[pinnahertz]

   
1. Agreed, and my "roughly equivalent to a 192Kbps" took some account of that fact. 448Kbps divided by 6 is about 75kbps, so if DD were a fixed bit rate, that would be 150kbps for stereo rather than the 192kbps I quoted.

That's absolutely not how it works...and I'm pretty sure you know that. DD doesn't take the whole bit rate and divide it up evenly per channel.

Even if we were to approximate DD as equivalent to a 320kbps CBR MP3, that's still not quite as high definition as CD, although we're at/very close to the limits of audibly indistinguishable.

The two are not directly comparable. They are different codecs, and a completely different channel count, and "bit budget" if you will. Yes, a 32kbps mp3 is pretty much indistinguishable from the CD, but you can't translate that to a comparison multichannel Dolby Digital AC3.

Considering the sheer amount of data compression occurring, DD is exceptionally good but it would be unreasonable to argue that it's actually higher definition/fidelity than "SD" CD and that it qualifies as HD audio.  

I would agree if you compare two channels of DD to two channels of CD, and DD/AC3 would not be equivalent, certainly not higher def. But that's as far as you can go with that comparison. When you throw 5.1 into a comparison with two channel CD, it all gets messy. For example, take your best fully active 5.1 mix and downmix it to 2.0 redbook. You had 5.1 at 48kHz, you throw out all the specific directional cues, localization, hard center channel, discrete LFE , etc., and end up with 44.1 stereo. Would you not agree that the result is lower definition than the 5.1 DD version? You lost channels and sampling frequency. It's not better at all, but you started with a lower total rate bitstream and mixed it to a higher total bitrate. See what I mean? Messy comparisons, pretty much pointless to go down that road.  And the resulting two-channel mix might sound "fine", but nobody would argue it's instantly and recognizably different than the 5.1 version.  I do realize content is king here, and you could do a 5.1 channel mix where artistic values dictate primarily two channel useage, but that's not what I'm talking about.  

2. For the consumer I would agree and of course comparing a surround mix to a stereo mix is an apples to oranges comparison. However, your statement is not an absolute truism. In my work I quite often have to constantly compare them and the "radically different" mix is somewhat different in my circumstance as I have complete control over both mixes. Maybe in my post it would have been more accurate to say "very roughly equivalent" or even "extremely roughly equivalent" rather than just "roughly equivalent" but I feel you're being a little over pedantic.
 
G

OK, fine, I'm over pedantic. But I hope my point is clear. Just comparing the bitrate of the total bitstream and generalizing, or comparing different channel counts within the same bitrate but totally different codecs is pointless, and misleading.

 

[icebear]

Can you please take the discussion on multi channel format intricacies to pm please?
Thanks.

 

[gregorio]

[1] That's absolutely not how it works...and I'm pretty sure you know that.

[2] You had 5.1 at 48kHz, you throw out all the specific directional cues, localization, hard center channel, discrete LFE , etc., and end up with 44.1 stereo. Would you not agree that the result is lower definition than the 5.1 DD version?

[3] I would agree if you compare two channels of DD to two channels of CD, and DD/AC3 would not be equivalent, certainly not higher def.

 
1. And I'm pretty sure you know that an "if [something] were ..." construct in English grammar is the subjunctive tense, used for counter-factual or hypothetical statements.
 
2. No, I wouldn't agree. What "you had", what you feed DD/AC3 is 6 channels of 24/48, which is 6,750kbps and what you end up with is a DD datastream of just 448kbps. That's a (lossy) compression ratio of over 15 times, which is more than double the approx. 7 times compression ratio of 16/44 to MP3 192. I'm not of course saying that MP3 192 is therefore more than double the "definition" of DD, for the reasons you mentioned, DD is adaptive, more efficient and additionally, this compression ratio comparison hasn't accounted for the fact that the vast majority of the LFE data can simply be discarded (we really only need a sample rate of a little over 240Hz for perfect reconstruction). However, I'm not sure that the additional channels of 5.1 (the additional localisation/positioning) can be equated to "higher definition", higher fidelity yes but not necessarily "higher definition" (higher resolution). In theory we can achieve a roughly (!) equivalent amount of localisation/positioning with binaural stereo (albeit only in headphones). I believe we're largely in the realm of just semantics here though.
 
3. As The DD specs for HDTV do allow for a native 2.0 mix and had I thought about it in greater detail, I might have used such a mix as comparison. In which case, a better (though still NOT accurate) comparison would have been a 2.0 DD mix with a 276kbps MP3. Not such a good practical example though, as native 2.0 DD mixes are relatively rare.
 
