Determining lightning polarity (positive or negative) by raw signal processing?
#1
Question 
Hello, my name is Oana Liviu from Romania and I am currently operating the station number: 951 from Resita, Romania (system RED).

It is well known that if you use two perpendicular antennas, if you position them to N-S/W-E and you wire them correctly, you can get an estimate of the direction of the strike by processing the raw signal data (see attached image). This can be acheived by calculating the arctangent of the ratio between the signals coming from each antenna among some other calculations.

Let's say that one is able to determine a relatively correct direction of the signal.

My question is: if the calculated direction by the means posted above is somewhat the opposite of the real direction (the strike is located to the south-west, but the calculation yields a north-east direction), that means that that strike have a positive polarity? Considering the calculations have been correctly done?

I am considering some experiments based on this and I want to know if I'm heading in the right "direction".

Thanks Wink


Attached Files Thumbnail(s)
   
Stations: 951
Reply
#2
If you know the right direction, the polarity tells you whether it is cloud-ground, or ground-cloud.

An E-field antenna tells you the polarity, but not the direction.

/Richo
Stations: 584, 585, 2017
Reply
#3
Thanks for the reply!

So basically by processing the direction component of the signals, one can tell if the lightning originates from the ground, but only an E-field antenna could detect the positively-charged lightnings that occures between an anvil and the ground?

Best regards,

Liviu.
Stations: 951
Reply
#4
Richo,
I don't quite agree with you: cloud-ground lightning, CG for short, comes in 2 polarities: about 90% of the strokes carry negative charge to earth, they are -CG lightning, the remaining 10% carry positive charge to earth, they are +CG. So the polarity has nothing to do with the difference between lightning in or between the clouds (intra- or inter cloud ligthning) and CG lightning. Cloud lightning produces different signals from CG lighting, they are more complicated and usually weaker. All lightning detection systems detect almost exclusively CG lightning with high efficiency; cloud lightning is sometimes detected but with low efficiency. 
The best way to determine the polarity of CG lightning is to use the E field antenna. I use the H antennas  for triggering and the E field antenna only for the signal; the trigger treshold is set so high that E field noise only rarely triggers.
Stations: 1836
Reply
#5
When I write "cloud-ground, or ground-cloud." I'm talking about main discharge between cloud and ground.
...and YES...lightning physics is quite complicated - I still do not understand much :-)

/Richo
Stations: 584, 585, 2017
Reply
#6
Yes, lightning physics is complicated and still an active field of research... I was surprised to learn that it is still not clear how lighting discharge starts at all.
For polarity detection the E field recording is needed, but many stations are not using the E field antenna, maybe because it generates to much false triggers, but it would be good if they could give it a try.
Stations: 1836
Reply
#7
For example, LINET system uses loop antennas and can detect all types (and other data) of discharges: CG, IC, polarity and lightning current values.
Below antenna for LINET system.


Attached Files Thumbnail(s)
   
Reply
#8
Kriu,
you are right: they use only the loop antennas. The webpage of the company is: https://www.nowcast.de/.
The high accuracy of their detection (they claim 100 m to 75 m) allows them also to check for the height of the discharge, but I am not sure what their efficiency for intra- or inter cloud lightning really is - what do they use as reference??
Anyway, this appears to be the best current system for lightning detection.
Stations: 1836
Reply
#9
Thanks for replies !

I will give a shot with e-field antenna, trying to mount it ASAP Wink
On the other side, I think my proposed approach can detect the polarity, even though with a smaller detection ratio, because data for both of the channels must be sent to server. I can make even a comparision.
I will post some results here,

Best regards,

Liviu.
Stations: 951
Reply
#10
Determining the polarity is easy in theory. What we need are a couple of stations where we now the parameters.

Here's that what you can do:

1. Select a stroke which is not too far and not too close (i.e. 100 - 300km) to your station. Too close one may cause some interference, too far might be a reflection.
a) The signal shape should be some sort of a single spike and not too weak. Don't use signals of GREEN stations.
b) The accuracy of the stroke should be good, let's say < 1km.
c) The signal should include an e-field channel and two horizontal H-field channels.
d) Check other stations signal of the stroke which have similar distances. They should be very similar shape.
d) Save the signal!

2. Search for the stroke in another archive like
a) Germany: https://kachelmannwetter.com/de/blitze
b) USA: https://weather.us/lightning
c) For other regions you can check the latest 60 minutes here https://weather.us/lightning/americas
--> Double check whether the selected stroke is really the one of the signal.

