Answers inline below.
> MAP65 is quickly becoming obsolete so I am looking at alternatives. This
> brings me to Linrad so I have questions about Xpol and polarity handling in
> Linrad. Since WSJTx is being actively updated we can always expect it to
> have up to date features and modes and performance levels. The first
> question is about the wide band waterfall. For a system where both vertical
> and horizontal signals are presented to Linrad, how does the wide band
> waterfall display signals of both polarities? Is it some sort of digital
> mix from both channels? Will it be as sensitive as a polarity matched input
> to a single receiver?
Linrad computes the power of the polarization that is present in each
frequency bin. This becomes powerful when there are several spectra
for each line in the waterfall. On the other hand narrower bin widths
improves S/N while making the waterfall slower. Optimum is to set
the bin width to match the line width of the signal of interest.
On 144 MHz, for CW, the bandwidth of the carrier is about 0.2 Hz due
to doppler spread but frequency stability (chirp) typically makes
the bandwidth larger. For the sync tone of jt65 the bandwidth is
5 Hz if I remember correctly. That means that each FFT spectrum spans
0.4 seconds when sine squared windows are used. Then spectra overlap
by 50 % so a new spectrum would arrive each 0.2 second. If averaging
is set to 25 a new line would appear every 5 th second and a complete
transmission would produce about 10 lines. Then sensitivity would be
extremely good. The 25 spectra would give two average power spectra
plus one correlation spectrum (real and imaginary part.) I wrote the
code long ago and I did not make it quite right - and when now looking
I did not find that particular code snippet, but the formula I
adopted was the sum of the horisontal power and the vertical power
plus a function of the correlation spectra that would give a signal
with 50% of the power in H and 50% of the power in V not the sum
of H^2 and V^ (50% of what a signal in either antenna would give)
but also the power of the correlation spectrum so that a signal
at 45 degree polarization would give the same power as a signal
in one or the other antenna.
Today I think a better choice would have been to make a linear
combination of the signals to get new polarizations that would
have zero correlation. Then display the strongest one only.
A frequency bin with only noise in it would have almost zero
correlation if many anough spectra are averaged for each waterfall
line so then only the strongest of H or V would be displayed.
With noise only they would be equal so noise only would be
reduced by 3 dB. Likewise, by use of the correlation spectra one
could do a transformation into two uncorrelated signals and then
skip the weakest.
> Maybe I missed where this is written about. A link would be welcome. There
> is little information on the internal workings of Map65. Does the wide band
> waterfall of Map65 simply copy Linrad or is it different?
I am pretty sure it is different.
> The next question is about adaptive polarity. I understand adaptive
> polarity was designed for CW EME using an Xpol system. What I have noticed
> while using an Xpol system with it and monitoring noise, it will seek to
> minimize one channel while maximizing the other channel. In my case it is
> I can see that for the case of CW one might use a narrow bandwidth
> perhaps 100 hz and the CW signal may then show a significant signal to
> noise ratio and the adaptive polarity will correctly maximize one channel
> while minimizing the other.
I made the algorithm for use with the other enhancements of Linrad.
For CW the bandwidth would be in the range 15 to 25 Hz. Seldomly above
18 Hz. The AFC that looks both backwards and forwards in time would
guarantee that the signal is centered in the passband even when it
is well below what can be copied even with the coherent detect.
(Equivalent to synchronous detection of AM.)
> I would be passing JT65B signals through it at
> a 600 to 1000 hz bandwidth.
That would be useless. The JT65B bandwidth is something like 400 Hz.
If you set 800 Hz you loose 3 dB. Linrad would compute power spectra
and channel correlations averaged over the time constant you have chosen.
Maybe 10 seconds. That means you would average 10000 times so the
S/N improvement would be sqrt(10000) or 20 dB. Assuming an S/N of
6 dB is needed for a reasonably accurate polarization one should
need -14 dB in 800 Hz or -17 dB in the JT65 scale at 2.4 kHz.
That is a pretty strong signal...
BUT - the main problem is that this works ONLY in case there is no
polarized interference - so forget it.
> Given that these signals are often much weaker
> than CW signals the adaptive polarity may simply peak on noise instead of
> signal? Is this correct?
> What would be nice is if adaptive polarity could
> be used on Linrad to drive WSJTx giving a simple single frequency solution
> to automatic polarity control. I wonder how Map65 does this polarity
> control? Does it simply copy the adaptive polarity control of Linrad or
> perhaps it is done differently.
It is different. Linrad has an equal sensitivity for linear and circular
polarization. For CW there is no loss associated with that. MAP65 has
a 3 dB loss for circular.
> Maybe it is done in digital and based on
> the sync signal. Is there a link detailing the workings of this? Failing
> the adaptive polarity control working for this, I need to duplicate the
> W3SZ experiment where multiple programs are run for various polarities. I
Yes. This is the only solution unless the adaptive polarization algorithm
has a decoder for each mode of interest. It is highly efficient however
and only four instances of WSJTx and one instance of Linrad would need
to be visible on your screens. The maximum polarization error would be
22.5 degrees - but you would have two largely independent channels to
detect a weak signal in one or the other so the loss would be less than
the the 0.7 dB associated with a 22.5 degree polarization mis-match.
You could also tell linrad to supply three instances of WSJTx with
polarizations H, +60 and +120 degrees. The maximum error would be 30
degrees wioth an associated loss of 1.3 dB. Somewhat less, because there would
be two channels with the same loss and tha chance to fail in both for
a marginal signal is less (probably far less) than 100%
> hope I don't have to do that. I'm not sure I have enough room for the
> monitors to keep track of all of that.
On some microwave bands people use circular....