Hi Leif,

I am still working with linrad and the SDR-IQ and trying to understand
things.  Is this is off topic?

In an earlier email, you wrote:

How do I get from (21000-4000)/1100 = 15.4545 to a sampling rate of 2048
or greater.  Is this because in the SDR produces I and Q and I need 2
bins per pixel?

Tom

Hi Thomas,

Of course not. Anything that helps to understand how things work is OK:-)

With 15.45 Hz per pixel and also 15.45 Hz per fft bin you will get a spectrum
with a resolution of 30 Hz (with a sine squared window.) That will be optimum
in some situations, but if you want to look at really weak signals that are
close to very strong signals you might want a sine to power 8 window and
then you might want to have two bins for each pixel.

Now, with one bin for each pixel you would need 1100 bins for the frequency
range that you want to display. Linrad can not compute 1100 bin FFTs. The
smallest one would be 2048. If you would choose 1024, the maximum frequency
coverage would be 1024*15-45 Hz = 15.82 kHz, but you would not be able to use
all of that. There has to be filters that prevent aliasing from resampling
so the useful bandwidth might be something like 14 kHz only. You said you want
17 kHz bandwidth so it will be necessary to use 2048 FFTs (or larger.)

If you sample at 31.65 kHz, the dynamic range will not be particularly good.
The SDR-IQ only transfers 16 bit over the USB. At 126 kHz you would get 6 dB
more dynamic range.

When sampling at 126 kHz you might ask for a bandwidth of 60 Hz and
display 2 bins for each pixel. It would be a bad idea for low order
windows, but with higher order windows you might like the result.
It all depends on what you are looking for.

By the way, I completed "newcomer install instructions for Ubuntu 10.4"
yesterday. One important thing is that "Visual Effects" is enabled by
default for all users. One has to disable it individually for each
user....
http://www.sm5bsz.com/linuxdsp/install/cd-dists.htm

73

Leif / SM5BSZ

Hi Leif,

I am keeping some notes.  When I get enough, I will send them to you.
May help someone in the future.

When I change the par files, I often get a waterfall display with large
negative frequencies across the top.  I have a hard time getting Fmin at
the top left and Fmax at the top right.

I have
xM_CIC2 [7]
M_CIC5 [25]
M_RCF [3]
points per pixel [1]
pixels per xpoint [0]
first xpoint [258]
xpoints [1024]
freq [0.025000000372529029846191406250]

When I start xlinrad and press B, I get
1st point -36.4846 KHz,  last point -28.5714KHz.

What determines the frequency displayed across the top?

Tom

Hi Thomas,

This generates a sampling frequency of 126.984 kHz.
To receive from zero to 127 kHz you would have to set the center
frequency to 0.063492 MHz

The fft size is 8192. your frequency range is 25-63.49 to 25+63.49 kHz=
-38.49 to 88.49 kHz. To get your scale starting at 4 kHz you need to set
the first pixel to (4000+38490)/15 = 2833
If you keep the center frequency 0.025 kHz you should set:
first xpoint [2833] to see from 4 kHz and up.

You could also set the center frequency to 60 kHz and
first xpoint [479] to see from 4 kHz and up.

Do not set the frequency to 64 kHz. At 4 kHz signals would then
become attenuated by the RCF filter. The useful bandwidth is
not 127 kHz but about 85% of that so you need to leave about
10 kHz unuset at both spectrum ends.

I suggest you try the arrow boxes. When you click inwards, you
will see more spectrum inside the window. Outwards will move
a fraction of what you see outside the window. If you make the window
width small, the amount of spectrum you move outside will become small.

The frequency scale on top is the antenna frequency. It is determined
by the center frequency C from the freq/att window and by the offset
in bins from (N/2)*w
F=C+(B-N/2)*w
F=freq on top
B=FFT bin number
N=FFT size
w=bin width
C=center frequency
w=S/N
S=sampling frequency.

73

Leif