SACD/DVD-A vs. Oversampling

[Buckle-meister]

Dear all,

Here is a question my curious mind has come up with, which some of you may know the answer to:

In theory, would music be more accurately represented (thereby sounding better ) with SACD's/DVD-A's high sampling rate, or by regular CD's which were then oversampled? In other words, which format has the higher sample rate?, or is it not quite as simple as this?

Regards

 

[sploo]

Buckle-meister,

There are many others on this forum with a far better grasp of the subject than I, but, here goes:

CDs are recorded at 44.1kHz, with each sample having 16bits of resolution. According to that nice Mr Nyquist (http://mathworld.wolfram.com/NyquistFrequency.html) the highest frequency that a CD can therefore store and reproduce is 22.05kHz (though I believe there's a lot more to it than that).

DVD-A can have many different sampling rates, up to 192kHz, with more than two channels, and at sample sizes of 16-24bits.

SACD uses a 1bit system with a sample rate in the MHz range (I don't pretend to understand it fully).

For more info, look here: http://www.audioholics.com/techtips/specsformats/SACDvsDVD-AudiovsCD.php

As for oversampling, info here: http://www.audioholics.com/techtips/specsformats/upsamplingvsoversampling.php

Concentrating purely on stereo sound, I'd argue that the recording, mixing and mastering of an album has more affect on the way it sounds than the difference between the formats. That's not to say any one particular format isn't theoretically better than the others (and there's plenty of argument on this too).

I'm not aware of any blind testing that has been performed with supposedly identical recordings on CD, DVD-A and SACD to determine which format is preferred. Anyone know of any?

 

[mtrycrafts]

Dear all,

Here is a question my curious mind has come up with, which some of you may know the answer to:

In theory, would music be more accurately represented (thereby sounding better ) with SACD's/DVD-A's high sampling rate, or by regular CD's which were then oversampled? In other words, which format has the higher sample rate?, or is it not quite as simple as this?

Regards

Sploo got it right. While the other formats have higher numbers, it doesn't necessarily translate to audibility by humans What we can hear is limited, pretty well known and music is very forgiving and masking, that is one reason perceptual coding like DTS and DD, MP3, etc, works so well. Ther eis just lots in that music that you will never hear, masked by other musical sounds

 

[tbewick]

The main benefit of SACD and DVD-Audio over audio CD is that they produce high quality surround sound.

Audio CD has excellent dynamic range and resolution. Most people cannot hear up to 20kHz, so I doubt the higher sampling rate of DVD audio/SACD will affect the sound quality. A well designed digital audio CD system is entirely transparent. The arguments against the Nyquist theorem never seem to be that specific or clear, which would make make believe that they're a load of rubbish.

DTS at its full data rate probably has sound comparable to SACD/DVD audio, and is a lot to get set up too.

 

[Buckle-meister]

'just can't stop pondering...

Hmm, interesting. Thanks for the replies ; sorry for the delayed response . In a somewhat related matter, may I further enquire about something on my mind?

I will more than likely soon be in receipt of Yamaha's S2500 DVD player which some of you may already know has the ability to play DVD-As/SACDs. I should note that I am buying the DVD player for DVDs, but am quite happy about the 'extra abilities' thrown in.

I am, in general, quite the traditionalist, and haven't much interest in multichannel audio, but had considered buying a few DVD-As/SACDs of my favourite CD albums, and then downmixing them to 2-channels in the belief that Id be getting the same album at a superior quality. Based on the previous posts comments however, I wonder whether that's true.

Has anybody actually compared a CD to the same album on DVD-A/SACD, and if so, can you hear the difference? Is it significantly better on DVD-A/SACD?

Bear in mind that I am well aware that the answer/s I recieve would represent biased opinions since they would not be based on double-blind listening tests etc. Still, I'd be really interested to know anyway.

Also, to cater for those of you in the 'double-blind ets tests' camp , if I am able to hear up to around 17kHz (I'm 30 years young ), then do you think that I would be unable to hear the difference from a CD to, say, the same CD oversampled?

Whats the consensus folks?

Regards

 

[WmAx]

I am, in general, quite the traditionalist, and haven't much interest in multichannel audio, but had considered buying a few DVD-As/SACDs of my favourite CD albums, and then downmixing them to 2-channels in the belief that Id be getting the same album at a superior quality. Based on the previous posts comments however, I wonder whether that's true.

