Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: 'Normalization' of PCM audio - subjectively benign? (Read 140719 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

'Normalization' of PCM audio - subjectively benign?

Reply #100

Compare the sound of your 'live' square wave to the one mangled by the 1 bit, over-sampled, noise-shaped processing of a typical 1-bit DAC. It WILL be different, I can guarantee it.

Where are your ABX results?
Exactly the same applies to the much more complex sound of musical instruments, especially those whose timbre is defined by an extended harmonic structure. What instrument's isn't, as a matter of fact?

Snare drum?


Snare drum? Most definitely.

edit>>  ABX results? Oh dear. Well -  disregard everything I've said.

'Normalization' of PCM audio - subjectively benign?

Reply #101
Quote
And the simple fact is that audio is data...If you hear music, then clearly sound is a form of data.
 
means that when i hear a acoustic guitar, my dog,one airplane,one pretty girl talking with me...i hear data? lol


Why don't you leave this thread to the grown ups?

Actually, if you really want to be a smart ass, we could say your ear is actually sampling the data that is being transmitted by air.  Electric signals are what are being sent to your brain.  So we have an DAC -> air <- ADC -> brain chain that is being innefficient.

But, then, I am just being stupid myself.

'Normalization' of PCM audio - subjectively benign?

Reply #102
There are many instruments (for e.g. almost any muted brass) which generate edges much faster than CD could ever hope to follow accurately.


... let's assume what you say is correct.  Even if that instrument was played back live in front of you, you couldn't hear the frequencies above about 18kHz (your an adult, right?).  That's below the 22.05kHz a CD can reproduce (well, assuming your equipment can!).

BTW, your vinyl table would not be able to follow that frequency either.

'Normalization' of PCM audio - subjectively benign?

Reply #103
facetiousness aside - how about any instrument  - Miles Davis's trumpet, Pinkas Zuckerman's violin, Murray Perahia's piano, Suzanne Vega's and Deborah Harry's voices ........

Yeh, I'm greedy, I want to hear them.


There are many instruments (for e.g. almost any muted brass) which generate edges much faster than CD could ever hope to follow accurately.


... let's assume what you say is correct.  Even if that instrument was played back live in front of you, you couldn't hear the frequencies above about 18kHz (your an adult, right?).  That's below the 22.05kHz a CD can reproduce (well, assuming your equipment can!).

BTW, your vinyl table would not be able to follow that frequency either.


Oh yes, it could.

edit >> LP is capable of (slightly wayward) response to 30KHz+. It's a whole different ball-game to CeeDee.


'Normalization' of PCM audio - subjectively benign?

Reply #105
[quote name='RockFan' post='426789' date='Sep 2 2006, 14:39']
These are the actual output of a perfectly respectable digital  tone-generator application.
[/quote]

Err, whats your point? 

[quote name='RockFan' post='426789' date='Sep 2 2006, 14:39']
They are are exactly what one one would get if a signal generator's output was captured by a 16/44 ADC.
[/quote]

Completely false.  Have you ever used an actual ADC?  I'm not talking about software for recording, but the actual hardware that will let you see the data and not an interpolated or abstracted view of it.  Try it sometime and save the PCM data to a disk.

[quote name='RockFan' post='426789' date='Sep 2 2006, 14:39']
Mathematical interpolation is what allows them to be 'reconstituted' and the aliasing removed by a typical DAC (at the expense of other aspects of fidelity).
[/quote]

This is correct.  However, the its also what your "perfectly respectable digital  tone-generator application" is doing when it displays that neat little output. 

