Noise...
Simon HB9DRV
So you’re watching 1MB of bandwidth, no signals visible. How would you detect the presence of RF Energy, possibly as weak as possible? Consider you’re watching the moon for the weakest of weak digital signals and you don’t know what the mode will be, you just want to know ‘Is there anybody out there’ … a bit like SETI.
Very long FFT?
I’m sure there must be some logic which I either haven’t found or have forgotten.
Simon Brown, HB9DRV
http://sdr-radio.com
Not sent from an iPhone: I prefer cappuccino to Cupertino.
Leif Asbrink
> So you're watching 1MB of bandwidth, no signals visible. How would you
> detect the presence of RF Energy, possibly as weak as possible? Consider
> you're watching the moon for the weakest of weak digital signals and you
> don't know what the mode will be, you just want to know 'Is there anybody
> out there' . a bit like SETI.
We must know what we look for. The FFT should be set to match the
bandwidth of the signal of interest.
As a radio amateur I might want to look for JT65 signals bounced off
the moon. Knowing that the bandwidth is about 5 Hz on 144 MHz I would use an
FFT with a bin separation of about 2.5 Hz. Perhaps a 524288 FFT with
a sine squared window. Bin separation would be 1.9 Hz and the bin bandwidth
would be 3.8 Hz. The transforms must overlap by 50% so one would
have to compute those transforms at a rate of 3.8 Hz.
The transforms have to be averaged over at least 5 seconds, preferrably
more. To display the spectrum on a screen with about 1000 pixels one has
to split the 524288 FFT into 1000 groups with 524 bins in each one.
One would find the bin with the largest averaged power in each
group and display the corresponding colour for the pixel representing the group.
The 5 seconds would contain 20 transforms, but with 50% overlap the averaging
is only a factor of 10. A signal with S/N in the order of -3dB in 4Hz bandwidth
will be visible. That corresponds to -28 dB in SSB bandwidth. If one zooms into
the spectrum to place one pixel on screen for each FFT bin one would be able
to see even weaker signals. The process of picking the strongest bin in each
group introduces noise. With more averaging that noise is reduced.
The problem in real life today is all the spurs. There will be a spur in almost
every group of 524 bins.....
73
Leif / SM5BSZ
stan...@aol.com
I really don't understand as well as I should. I use Linrad every day. I operate in the "Expert Mode".. I have been using Linrad for many, many years with WSE. My brain behaves an a newcomer... My apology but I am not able to follow your explanation below..
I see JT65b signals from the moon in the WSJT waterfall. What parameters should I use in Linrad that would make my waterfall match the WSJT waterfall?
I think your answer will be about bins, FFT and transform rates..... Is there a simple way to help me set numbers like: Main Spectrum==Lower left Box=? or perhaps numbers in par_ssb that would be better??
First FFT bandwidth (Hz) [300]
First FFT window (power of sin) [2]
First forward FFT version [4]
First FFT storage time (s) [1]
First FFT amplitude [450]
Main waterfall saturate limit [1]
Enable correlation spetrum [0]
Enable second FFT [1]
First backward FFT version [1]
Sellim maxlevel [6000]
First backward FFT att. N [6]
Second FFT bandwidth factor in powers of 2 [2]
Second FFT window (power of sin) [2]
Second forward FFT version [0]
Second forward FFT att. N [7]
Second FFT storage time (s) [5]
Enable AFC/SPUR/DECODE (2=auto spur) [2]
AFC lock range Hz [150]
AFC max drift Hz/minute [100]
Enable Morse decoding [0]
Max no of spurs to cancel [25]
Spur timeconstant (0.1sek) [5]
First mixer bandwidth reduction in powers of 2 [4]
First mixer no of channels [1]
Third FFT window (power of sin) [2]
Baseband storage time (s) [2]
Output delay margin (ms) [5]
Output sampling speed (Hz) [8000]
Default output mode [131]
Audio expander exponent [3]
Baseband waterfall saturate limit [0]
A/D speed [96000]
Check [321]
Leif Asbrink
> I see JT65b signals from the moon in the WSJT waterfall.
