Re: Why bubbling mud?

Richard Evans wrote:
DAB is the Betamax of digital radio wrote:
Richard Evans wrote:
DAB is the Betamax of digital radio wrote:
Richard Evans wrote:
I'm not certain whether mp2 converts the lower frequency sub bands
to a lower sampling rate, but it would seem odd to me if it
didn't, as that would save a lot of bits in the low frequencies.

The 1152 input samples are split up into 36 samples per subband,
and there's 32 subbands (36 x 32 = 1152), so the sampling rate for
each subband is lower, but ultimately it's the same sampling rate,
if you know what I mean.
You completely lost me on this part. I would however be interested
to know how this works. Perhaps I'll try a bit of Googleing and see
if I cn find a bit more about it.



An audio frame has 1152 input audio samples. These 1152 samples are
split up into 32 subband signals that consist of 36 samples each (32
x 36 = 1152).

OK I can see how that would work for the low frequencies, but I can't
see how the higher frequency sub bands would work with such a low
sampling rate. Do they some how make use of the information in the
lower frequency bands?


Nyquist's sampling theorem requires that the sampling frequency be twice the
bandwidth, NOT twice the highest frequency present. You see the sampling
theorem defined as being twice the highest frequency present all the time in
books and so on, but it's not technically correct.

For example, take an FM radio station with a bandwidth of 300 kHz
transmitting at 100 MHz. If you sampled at twice the highest frequency then
the sampling frequency would have to be 200 MHz, but the sampling theorem
actually only requires you to sample at 600 kHz.


Actually there is a method I was thinking about the other day, perhaps
they do something similar.

The idea I was thinking about was that the lowest frequency band could
be generated by taking only some of the samples, (in this case it
would be every 32nd sample). The next band up could then take this
value and and one difference value to each sample from the lowest
band, hence adding one extra sample for each sample in the lowest
band. Then each band would simply add a difference value for each
sample in the previous band.

Is it anything similar to that?


But even a difference value would need to be encoded, so by the sounds of
what you've described there would be more than 1152 samples altogether - if
you count difference values as samples.

The thing about the PQMF filter that's used to generate the 32 subband
signals is that the inverse filter perfectly recreates the original signal,
so there's no point in generating new samples or difference values that
aren't needed, because they'd have to be encoded so the bit rate would
increase unnecessarily.


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/freeview/freeview_receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php


.

Relevant Pages

  • Re: Amplitude response of a Biquad Filter
    ... I put a gain of +12 dB for all the bands. ... sampling frequency of the system. ... I am experiencing even in the floating point implementation. ... low band with +12 dB boost that results in +9 dB instead? ...
    (comp.dsp)
  • Re: The Obesification of America
    ... band beyond description wrote: ... check out this sampling of southern delight served at a place called ... chili,cheese and onions and served on a hoagie bun, topped with a fried egg and two fistfuls of fries." ...
    (rec.music.gdead)
  • Re: The Obesification of America
    ... band beyond description wrote: ... check out this sampling of southern delight served at a place called ... to quote the article: ... egg and two fistfuls of fries." ...
    (rec.music.gdead)
  • Re: bandpass sampling
    ... Is it the center of your band of interest for example? ... to visulaize the signal after sampling.As after sampling my information ... should lie in first Nyquist zone ie between 0 to Fs/2. ...
    (comp.dsp)
  • Re: blind demodulation - sound card - lucent winmodem
    ... A higher sampling rate won't do you too much good, in fact, it's the ... > This list is provided by the SecurityFocus Security Intelligence Alert ... For more information on SecurityFocus' SIA service which ...
    (Pen-Test)