3. Try to match these parameters with your signals:
a) The strokes polarity should always match the polarity of the spike in the E-field channel (either always equal, or always the opposite). If you also consider the directions, you can check the H-field antennas as well (see first post).
b) The current should be propotional of the spikes amplitude multiplied with gain on equal distances from stroke to station.
c) The type (cloud to ground or intra cloud) might have some influence on the shape of the signal.

4. Post your results here.

After we have some results, we can implement algorithms to compute those parameters. Then signals of other stations can be used as well, as their antenna parameters can be calculated from the calibrated stations.

It is also possible to recompute past events again and again to improve all the algorithms. But we have to record them manually. For example, I have world wide signal data from 2017-08-18 15:00 UTC +6 hours which is around 60GB.

It is important not to change any antenna positions/directions or filter settings since you've recorded the signals.

Note: All this can take some long time! I hope that we can implement some of the things from above within the next months, but I can not guarantee it.
Stations: 538, 672, 1534, 1555, 1696, 1712
Reply
#11
Good idea Tobi  Idea Wink
Stations: 584, 585, 2017
Reply
#12
Hi All, I agree with Richo. They are selling their system and using semi-technical "explanations" to sell this to people that have even less knowledge.

"Though the antenna shows receiving capability up to about 1 MHz, an upper limit is
set to 400 kHz in order to serve anti-alias properties, and a digital filter can be adjusted
in the range 100-400 kHz in order to reduce influence of radio signals; the wave form
distortion due to the actually chosen band-width limitation remains acceptable for the
present purposes.
Signal timing is achieved by means of a separately mounted commercial GPS clock
with an accuracy of 100 ns (initially 300 ns, module 2). Special measures are taken
to transport this level of basic accuracy as far as possible to the actual event timing.
Signal amplification, filtering, AD-conversion and data processing are performed with
a single plug-in card (module 3) in a remotely positioned standard PC (module 4).
The effective sampling rate was set to 1 MHz and incoming signals are recorded with
14 bit resolution in a continuous mode. Triggered events are transferred to and processed
in a parallel unit so that no data loss occurs and no allowance for rearm time
must be provided (zero deadtime). Signal rates of up to approximately 1 kHz can be
handled. Each station collects data and transmits packets of condensed information
to the central station at Garching. Depending on the type of chosen line-connection
this transmission can take place immediately after completion of an individual signal
analysis, or within pre-selected time intervals for a group of signals which has been
accumulated within this interval. Due to transmission line band width the transfer is
presently limited to some 100 signals/s which is plenty even during strong thunderstorm
activity."
(Geophysical Research Abstracts, Vol. 7, 00685, 2005
SRef-ID: 1607-7962/gra/EGU05-A-00685
European Geosciences Union 2005
Modern Lightning Detection and Implementation of a
New Network in Germany
H.-D. Betz, W. P. Oettinger, K. Schmidt, and M. Wirz
Physics Department, University of Munich, D-85748 Garching, Germany)

We are actually doing things far better, our timing is order of magnitude better, 10ns, and our reduced bandwidth does not have to deal with broadcast and military stations and other sources of interferrence that exist in the LF/VLF band that they profess to be using (Up to 1 MHz?)
I am also not certain that the vertical radiation pattern of a loop type antenna is sufficient to deal accurately with the difference in height between C to C and C to G type lightning.
A multi-antenna system is better.
For Europe especially Germany we have a better network density with more stations used in the detection and location of strokes, in another part of this document they talk of a delay time between being able to calculate individual strokes, therefore they are not actually recording all the strokes.
The internet speed and the computing power and timing accuracy has also increased and we are constantly updating our algorithm.
We do not suffer from commercial inertia, therefore if we have not already done so, eventually we we will overtake them and our system will be better.
Analyzing data is always the bottleneck, I read that NASA has accumulated data from the various space probe missions that even if it stopped now would take over 15 years to be reviewed and analyzed.

Brian.
Reply
#13
It will pass some time until I can mount a E-field antenna on my system red in Romania, but I just started deploying a linux-based virtual machine where I can do some tests.

For now, I started downloading the signals from all red and blue stations in Germany (to record at least 2-3 days of signals and strikes) and will start analyzing the strikes/signals with Tobi recommendations.


Attached Files Thumbnail(s)
   
Stations: 951
Reply
#14
I have received a signal from a lightning which saturated one of the H channels. Data from the european lightning detection network EUKLID could be correlated with this ligthning, it was a very strong negative lightning with -144 kA strength. The E field channel (magenta trace) goes POSITIVE in the first peak of the signal, so this is the signature for negative CG lightning. Positive CG lightning will first go negative in the E field channel.


