Thinking about high frequency accuracy and channel discipline with fldigi
Hello Linuxhams, this is Lee, AG6CB. I just finished field day where I
did 5 fldigi digital contacts.
I came up with a whole bunch of speculative ideas for digital radio, but
I am going to write here with thoughts and questions about how to do
high frequency accuracy fldigi.
What I am thinking about is a way for PSK31 radios to do high accuracy
tuning and how to make a simple channel scheme to give every ham a clear
frequency if it exists.
Some of these Ideas have been worked out already. I welcome reference to
previous articles. Also, if you have got the WWV tuning feature working
well, tell me about how to do it.
1. The first thought is, I wish I could work out a way to get the radio
I am using, the soundcard I am using and fldigi adjusted using WWV so I
could receive and transmit PSK31 signals with a numerically exact
frequency. It would be really nice when working on 20M to get a 14.070
Mhz PSK31 signal accurate to 1 Hz which is 71 parts per billion. I
understand PSK31 is 31Hz wide. For example, my radio is described as
having 10 PPM tuning accuracy. That is 14Hz on 20 meters.
Because fldigi uses a soundcard with a 8-15khz bandwidth, I don't
expect that 71 ppb frequency accuracy will result in much better
filtering than is accomplished at present.
2. One of the things we could do with a high level of accuracy is
experiment with channelization of the 6.2 khz bandwidth of
a typical 20 meter 14.070 PSK usage.
Suppose we think about setting up channels with all the channels spread
out across the entire 6.2 khz SSB bandwidth. We can use an alternating
binary division scheme to spread out all the signals so that as many
signals as possible miss the harmonics created by other signals. I
understand a square wave generates odd numbered harmonics, like 1, 3,5.
We would put signal 0 at -3.1khz, signal 1 at -3.1+(3.1/2), signal 2 at
3.1-(3.1/2), signals 4,5,6 and 7 in between and so on. It is a fractal
process where the linear space is divided in 1/2, then 1/4, then 1/8
A channel scheme would divide 6.2 khz into 64 pieces with 96 Hz between
each. The first 32 pieces would have 193 Hz each.
Channels could lead to a systematic way for looking for a clear
frequency. You take your radio Grid square modulus 64. The result of the
modulus calculation is between 0 and 63. Check that channel is it free?
If no, try the next channel. As you search for a quiet channel, you hop
evenly all around the PSK31 band.
A channel scheme should also be studied by W1HKJ so it would fit well
with the existing fldigi signal decoding software.
One of the things I would like to explore as a follow-on to
channelization is whole band signal decoding and storage together with a
follow on system of recovering and continuing QSOs after a band dropout
The discussion of frequency accuracy has another problem. I see some
PSK31 stations pull a few Hertz when the station begins transmitting. I
can usually deduce that it is the distant station, not my radio, when I
am listening or watching the waterfall. I have found it very difficult
to determine if my station pulls when transmitting. The best I can come
up with is ask another station to closely watch my signal.
I bought a bunch of crystals and oscillators at an electronic surplus
store. The best I can say so far from my playing around is getting
better than 10 or 20 PPM accuracy or stability takes very careful
fiddling. The TXCO oscillators sold as an accessory for some ham radios
cost money because high accuracy and stability are not easy to achieve.
Cordially yours, Lee AG6CB
Posted by: Lee McKusick <lmckusic@...>