[quote name='RockFan' post='426789' date='Sep 2 2006, 14:39']
[quote name='Mike Giacomelli' post='426774' date='Sep 2 2006, 12:57']
[quote name='RockFan' post='426724' date='Sep 2 2006, 10:26']
[quote name='kjoonlee' post='426710' date='Sep 2 2006, 08:51']
No, for people who can't hear anything above 17kHz, a change at 18kHz won't mean a thing.
[/quote]
Why do think I mentioned "roll-off"? Filters don't just 'brick-wall' over a few Hz or 100's of Hz, even digital ones. If, say, 3 dB of attenuation is seen at 18KHz, the HF roll-off will probably have started a KHz or more lower.
[/quote]
You really don't know what you're talking about.  Try FFTing the output of a sinc function on a DAC.  Even the cheap DAC will have the 3 dB point above 19.5k.  Maybe even above 20-21k.  I've done the measurements.  A typical sound card can produce essentially 100% amplitude up to around 20k, and good ones go higher then that.  Thats the whole idea of oversampling, it allows you to build DACs with extremely sharp cutoff filters very easily, and its why you can have $5 DACs with transistion bands above 20k.
[quote name='RockFan' post='426724' date='Sep 2 2006, 10:26']
But anyway, that's missing the point. Lets assume a filter that starts rolling off at 14 KHz,  is that better?
[/quote]
Why stop there?  If we're making things up, why not assume it starts at 10k?  Or 1K?  Hell lets assume there is no pass band 

Look, I'm sorry if this makes me sound like an ass, but get a decient scope and measure these things.  You're making a lot of assumptions that have no basis in reality.  If you take a few minutes to do the experiments and see for yourself what equipment is actually capable of, I think you will realize how silly this issue really is.
[/quote]
Here's a simple "thought experiment", but anyone who happens to have the hardware could do it for real.

Connect a signal generator up to a good amp and monitors, or headphones, and have a variable slope low-pass filter in line.

Configure the generator to produce, say, an 8KHz square wave and listen to it. Now start rolling HF off with your filter.

When do you start to hear a difference in the sound of the square wave???? No earlier than when the knee in the filter slope starts at 8KHz or lower -  above that it will make NO difference at all.

If you want to confirm it with measurements? Run the signal into a spectrum analyzer, or capture it (digitally, shall we say) and do the same.

The spectrogram will reveal that the the 8KHz square wave has harmonics extending in to the ultrasonic, even if practically everything above 8KHz in the actual signal has been has been rolled off.

How strange! Are we learning yet?



[/quote]

I think you confused me with someone else since that reply doesn't appear to have anything to do with my post.

[quote name='RockFan' post='426856' date='Sep 2 2006, 19:30']
facetiousness aside - how about any instrument  - Miles Davis's trumpet, Pinkas Zuckerman's violin, Murray Perahia's piano, Suzanne Vega's and Deborah Harry's voices ........

Yeh, I'm greedy, I want to hear them.

[quote name='Radetzky' post='426855' date='Sep 2 2006, 18:27']
[quote name='RockFan' post='426316' date='Sep 1 2006, 02:45']
There are many instruments (for e.g. almost any muted brass) which generate edges much faster than CD could ever hope to follow accurately.
[/quote]

... let's assume what you say is correct.  Even if that instrument was played back live in front of you, you couldn't hear the frequencies above about 18kHz (your an adult, right?).  That's below the 22.05kHz a CD can reproduce (well, assuming your equipment can!).

BTW, your vinyl table would not be able to follow that frequency either.
[/quote]

Oh yes, it could.

edit >> LP is capable of (slightly wayward) response to 30KHz+. It's a whole different ball-game to CeeDee.
[/quote]

Can you hear above 20kHz?

'Normalization' of PCM audio - subjectively benign?

Reply #106
I've only read the most recent posts in this thread, but now it seems to be very off-topic from normalization.

It's probably rehashing stuff from the FAQ, so I humbly propose a thread split or a thread lock.

'Normalization' of PCM audio - subjectively benign?

Reply #107
You just don't get it

I'm really sorry.

'Normalization' of PCM audio - subjectively benign?

Reply #108
"The spectrogram will reveal that the the 8KHz square wave has harmonics extending in to the ultrasonic, even if practically everything above 8KHz in the actual signal has been has been rolled off."