> What parameters should I use in Linrad that would make
> my waterfall match the WSJT waterfall?
>
> I think your answer will be about bins, FFT and transform
> rates..... Is there a simple way to help me set numbers
> like: Main Spectrum==Lower left Box=?
> or perhaps numbers in par_ssb that would be better??
from telling just a number for something.
Now, you give me a precise question so I think I can give
a precise answer. You want to see JT6565 so you need a bandwidth
of about 5 Hz in the waterfall. I write comments below each
parameter that matters.
> First FFT bandwidth (Hz) [300]
You want 5Hz. This bandwidth is 150 times above that so keeping
this value would mean that the second FFT should be run with
a bandwidth that is 128 or perhaps 256 times narrower. That
will be perfectly OK under normal circumstances, but it could
create problems with the blanker in case you have very strong
and very stable signals in the passband.
The advantage of a wide bandwidth here is that the sudden
appearence of a strong signal is detected more rapidly so
the time before it becomes routed outside the blanker becomes
smaller.
You may set different values here, but the value does affect
another parameter. See below.
> First FFT window (power of sin) [2]
This window should ALWAYS be used with MAP65. The reason is that
the other windows cause loss of one bit. In other contexts
that does not matter, but MAP65 does not like the noise level to
be set 6 dB higher to compensate.
(If you actually need a higher order window you should set
the timf2 level higher than normal by 10 dB and then use the timf2
attenuator to make MAP65 see a signal level that it likes.)
> First forward FFT version [4]
This could affect CPU load. On most systems it does not matter.
> First FFT storage time (s) [1]
This limits the averaging of the main spectrum.
It could affect the noise blanker. In case you want
to route somewhat weaker signals outside the blanker you need more
averaging and this parameter may set a limit. You might
make it 5 or 10 seconds. the extra memory allocated will
be small.
> First FFT amplitude [450]
This is very important. It affects MAP65 as well as the
Linrad waterfall. Read about it here:
http://sm5bsz.com/lir/map65/map65.htm
Particularly the link "The first FFT level"
(The pages are updated. Several errors have been corrected
and the text is improved due to feedback from Guy, VK2KU)
> Main waterfall saturate limit [1]
> Enable correlation spetrum [0]
> Enable second FFT [1]
This is a must for MAP65
> First backward FFT version [1]
This will set the 16 bit format for the data sent to MAP65.
As far as I know that is required by MAP65.
> Sellim maxlevel [6000]
This parameter is usually of no significance for MAP65.
In case you select a wide bandwidth for the first FFT
while setting 5 Hz for the second FFT you might have to reduce
this parameter to avoid saturation in "fft2 strong" That is
one of the things you will see after pressing "A" on the
Linrad screen when 16 bit integer is selected for the
second fft.
> First backward FFT att. N [6]
This is very important. It affects MAP65 as well as the
Linrad waterfall. Read about it here:
http://sm5bsz.com/lir/map65/map65.htm
Particularly the link "The timf2 level"
(The pages are updated. Several errors have been corrected
and the text is improved due to feedback from Guy, VK2KU)
> Second FFT bandwidth factor in powers of 2 [2]
This is totally wrong!!
With a desired bandwidth of 300 Hz and a sine squared window you
presumably have a bandwidth of 375 Hz for the first FFT.
Press X, then P and check. It is the Bw in yellow on the top line.
With a bandwidth factor of 2 to power 2 (=4) your second FFT
bandwidth should be 93 Hz. That is FAR too wide. In case you keep
the first FFT bandwidth parameter 300 you need 16 times narrower
bandwidth than that. At least.
For 6 Hz bandwidth you can use these combinations:
First FFT Bw Second Bw factor in pwr of 2
300 6
150 5
75 4
37 3
18 2
10 1
What to choose depends on your interference situation.
