MQA alternative - open source
 

On 06/09/2016 09:24, Jim Lesurf wrote:
I've tended to use 96k/24 and then leave it as such after declicking. Saves
the bother of then resampling, and pushes any DAC reconstruction problems
up well above 20kHz. As things stand I have plently of disc space. But I
realise this is wasteful.

I'm not sure I really believe in Shannon; while it's theoretically
possible to sample a 20kHz wave at 44.1, and reproduce it, it's damn
hard at anything over Fs/3.

Sadly this no longer matters :( I've reached that time of life where my
hifi is better than my ears. 44.1 is enough.

Andy

MQA alternative - open source
 

On 07-09-16 22:28, Vir Campestris wrote:

I'm not sure I really believe in Shannon; while it's theoretically
possible to sample a 20kHz wave at 44.1, and reproduce it, it's damn
hard at anything over Fs/3.

It is one thing to say "While Shannon is true in theory, it is tricky
to make very steep filters, so in practice you need a bit of margin",
and another to state "I am not sure I believe in Shannon". It is kind
of like saying "a falling object is affected by air resistance, so
I am not sure I believe in Newton's laws of gravity".

Fs/3 is a rather conservative measure with modern knowledge about
filter design.

MQA alternative - open source
 

In article , Johan Helsingius
wrote:
On 07-09-16 22:28, Vir Campestris wrote:

I'm not sure I really believe in Shannon; while it's theoretically
possible to sample a 20kHz wave at 44.1, and reproduce it, it's damn
hard at anything over Fs/3.

It is one thing to say "While Shannon is true in theory, it is tricky to
make very steep filters, so in practice you need a bit of margin", and
another to state "I am not sure I believe in Shannon". It is kind of
like saying "a falling object is affected by air resistance, so I am not
sure I believe in Newton's laws of gravity".

Fs/3 is a rather conservative measure with modern knowledge about filter
design.

Agreed. Even early generations of CD players could produce reasonable 20kHz
sineusoids and waveforms that had components up to that. And the better the
recording and reconstruction filtering, the closer you can get to the
nominal ideal.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

MQA alternative - open source
 

On Wed, 7 Sep 2016 21:28:30 +0100, Vir Campestris
wrote:

On 06/09/2016 09:24, Jim Lesurf wrote:
I've tended to use 96k/24 and then leave it as such after declicking. Saves
the bother of then resampling, and pushes any DAC reconstruction problems
up well above 20kHz. As things stand I have plently of disc space. But I
realise this is wasteful.

I'm not sure I really believe in Shannon; while it's theoretically
possible to sample a 20kHz wave at 44.1, and reproduce it, it's damn
hard at anything over Fs/3.

Sadly this no longer matters :( I've reached that time of life where my
hifi is better than my ears. 44.1 is enough.

Andy

You don't sample at 44.1kHz, you oversample; 16 x would be typical. In
other words about 700kHz. It is trivially easy to make a filter that
gives you adequate rejection at that frequency, so aliasing is not an
issue.

Having done that, you filter digitally with a multiple tap filter,
which is easy to design, and gives you all the rolloff you need
between 20kHz and 22.05kHz.

Finally you decimate down to the wanted 44.1kHz sampling rate, still
with a perfect signal.

Playback is the reverse of this process. You recreate the intermediate
values at your oversampling rate with a reconstruction filter, pop it
out through the DAC and you have the same trivial lowpass filter to
give you your final analogue output.

As for Shannon. His equation is correct - it has worked for years, and
is showing no sign of giving up yet.

d

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

MQA alternative - open source
 

On 08/09/2016 13:11, Don Pearce wrote:
You don't sample at 44.1kHz, you oversample; 16 x would be typical. In
other words about 700kHz. It is trivially easy to make a filter that
gives you adequate rejection at that frequency, so aliasing is not an
issue.

I read Jim's web page. Turns out there's a lot of high frequency mush in
supposedly high quality signals. It didn't get filtered out, which is
why "bit freezing" works.

snip
As for Shannon. His equation is correct - it has worked for years, and
is showing no sign of giving up yet.

OK, I handle digital data and I don't really have a problem there. It's
just when someone throws a 22kHZ signal at a 44.1kHz sampler I can't
really see how you don't get some kind of aliasing. The signal and the
sample rate will beat, just like a moiré fringe.

A quick rummage in my tracks seems to show brick wall at 16kHz.

Andy

MQA alternative - open source
 

On Thu, 8 Sep 2016 22:00:54 +0100, Vir Campestris
wrote:

On 08/09/2016 13:11, Don Pearce wrote:
You don't sample at 44.1kHz, you oversample; 16 x would be typical. In
other words about 700kHz. It is trivially easy to make a filter that
gives you adequate rejection at that frequency, so aliasing is not an
issue.