You'll get it if you really try.

 

'Normalization' of PCM audio - subjectively benign?

Reply #109
No, it's you. You just don't understand what you're criticising, so by definition, you can only resort to straw-man attacks.

I hope you enjoy your digital music as much as I do mine.

'Normalization' of PCM audio - subjectively benign?

Reply #110
These are the actual output of a perfectly respectable digital  tone-generator application.

They are are exactly what one one would get if a signal generator's output was captured by a 16/44 ADC. Mathematical interpolation is what allows them to be 'reconstituted' and the aliasing removed by a typical DAC (at the expense of other aspects of fidelity).


<sarcasm>
I really should ask for a refund of my electrical engineering courses.  Those Universities in Canada really teaches us crap.

Would you mind postulating for a position as a professor in signal processing?
</sarcasm>

You really believe this is what is being sent to your speakers?  This is what a DAC, even a cheap one, outputs?  Your cute.

I noticed many smart users stopped participating to this thread.  I think I should do the same.  This is really getting pathetic.

edit >> LP is capable of (slightly wayward) response to 30KHz+. It's a whole different ball-game to CeeDee.


I propose to add a section in the Wiki.  An all of fame for the funnies.

'Normalization' of PCM audio - subjectively benign?

Reply #111
No, it's you. You just don't understand what you're criticising, so by definition, you can only resort to straw-man attacks.

I hope you enjoy your digital music as much as I do mine.


I have very rewarding digital playback;

CD's ripped via a Plextor drive and Plextools, to Monkey's Audio, output via M-Audio DIO2448 card, Foobar2000 kernel-steaming, a video-broadcast quality coax, to an Audio Alchemy DAC , a nice Rotel RA 820BX4 amp and JPW Gold-monitors + passive sub.

Impressed? You would be if you heard it.

But true analogue is for those who make the effort.


'Normalization' of PCM audio - subjectively benign?

Reply #113
I've only read the most recent posts in this thread, but now it seems to be very off-topic from normalization.

It's probably rehashing stuff from the FAQ, so I humbly propose a thread split or a thread lock.


Yeah, but it's intersting all the same, No?


'Normalization' of PCM audio - subjectively benign?

Reply #115
These are the actual output of a perfectly respectable digital  tone-generator application.

By "output" you mean the digital signal or the visualization of it?

case: "visualization"
Mike already pointed out this is an easy way to visualize the set of sample values that has been generated. If you think it looks ugly then yes, we can all agree on it. However, it's just a simple visualization which obviously serves other purposes than to be close to what a DAC should output.

case: "digital signal"
This is probably the root of all the problems here. You think that the continues function you see plotted (stair steps) is somehow coded in your previously generated WAVE file and that the program you used to visualize the wave "just plots" it the way it's in the WAVE file. Well, you're wrong. Your digital PCM signal is just a bunch of measurements (sample points) for certain equidistant points in time. It does not dictate any visualization program to draw stair steps. The "values in-between" (between the sample points) have to be properly reconstructed (bandlimited interpolation). Even the author of the page you referenced says something about proper reconstruction (does "sin(t)/t" ring a bell?). But he seems to prefer showing stair steps for his visualizations and thus misleads the reader.

Cheers!

'Normalization' of PCM audio - subjectively benign?

Reply #116
For sure, the way samples are visualized in wave editor softwares must have misled quite a few.
It seems Adobe Audition is the only one able to show the waveform as is obtained by bandlimited interpolation rather than joining the dots, so I suggest Mr Rockfan to download some demo of this software and create various signals and see what happens. (the waveforms won't be perfect fit to squares / sawtooths - they don't need to be...- but they'll be a lot closer than what he got previously).
The reconstructed waveform is just like the original one, stripped from its high-frequency content.
A waveform is pretty, but unless you can mentally compute its fourier transform, it doesnt say much about what's actually heard by human ears.