You seem to have missed the actual point I was making though. I was not trying to make an accurate comparison or evaluation of DD, MP3 and CD. Just provide a "rough estimate", an estimate with only enough accuracy to show that DD is somewhat lower definition than uncompressed 44.1/16. This point is valuable to make in the context of this thread as it demonstrates that arguably the widest use of the term "high definition" (in HDTV), effectively equates HD visuals to audio which is somewhat below a definition/resolution accepted as "standard definition". My argument being, that what is commonly referred to as "standard def" digital audio (uncompressed 16/44.1) is in practise effectively "infinitely high def" digital audio and superior to even the latest extensions to what is called "ultra high def" in visuals.
 
G

 

[pinnahertz]

   
1. And I'm pretty sure you know that an "if [something] were ..." construct in English grammar is the subjunctive tense, used for counter-factual or hypothetical statements.
 
2. No, I wouldn't agree. What "you had", what you feed DD/AC3 is 6 channels of 24/48, which is 6,750kbps and what you end up with is a DD datastream of just 448kbps. That's a (lossy) compression ratio of over 15 times, which is more than double the approx. 7 times compression ratio of 16/44 to MP3 192. I'm not of course saying that MP3 192 is therefore more than double the "definition" of DD, for the reasons you mentioned, DD is adaptive, more efficient and additionally, this compression ratio comparison hasn't accounted for the fact that the vast majority of the LFE data can simply be discarded (we really only need a sample rate of a little over 240Hz for perfect reconstruction). However, I'm not sure that the additional channels of 5.1 (the additional localisation/positioning) can be equated to "higher definition", higher fidelity yes but not necessarily "higher definition" (higher resolution). In theory we can achieve a roughly (!) equivalent amount of localisation/positioning with binaural stereo (albeit only in headphones). I believe we're largely in the realm of just semantics here though.
 
3. As The DD specs for HDTV do allow for a native 2.0 mix and had I thought about it in greater detail, I might have used such a mix as comparison. In which case, a better (though still NOT accurate) comparison would have been a 2.0 DD mix with a 276kbps MP3. Not such a good practical example though, as native 2.0 DD mixes are relatively rare.
 
You seem to have missed the actual point I was making though. I was not trying to make an accurate comparison or evaluation of DD, MP3 and CD. Just provide a "rough estimate", an estimate with only enough accuracy to show that DD is somewhat lower definition than uncompressed 44.1/16. This point is valuable to make in the context of this thread as it demonstrates that arguably the widest use of the term "high definition" (in HDTV), effectively equates HD visuals to audio which is somewhat below a definition/resolution accepted as "standard definition". My argument being, that what is commonly referred to as "standard def" digital audio (uncompressed 16/44.1) is in practise effectively "infinitely high def" digital audio and superior to even the latest extensions to what is called "ultra high def" in visuals.
 
G

The problem we're having is with the definition of "definition".  
 
I agree, Dolby Digital AC3 is lossy.
 
I agree CD Redbook is lossless (assuming the original is 16/44.1). 
 
I don't mean to synthesize a definition here.  I look at Definition and High Definition as an evaluation of transport system to reproduce the original signal.  If we consider the original signal to be a two channel 16/44.1 bitstream, then Redbook is as high as definition can go, it replicates the original exactly.   If we consider the original to be a 24/48 5.1 mix, then Dolby Digital is lossy, and not High Definition.  It was higher definition than the mutli-channel delivery methods available at the time it was introduced, and at that time it could have been called High Definition if we'd been using that term.  It beat Dolby Stereo optical (matrix), and pretty much knocked heads with six channel magnetic with Dolby SR, though that was not a common delivery format for very long.  However, it's not high definition by today's standards.
 
The problem is stating the CD is Standard def relative to Dolby Digital, the latter being below standard def.  You've got your standards mixed up.  Look at the goals.  The CD's goal is to reproduce two channels.  It does that losslessly relative to 16/44.  Dolby Digital's goal is to reproduce 5.1, it does that but via a lossy coded.  But, if you took that 5.1 signal and tried to get it through the CD, about all you have is the Dolby matrix.  And, while the CD's portion of that chain is lossless, or slightly lossy if the original was 24/48, passing the total mix through the matrix is a lossy process, so much so that the mix must be modified somewhat for the intent to make it all the way through.  What you end up with is less like the original 5.1 master than the Dolby Digital version.  
 
So, what's "standard def" 5.1 audio for video and digital TV?  I would suggest from the consumer's viewpoint it's Dolby Digital.  That's what we have had on DVD for some time, digital broadcast, Blueray, and most streaming.  Yes, it's lossy, but it's THE common bitstream, that makes it the "standard".  Now we have TrueHD and Master Audio, both capable of 24/48 multichannel lossless.  That's the new HD audio, but not for broadcast.  
 