I read Jim's web page. Turns out there's a lot of high frequency mush in
supposedly high quality signals. It didn't get filtered out, which is
why "bit freezing" works.

snip
As for Shannon. His equation is correct - it has worked for years, and
is showing no sign of giving up yet.

OK, I handle digital data and I don't really have a problem there. It's
just when someone throws a 22kHZ signal at a 44.1kHz sampler I can't
really see how you don't get some kind of aliasing. The signal and the
sample rate will beat, just like a moiré fringe.

A quick rummage in my tracks seems to show brick wall at 16kHz.

Andy

Brickwall at 16kHz usually means lossy compression - MP3 or somesuch.
Limiting the top end this way makes compressing the rest a much easier
job.

d

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

MQA alternative - open source
 

In article , Vir
Campestris wrote:
On 08/09/2016 13:11, Don Pearce wrote:
You don't sample at 44.1kHz, you oversample; 16 x would be typical. In
other words about 700kHz. It is trivially easy to make a filter that
gives you adequate rejection at that frequency, so aliasing is not an
issue.

I read Jim's web page. Turns out there's a lot of high frequency mush in
supposedly high quality signals. It didn't get filtered out, which is
why "bit freezing" works.

The frequency distribution of the 'mush' is largely irrelevant. It is the
total level of it summed over all frequencies that matters. So this simply
isn't really relevant to any argument about if 44.1k sample rate is
sufficient to carry 20kHz components accurately.


That the 'mush' level has higher spectral density at HF just tells you
something about the recording/ADC systems employed. But in general, the
old maxim rules: The wider you open the window, the more muck blows in!

However as above all that really matters for bitfreezing is the total
level of the 'sea of noise'.

The above *does* however show that starting off from using systems like
DSD is actually bad news as it inherently has a poor SNR when considered
wideband. So it is actually a lousy basis for being converted and used
wideband IMHO. But DSD and low-bit ADCs are often used for making the
source material that may then end up being sold as high rez lpcm.


snip
As for Shannon. His equation is correct - it has worked for years, and
is showing no sign of giving up yet.

OK, I handle digital data and I don't really have a problem there. It's
just when someone throws a 22kHZ signal at a 44.1kHz sampler I can't
really see how you don't get some kind of aliasing. The signal and the
sample rate will beat, just like a moiré fringe.

You started off talking about 20k, now you have moved camp to 22k. But the
principle is the same.

Shannon and the relayed maths shows us quite clearly that a series of
samples at 44.1k can represent signal frequency compoents up to a 'tad
under' 22.05k.

The meaning of 'tad under' is defined by the duration of the recording
(number of samples). So a recording lasting, say, a few minutes, is quite
capable of conveying components of 22k.

A quick rummage in my tracks seems to show brick wall at 16kHz.

Which tells you about how the people creating those tracks recorded them.
It doesn't tell you about what 44.1k can do.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

MQA alternative - open source
 

In article ,
Vir Campestris wrote:
A quick rummage in my tracks seems to show brick wall at 16kHz.

Just for info, 16kHz was regarded by the BBC as the upper limit needed for
the highest quality radio transmissions.

On any true programme material, there was no useful information above this.

--
*I finally got my head together, now my body is falling apart.

Dave Plowman London SW
To e-mail, change noise into sound.

MQA alternative - open source
 

On Fri, 09 Sep 2016 13:32:50 +0100, "Dave Plowman (News)"
wrote:

In article ,
Vir Campestris wrote:
A quick rummage in my tracks seems to show brick wall at 16kHz.

Just for info, 16kHz was regarded by the BBC as the upper limit needed for
the highest quality radio transmissions.

On any true programme material, there was no useful information above this.

And of course they had to leave room for an analogue notch filter to
take out the 19kHz stereo pilot tone. But of course the distribution
standard to all the BBC FM transmitters is NICAM, which is sampled at
32kHz, so 16kHz is very much the upper limit of possibility.

d

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

MQA alternative - open source
 

In article , Don Pearce
wrote:

And of course they had to leave room for an analogue notch filter to
take out the 19kHz stereo pilot tone. But of course the distribution
standard to all the BBC FM transmitters is NICAM, which is sampled at
32kHz, so 16kHz is very much the upper limit of possibility.

Yes. Although I think I recall reading that in the early stereo days when
only Wrotham was occasionally used for stereo on Radio 3 they had to fit
filters for the first time to fix some problems. Before that they hadn't
explicltly added them. Probably relied on the feeds not to have much above
about 15kHz and the - then analogue transmission lines to TX - not actually
carrying it!

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html