Next step is to understand that PCM44 works by assuming the stripped content is not audible, whence the "roundering" of sharp edges in the reconstruction process is a non-issue given that assumption. It's a bit of shock to learn the ear can't tell the difference between a neat, sharp square-wave and a wavy rounded-one, but after a while one gets used to it and doesn't give the same importance to waveforms views. (just realize the difference between those 2 sounds is a sound whose energy only lies in very high frequencies, and thus unaudible).

Then arguments can only be made on :
1. validity of the assumption, i.e. whether components above 20kHz can have any effect on human ears. (and how much does the ear behave in a *linear* way ?)
2. how good hardware do that reconstruction (whether 22050 Hz sampling gives enough roll-off for cheap DACs to work with, etc..)

Both those points are valid, and can lead to interesting debates.
But for now, Rockfan is defending wrong asumptions, and no-one has been able to bring him to debate on constructive subjects.

'Normalization' of PCM audio - subjectively benign?

Reply #117

What is "the audio band"?

Most people would define it as the range of frequencies which are audible to the human ear. Typically 20Hz-20kHz for young listeners, though there are more careful definitions. What's your definition?


No sh*t! You could probably be a little more pedantic if you really tried. (you'll have to forgive the sarcasm, but I take a dim view of being patronized gratuitously )


RockFan,

Let's be clear here. You're wrong about the topic of this thread. Every engineer and interested amateur who has dropped in to try to help has told you you're wrong.

The thought experiments you have proposed are wrong, and quite bizarre. The things you've stated are self contradictory, but you don't seem to notice.

This suggests to me that there's at least one fundamental belief that you've picked up from somewhere which you believe to be true, but which isn't. I don't mean the whole "CD vs LP" thing - I mean something about human hearing, or sound, or filtering, or something.


I have come close to patronising you when I thought you were a troll on the first page of this thread. But we're way past that now.


So the post of mine that you replied to wasn't patronising at all. I looked at what you'd posted, I tried to figure out how you and I could have such a completely different view of it.

Unfortunately, you've got very angry, and haven't posted anything helpful, so we're not getting anywhere.



Quote
Band limiting or low-pass filtering does NOT, contrary to what you appear to be saying, "change the shape of the waveform" or the timbre of that recorded trumpet (or violin, or soprano voice) , but digital filtering in a 16/44 DAC certainly does.


Here's a simple "thought experiment", but anyone who happens to have the hardware could do it for real.

Connect a signal generator up to a good amp and monitors, or headphones, and have a variable slope low-pass filter in line.

Configure the generator to produce, say, an 8KHz square wave and listen to it. Now start rolling HF off with your filter.

When do you start to hear a difference in the sound of the square wave???? No earlier than when the knee in the filter slope starts at 8KHz or lower -  above that it will make NO difference at all.

If you want to confirm it with measurements? Run the signal into a spectrum analyzer, or capture it (digitally, shall we say) and do the same.

The spectrogram will reveal that the the 8KHz square wave has harmonics extending in to the ultrasonic, even if practically everything above 8KHz in the actual signal has been has been rolled off.


Quote
The somewhat misleadingly named 'digital filtering' of an D/A covnvertor is another matter.



Those three quotes suggest that you don't know what filtering is, how it's done, or any of the implications - in the analogue or digital domain.

There was a fourth quote about how accurate a digital filter can't be in the frequency domain, which was also wrong. A digital filter can do anything you care to define which doesn't violate the time/frequency equivalent of the uncertainty principle. I would put "given enough processing power", but I can make a filter to slice a single 1Hz-wide signal from a 44.1kHz sampled signal here on my PC (not that you'd want to!), so I don't think "given enough processing power" is relevant these days (at least at the level of proving things - you still wouldn't do some of this stuff in cheap commercial real-time devices yet).


Later, in the thread, you quoted yourself ...

Quote
"The spectrogram will reveal that the the 8KHz square wave has harmonics extending in to the ultrasonic, even if practically everything above 8KHz in the actual signal has been has been rolled off."

