Antenna for receiving pulsar signals

Given a quiet local environment and several hours of
integration time, what rough size of antenna would be
needed to receive millisecond pulsar signals?

George

On a sunny day (Thu, 15 Jan 2004 19:57:06 -0000) it happened "George Dishman"
wrote in :

Given a quiet local environment and several hours of
integration time, what rough size of antenna would be
needed to receive millisecond pulsar signals?

George
Not sure what you mean, but if the 'integration time' is longer
then milliseconds, you for sure will not be able to detect those milliseconds
(as modulation in amplitude).
You will detect there is something there (average power).
So you mean measurement time?
No idea, depends on how strong the signal is, and I dunno that.

"Jan Panteltje" wrote in message
...
On a sunny day (Thu, 15 Jan 2004 19:57:06 -0000) it happened "George
Dishman"
wrote in
:

Given a quiet local environment and several hours of
integration time, what rough size of antenna would be
needed to receive millisecond pulsar signals?

George
Not sure what you mean, but if the 'integration time' is longer
then milliseconds, you for sure will not be able to detect those
milliseconds
(as modulation in amplitude).
You will detect there is something there (average power).
So you mean measurement time?
No idea, depends on how strong the signal is, and I dunno that.

If you already know the period of the pulsar it is possible to integrate the
signal and raise it above the noise background. Probably would employ
an electronic device known as a phase-locked loop, which can "lock-on"
to a weak signal if it is set close to the signal frequency ( the weaker the
signal, the closer the PLL has to be set to the exact signal frequency ).

Andy

On a sunny day (Fri, 16 Jan 2004 00:37:24 -0000) it happened "andrewpreece"
wrote in :


"Jan Panteltje" wrote in message
...
On a sunny day (Thu, 15 Jan 2004 19:57:06 -0000) it happened "George
Dishman"
wrote in
:

Given a quiet local environment and several hours of
integration time, what rough size of antenna would be
needed to receive millisecond pulsar signals?

George
Not sure what you mean, but if the 'integration time' is longer
then milliseconds, you for sure will not be able to detect those
milliseconds
(as modulation in amplitude).
You will detect there is something there (average power).
So you mean measurement time?
No idea, depends on how strong the signal is, and I dunno that.

If you already know the period of the pulsar it is possible to integrate the
signal and raise it above the noise background. Probably would employ
an electronic device known as a phase-locked loop, which can "lock-on"
to a weak signal if it is set close to the signal frequency ( the weaker the
signal, the closer the PLL has to be set to the exact signal frequency ).

Andy
Yes that is one method, to try to lock a PLL to it,
however PLL does not like noise at all.
So then your loop filter for the PLL will have to have a low bandwidth.
Normally, the way I see it, if a signal (noise with some of the pulsar
pulsing in it at frequency x), you would take n samples (so for a fixed
sample frequency a specific time), run a FFT on it.
In the resulting frequency spectrum you would see a peak at say 1 kHz
(if the pulsar pulsed once every millisecond).
The more samples you have (the longer your data, *receiving time*), the
better the result of the fft.
You need at least a (Nyquist) 2 x the pulsar frequency to make anything
out.
So integration is (in this example) not the right word.
This is the way *I* would look for a signal (more advanced algos exist).
When using a non digital setup, you are heterodyning , down mixing from some
short wavelength, and then use a AM detector, it would show a 1 ms pulsar as
a 1 kHz beep (on a speaker), slower ones do 'phhs', 'phhs', 'phhs',
with each 'phhs' for one revolution of the pulsar (not sure if it pulses
twice or once per revolution, may depend on the angle dunno).
It is a typical sound, I have heard it.
And noise...
The filter after the AM detector is a lowpass (you can call that an
integrator), and it must be able to pas the 1000 Hz as in this example, a
1 Hz low pass would give you no pulsar 'sound'.
It WOULD give some DC level that you could detect.
As good narrow band filter (once you find the frequency) (parametric equalizer
for example) could filter out more noise and you'd have a nice reference that
you could then sync you watch to..
Pulsar is supposed to be very stable....
Note no PLL here, but could be used of cause.
This is all I know about pulsars, but the electronics I can expand on if you
wish.
JP

"Jan Panteltje" wrote in message
...

