[Lf] Re: lf digest, Vol 1 #146 - 1 msg

[Vittorio De Tomasi]

Terry Fox wrote:
> 
> As we talked about at tacos last Saturday, you will probably need a lossless
> compression system.  They are generally limited to a 2:1 compression ratio at
> best.  Any other system will throw away data that is needed.  


Terry,

this probably does not work: if your input signal is close to white
noise (and VLF signals are very similar to noise!), you cannot find a
lossless compression scheme, because there is no way to reduce the data
entropy using a different coding system.

> As an alternative,
> we can reduce the samples in frequency.  Since you want a narrow bandwidth
> result,  if you resample at a lower frequency, you can drop the audio frequency
> range to close to the bandwidth.  This gives the same frequency range
> (remembering Nyquist), but much fewer samples.
> 
> For example, a 200Hz bandwidth centered around 1KHz would require at least a
> 2.2KHz sample rate.  Moving the samples down to 300Hz would only require a 800
> Hz sample rate (I think).
> 

Not exactly. If you have a signal with a bandwidth of B hertz, whose
bandwidth extends from Bo to Bo + B hertz, you can sample it using 2B
samples/second without introducing aliasing (you actually sample the
analytic representation of the signal). But to reconstruct the signal as
the original one, you need to filter the sampled sequence with a
suitable reconstruction filter. That reconstruction filter is a bandpass
with bandwidth [Bo, Bo + B], where Bo is the lowest frequency contained
in your signal. Now put Bo = 0, and the bandpass filter becomes...a
lowpass, and you get the good old Nyquist theorem described in
textbooks. If you have the famous Oppenheim Shaefer textbook on DSP at
hand, sampling methods for bandpass signals are described in the chapter
about Hilbert transform (and not all DSP textbooks unfortunately tell
this...).

> Now, apply a lossless compression of about 2:1.  That would result in 400
> samples per second.  If those samples were reduced from 16 bits to 12 or even 8,
> then more data can be sent.  While frequency compression may lose too much data,
> a non-linear amplitude compression method may be successful.
> 

Yeah, it should be successful,  and it is where we should experiment
with... we could also think about moving amplitude compression into the
AGC stage, using a DSP-controlled AGC system. Any idea ?!?

> I think we can achieve some compression on the audio data without losing data
> integrity.  If we can achieve a low enough data rate, we can also use radios to
> connect to remote receivers in addition to the Internet.  Sending the audio as
> data would be much better than just sending the audio itself.  It would not be
> distorted by the retransmission method.
> Terry
> 

We should use this method also with our FM repeaters: will we have one
day an ham-radio digital communication system similar to Ericsson's
TETRA ?!? We could save a lot of bandwidth, and use the saved radio
spectrum for something more interesting...

vy 73

Vittorio


-- 
*************************************************************************
Vittorio De Tomasi         ik2czl at amsat.org
Home page:                 http://space.tin.it/scienza/vdetomas 
My DSP page:               http://www.freeyellow.com/members/padan

"Wir muessen wissen; wir werden wissen" (David Hilbert)