You'll get it if you really try.


The fact that you quoted this is quite disturbing! I get the feeling you think this was your finest hour, revealing the something about the true nature of digital recording - something which every audio engineer failed to grasp, but RockFan uniquely understood.

I'm almost speechless.

Isn't is possible that the hundreds of thousands of people who understand this subject are right, and you (who by your own admission don't actually understand filtering) are wrong?


What you've written above is actually "The spectrogram will reveal that the 8KHz square wave has harmonics extending in to the ultrasonic, even if there are no harmonics extending in to the ultrasonic."

You're either trying to invent a new kind of spectrogram, or invent a new kind of "filter" that doesn't "filter".

Try visiting reality - it makes a surprising amount of sense. Filers filter, and spectrograms show what is there (not what isn't). It's great!

(OK, that was patronising, but the rest of the post was serious.)

If you want to work through this, and figure out what someone has told you which has skewed your thinking so badly on this, fine. If you want to read a DSP text book and learn something, fine.

If you want to come back, say that we're all wrong, and propose incorrect thought experiments which you won't let anyone challenge or dissect, then I think you'll wear out your welcome very quickly.

Cheers,
David.

'Normalization' of PCM audio - subjectively benign?

Reply #118
RockFan, this is for you!
I can do plots, too. 
The name of this pic is "reconstruct 18 kHz"
Drop me 20 bucks and I'll make you a poster.



Magic? Not really. This approach only corresponds to a reconstruction filter whose impulse response is a good deal closer to the normalized sinc function (see Whittaker–Shannon interpolation formula) than your stair step thingy.

Of course, this isn't the only practical way to get a decent reconstruction -- but one of the better ones and IMHO well-suited for visualisation within a Wave editor.

Cheers!

'Normalization' of PCM audio - subjectively benign?

Reply #119
What I'd like to ask people here is; in their experience, is normalization completely 'benign', sonically? Are the algoritms used in different applications much the same, or are some better than others?

R.


Why ask this question, only to argue against everybody's answers received for the next five pages? If you (think you) knew the answer already, why bother asking, unless you are trolling?

IMO, people have been more than patient enough, and going out of their way trying to explain it to you.

'Normalization' of PCM audio - subjectively benign?

Reply #120
Quote
...people have been more than patient enough...

for me is good,i'm learning lots with their patience answers.
 
go ahead boys.


'Normalization' of PCM audio - subjectively benign?

Reply #122







Top Trace : 14khz tone as 44.1k 16bit PCM file. This is the view of the actual file when loaded into "Audacity".

Middle Trace : Actual analog out when above PCM file is played (foobar) without any resampling or DSP. Soundcard is onboard "soundmax" AC97 of cheap motherboard. (Photo of osciloscope image)

Bottom Tave : As above but zoomed in.

'Normalization' of PCM audio - subjectively benign?

Reply #123
Top Trace : 14khz tone as 44.1k 16bit PCM file. This is the view of the actual file when loaded into "Audacity".

Middle Trace : Actual analog out when above PCM file is played (foobar) without any resampling or DSP. Soundcard is onboard "soundmax" AC97 of cheap motherboard. (Photo of osciloscope image)

Bottom Tave : As above but zoomed in.


Magic!*

-brendan

* or is it just reality conforming to the theory... 

'Normalization' of PCM audio - subjectively benign?

Reply #124
I get stressed when I see this thread... clearly remembered those Digital Signal Processing classes which I failed the first time I took it...

Why? Because I was thinking like you, RockFan. The flawed "Connect-the-dots" point-of-view.

But when someone gently and patiently guide me... showing me things that are basically what SebastianG showed...

Ah! The light comes out!

I retake the course and scored an 'A' there...

So, please RockFan. Before commenting further, do buy some introductory and intermediate books on Digital Signal Processing instead of relying on your (flawed) point of view.