I suggest you try a couple of different combinations. The sliders for
the blanker will need different settings and you may have to adjust
the first backwards att. N. Under normal circumstances I do not
think you will notice any difference in the MAP65 performance
because normally the details on how strong signals are routed
outside the blanker is not important.
Press A and check that nothing is saturated.
Set the waterfall averaging for a new line about every 10 second.
Adjust colour and gain for the noise to become visible in the
waterfall.
As an alternative you might try this:
For 3Hz bandwidth you can use these combinations:
First FFT Bw Second Bw factor in pwr of 2
300 7
150 6
75 5
37 4
18 3
10 2
I can not say for sure whether 3 or 6 Hz will work best.
> Second FFT window (power of sin) [2]
> Second forward FFT version [0]
This means using float arithmetics for the second fft.
This way you will not have to worry about saturation.
In case you set the next parameter right you can get
equally good performance with considerably smaller cpu load.
> Second forward FFT att. N [7]
Only relevant in case the previous parameter is set to 2.
> Second FFT storage time (s) [5]
> Enable AFC/SPUR/DECODE (2=auto spur) [2]
You have enabled auto spur and AFC.
> AFC lock range Hz [150]
> AFC max drift Hz/minute [100]
These do not affect MAP65, but the AFC will behave very differently
now when you have set a narrow bandwidth for the second FFT.
Spur removal will be totally different also.
> Enable Morse decoding [0]
> Max no of spurs to cancel [25]
This is a very small number.
You might try 1000
> Spur timeconstant (0.1sek) [5]
In case you want to have spurs removed you should try different
values here.
> First mixer bandwidth reduction in powers of 2 [4]
> First mixer no of channels [1]
> Third FFT window (power of sin) [2]
> Baseband storage time (s) [2]
> Output delay margin (ms) [5]
> Output sampling speed (Hz) [8000]
> Default output mode [131]
> Audio expander exponent [3]
> Baseband waterfall saturate limit [0]
> A/D speed [96000]
> Check [321]
In case you click a signal, the frequency range that corresponds
to the selected passband will be routed through the blanker and
the gain would be forced to be the same for the selected passband.
In case you click a very strong JT65 signal or a frequency
with a very strong spur with a weak JT65 signal nearby this
group of parameters could affect MAP65. Normally none of them
will affect MAP65 or the Linrad waterfall at all.
73
Leif / SM5BSZ
VK2KU
I am actually starting to understand more and more :-)
73 Guy VK2KU
stan...@aol.com
This helps. I have changed below. I like what I see. I do not use Map65.
What would be a good First FFT bandwidth for CW?? and SSB?
73 Stan
From: Leif Asbrink <le...@sm5bsz.com>
To: linrad <lin...@googlegroups.com>
Sent: Tue, Feb 7, 2012 3:39 pm
Subject: [Linrad] Re: Noise...
For 6 Hz bandwidth you can use these combinations:best for SSB/CW?
First FFT Bw Second Bw factor in pwr of 2
300 6
150 5
75 4 I am experimenting with these combinations.
37 3
18 2
10 1You might try 1000 I changed to 1000 > Spur timeconstant (0.1sek) [5] In case you want to have spurs removed you should try different values here.I need help. How do I determine best?
Leif Asbrink
> This helps. I have changed below. I like what I see. I do not use Map65.
> What would be a good First FFT bandwidth for CW?? and SSB?
That depends on your interference situation.
It also depends on whether you want fast response (to run QSK) and
if you are interested in listening very close to very strong
signals and on other things as well.
In all, your question is far too unspecific....
What you can do - something that I repeatedly ask for on all the
mailing lists - is to make a wideband recording of a situation
where you have a signal that is not quite copyable. CW, SSB or
something else.
I usually ask for such recordings made by users of other SDRs
than WSE/Linrad, but it is of interest with Linrad recordings also.