My only real issue is the method of comparison used to establish if something is "standard" or below.  You can't compare a two channel lossless transport system with a 5.1 channel lossy transport system, the goals aren't the same, and one can't do what the other can.  Throwing a reference to .mp3 into the discussion further confuses things.  MP3 can't do what either DD or CD can.  It's yet another animal.  Any comparison of raw bit rate is also irrelevant, because it ignores codec performance and efficiency as well as channel count. 
 
I'm pretty much leaving it here.  I guess someone doesn't feel our discussion of the resolution of different audio formats fits into a thread titled "Is there a meaningful limit to resolution?"  Go figure.
 
I'll just close with this comment.  A few years ago THX inventor Tom Holman did a series of presentations called "The Bit Rate of Reality".  That presentation demonstrated that two channels of "high resolution" audio like 24/96 didn't come close to representing "reality" at all, where his 10.2 channel system with non-exotic audio bitrates (though the demo used discrete lossless channels) could replicate an acoustic space well...reality.  Holman has also found that, starting with single channel mono, every time the channel count is doubled, every listener can clearly hear the improvement. The point of diminishing returns relates to the size of the audience, but in smaller spaces, it lands between 10 and 20, given the proper speaker layout. 
 
So my response to the OP and thread title would be, yes, there is a meaningful limit, it changes with technology and application, but relates more to channel count than bit rate per channel.  And, since we're in Head-Fi, and stuck at two channels, it would seem we've past the meaningful limit already.

 

[icebear]

Originally Posted by pinnahertz /img/forum/go_quote.gif

...
 
I'm pretty much leaving it here.  I guess someone doesn't feel our discussion of the resolution of different audio formats fits into a thread titled "Is there a meaningful limit to resolution?"  Go figure.
 
....
 
So my response to the OP and thread title would be, yes, there is a meaningful limit, it changes with technology and application, but relates more to channel count than bit rate per channel.  And, since we're in Head-Fi, and stuck at two channels, it would seem we've past the meaningful limit already.

 
Indeed

, since it's already complicated enough to compare just different formats for two channel, kicking it up a notch and bringing in the effects of even more channels and arguing which value of more format resolution equals how many more channels will not get us anywhere. Just my $0.02.
 
At least your final statement seems to be in line with my assumption that we are past any meaningful improvement. Yes there might be special "examples of sounds" , not necessarily actual music that make it possible to distinguish that some formats sound different. If this different is always preferred as better, is a different question.
 
The Norwegian label 2L has a format comparison page and also some info on the latest and greatest MQA format. There is a diagram on the 2nd page of this document:
http://www.2l.no/hires/documentation/2L-MQA_Comparisons.pdf
 
The max peak level of the example seems to be somewhere at -18dB. If you try to maximize the dynamic range of your recording medium you try to capture as much of the original signal as possible.
Typical analog recordings had been driven a little "hot" i.e. allowing for occasional +3dB peaks but getting a good S/N ratio for the major part of the recording. With digital you can't go beyond 0 dB and there are conventions to not go beyond -5dB I think but choosing examples at -18dB then in turn make it possible to point out that above 15kHz the CD 16/44 noise floor is above the signal level is obviously driven by some marketing drivel

 

[watchnerd]

 
 
The Norwegian label 2L has a format comparison page and also some info on the latest and greatest MQA format. There is a diagram on the 2nd page of this document:
http://www.2l.no/hires/documentation/2L-MQA_Comparisons.pdf
 

 
So they've shown that the 16bit noise floor is at -96 dBFS, and the 24bit noise floor is >-120 dBFS.  
 
But we already knew that.
 
And we also know that 96 dB range is more than sufficient for domestic playback.
 
Seems like we're covering well-worn ground that is already known.

 

[pinnahertz]

 
 
The max peak level of the example seems to be somewhere at -18dB. If you try to maximize the dynamic range of your recording medium you try to capture as much of the original signal as possible.
Typical analog recordings had been driven a little "hot" i.e. allowing for occasional +3dB peaks but getting a good S/N ratio for the major part of the recording. With digital you can't go beyond 0 dB and there are conventions to not go beyond -5dB I think but choosing examples at -18dB then in turn make it possible to point out that above 15kHz the CD 16/44 noise floor is above the signal level is obviously driven by some marketing drivel

The graph is the result of an FFT spectrum display.  It's not true peak, the response time of an FFT depends on resolution bandwidth, FFT length, etc.  So the -18dB(fs?) would not represent true maximum peak.  
 
