Sferic detection and localisation with DSP
 

Hi,

My project is to realise an handheld lightning locator designed around
a Digital Signal Processor.
From what i learnt about lightning, i think i can detect storm by
detecting sferic signals. For the detection, i think that just an
antenna+ADC+ FIR filter will be enough but for the localisation i
don't know how to do. Should i make a FFT? I thought that as the high
frequencies of the sferic signal (about 7-8 KHz) move more rapidly
than the lowest, i could determine the distance of the storm from the
determination of the time between the arrival of the high frequencies
and the lowest ones. What do you think about that? Does a FFT allow me
to do that?

I'll be grateful for your help!!

Thanks.

phunkyman wrote:

Hi,

My project is to realise an handheld lightning locator designed around
a Digital Signal Processor.
From what i learnt about lightning, i think i can detect storm by
detecting sferic signals. For the detection, i think that just an
antenna+ADC+ FIR filter will be enough but for the localisation i
don't know how to do. Should i make a FFT? I thought that as the high
frequencies of the sferic signal (about 7-8 KHz) move more rapidly
than the lowest, i could determine the distance of the storm from the
determination of the time between the arrival of the high frequencies
and the lowest ones. What do you think about that? Does a FFT allow me
to do that?

I'll be grateful for your help!!

Thanks.


A Fourier transform is anti-arrival time, replacing time with frequency.
The less precisly you determine frequency, the more precisely you
determine time.

A single spike transforms into a constant (complex) fourier transform
whose phase change wrt. frequency gives the (single) arrival time
of the spike. Differing arrival times with frequency is could be
gotten by plotting the curve of phase against frequency, the slope
giving arrival time. But that's going to be noisy unless you have
a single spike over the time window that you're doing the Fourier
transform (and a FFT has a problem with being periodic as well, so
you have to window and shade). A double spike would ruin the entire
computation.

Probably easier just to build a few bandpass filters and measure the
time signal output of each, and pick off peaks in each.

I haven't heard of any time delay with frequency but I suppose it's
possible once the frequency is low enough that the signal is
really a guided wave in a duct under the ionosphere; but on the
other hand you're in the near field of the thing and perhaps that
wipes out the effect entirely with evanescent responses.
--
Ron Hardin


On the internet, nobody knows you're a jerk.

Ron Hardin wrote:
Probably easier just to build a few bandpass filters and measure the
time signal output of each, and pick off peaks in each.

Or use identical low pass filters and multiply each band you want
down to DC (multiply by sine and cosine, and use a complex LPF);
a trivial computation. The LPF determines the filter shape, impulse
response, etc, and you get a complex signal (and so a phase for each
band) in addition to a peak from the envelope.
--
Ron Hardin


On the internet, nobody knows you're a jerk.

phunkyman wrote in message
m...
Hi,

My project is to realise an handheld lightning locator designed around
a Digital Signal Processor.
From what i learnt about lightning, i think i can detect storm by
detecting sferic signals. For the detection, i think that just an
antenna+ADC+ FIR filter will be enough but for the localisation i
don't know how to do. Should i make a FFT? I thought that as the high
frequencies of the sferic signal (about 7-8 KHz) move more rapidly
than the lowest, i could determine the distance of the storm from the
determination of the time between the arrival of the high frequencies
and the lowest ones. What do you think about that? Does a FFT allow me
to do that?

I'll be grateful for your help!!

Thanks.

Heck... I'd just go to an avionics shop and pick up a circa 1970's ' Ryan
Stormscope' out in the back. It was great... at the approach of a storm,
jump in an airplane, throw the master switch and watch the display---on the
ground!

RG

why bother ? ........
what would your's offer that all the current model doesn't ?
these sell for $25 used and $54 new and work out to 40 miles
away from the storm. these have proven to be VERY
dependable.

http://www.strikealert.com/
http://www.dwquailgolf.com/access/misc_strikealert.html


On 29 Mar 2004 00:17:07 -0800, (phunkyman)
wrote:

Hi,

My project is to realise an handheld lightning locator designed around
a Digital Signal Processor.
From what i learnt about lightning, i think i can detect storm by
detecting sferic signals. For the detection, i think that just an
antenna+ADC+ FIR filter will be enough but for the localisation i
don't know how to do. Should i make a FFT? I thought that as the high
frequencies of the sferic signal (about 7-8 KHz) move more rapidly
than the lowest, i could determine the distance of the storm from the
determination of the time between the arrival of the high frequencies
and the lowest ones. What do you think about that? Does a FFT allow me
to do that?

I'll be grateful for your help!!

Thanks.

Hi,

First of all, thanks for your answer.


A single spike transforms into a constant (complex) fourier transform
whose phase change wrt. frequency gives the (single) arrival time
of the spike. Differing arrival times with frequency is could be
gotten by plotting the curve of phase against frequency, the slope
giving arrival time. But that's going to be noisy unless you have
a single spike over the time window that you're doing the Fourier
transform (and a FFT has a problem with being periodic as well, so
you have to window and shade). A double spike would ruin the entire
computation.

I can't plot anything since i must design an handheld device.
Do you know something about sferics? If youd don't, i can say that
sferic signals are aperiodic signals, brief in time (between 1 or 2 ms
i think) and their spectrum shows that their main energy is located in
the 0-10KHz band.So i think there are many spikes. (I'm french so i
don't really understand the word spike. Does it mean "transitoire" in
french?).


Probably easier just to build a few bandpass filters and measure the
time signal output of each, and pick off peaks in each.

If i make filters, they will be digital filters since i have to do it
with a DSP.
If i make it, how can i measure the time? Could you be clearer ,
please.

I haven't heard of any time delay with frequency but I suppose it's
possible once the frequency is low enough that the signal is
really a guided wave in a duct under the ionosphere; but on the
other hand you're in the near field of the thing and perhaps that
wipes out the effect entirely with evanescent responses.

There does exist time delay : it's called dispersion (in french). Of
course if the storm is near i won't "feel" this phenomenon, but the
aim is to predict the arrival of storms and so to study storms which
are for the moment at a minimum distance from me.

Thanks again.

Hello,

Finally i think that may be i could determine distance of the storm
from the intensity of the sferic signal recieved on the antenna. What
do you think about that? But if i do that, how the intensity of the
sferic signal and the distance of the storm are linked?

Thanks again.