Hi Jose and all, > I have not seen the file you mention, but I would take an expanded look > to the carrier frequency using Spectran before blaming Winrad. > > I have seen some severe multipath and doppler spread that might very > well confuse the PLL. I have recorded one case of some five threads > moving sort of like snakes, in a 5 Hz bandwidth, fading, appearing and > dissapearing, and with similar strenghts...where does the PLL should > lock there? If the "carrier" it is using as reference to lock, fades, > what happens? It must unlock and jump...There, the dynamics of the PLL > play an important part, and as everything in engineering, there must > exist compromise choices that do not always solve all the possible > scenatrios...unless you go to multiple choices, which might be even more > confusing to non experts. > > Multipath with doppler spread is a tough case for an ECSS detector. > An envelope detector is much "nicer" then, but with the traditional > changing audio response and distortion associated to diode envelope > detectors. Possibly, a thereshold-less detector, or a mathematical > procedure like extracting the square root of the squared value might > "behave better" in a case as the one I describe. There is another alternative which is implemented in Linrad. The frequency band where the carriers are contained is filtered out with a narrow filter. In your example it would have to be 5 Hz wide. Then the phase of whatever signals that passes the filter is used as the carrier for synchronous AM and FM with the AM modulation to one ear and the FM modulation to the other ear.
I would be interested in a recording where the ECSS detector has difficulties. Linrad currently does not provide the corresponding function with a filter to remove one sideband but it would be a fairly easy add-on in case it would give any benefit. > Comparing the ionosphere with a boiling water pot, the water - air > boundary is constantly moving, and thus, the reflection / diffraction > height is a varying one. If there are more than one reflection from > multiple reflection points, locking to reflected/difracted doppler > spread carriers with varying amplitudes is anything but trivial. The > speed of "bubbling" in the ionosphere between two ionized gas regions is > much slower than the boiling water example. > > Regarding to the LF response, I would like to have the lowest possible > frequency cutoff available, 20 Hz might be a good choice, if it is > possible to go that low. Do the BC stations really transmit anything that low (close to the carrier)? What is "lowest possible" is determined by where you want the AGC time constant. The purpose of the AGC is to remove the variations in signal amplitude due to the the interference between signals that travel different distances. Sometimes they are in phase and create a strong signal, sometimes they are out of phase and nearly cancel out. When carriers are spaced at 5 Hz, the QSB rate will be 5 Hz and to make the AGC tune up the gain fast enough during a dip it may be reasonable to allow the AGC to absorb a substantial part of the AM modulation at 20 Hz. Another thing is that the fading may be selective for the amplitude of the entire signal to vary less rapidly or not much at all. I had no responses on the Linrad AM detector so far. I do not know whether it is because starting to use Linrad is too difficult or whether Linrad does not provide any improvement over existing AM detectors. You may download linrad.exe and run it on a wideband .wav file under any Windows system. Linrad is free and open source and it compiles under Linux Windows as well as under Windows. http://www.sm5bsz.com/linuxdsp/linrad.htm 73 Leif / SM5BSZ
