On 2/25/2015 6:05 AM, Roger Hayter wrote:
Jerry Stuckle wrote:

On 2/24/2015 7:03 PM, rickman wrote:
On 2/24/2015 6:37 PM, Jerry Stuckle wrote:
On 2/24/2015 5:47 PM, rickman wrote:
On 2/24/2015 12:00 PM, Jerry Stuckle wrote:
On 2/24/2015 11:32 AM, FranK Turner-Smith G3VKI wrote:
"AndyW" wrote in message
...
On 24/02/2015 12:47, gareth wrote:
What is the point of digital voice when there are already AM, SSB
and FM for those who want to appear indistinguishable from CBers?

Perhaps it is cynicism from the manufacturers who introduce such
things
as they see their traditional highly-priced corner of the market
being wiped away by SDR technologies?

Bandwidth reduction for one.
If you can encode and compress speech sufficiently then you can use
less bandwidth in transmission.

That's the bit I have trouble getting my head around. Back in the
1970s
and 1980s digital transmissions used a much greater bandwidth than
their
analogue equivalents. Sampling at 2.2 x max frequency x number of bits
plus housekeeping bits etc. etc.
A UK standard 625 line PAL video transmission would have used a
bandwidth of over 400MHz!
Times have changed and left me behind, but I've still got me beer
so who
cares?

But you forget compression. For instance, unless there is a scene
change, the vast majority of a television picture does not change from
frame to frame. Even if the camera moves, the picture shifts but
doesn't change all that much. Why waste all of that bandwidth
resending
information the receiver already has?

And voice isn't continuous; it has lots of pauses. Some are very
noticeable, while others are so short we don't consciously hear them,
but they are there.

And once you've compressed everything you can out of the original
signal, you can do bit compression, similar to zipping a file for
sending.

There are lots of ways to compress a signal before sending it
digitally.
About the only one which can't be compressed is pure white noise -
which, of course, is only a concept (nothing is "pure").

I think that depends on what you mean by "pure". Sounds very
non-technical to me. Even noise can be compressed since if it is truly
noise, you don't need to send the data, just send the one bit that says
there is no signal, just noise. lol


Pure white noise is a random distribution of signal across the entire
spectrum, with an equal distribution of frequencies over time. Like a
pure resistor or capacitor, it doesn't exist. But the noise IS the
signal. To recreate the noise, you have to sample the signal and
transmit it. However, since it is completely random, by definition no
compression is possible.

Why does it not "exist"? That is not at all clear. You don't
understand compression. Compression is a means of removing the part of
a signal that is unimportant and sending only the part that is
important. In most cases of "pure" noise, you can just send a statement
that the signal is "noise" without caring about the exact voltages over
time. So, yes, even noise can be compressed depending on your
requirements.


Pure white noise is a concept only. There is no perfect white noise
source, just as there is no pure resistor or capacitor.

And yes, I do understand compression. One of the things it depends on
is predictability and repeatability of the incoming signal. That does
not exist with white noise. The fact you don't understand that pure
white noise is only a concept and cannot exist in the real world shows
your lack of understanding.

Some compression algorithms (i.e. mp3) remove what they consider is
"unimportant". However, the result after decompressing is a poor
recreation of the original signal.

But for perfect recreation, nothing is "unimportant". Voice/video
compression is no different than file compression on a computer. Can
you imaging what would happen if your favorite program was not perfectly
recreated?


A friend worked in sonar where the data was collected on ships and
transmitted via satellite to shore for signal processing rather than
doing any compression on the data and sending the useful info. As the
signal was nearly all "noise" trying to do any compression on it, even
the aspects that weren't "pure" white noise, would potentially have
masked the signals. Sonar is all about pulling the signal out of the
noise.


You mean the signal can't be compressed? No way. Any non-random signal
can be compressed to some extent. How much depends on the signal and
the amount of processing power required to compress it. However, in
your example, the processing power to compress the signal would probably
have been greater than that required to process the original signal. So
if there wasn't enough power to process the signal on the ship, there
wouldn't be enough power to compress the near-white noise signal, either.

You really like your all encompassing assumptions. No, all signals can
not be compressed, even non-noise signals can't be compressed if the
signal is not appropriate for the compressor. This is really a very
large topic and I think you are used to dealing with the special cases
without understanding the general case.


Which is just the opposite of what you claimed above. Please make up
your mind.

Try visiting comp.compression and offering them your opinions. There
are many there who are happy to explain the details to you.


I understand the details, thank you. Much better than you do,
obviously. But that's not surprising, either.

You are both talking at cross-purposes. One of you is talking of taking
a sample of white noise and storing it as data. Because of its
statistical properties I would not be surprised if it were impossible to
compress. The other is assuming that by definition noise is not data
and compression would only be usefully applied to a hypothetical signal
added to the white noise, when no properties of the noise would be
relevant for the compressed signal.

I can't think why one should want to record and store a sample of white
noise, but that does not prevent it being used as a hypothetical
example.

I doubt you really have any disagreement, just a misunderstanding.

No, it is a fundamental issue in compression theory. *Any* signal can
be compressed if you use the right compressor. Likewise there is *no*
compressor that will compress every signal. They call this the counting
theorem. Using N bits you can represent 2^N possible signals.
Compression by definition uses a smaller number of bits, say M, to
represent the data. There will only be *some* of the possible input
combinations from the N set that can be represented by the M set. The
remaining combinations (2^N - 2^M) will require *more* bits to represent
them.

Conventional compression algorithms take advantage of redundancy in the
input signal to represent them with fewer bits, usually a lot fewer
bits. But by the same token there are the 2^N - 2^M possible signals
that these compressors will not compress and will either not reproduce
the input exactly or will require extra bits.

There was nothing in the above that says anything about which bit
patterns can be compressed or not compressed. Some people get confused
about the fact that most compression algorithms work on removing
redundancy and think that is the only way to compress a signal. When
discussing the theoretical we need to distinguish the things that are
possible from the things that are useful.

Then there is the side discussion of what white noise is and if it is a
concept or possible. A rather pointless discussion in the context of
compression, but there it is.

--

Rick