No need to 'downmix' in most cases. Usually a 2 channel mixed track available on the releases. But you may very well find better sounding versions of some albums on DVD-A and/or SACD as opposed to the CD[or CD layer] counterparts, due to apparently intentional bad quality that is bestowed upon CD masters in the apparently misguided/ignorant view(s) that highly compressed[loud] CDs will sound better on radio airplay. Audioholics.com has [in the past] reviewed/analysed a few albums and found DVD-A or SACD version(s) of the same album to be less compressed than the CD version(s). In additional support of that conclusion, I had a Telarc recording engineer admit to me that the CD version of a particular jazz release was more compressed than the SACD version. Now that DVD-A ripping has become possible, I am considering ripping the DVD-A[when such is available] just to downsample to 44.1/16 and create MP3s, instead of using the likely compromised CD versions. In summary, you may ultimatey get better quality 2 channel sound with DVD-A or SACD, but it's not due to an inherant technical ability as compared to CD, but due to the mastering differences.

-Chris

 

[sploo]

I am, in general, quite the traditionalist, and haven't much interest in multichannel audio, but had considered buying a few DVD-As/SACDs of my favourite CD albums, and then downmixing them to 2-channels in the belief that Id be getting the same album at a superior quality. Based on the previous posts comments however, I wonder whether that's true.

I can't give you a definitive answer, except that some albums being re-released onto the newer formats may have been 'tweaked', thus resulting in better sound (but the same would probably be true for a CD re-release).

WmAx seems to have alluded to this here (post 2): http://forums.audioholics.com/forums/showthread.php?t=11831

EDIT: In the time it took me to submit this post, I see that WmAx has already given the above answer (but in more detail).

Has anybody actually compared a CD to the same album on DVD-A/SACD, and if so, can you hear the difference? Is it significantly better on DVD-A/SACD?

Now that depends on what you call better. I've learnt (mainly due to this forum) that your room acoustics and speakers will have massively more affect on what you hear than anything else. Would you really hear the difference in quality (given the limitations of human hearing)? I don't know, though I'd like to see some research on the subject.

What does make a difference (at least to my ears) is a good surround mix. OK, it's a little flash, but I've found it can add significant 'fun' to an old recording. For instance, there's a multi-channel mix of Mike Oldfield's Tubular Bells. Hearing guitars over your head, and bass rotating around you, is a guilty pleasure (as it's a little over the top), but a pleasure nonetheless . Totally off topic note: someone, somewhere, PLEASE do a high quality surround DVD of Pearl Jam's MTV Unplugged appearance .

Also, to cater for those of you in the 'double-blind ets tests' camp , if I am able to hear up to around 17kHz (I'm 30 years young ), then do you think that I would be unable to hear the difference from a CD to, say, the same CD oversampled?

Ok, that doesn't actually fit with my understanding of oversampling/upsampling. I may have misunderstood the technology, but I thought it was about making the filters required (to remove frequencies above Nyquist) easier to implement (and sound less harsh). The fact that it will produce a higher sample rate is not the goal in itself, and will do nothing to improve the signal. If I'm wrong, someone please shout.

 

[Buckle-meister]

Ok, that doesn't actually fit with my understanding of oversampling/upsampling. I may have misunderstood the technology, but I thought it was about making the filters required (to remove frequencies above Nyquist) easier to implement (and sound less harsh). The fact that it will produce a higher sample rate is not the goal in itself, and will do nothing to improve the signal. If I'm wrong, someone please shout.

I agree Sploo, however, consider this:

Here is a (rather long; sorry) quote from the Audioholics paper "Upsampling vs. Oversampling for digital audio":

"Whether we upsample or oversample, the effect on the spectrum of our audio signal is similar. Instead of our signal of interest occupying almost our entire bandwidth all the way up to fs/2 (22.05 kHz), it now only occupies 1/8 th of that. This allows us the use of a very simple analysis filter at the head of our processing chain, the internal digital filters, or the reconstruction filter at the end of the chain. Both the processes of upsampling and oversampling give us this benefit. Without increasing the sample rate, we would need to design a very sharp filter that would have to cutoff at just past 20kHz and be 80-100dB down at 22kHz. Such a filter is not only very difficult and expensive to implement, but may sacrifice some of the audible spectrum in its rolloff. If we examine the spectrum at the increased rate, we can see that the filter can roll off gently well past 22kHz and as long as it is down in the cutoff region at 176.4kHz, the image created by the sampling process will easily be removed. The analog filter after the D/A converter is responsible for removing the audio signal's image as well as the frequency spurs caused by the DACs integration steps. An analog filter with a smooth roll off will have nicer phase characteristics as well"