I need the file plus the best loudspeaker output that you can
extract from it. In case you use Linrad I would of course be
interested in the complete set of parameter files as well.
I would use such recordings to try to find optimum parameters.
I actually do not know what will be optimum, I only make a reasonable
guess based on theory and I would then try to find something
better by trial and error. I would then make the files available
on the internet together with a description how to best use Linrad
copy the difficult signal.
73
Leif
Jeffrey Owen Katz
Yes. It seems spurs and birdies are a far more serious issue today than
in the days of the SP-600, RBC-3, and HRO-60. DDS chips seem to be the
primary offenders (along with far less pre-selection than in the distant
past).
Some people have told me that the si570, which uses a digital phase locked
loop, is virtually spur-free (except for simple harmonics). Has anyone
had experience using this as a local oscillator in weak signal work? It
looks like the si570 could be stabilised to better than a few ppb by the
addition of same extra hardware, so the few Hz drift at 144 MHz of the
unstabilised part could be easily eliminated.
Jeff
Simon HB9DRV
http://www.box73.de/catalog/product_info.php?products_id=1869
Look at the PDF
http://www.box73.de/catalog/pdf/BX-026.pdf?osCsid=1rqfi1lv0m3kr0fi68h0hkb2g6
and the picture on page 1.
Simon Brown, HB9DRV
http://sdr-radio.com
Not sent from an iPhone: I prefer cappuccino to Cupertino.
Leif Asbrink
The below old posting appeared in my mailmox today. I have
no idea why it was not delivered before.
On Sun, 5 Feb 2012 20:17:30 -0500 (EST)
Jeffrey Owen Katz <jeff...@scientific-consultants.com> wrote:
> Yes. It seems spurs and birdies are a far more serious issue today than
> in the days of the SP-600, RBC-3, and HRO-60. DDS chips seem to be the
> primary offenders (along with far less pre-selection than in the distant
> past).
To me spurs and birdies are ill defined concepts. In this thread
my use of the word was for undesired narrowband signals that enter
through the antenna. (External spurs)
There are false signals that may appear even when a dummy
load replaces the antenna and there are false signals that
are caused by strong signals.
Some false signals vary in frequency with the same rate
as the desired signal. Others may move at much higher
rates. I have a feeling that those were originally
called birdies (=fast moving spurs) regardless of
whether they were present withoout an antenna or not.
Anyway, false responses, is a problem caused by spurs on
LO signals and DDS chips can be a big problem. One should
avoid certain frequency ranges and use filters that make
spurs at wide separations harmless.
There is a lot written on the subject but I have no
personal experience. I just decided to stay away from the
technology.
> Some people have told me that the si570, which uses a digital phase locked
> loop, is virtually spur-free (except for simple harmonics).
This is not quite true. There are some spurs at wide
separation so it will be a good idea to apply some selectivity.
A design like the IQ+ for 144 MHz by HB9DRI has expensive
filtering that looks needless for a direct conversion
radio. Ideally it should be enough to attenuate overtones
and provide a little attenuation to avoid overload by
out of band signals. In real life the filters are not
wasted money because the Si570 has some rather strong
spurs well below 144 MHz that would create some nasty false
responses if narrow filters were not used.
> Has anyone
> had experience using this as a local oscillator in weak signal work? It
> looks like the si570 could be stabilised to better than a few ppb by the
> addition of same extra hardware, so the few Hz drift at 144 MHz of the
> unstabilised part could be easily eliminated.
The IQ+ is used for 144 MHz EME by several operators.
http://www.linkrf.ch/IQ+.html
The stability of the Si570 is good enough without any
stabilization as far as I know. There is some thermal drift
and I would guess operators have to keep the unit running
for a while before it is stable enough.
There are similar chips from Silicon Lab that use an external
clock signal with the same technology. They promise the
same phase jitter so it should be fairly straightforward
to build a GPS stabilized radio without spurs and with
excellent phase noise.
73
Leif / SM5BSZ