[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Linrad parameters

Hi All,

Jeffrey Pawlan, WA6KBL wrote:

> Need help:
>   I have printed out Leif's pages and also have the Dubus 
> article here. Yet I am still unclear how to setup all the 
> parameters and use the program as effectively as Leif or 
> Roger.  I would be very appreciate of help.

This is very difficult because in order to help I would
need to know what you want to do.

Linrad is an SDR toolbox. It can be used for many things:)

It is something like this.

1) I have this hardware: xxxxxxxxxxx
2) I want to receive this kind of signals yyyyyyyyyyyyy in 
an environment limited by this kind of interference zzzzzzzzzzzzzzz.

Question 1. Is my hardware suitable?

Question 2. How do I set my parameters to do the best I
can with the hardware I have now? 

For question 2 there some answers that are always valid.

1. Some parameters are independent on what you want to
do, they have to be set according to your hardware.
First of all, sound, screen and mouse have to be set
once and for all. Typically you would want the biggest
screen supported by your hardware. The mouse parameters
are set in /etc/vga/libvga.config, the audio mixer is
set from some mixer program. The mixer program has to
be used to set the sampling speeds in some cases (Delta 44)

2. Some parameters have to be set according to your
analog hardware, but they may have to be compromised
if your digital hardware is not fast enough. With a Delta
44 you may select 16 or 32 bits. If you select 32 bits
the raw data files you may save with the S command become 
slightly bigger. (they then save 18 bit) As it happens on
my Pentium III Linrad could save 16 bits but not 18,
overrun errors occurred after a while. Since then both
Linrad and the Linux kernel have improved and I have
no longer any reason to select 16 bit for my Delta44.
To set the sampling speed, make a guess, then look at 
a strong signal that is slowly tuned upwards in frequecy.
If the signal disappears completely into the noise before
you have reached the end of the spectrum you can safely
set a lower sampling speed. If you can follow the signal 
to the end of the spectrum and see if fold back towards
lower frequencies while you tune still higher you should try
a higher sampling speed. 

3. Some parameters depend on the mode. What is the
modulation characteristics of the signal you are going 
to look for. These parameters always represent a compromise
because as the operator of a radio you want impossible things.
The fundamental compromise is between processing delay and
filter steepness. If you ask for a filter with a bandwidth
of 0.25Hz the delay will be 4 seconds. You are allowed to 
select more or less anything because you may have perfectly
valid reasons. Another compromise is between filter shape
and cpu load. A higher N in the fft window functions will
give filters that suppress alias spurs better. Alias spurs
are typically not any problem, you need very pure signal
to see them at all unless you decide to use no window at all.
There will be an article about these things in QEX soon,
read about the mode depending parameters at the Linrad Home 
Page in the meantime.

The mode depending parameters are set from the menu you 
reach by pressing X from a mode.

4. Some parameters are set on the fly to fit the currently
received station. These parameters are set with mouse clicks
on the screen and in some cases by entering numbers from the 


I know well how to set Linrad up for 144MHz EME to give me
best performance. Some things are a matter of taste and others
may find other choices better even for this particular
purpose. I typically use a delay of 8 seconds, most of it
because of the AFC delay I find optimum. When I use coherent
CW (for the really weak stations) the delay may be up to 12 

If you want to use Linrad for HF contesting you would obviously
not want more than a few tenths of a second for the delay.

If you want to use Linrad to monitor 90kHz at 10GHz  something
completely different is optimum.

In general terms, the fft bin width in the waterfall should
not be made smaller than 20Hz unless the spectral width of the 
signal you are going to look for is narrower. You would introduce
delay and gain nothing. 

For 144MHz the width of a CW carrier is about 0.25Hz. There is 
no reason to set the bin width this narrow with current Linrad 
versions because there is currently no way of retrieving the
keying information. Coherent averaging of CW signals is a very 
interesting way of making use of the extremely high sensitivity
one would get with stable enough signals but until it is implemented
there is no reason to set the bin width narrower than 5Hz or so.
You will be able to find and lock to signals too weak to decode
already at this bandwidth.

Linrad has some help available with the F1 key. Use it and try to
get a feeling for what you can do. Then when you think you have
settings that seem good for your purpose, make a raw data file
(Press S) and record a signal that is close to the threshold of
what you can copy. Send the S-file to me together with the parameter 
files of your system. I will look at it and suggest different 
parameters if I think it makes a difference. I will also make
changes to the algorithms within Linrad if I can see some 
phenomenon I have not encountered before and that I can 
do something about. BUT please do not send huge data files via 
E-mail unless we made an agreement about it;)


Leif  /  SM5BSZ