Clearly, the process allows for the use of a gentle filter, to hopefully make the sound, as you said, 'less harsh', but wont that result in 'better sound'; in this case, a term which (I'm sticking my neck out here) surely cannot be disputed. After all, if a gentle filter (combined with a large sampling frequency) takes away the harshness of sound which results from 44.1kHz sampling, then isn't the harshness an artifact which was never 'there' in the live recording in the first place?

In addition, the above paper notes that steep filters are expensive, yet the vast majority of CD players, as far as I'm aware, do not oversample; this process tending to be confined to far pricier models. So why are (cheaper filter) oversampling CD players more expensive than (expensive filter) non-oversampling CD players? Please dont tell me its the evil marketing thing again . Or is it that even though they use cheaper filters, oversampling CD players sell for more because of the 'better (i.e. less harsh) sound'?

In any event, this still leaves my original question unanswered; namely that if I can only (currently) hear up to about 17kHz, then can I even hear the 'harshness' of CD's which are not oversampled, i.e., is oversampling a complete waste of money (for me) because I'd never hear the difference anyway?

Thanking you

 

[WmAx]

In addition, the above paper notes that steep filters are expensive, yet the vast majority of CD players, as far as I'm aware, do not oversample; this process tending to be confined to far pricier models.

Huh? As far as I know, just about every CD player every made oversamples... it's just not a feature that is listed on things anymore. Remember the old 'oversampling' wars back in the late 80s? Consumer CD players with 8x oversampling were common consumer items years ago.

-Chris

 

[Buckle-meister]

Humble pie

You see, just when you think you know something; Bam! you find out that you dont'!

Oops. What can I say? I'm................learning...........................all the time.

Slowly.

Real.................real...................................................slowly.

EDIT: I've just taken a look at my CD players specification: "D/A converter: 8-times oversampling 20-bit digital filter."

Damn it WmAx you evil genius! Now I feel like a complete moron.

Still..........................learning.

Slowly.

 

[mtrycrafts]

You see, just when you think you know something; Bam! you find out that you dont'!

Oops. What can I say? I'm................learning...........................all the time.

Slowly.

Real.................real...................................................slowly.

EDIT: I've just taken a look at my CD players specification: "D/A converter: 8-times oversampling 20-bit digital filter."

Damn it WmAx you evil genius! Now I feel like a complete moron.

Still..........................learning.

Slowly.

Well, don't feel like a moron. You have shown your ability to learn and adjust, not some folks we have run into over the years on the net

 

[sploo]

I must admit I thought that most players oversampled. I think it's upsampling that's currently the domain of more expensive players.

Whether it actually makes a difference to the sound is debatable (usually loudly, and by lots of people )

I believe that some of the very early CD players didn't have oversampling, and thus produced the harsh sound which some people associate with digital (never heard one, I was barely out of short pants in 82/83 ).

As this thread has the attention of both mtrycrafts and WmAx, I'd like to ask my own question (something of a thread highjacking, but hopefully it'll be of interest).

The argument of the anti-snakeoil brigade seems to be that any reasonably decent CD player can output a very low distortion signal with a pretty flat frequency response. The 'superiour' sound produced by many more expensive units is therefore usually the result of deliberate frequency response adjustment, and possibly the addition of some 'pleasant' distortion (stop me here if I've misunderstood).

I've been playing with Doug Plumb's ETF software, examining room frequency response, amongst other things. This got me thinking - given a PC with a good (flat frequency response) sound card, and a CD containing a frequency sweep, could you sample the output of a player to determine its response?

After examining the results, you could then use your PC's CD player + equalizer (or a flat response CD player and an external parametric equalizer) and mimic the sound of another CD player.

I appreciate this ignores distortion (both the measurement, and the possibility of it affecting your results) but am I on the right lines here?

 

[mtrycrafts]

I must admit I thought that most players oversampled. I think it's upsampling that's currently the domain of more expensive players.