You're confusing peak reference levels and average reference levels.  Analog tape used several different "reference levels", specific magnetic fluxivity, and measure with a Volume Indicator (VU meter).  Since tape systems saturate slowly, there is no 0dBFS, it's a trade-off between maximum level and tolerable distortion (relating to duration and frequency).  It's not actually correct to say analog recordings were driven a little hot, the entire reference level issue on tape was a bit spongy. There should be way more than a few +3dB peaks, they should be around +8 to +10 (the difference between using a VU and true peak meter). 
 
The original CD standard was to insure that the highest peak got close to, but not beyond 0dBFS. There's no need to keep peaks at -18dBFS, and good reasons not to.  However, a VU meter used with a digital system may have been calibrated to read 0dB at -18dBFS, a conservative calibration.  We also used -15 and -12, perhaps a few others. But at no time did we hold true peaks to -18dBFS.  

 

[castleofargh]

+1, true peak(depending on the oversampling value for digital stuff)), VU meter, independent settings to have 0 at 0 or at -0.1 or whatever, dbV dbu dbeverything, good guy Bob Katz and his K-system... it's easy to lose ourselves in references.
 
didn't know about tapes, I assume it's basically a non linear dynamic when reaching maximum values? more energy resulting in smaller and smaller change past some point with an actual clipping just too far away to reach in practice.  first time I came face to face with such concept was in a dark room with B&W paper that a guy adored even though it was known to have one of the worst dynamic range and linearity. the "secret" was that he didn't need to do a good job, over exposure of that paper would almost never result in a "burnt" area. ^_^

 

[pinnahertz]

 
 
didn't know about tapes, I assume it's basically a non linear dynamic when reaching maximum values? more energy resulting in smaller and smaller change past some point with an actual clipping just too far away to reach in practice.  first time I came face to face with such concept was in a dark room with B&W paper that a guy adored even though it was known to have one of the worst dynamic range and linearity. the "secret" was that he didn't need to do a good job, over exposure of that paper would almost never result in a "burnt" area. ^_^

Tape is perhaps a bit more complex that B&W paper, which is basically responding to white light exposure and development time/temp. Each tape type has a somewhat different overload (actually, saturation) characteristic.  To make it more complex, it also changes with tape speed, and varies with frequency.  Compared to a digital system it's a very soft mushy overload, so where to define your reference point is also a matter of tradeoffs.  There are industry standard reference levels to help with interchange between studios, establish accurate reference levels for noise reduction systems, etc., but it's only a reference level, and what you do with it is pretty much only standardized within systems, studios, etc.  And so how you figured S/N depended on how much headroom you wanted, and of what type, tape formula, track width, and linear speed.  You could reduce percieved noise by raising the speed, but at the expense of low frequency response.  You could reduce noise by using wider tracks at the expense of more critical alignment.  Or, you could just pick your favorite speed and apply a double-ended noise reduction system.  The all-time winner was Dolby SR, just before digital swamped the market. 

 

[gregorio]

  The problem we're having is with the definition of "definition". 

 
Partially yes, which as I mentioned is largely semantics. However, the point I was trying to make was not about the problem of the definition of "definition" (resolution) but the problem of "standard" and therefore it's relationship to "high" (and even "ultra-high"). In the days of analogue TV broadcast, I don't remember the audio standard off the top of my head but it was a significantly smaller dynamic range and a freq bandwidth limited to 15kHz. With the advent of DVD, the resolution of the visuals remained the same (though with less visual noise) but the resolution of the audio improved significantly, with DD/DTS providing full bandwidth, a greatly enhanced dynamic range and of course the same significant reduction in noise (plus additional channels but we'll ignore that for now). This was still called "standard definition" although in comparison, the audio was very significantly higher definition than what was then "standard". Then along came HDTV with obviously far higher visual resolution but absolutely no difference at all in the audio definition compared to the standard definition DVD.
 
As you mentioned, we now have audio formats like TrueHD, capable of uncompressed 24/48 but we always use a dynamic range very significantly less than 16bit (even in theatrical mixes, let alone TV mixes!), so we have audio formats in the TV/video world called High Definition which are in effect the same as audio formats in the music industry which are called standard definition (as I'm assuming HD audio in the music industry to be 24/88.2 or greater). And, if this weren't confusing enough, TrueHD can be, and quite often is, used on BluRay (which is HD) but current UHD (4k) streaming, like NetFlix for example, is limited to Dolby Digital Plus which is a lossy compressed format. So in practise currently, the max audio resolution of UHD (broadcast/streaming) is lower than HD (bluray video)!
 
Audio "definition" has for many years been independent of visual "definition", making analogies between them is potentially (highly) misleading.
 
G