In article , DAB sounds worse than
FM wrote:
Jim Lesurf wrote:
In article , DAB sounds worse
than FM wrote:

If you want to compare results with me then do the following:

Erm... what OS, etc, are you assuming I am using? :-)

* download Lame v3.90.3:

[snip]


You would need to direct me at a version that runs under RO on an
ARM-core CPU and which can be verified to act identically to the one
you use. However even if that were done and I had the time to try it...


Okau, I'll just tell you the results I've had when I've done this in the
past. I recorded Radio 1 and Radio 3 off DTT and re-compressed
(transcoded) to VBR MP3 at a given quality and the classical music on R3
required a lower bit rate than R1.

Afraid I don't know what definition you are using for 'quality' in the
above in terms of the actual amount of info discarded, and the judgement
rules used to discard components or increase their quantisation. Hence I'm
not clear how it would relate specifically to what the BBC do for R1/2/3
and the previous statements you have made.

Also, yesterday I encoded Morrissey's and Snow Patrol's most recent
albums, the latter because it's got quite a bit of the loud electric
guitars I mentioned, and the former because I've been playing about with
the new HE AAC codec recently and it suffers badly when starved of
bits. Both albums required a bit rate significantly higher than the
norm for that VBR quality setting.

What basis apart from your subjective judgement are you using for the
comments about "suffering" and "required a bit rate"?...

It isn't clear to me what value the process you suggest would have
w.r.t the issues we were discussing in terms of the differences
between what the BBC do for R1/2 and R3.


If you had paid attention,

sigh More irrelevant personal comments...

I described a way of comparing difficulty of encoding by comparing NMR
(noise to mask ratio), and encoding using VBR for a given quality
setting will try to set the NMR to be relatively constant -- MP3
encoders iteratively allocate bits to minimise the NMR (or equivalently,
maximise the mask to noise ratio).

But you seem not to have explained how this establishes your assertions
about R/1/2/3 specifically are correct.

If you wish to support your argument for that then I'd assume you need
to do so in terms of the specifics of the BBC signals and waveforms.


To prove this you don't need BBC signals or waveforms, you just have to
show that flatter spectra are more difficult to encode.

Again, this depends on the specific definition of 'flatter' and 'more
difficult'. As I think I pointed out in previous postings.

The snag being that to do so you ideally may need access to the
originals before they were level compressed and data reduced.


No, I wouldn't need access to the originals.


AIUI your point was based on argueing that the R1/R2 typical signals
have a more unform and flatter spectrum. My point was that it may also
depend on the number of spectral components, not just the uniformity
and range of those present.


A totally flat and very wide spectrum is bound to end up with more
frequency components post-masking than one that tails off far faster and
isn't as broad. It's an assumption, but a very, very good one. I'm sure
you've used far weaker assumptions before than the one I'm using.....

I would agree if by "totally flat" you mean much the same level at *every*
frequency in the spectrum for the relevant time frame. However the point I
was asking about was how you know this is the case for the R1/2 signals
before they are data reduced. Again, I was trying to distingish this from
the orginal having a number of components of a given, similar level, but
spread across the band with other components at lower levels. So far as I
can tell, you have not so far dealt with this distinction in your replies.
My apologies if I have missed it. If so, please indicate the relevant
posting and I'll re-read it if it is still in scope of my newagent's time
limit.

Without access to the BBC original waveforms and their level
compression, etc, how would my attempting to use Lame establish this?


See above.

See above. :-)

I'm going away for the weekend, so I won't be able to reply to any
further posts until I'm back.

OK.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html
Barbirolli Soc. http://www.st-and.demon.co.uk/JBSoc/JBSoc.html