Attached Files Thumbnail(s)
   
Stations: 1836
Reply
#15
(2017-08-28, 09:45)pasense Wrote: I have received a signal from a lightning which saturated one of the H channels. Data from the european lightning detection network EUKLID could be correlated with this ligthning, it was a very strong negative lightning with -144 kA strength. The E field channel (magenta trace) goes POSITIVE in the first peak of the signal, so this is the signature for negative CG lightning. Positive CG lightning will first go negative in the E field channel.

That would be an interesting observation... if it proves consistent. Using the attached correlation E with Channel C ...(E signal in phase with A and B, trust me,)  which of the following random "sferics" might be "Positive" sferics? Rolleyes Why?  (Just assume they're all strokes, not junk)...


Attached Files
.pdf   Positive query huh.pdf (Size: 205.61 KB / Downloads: 14)
.pdf   Positive query huh2.pdf (Size: 206.26 KB / Downloads: 7)
                   TwinHollies WeatherCenter  Frankfort KY, USA
  Americas Operators at Sferics.us
        Stations: 689, 1439
Reply
#16
Observations from long distance are not relevant, as they can be reflections :-)

/Richo
Stations: 584, 585, 2017
Reply
#17
Photo 
(2017-08-28, 13:24)RichoAnd Wrote: Observations from long distance are not relevant, as they can be reflections  :-)

/Richo

Exactly.... and which channel is referenced for phasing?
[Image: evsh.png]
http://onlinelibrary.wiley.com/doi/10.10...10294/full

Positive Ground strokes tend to single discharge events... negative tends to multiple discharges. even tthat rule of thumb isn't always valid....
[Image: polaritychart.png]
https://www.ikhebeenvraag.be/mediastorag...liksem.pdf
                   TwinHollies WeatherCenter  Frankfort KY, USA
  Americas Operators at Sferics.us
        Stations: 689, 1439
Reply
#18
Richo, 
I used the very beginning of the signal, so this cannot be a reflection which is contributing only later to the signal. The line of sight from the station to the stroke was only over the sea, so the ground wave had little attenuation. I will try to get the opinion of lightning location specialists on this topic (identification of polarity for distant strokes).
Cutty, there was another stroke only 40 msec later at the same postition, so this is consistent with a negative stroke. Positive CG has only very rarely a second stroke at exactly the same position. A problem with +CG is that these waveforms are more complicated than -CG because very often there is significant cloud discharges before the return stroke. Sometimes this even leads to strong +CG not being identified correctly, instead a cloud discharge is assumed by the system.
What do the colors in your attachments stand for?
Stations: 1836
Reply
#19
Assuming  the magenta channel is the E field and the yellow one the H field, this is my tentative identification:

Positive query huh
unclear -CG
unclear -CG
+CG     -CG
-CG ??  -CG
unclear +CG
-CG      -CG

positive query huh2
+CG    +CG
unclear -CG
+CG    +CG
-CG     +CG
unclear unclear
unclear +CG

I used the polarity of the first part of the E field signal, when I could identify the start. When the first pulse of the E field is positive I identify it as -CG, when it goes negative as +CG. This is then in line with the signal I presented above.
I would have prefered to see the horizontal H signals as well to get a better idea of the start, the signal of the  C channel is a bit weak.
Stations: 1836
Reply
#20
(2017-08-28, 16:33)pasense Wrote: Assuming  the magenta channel is the E field and the yellow one the H field, this is my tentative identification:

Positive query huh
unclear -CG
unclear -CG
+CG     -CG
-CG ??  -CG
unclear +CG
-CG      -CG

positive query huh2
+CG    +CG
unclear -CG
+CG    +CG
-CG     +CG
unclear unclear
unclear +CG

I used the polarity of the first part of the E field signal, when I could identify the start. When the first pulse of the E field is positive I identify it as -CG, when it goes negative as +CG. This is then in line with the signal I presented above.
I would have prefered to see the horizontal H signals as well to get a better idea of the start, the signal of the  C channel is a bit weak.

Correct on Yellow being H C, and Magenta is E... this should be the standard...
I threw a curve, "Huh2" is simply 'Huh" inverted and mirrored.,... same signals... just diddled. Done to illustrate the possible difficulty of the premise....and if we only 'compare' one H channel,  than the 'missing' A and B channels are irrelevant for this illustration of channel vs channel... My experience with both RED and BLUE indicates that E is almost always In phase with one or more channels...

Sol, if we're looking for the 'out of phase' H vs E.... which H channel should we use as H ref?
                   TwinHollies WeatherCenter  Frankfort KY, USA
  Americas Operators at Sferics.us
        Stations: 689, 1439
Reply




Users browsing this thread: 1 Guest(s)