didn't make it more expensive. Second, I'll admit I'm not an RF guy, so for me the loop antenna seemed intuitive; it looked inductive and I could tune it easily with a single capacitor.

10.27 MHz was chosen because I knew for a fact that the OPA347 had terrible EMI rejection at 10 MHz. A customer requested I measure that part on the E2E forum several months prior. 10.27 MHz just happened to be the tuning frequency when using a standard cap value. Most lightning detectors operate at lower frequencies where lightning actually has stronger emissions. In fact, the first iteration of this detector was tuned to 300kHz. However, it is very difficult to find an op amp with poor EMIRR at 300kHz. So a compromise was made.

Snagging pictures with this board has been a challenge and normally results in me getting rained on. The first hurdle to overcome is that most DSLR camera's have some debouncing circuitry on the shutter release. If the detector does not output a pulse for a long enough duration then it won't trigger the shutter release. I used a monostable multivibrator circuit (one-shot) to overcome with this. The second hurdle is that these detectors, by their very nature, prevent thunderstorms! It seems like after I got the detector up and running there was never a cloud in the sky!

3. TheMeasurementBlues
September 17, 2013

What did you use to measure the amplifier's output? I assume some type of oscilloscope. What was the sample rate and banwidth? Were you able to measure the rise time of the lightning pulse or did the risetime/BW of the scope slow down that measurement?

We had some lighting the other day in Boston. Now on the back side, it's much cooler. Feels like Fall tonight.

4. rfec1356
September 18, 2013

People have been using RF emissions to detect lightning for at least 75 years.  In the late 1930s, before radar, there was a system called Spherics (short fo atmospherics)  It employed a pair of crossed loop antennas feeding respective receivers and orhtogonal axes of a CRT.  A diametric line across the screen showed the aximuthal bearing to the lightning strike.  Two such stations could triangulate on a distant storm.

10 MHz is a bit high since most lightning emissions lie in the LF band (below AM broadcast).  I use the low frequency ARC-5 receiver from WWII (BC-543 or R-23)  for lightning detection.  I tune to about 250 KHz.  The set is sensitive, reliable, and being all vacuum tube it is relatively immune to overload in case of a nearby strike.

5. WKetel
September 18, 2013

That is a very impressive photo in the posting, it certainly matches the topic. And while an amp having poor rejection at a frequency is an adequate reason, it certainly appears to me that the circuit is not dependant on the poor rejection for good operation. And with all of that circuit board area you could have had a lot more turns in the loop, especially since it is doublke sided. Probably you could have had it tuned to below 2 MHz, which is a better frequency for detecting lightning, but then 200KHz is a better yet choice. But that would have needed a lot more turns, or adding some ferrite, or lots more capacitance.

But thanks for an interesting article.

6. kvasan
September 23, 2013

Awesome post. Very clear in narration.

I too am not an RF guy and am keen in using a loop antenna as it is easily realisable without much cost.

I also would like to know about the details of the test equipment you have used and curious to know if you had a data logger of some sort.

7. jkvasan
September 23, 2013

Awesome post. Very clear in narration.

I too am not an RF guy and am keen in using a loop antenna as it is easily realisable without much cost.

I also would like to know about the details of the test equipment you have used and curious to know if you had a data logger of some sort.

8. Vishal Prajapati
September 23, 2013

Designing with this much details without being non RF guy is really amazing. And at the end you made it work that is amazing. Very good analysis of the circuit explained. What problems did you face before making this final version? How many iterations had to be done?

September 25, 2013

@rfec1356 – I recall those ARC-5 receivers from when I was a kid. Nice equipment.

That's a nice technique for locating the storms using some technology that was pretty clever for its time.

10. jcaldwell
October 7, 2013

Thanks for your comment! Jayaraman: The data for this article was taken using a digital oscilloscope. One channel of the oscilloscope was connected to a comparator output of the detector which provided the triggering signal for the scope. A second channel of the oscilloscope monitored the output of the amplifier.

Data was taken “manually”. That is to say, my data acquisition system was me on a lawn chair in my backyard, watching the output of the detector on the oscilloscope and saving the waveform data to a USB stick as a thunderstorm approached.

Sure, the proper way to do this would have been to write a data logging program to take the data for me. But sometimes the “brute force” method is adequate.

Vishal: The initial version of this detector was tuned to 300kHz, but I found that it did not give adequate output during lightning storms. Also, it used a bipolar op amp, and the bias current drift and 1/f noise caused the output to be much less stable. Changing to a CMOS amplifier and tuning the circuit to a higher frequency where the op amp had a much lower EMI-RR improved the response. Because testing the board with a thunderstorm every time I change the circuit was not practical, I used a handheld power drill to generate EMI to test the circuit.

One mistake I also made early on was not limiting the bandwidth of the signal from the loop antenna. When I first built the circuit the output was swamped with 60Hz noise, which is not surprising in the least, I just overlooked it while assembling the board.

There were two iterations of the PCB. The second iteration added a monostable multivibrator circuit to trigger the shutter release on my camera. I found that the comparator output from the first board was not held low long enough to trigger the shutter. This is probably due to debounce circuitry on the shutter release.

11. jkvasan
October 15, 2013

@jcaldwell,

That is a concise account of what you did, thanks.

“The second iteration added a monostable multivibrator circuit to trigger the shutter release on my camera. I found that the comparator output from the first board was not held low long enough to trigger the shutter. This is probably due to debounce circuitry on the shutter release. “

A re-triggerable monostable multivibrator could have helped more, perhaps?

12. jcaldwell
October 15, 2013

That's a good point, it would be adventageous to hold the shutter release low during multiple strikes. My current solution was just to set a long exposure time to capture multiple bolts.