You might also wish to look at software methods to pull
the signal out of the collected signal. See, for
example, Scargle Periodogram methods.

I did an article on this for The Orrery newsletter a while
back.


--
-----------------------------------------------------------------------
Greg Neill, Editor
The Orrery: Models of Astronomical Systems
http://members.allstream.net/~gneill/
-----------------------------------------------------------------------

On a sunny day (Fri, 16 Jan 2004 13:02:49 -0500) it happened "Greg Neill"
wrote in
:

"Jan Panteltje" wrote in message
...

You might also wish to look at software methods to pull
the signal out of the collected signal. See, for
example, Scargle Periodogram methods.

I did an article on this for The Orrery newsletter a while
back.


I just did a Google for Scargle Periodogram pulsar.
Very interesting, finding planets around pulsars with it.
Even with missing data points.
http://astrosun.tn.cornell.edu/~akgun/Grad/pulsarz.ps
Nice paper too.
Thank you.

--
-----------------------------------------------------------------------
Greg Neill, Editor
The Orrery: Models of Astronomical Systems
http://members.allstream.net/~gneill/
-----------------------------------------------------------------------

First I'd like to thank all who have replied, the information
provided has been very helpful.

"Jan Panteltje" wrote in message
...
On a sunny day (Fri, 16 Jan 2004 00:37:24 -0000) it happened
"andrewpreece"
wrote in :


"Jan Panteltje" wrote in message
...
On a sunny day (Thu, 15 Jan 2004 19:57:06 -0000) it happened "George
Dishman"
wrote in
:

Given a quiet local environment and several hours of
integration time, what rough size of antenna would be
needed to receive millisecond pulsar signals?

George
Not sure what you mean, but if the 'integration time' is longer
then milliseconds, you for sure will not be able to detect those
milliseconds (as modulation in amplitude).
You will detect there is something there (average power).
So you mean measurement time?
No idea, depends on how strong the signal is, and I dunno that.

If you already know the period of the pulsar it is possible to integrate
the
signal and raise it above the noise background. Probably would employ
an electronic device known as a phase-locked loop, which can "lock-on"
to a weak signal if it is set close to the signal frequency ( the weaker
the
signal, the closer the PLL has to be set to the exact signal frequency ).

The weaker the signal, the narrower the bandwidth of the
loop filter needed to reject the noise to below the signal,
and of course the initial PLL frequency has to be within
about a bandwidth of the actual signal.

However, AIUI, pulsar signals are broadband. As a result I
had assumed the signal would also be non-coherent (i.e. not
a modulated sine wave but amplitude modulated noise) but a
page on "Phase-coherent De-dispersion" has made me wonder
about that now.

Dispersion arises because different frequencies are delayed
by different amount by the interstellar medium (ISM) so if
you had a wide enough bandwidth on the receiver, it could
be delayed a whole cycle and the sum would be a continuous
constant level.

http://www.jb.man.ac.uk/~pulsar/Educ...00000000000000

Yes that is one method, to try to lock a PLL to it,
however PLL does not like noise at all.
So then your loop filter for the PLL will have to have a low bandwidth.
Normally, the way I see it, if a signal (noise with some of the pulsar
pulsing in it at frequency x), you would take n samples (so for a fixed
sample frequency a specific time), run a FFT on it.
In the resulting frequency spectrum you would see a peak at say 1 kHz
(if the pulsar pulsed once every millisecond).
The more samples you have (the longer your data, *receiving time*), the
better the result of the fft.
You need at least a (Nyquist) 2 x the pulsar frequency to make anything
out.

Imagine a simple wideband receiever with a fast amplitude
detector on the output. The 'signal level' would vary on
sub-millisecond timescales. Put that amplitude rather than
the RF signal itself into the FFT and you would get what you
describe. Note the "Square law detectors" in this:

http://tucanae.bo.astro.it/pulsar/32mt/

The use of the term "Phase-coherent" when talking of de-
dispersion puzzles me since RF phase information is discarded
by such a detector.

So integration is (in this example) not the right word.

The key is what Andy said, "If you already know the period
of the pulsar ..."

