Ask questions to Audio insider Sir Terrence

[Sir Terrence]

Originally Posted by dialog_gvf

But, it isn't ONE wave, it's a mix of N different simultaneous sine waves.

Consider this:

1 (48Khz): 10000 0 10000 0 10000 0 10000 0 ... (pure 24Khz tone)

2: (98 Khz): 10000 1000 0 2000 10000 3000 0 4000 10000 5000 0 6000 10000 7000 0 ...

Contrived, of course, but it illustrates that information can be lost.

Gary

I am not so sure I would cry if there were losses at 24khz. My dog probably would though

Greg, wouldn't a 24khz pure tone hit the anti-aliasing filter at about 24khz with a 48khz sample rate? It would have to in order to satisfy the Nyquist/shannon theory. And wouldn't the signal be highly distorted if the filters didn't kick in at 23.050khz? If you recorded in 96khz, and downsampled to 48khz, you still wouldn't get any losses, because again, you wouldn't hear the untruncated data anyway.(that is the frequencies from 24khz to 48khz)

I think you mean 96khz instead of 98khz.

[Simplayer]

Originally Posted by dialog_gvf

But, it isn't ONE wave, it's a mix of N different simultaneous sine waves.

Why does that matter? We can use the same points to define multiple equations. Anyways, the theorem states:

If a function f(t) contains no frequencies higher than W cps, it is completely determined by giving its ordinates at a series of points spaced 1/(2W) seconds apart.

The analog signals that we're analyzing may have frequencies above double our sampling rate, but we can't hear those anyways, so that won't matter.

Consider this:

1 (48Khz): 10000 0 10000 0 10000 0 10000 0 ... (pure 24Khz tone)

2: (98 Khz): 10000 1000 0 2000 10000 3000 0 4000 10000 5000 0 6000 10000 7000 0 ...

Contrived, of course, but it illustrates that information can be lost.

Gary

I'm not really sure what that's supposed to mean. Can you explain?

[Sir Terrence]

Originally Posted by Simplayer

Why does that matter? We can use the same points to define multiple equations. Anyways, the theorem states:

If a function f(t) contains no frequencies higher than W cps, it is completely determined by giving its ordinates at a series of points spaced 1/(2W) seconds apart.

The analog signals that we're analyzing may have frequencies above double our sampling rate, but we can't hear those anyways, so that won't matter.


I'm not really sure what that's supposed to mean. Can you explain?

He is using binary numeric language(which is what a laser would see)to describe a waveform.

[dialog_gvf]

Originally Posted by Sir Terrence

I am not so sure I would cry if there were losses at 24khz. My dog probably would though

Greg, wouldn't a 24khz pure tone hit the anti-aliasing filter at about 24khz with a 48khz sample rate? It would have to in order to satisfy the Nyquist/shannon theory. And wouldn't the signal be highly distorted if the filters didn't kick in at 23.050khz? If you recorded in 96khz, and downsampled to 48khz, you still wouldn't get any losses, because again, you wouldn't hear the untruncated data anyway.(that is the frequencies from 24khz to 48khz)

You're right, you really wouldn't care about 24Khz. But, it was meant to be illustrative.

People are claiming 96Khz offers SOMETHING, and to mean it is in the highs from 24Khz-48Khz doesn't make any logical sense (to me). So, it must have something to do with there being some loss in the lower frequency information at 48Khz that is captured at 96Khz.

Gary

[Sir Terrence]

Originally Posted by dialog_gvf

You're right, you really wouldn't care about 24Khz. But, it was meant to be illustrative.

People are claiming 96Khz offers SOMETHING, and to mean it is in the highs from 24Khz-48Khz doesn't make any logical sense (to me). So, it must have something to do with there being some loss in the lower frequency information at 48Khz that is captured at 96Khz.

Gary

I am going to offer this up to you, and lets see if you bite. That SOMETHING really has nothing to do with the sample rate, but more to do with the filter response of the sample rate. The further you move the filter away from the highest point we actually hear, the better the audio will sound. The closer you move the filter into our hearing threshold, the worse the audio will sound, because the filters side effects are more easily heard. If the effects of the filter are only audible at 48khz, this is more than one octave above our hearing threshold, so imaging will improve over 48khz (who's filter is much closer to our threshold at 24khz) the timbre will be more accurate, and the signal will sound cleaner overall. This is why oversampling works so well when the algorythms for it are well designed.

So its not the sample rate specifically, its the audiblility(or lack of) of the side effects of the anti-aliasing filters.