I am not sure when the first player used over sampling


I believe that some of the very early CD players didn't have oversampling, and thus produced the harsh sound which some people associate with digital (never heard one, I was barely out of short pants in 82/83 ).

Or, more likely, it was due to the newness of the format and the total inexperience of the engeneers to adopt to a medium as flat as the CD was/is vs the vinyl mixing

As this thread has the attention of both mtrycrafts and WmAx, I'd like to ask my own question (something of a thread highjacking, but hopefully it'll be of interest).

The argument of the anti-snakeoil brigade seems to be that any reasonably decent CD player can output a very low distortion signal with a pretty flat frequency response. The 'superiour' sound produced by many more expensive units is therefore usually the result of deliberate frequency response adjustment, and possibly the addition of some 'pleasant' distortion (stop me here if I've misunderstood).

This certainly would apply to the digital output

I've been playing with Doug Plumb's ETF software, examining room frequency response, amongst other things. This got me thinking - given a PC with a good (flat frequency response) sound card, and a CD containing a frequency sweep, could you sample the output of a player to determine its response?
After examining the results, you could then use your PC's CD player + equalizer (or a flat response CD player and an external parametric equalizer) and mimic the sound of another CD player.
I appreciate this ignores distortion (both the measurement, and the possibility of it affecting your results) but am I on the right lines here?

I will leave WmAx to respond He needs to have something to input on

 

[WmAx]

I must admit I thought that most players oversampled. I think it's upsampling that's currently the domain of more expensive players.

Oversampling is a term that is commonly used in reference to the filtering, as in the signal is upsampled for the purposes of more precise filtering. Upsampling may refer only to increasing the sample rate, not specifically for reasons of improved filtering accuracy. But really, it's just word play, you could exchange either term for the other. If your CD player oversamples, then it upsamples. Vice versa. Usually, this is done to have precise filters, but the term itself is not specific.

This got me thinking - given a PC with a good (flat frequency response) sound card, and a CD containing a frequency sweep, could you sample the output of a player to determine its response?

Just about any full duplex soundcard can be used to analyse frequency response. You can test the soundcard and cd player frequency response with a program such as RMAA( www.rightmark.org ).

After examining the results, you could then use your PC's CD player + equalizer (or a flat response CD player and an external parametric equalizer) and mimic the sound of another CD player.

Assuming that the frequency response is the audible difference, yes. Some player DACs may have purposely induced high distortion, however, such as the Audio Note DAC. It actually has no anti-alias filter. An audiophile idea. Some players may have other problems, due to defect or insufficient design[in a 1st generation player, for example].

-Chris

 

[MACCA350]

Originally Posted by WmAx
Now that DVD-A ripping has become possible

WmAx, could you point me to the software that can rip DVD-A. I was under the impression that DVD-A has'nt been cracked yet. I have a number of DVD-A's I'd like to backup

cheers

 

[MACCA350]

Found it

 

[sploo]

Just about any full duplex soundcard can be used to analyse frequency response. You can test the soundcard and cd player frequency response with a program such as RMAA( www.rightmark.org ).

That looks useful. I'll have a play when I get time.

Assuming that the frequency response is the audible difference, yes. Some player DACs may have purposely induced high distortion, however, such as the Audio Note DAC. It actually has no anti-alias filter. An audiophile idea. Some players may have other problems, due to defect or insufficient design[in a 1st generation player, for example].

Ok, assuming that jitter is below audible levels, and we're not dealing with an old player, is there anything other than frequency response and distortion that would be the difference?

It does seem, well, very audiophile, that the difference between a decent cheap player and one costing many thousands of pounds/dollars/camels (depending on your location) is the equivalent of a few tens of £/$/ of equalizer!

 

[Buckle-meister]

Ok, here's another one for you (sorry Sploo, but I need to borrow the thread again ). If I understand things rightly (don't even go there!):

Once we have sampled at a given frequency, we cannot further increase the amount of information we have of the source material. With regard to CD's, at frequencies higher than 44.1kHz, we have the problem of aliasing.

Step up SACD/DVD-A's with their sampling rates far beyond that of CD's. Now we have more source material information recorded at the outset, and because we sampled at a larger frequency, aliasing, although it will still exist, will only occur at frequencies that are far beyond the threshold of even the most golden-eared person.