Suppose the period is known to be exactly 1ms. By sampling
the RF amplitude every 100us and adding the value to a set
of ten 'bins', the pulse shape emerges from the noise:

http://www.radiosky.com/rspplsr.html

The trick is that the free-running clock doing the 100us
timing has to stay exactly in step with the unknown signal
for the integration period or the pulse will just smears
over the bins and give the average in all. This is where
I was wondering if timing the sampling with an atomic clock
could allow longer integration periods and hence a smaller
antenna.

This is the way *I* would look for a signal (more advanced algos exist).
When using a non digital setup, you are heterodyning , down mixing from
some
short wavelength, and then use a AM detector, it would show a 1 ms pulsar
as
a 1 kHz beep (on a speaker), slower ones do 'phhs', 'phhs', 'phhs',
with each 'phhs' for one revolution of the pulsar (not sure if it pulses
twice or once per revolution, may depend on the angle dunno).
It is a typical sound, I have heard it.
And noise...
The filter after the AM detector is a lowpass (you can call that an
integrator), and it must be able to pas the 1000 Hz as in this example, a
1 Hz low pass would give you no pulsar 'sound'.
It WOULD give some DC level that you could detect.
As good narrow band filter (once you find the frequency) (parametric
equalizer
for example) could filter out more noise and you'd have a nice reference
that
you could then sync you watch to..
Pulsar is supposed to be very stable....

They can be more stable than atomic clocks!

Note no PLL here, but could be used of cause.
This is all I know about pulsars, but the electronics I can expand on if
you
wish.

If you have any information how broadband de-dispersion
might be performed in the RF stage, say by a configurable
flat gain frequency-dependent phase shifter, I would be
very interested.

George

On a sunny day (Sat, 17 Jan 2004 09:29:28 -0000) it happened "George Dishman"
wrote in :

the
signal, the closer the PLL has to be set to the exact signal frequency ).

The weaker the signal, the narrower the bandwidth of the
loop filter needed to reject the noise to below the signal,
and of course the initial PLL frequency has to be within
about a bandwidth of the actual signal.
In some video applications and servo applications, you
let it lock in frequency first (high bandwidth), then phase lock with
lower bandwidth.
I have even seen servo system where a third even lower
bandwidth / high gain loop then takes over (if phase locked),
to further reduce phase errors.
If some transient kicks in, the whole sequence starts again...

However, AIUI, pulsar signals are broadband. As a result I
had assumed the signal would also be non-coherent (i.e. not
a modulated sine wave but amplitude modulated noise)
eh, no, it wil always be a 'superposition', unless there is some
effect in the RF or IF stages like amplitude dependent gain
(say non-linearity).
There will be a whole spectrum of stuff around that center
1 KHz (in this case), if it comes from a jet, it may vary over time
(not likely that jet, when sweeping across your receiver will be
a smooth thing), but that spectrum could be anything: say the jet
consists of some small sub-beams, then you get 1KHz, but also
harmonics, and the phase and frequency of these may well change
over time.


but a
page on "Phase-coherent De-dispersion" has made me wonder
about that now.

Dispersion arises because different frequencies are delayed
by different amount by the interstellar medium (ISM) so if
you had a wide enough bandwidth on the receiver, it could
be delayed a whole cycle and the sum would be a continuous
constant level.
Then the higher the pulsar freq, the more that could happen?
I will have to read that, OK.
http://www.jb.man.ac.uk/~pulsar/Educ...00000000000000
OK, later now, read that.
Would it make sense that if you know the phase versus frequency
curve, then you can make a compensator for that?
So that would be a filter with a phase characteric that shifts in
such a way that (for the bandwidth used) zero phase error remains?
Makes that sense?

Imagine a simple wide band receiver with a fast amplitude
detector on the output. The 'signal level' would vary on
sub-millisecond timescales. Put that amplitude rather than
the RF signal itself into the FFT and you would get what you
describe. Note the "Square law detectors" in this:

http://tucanae.bo.astro.it/pulsar/32mt/

The use of the term "Phase-coherent" when talking of de-
dispersion puzzles me since RF phase information is discarded
by such a detector.
Have not read it (yet), but I think I pointed out that 'phase coherent'
here refers to the sidebands, well, for all I know, over a relatively
short period of time (in the pulsars life, so for a long time for us),
if nothing physically changes (in that time) for the beam(s) emitted by
the pulsar, the sidebands should be phase coherent.
Not sure that is what you mean, I will read that stuff.