[dialog_gvf]

I'm not sure we aren't talking about the very same thing, from different perspectives. Consider:

Imagine a 1 Khz tone.
Now, apply a 48Khz amplitude modulation (that is, the 1Khz sine wave would appear to be made up of tiny modulating wave)

That would be a simple example of a specific timbre of a 1Khz tone, would it not?

How do you encode it properly at 48Khz?

The problem is the aliasing of the 48Khz modulation in the 48Khz sampling, is it not?

Gary

[Sir Terrence]

Originally Posted by dialog_gvf

I'm not sure we aren't talking about the very same thing, from different perspectives. Consider:

Imagine a 1 Khz tone.
Now, apply a 48Khz amplitude modulation (that is, the 1Khz sine wave would appear to be made up of tiny modulating wave)

That would be a simple example of a specific timbre of a 1Khz tone, would it not?

How do you encode it properly at 48Khz?

The problem is the aliasing of the 48Khz modulation in the 48Khz sampling, is it not?

Gary

Gary, I am not understanding your example. A 48khz sampling would imply a frequency limit of 24khz. To prevent aliasing, you need to have a filter that stops all output above 24khz. So there is no signal at 48khz to modulate, it would have to be filtered out or severe aliasing would occur. Your example is conflicting with the nyquist theory.

[AlexBC]

Sir Terrence,

thank you very much for all you thoughfull insights.

I have a question regarding the relationship between sampling rates and bitrates.

For DTHD and DTS-HD MA tracks, what would be the average and peak bitrate for 48/16, 48/20 and 48/24 tracks?

[Sir Terrence]

Originally Posted by AlexBC

Sir Terrence,

thank you very much for all you thoughfull insights.

I have a question regarding the relationship between sampling rates and bitrates.

For DTHD and DTS-HD MA tracks, what would be the average and peak bitrate for 48/16, 48/20 and 48/24 tracks?

Since the bitrate would be variable from moment to moment, this is an impossible question to answer. With variable bitrate codecs the bitrate is determined squarely on how many bits it takes to represent the signal tranparently from second to second. The more demanding a soundtrack is, the more bits required. In the absence of sound effects and music, and with just dialog the bitrate drops substantially. This is unlike the lossy DD and Dts which use a constant bitrate and code and allocate what bits are available in that steady pool.

[Teazle]

Sir Terrence,

Thanks for your fascinating and informative posts on this forum. Picking up the thread re: multichannel audio as it promises to take over from stereo, I have a question about 5.1 vs. 7.1 surround.

People say that a true 7.1 recording played on a 7.1 setup is more immersive, more “surround-sounding” etc. than the same recording downmixed to 5.1 and played back on a 5.1 setup. Is there any way to quantify the difference to make such claims precise?

After all, 7.1 is 33% more channels than 5.1. But it seems doubtful to me that the “surroundedness” of 7.1 is 33% better. Is there some objective / scientific way to settle the matter? E.g. is there a way to measure (given some ideal speaker setup and listener position) the surround radius apparent to the listener, so it could be checked just how much wider a surround experience 7.1 offers? With 7.1 you can position a sound more precisely (since you can stick a sound exactly at x o’clock if a speaker happens to be there and the more speakers you’ve got the more likely it is that a speaker will be at x o’clock) but how much more accurately can the sound-field be manipulated?

No doubt 7.1 is better than 5.1; I’m just curious whether as far as surround sound effects go, 5.1 represents a point of diminishing returns.

[Big Daddy]

Originally Posted by AlexBC

Sir Terrence,

thank you very much for all you thoughfull insights.

I have a question regarding the relationship between sampling rates and bitrates.

For DTHD and DTS-HD MA tracks, what would be the average and peak bitrate for 48/16, 48/20 and 48/24 tracks?

The maximum uncompressed bit rates for a movie soundtrack are approximately:

48,000(samples per second) x 16(bits per sample) x 6(channels) = 4.6Mbps
48,000(samples per second) x 20(bits per sample) x 6(channels) = 5.8Mbps
48,000(samples per second) x 24(bits per sample) x 6(channels) = 6.9Mbps
48,000(samples per second) x 16(bits per sample) x 8(channels) = 6.1Mbps
48,000(samples per second) x 20(bits per sample) x 8(channels) = 7.7Mbps
48,000(samples per second) x 24(bits per sample) x 8(channels) = 9.2Mbps

Dolby TrueHD and DTS HD MA can go up to
96,000(samples per second) x 24(bits per sample) x 8(channels) = 18.4Mbps

However, Sir Terrence is correct. Since both CODECs use variable bit rates, we cannot calculate the average bit rate of a typical soundtrack. In addition, please note that Dolby TrueHD and DTS MA use different compression algorithms and on the average use less than the maximum numbers.