13. etnapowers
October 25, 2013

Is the peak value of 100uA that you used to simulate the current induced in the loop from the magnetic field of a lightning bolt, a typical value? what happens modifying this value?

14. jcaldwell
October 25, 2013

100uA was just a value I chose for the simulation. Modifying this value will change the amplitude of the output waveform from the antenna circuit, causing a larger or smaller offset in the output amplifier. This is exactly what happens with lightning that is farther away rather than close to the detector.

15. etnapowers
October 28, 2013

@jcaldwell: thank you for your clarification, so I guess that a simulation of a peak current sweep could be  interesting because it would show the impact of the distance between lightning source and detector on the output waveform from the antenna circuit.

16. etnapowers
November 11, 2013

Does the fall time of the pulsed current, that you used to simulate the current induced in the loop from the magnetic field of a lightning bolt, influence the result of the simulation?

17. etnapowers
November 11, 2013

@Martin:this is a good point, because measuring the fall time of the lighting pulse might give indications on how well the simulation reproduces the current induced from a lightning bolt .

18. etnapowers
November 26, 2013

It's very interesting to check the frequency response of the antenna, by modifying the LC constant the output frequency of the antenna will vary accordingly, so this frequency can be tuned.

19. etnapowers
November 26, 2013

By selecting a good capacitor value to smooth the input current of the antenna, hence the resonance frequency is fixed too , because the value of the inductance of the antenna is selected by the designer. The conditioning system to elaborate the signal has to perform a good response at that frequency.

20. etnapowers
December 4, 2013

@Wketel: I agree with you this is really interesting post, and I think that also a sweep of the resonance frequency, selected by the LC tank, could be interesting to model the behaviour of the antenna.

21. WKetel
December 4, 2013

Lightning detectors have always interested me, and the comment about using an old ARC5 LF receiver was also intersting, since I recently purchased one of them in almost perfect condition. The challenge of using a lower frequency range detector is that the storms produce LOTS of energy in that area and so the detector may be overwelmed, or at least it will be detectingstorms that may be a long way off. So a lower sensitivity detector could be used to advantage, possibly. The narrow bandwidth, whilebeing counter-intuitive, would help to reduce the effect of other noise sources,which there are a lot of them in the LF range. Just pick a quite spot and then wait for a storm to develop and see the signals. Add an optical flash detector and a thunder noise detctor and the system would be complete, and able to provide ranging as well as direction of approach information. That might not be a Planet Analog project, but it would certainly fit in “Nuts and Volts”, except that they would delete all of the analog circuits and stick in an Arduino. That is OK if you want to be a programmer, but it teaches NOTHING about electronic circuits functioning. (please excuse my rant there, but doing everything in digital is seldom the best choice).

22. etnapowers
December 5, 2013

@WKetel: really a nice comment, I appreciate it. Will you make a board or a BLOG about the ARC5 LF receiver program? It should be really interesting to see the results. About the difference between analog and digital designing,I think that a good engineer should have the fundamentals of both of these.

23. WKetel
December 5, 2013

@etna, my point about analog versus digital was that it makes no sense to use a processor board package to do the same thing that could be done with just a compoarator and a transistor. IT often seems like that entire publication is just one big advertisement for the Arduino.

Of course engineers need to not only understand both analog systems and digital systems, they often need to be skilled in using both of them. But in a whole lot of instances it is both cost effective and far more energy eficient to use a few discrete parts instead of a processor controller package.

24. WKetel
December 5, 2013

That ARC5 low frequency storm detection system will take a while for me to organize, since the functionality would be fairly complex. It would actually be a sensor fusion type of project made more interesting by using sensors from different eras. Probably a replacement for that receiver using more current technology would be an interesting project all in itself.

25. etnapowers
December 6, 2013

@WKetel: you're exactly correct, many times is preferred to adopt a semiautomatic system which needs only a few parameter  to be setted instead of a complex system which requires a deeper knowledge of electronics design . I think that a good engineer should have both digital and analog design skills.

26. etnapowers
December 6, 2013

@WKetel: yes it's a complex blog but I think it's really interesting, and could lead to very interesting results, thank you for your effort.

27. WKetel
December 8, 2013

Here are some thoughts relative to a “hobby level” lightning detection system using what some woulod call sensor fusion. The challenge winds up being in doing the task using inexpensive materials. The easy part is the RF component detection portion, which would utilize an old ARC5 lf receiver tuned to a quite spot someplace near 300Khz. A simple diode detector connected to the audio output, but using perhaps two diodes would primarily respond to bursts of noise, which is the classic lightning signature, as the oscillations decay along the connecting plasma trail. Multiple diodes will assure that it does not respond to lower amplitude disturbances. Determining the direction of the flash bu optical means is a bit more complex becausenot much light is available from lightning several miles away. That sort of rules out the in3expensive detectors without lenses. The choice would be to have eight segments with a sensor for each one. While detection inthe dark can be done, it is a real challenge to detect the small difference in intensity during daylight. The sensors are available but they are not cheap. Of course since only the vary fast risetime of lightning does enable the use of rate-of-change circuitry to assist in deciding what is lightning and what is not. In addition the flash happens at thge same time the radio noise burst is detected, so that correlation should be useful. The last part is the thunder detection, which, given the reverberation of much thunder it would have to detect the first sound to arrive. That should not be extremely difficult. The expensive part would be microphones that are directional at the low audio frequencies and at the same time rugged enough to not be degraded by a rainstorm. But the parts wind up bringing the price up above the hobby experimenters level, at least I think that they do.

This site uses Akismet to reduce spam. Learn how your comment data is processed.