The thing is:

1. If oversampling effectively removes the audible effects of aliasing from CD's anyway (ignoring the reduced manufacturing cost of the low pass filter), then how are SACD/DVD-A's better? Granted; their larger word lengths should give them greater dynamic range than that of CD's, but don't CD's have more than enough (90dB+) to cater for a properly mastered album as it stands?.

2. Nyquist's theorem tells us that even with just two samples of a 22.05kHz frequency, we can reconstruct the original waveform precisely. So, ignoring loss of data from scratches etc., why have an increased number of samples for SACD/DVD-A? We appear to have the requisite number to reconstruct all frequencies anyway, and we cant hear above approximately 20kHz.

I'm guessing that I'm missing something. Anyone care to enlighten me?

 

[MDS]

You aren't really missing anything, but I can add a bit.

Once we have sampled at a given frequency, we cannot further increase the amount of information we have of the source material. With regard to CD's, at frequencies higher than 44.1kHz, we have the problem of aliasing.

Correct. Aliasing is always a problem and is the reason the Nyquist theorem specifies a sampling rate of twice the highest frequency you want to reproduce. It isn't that aliasing occurs at frequencies higher than 44.1 kHz - it's that with a sampling rate of 44.1 kHz, the highest frequency you can reliably replicate is 22.05 kHz. If you wanted to be able to reproduce 22.05 kHz accurately, but sampled at only 40 kHz, you would have aliasing.

1. If oversampling effectively removes the audible effects of aliasing from CD's anyway (ignoring the reduced manufacturing cost of the low pass filter), then how are SACD/DVD-A's better? Granted; their larger word lengths should give them greater dynamic range than that of CD's, but don't CD's have more than enough (90dB+) to cater for a properly mastered album as it stands?.

SACD uses 1 bit sampling so is not valid for this topic. DVD-A is PCM so is directly comparable. Increased word length is the true benefit. The increased bit depth gives you finer grained samples. With CD 16 bit word length, the values for each sample (44.1 K per second) range between -32767 and +32,768 [LPCM is signed and the waveform has both positive and negative components]. With 24 bit, you get a range of -8,388,607 to 8,388,608 - so the accuracy of the amplitude of each sample is better.

Sillly example that illustrates the point. I am 6 feet tall, but let's say I am actually 6' 1/8" tall. If my height were sampled in 'inches' you could only obtain a value of 72 inches or 73 inches to represent my height because the smallest unit of measurement is one inch. If you were instead to sample my height with smaller increments of 1/8" inch you can get the actual value for my height.

The second benefit you have already alluded to - the higher sampling rate moves the Nyquist cutoff frequency way above the range of human hearing. So after the reconstruction filter, much more of the original frequency content is preserved.

Now is it really all that? Professional recording/mastering engineers seem to be split on the issue, at least from what I've read over the years. Most still think a well mastered CD that truly utilizes the full 16 dynamic range can sound just as good. The higher bit depth is great, but the higher sampling frequencies (more samples per second) doesn't necessarily buy that much.

2. Nyquist's theorem tells us that even with just two samples of a 22.05kHz frequency, we can reconstruct the original waveform precisely. So, ignoring loss of data from scratches etc., why have an increased number of samples for SACD/DVD-A? We appear to have the requisite number to reconstruct all frequencies anyway, and we cant hear above approximately 20kHz.

Whether it's two samples or N samples, it is an approximation of the original analog signal. Analog is continuous, whereas digital is discrete. The reconstruction filter essentially does 'curve fitting' to recreate a continuous waveform from a series of discrete samples. The more samples you have the easier it is to approximate the original, but of course only up to a point - and that is where the debate goes on as to whether extreme sampling rates like 192 kHz is overkill.

 

[Buckle-meister]

Increased word length is the true benefit. The increased bit depth gives you finer grained samples. With CD 16 bit word length, the values for each sample (44.1 K per second) range between -32767 and +32,768 [PCM is signed and the waveform has both positive and negative components]. With 24 bit, you get a range of -8,388,607 to 8,388,608 - so the accuracy of the amplitude of each sample is better.

Understood. However, if a well mastered album utilized the full 16 bits, then the maximum error in a CD sample's amplitude would be 1 part in 32767 as opposed to DVD-A's 1 part in 8,388,608. Correct? But whilst this endows DVD-A with 256 graduations for every single 32767th part , an astounding degree of accuracy in its own right, can people even hear 1 part in 32767? Sounds dubious to me.

Regards