[lgans316]

You forgot to mention this

48,000(sample per second) x 16(bits per sample) x 8(channels) = 6.1Mbps

[Big Daddy]

Originally Posted by lgans316

You forgot to mention this

48,000(sample per second) x 16(bits per sample) x 8(channels) = 6.1Mbps

Thank you. I fixed it.

[HDJK]

Originally Posted by Sir Terrence

...

Lavry makes the best digital equipment I have ever heard in 20 years of recording audio. When that man speaks, this audio engineer listens.....very carefully.

I have read so many papers on the topic I don't remember where I read it, but I think it was Dan who said that a converter who can do 192k is compromised at 96k, so a comparison between 192k and 96k on the same converter would not be scientific; another reason why he only builds 96k converters.

Is this true for any converter out there? Or does this only apply to certain models that are 'optimized' for 192k?

Do you have any experience with the LavryBlack AD10/DA10's? How would they compare to a RME Fireface 800? Sadly they are the only Lavrys I could afford in the near future.

[dialog_gvf]

Originally Posted by Sir Terrence

This is unlike the lossy DD and Dts which use a constant bitrate and code and allocate what bits are available in that steady pool.

And if the entire surround is complicated, do the lossy codecs rob bits from the rears to feed the front?

[Sir Terrence]

Originally Posted by dialog_gvf

And if the entire surround is complicated, do the lossy codecs rob bits from the rears to feed the front?

Dolby digital does, Dts does not. Its called global bit pooling.

[Sir Terrence]

Originally Posted by Teazle

Sir Terrence,

Thanks for your fascinating and informative posts on this forum. Picking up the thread re: multichannel audio as it promises to take over from stereo, I have a question about 5.1 vs. 7.1 surround.

People say that a true 7.1 recording played on a 7.1 setup is more immersive, more “surround-sounding” etc. than the same recording downmixed to 5.1 and played back on a 5.1 setup. Is there any way to quantify the difference to make such claims precise?

I think the only way to do it is to install it in a room, and let folks listen to both and decide. It is all room dependent. Large rooms benefit more from 7.1 than small rooms do. My experience with 7.1 and small rooms points me to the impression that its benefits are pretty limited, and can easily be mimicked with properly placed surrounds, and equidistant seating in between them. As the room gets larger, you begin to hear the benefits of 7.1 much more clearer.

After all, 7.1 is 33% more channels than 5.1. But it seems doubtful to me that the “surroundedness” of 7.1 is 33% better. Is there some objective / scientific way to settle the matter? E.g. is there a way to measure (given some ideal speaker setup and listener position) the surround radius apparent to the listener, so it could be checked just how much wider a surround experience 7.1 offers?

I am sure this can be done, it just hasn't because it is so subjective. Using someone perception is always a subjective matter.

With 7.1 you can position a sound more precisely (since you can stick a sound exactly at x o’clock if a speaker happens to be there and the more speakers you’ve got the more likely it is that a speaker will be at x o’clock) but how much more accurately can the sound-field be manipulated?

Well, there is always phantom information in between speakers as well.

No doubt 7.1 is better than 5.1; I’m just curious whether as far as surround sound effects go, 5.1 represents a point of diminishing returns.

In very small room, anything beyond 5.1 does yield diminishing returns. 7.1 sounds crowded in small rooms, as you need some seperation of the speakers to obtain the clarity that 7.1 can give.

If you are speaking from a mixing perspective, it is always better to have a speaker to pan to, rather than relying on phantom imaging between speakers. You are more sure of the results from room to room with 7.1 played by through a 7.1 system. Results from a 7.1 mixdown to 5.1 will be variable, and in some cases imaging could break down because of the way the 5.1 system is set up.

[hanser]

Sir Terrence,

could you give a rough number where you would locate the border between "small room" and "large room"?

[Sir Terrence]

Originally Posted by hanser

Sir Terrence,

could you give a rough number where you would locate the border between "small room" and "large room"?

Using one of my viewing rooms as a reference, I would estimate a room that is 12ft wide and less would benefit less from 7.1. 6.1 with a single center rear would probably work well, but two center rear speakers would just turn to smush. This is just my opinion, and my experience in one of my viewing rooms.

[lgans316]

Please excuse me if the below had already been asked or answered.

I would like to know if there will be any difference between lossless audio bitstreamed to AVR as-is vs lossless audio internally decoded and bitstreamed to AVR as LPCM by the PS3 ? I am sure there will be some kind of manipulations being enforced during the internal decoding but I would like to know if it will impact the overall sound quality ?