So integration is (in this example) not the right word.

The key is what Andy said, "If you already know the period
of the pulsar ..."

Suppose the period is known to be exactly 1ms. By sampling
the RF amplitude every 100us and adding the value to a set
of ten 'bins', the pulse shape emerges from the noise:

http://www.radiosky.com/rspplsr.html
Yes this is well known.
Now later: Here your clock may help?

The trick is that the free-running clock doing the 100us
timing has to stay exactly in step with the unknown signal
for the integration period or the pulse will just smears
over the bins and give the average in all.
Yes a constant clock (you want atomic clock), will reduce error.
Interesting thing is all the time in this discussion I was thinking about
the thread (some year ago?) in sci.crypt (or sci.physics), where
NON constant sampling was used to get almost undetectable info from
pictures (like Roentgen.
Some prof could detect images in almost TOTAL noise with a new algo,
using VARIABLE sampling.
They were a bit secretive about the math, but it seemed to make sense.
Applications were of cause military etc...
I was thinking if this could be used to get better resolution.
If an atomic clock would improve results in this (pulsar) case, I thing
stability we think about here over the time of a measurement....
In the planet finding paper they talk about 200 days... (now you look for the
frequency variation of the pulsar), maybe to have absolute time will help.
The drift in 200 days should be factor n of a period less then period
time of the pulsar?
So if 1000 pulses / second, 3600 x 24 x 200 periods = 17 280 000 pulses.
For 1 degree phase error x 360 = 6 220 800 000
So 1 part in 6 10^9, yes :-)
I really dunno if this makes sense.

This is where
I was wondering if timing the sampling with an atomic clock
could allow longer integration periods and hence a smaller
antenna.

If you have any information how broadband de-dispersion
might be performed in the RF stage, say by a configurable
flat gain frequency-dependent phase shifter, I would be
very interested.
I will read those links first, if I can think of anything that makes
sense I will post it here.

OK I have read those now, what I wrote stands.
Your freq dependant phase shifter is the same idea, I do
not see whay this should not work.
To make such a filter would not be that difficult, if provided
with actual degrees and MHz.
Spice
It does not have to work at (for example) 400 MHz (as in that link),
but the receiver will have a much lower IF amplifier, with known
bandwidth, and at say 40 MHz and 10 MHz bandwidth you are in normal
analog TV stuff, and filters at that frequency are easy.
At least I have experience with that.
If it all helps, I dunno, the proof of the pudding is in the eating really.
For the avid experimenter here is a great area I am sure.
Digitizing at for example at 32 mega samples per second, like I am doing
here for video, you could do the compensation with a DSP, or later with
some algo?
Maybe I got this wrong? Seems too simple, some catch somewhere?
JP

"George Dishman" writes:
Dispersion arises because different frequencies are delayed
by different amount by the interstellar medium (ISM) so if
you had a wide enough bandwidth on the receiver, it could
be delayed a whole cycle and the sum would be a continuous
constant level.

That's true. The "professionals" often sample a bank of narrow
bandwidth channels, and then recombine these channels later using the
dispersion correction formula.

The use of the term "Phase-coherent" when talking of de-
dispersion puzzles me since RF phase information is discarded
by such a detector.

George, phase coherent here is refering to the pulsar *pulse* phase,
not the RF signal phase. When pulsar people talk about a phase
coherent solution, they mean they have a model which can account for
every pulse cycle in the data arc. Of course, just as for RF, it's
possible to lose track of the pulsar cycle count, especially over
large data gaps.

Craig

--
--------------------------------------------------------------------------
Craig B. Markwardt, Ph.D. EMAIL:
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
--------------------------------------------------------------------------

In message , George Dishman
writes
Given a quiet local environment and several hours of
integration time, what rough size of antenna would be
needed to receive millisecond pulsar signals?

An acre or two of phased array dipoles should be enough to get started.
Though you may need more than that to get down to the millisecond
pulsars even with sophisticated integration techniques.

ISTR about 20000m^2 of aerial discovered the Crab nebula at ~80MHz

Regards,
--
Martin Brown