AVS › AVS Forum › Audio › Surround Music Formats › 'hi rez' DVD-A? Sometimes.
New Posts  All Forums:Forum Nav:

'hi rez' DVD-A? Sometimes. - Page 3

post #61 of 95
3/24/08 at 10:18am
Thread Starter 
You might consider offering some content of your own then. Or at least suggesting what content you'd LIKE to see.
Reply
Reply

AVS Top Picks

post #62 of 95
3/24/08 at 10:25am
Quote:
Originally Posted by Jim85IROC View Post

Yet another thread that starts out with so much promise, just to turn into 3 pages of termonology pissing.

No, not here. Never.
Reply
Reply
post #63 of 95
3/24/08 at 11:41am
I respectfully disagree that the thread turned into a pissing match. It did start with promise, and I learned some things worth knowing from krabapple's graphs. I learned a lot more when a highly experienced audio engineer spoke up. PenteoSurround patiently explained the nomenclature of his profession and also showed why accurate definitions matter (don't they always?). This is a good thing.
Reply
Reply
post #64 of 95
3/24/08 at 11:42am
Quote:
Originally Posted by krabapple View Post

You might consider offering some content of your own then. Or at least suggesting what content you'd LIKE to see.

I've been meaning to post my thoughts on a particular aspect, so here I go:

Forget for a while that the signal is being processed digitally - pretend that we're dealing with some magic analog format like tape, but without hiss, etc, and an essentially undefined noise floor.

But also assume that said analog format has some kind of maximum signal strength inherent in the design (player limited, industry convention to limit max signal to 1VDC, whatever).

The abuse of compression seem in the first post, and also experienced with Rush's Vapor Trails, would likely still exist, due to the record company's desire to have the album sound as "loud" as possible, assuming stupid listeners and radio stations do not adjust volume between albums/songs and therefore the loudest album/song is the best.

This kind of abuse was "excused" with CD because of the so-called limitation of 16 bits of depth, but really that doesn't hold much water: a well mastered digital encode of 16 bit depth should have room for all the required dynamics (96 dB worth). I'm not sure what the maximum dynamic range of vinyl is, but I'm pretty sure it doesn't exceed 96dB. Whether 16 bits at 44.1 kHz reproduces the sound as warmly as vinyl is not the point: the point is that 16 bits is more than enough to represent the required dynamics. Compression-abuse to "provide enhanced dynamics" is a crock.

So to see overly-compressed DVD-A encodes means the record company just re-sampled the overly compressed master used for a botched CD release, which didn't need to be over-compressed in the first place.

See the wikipedia entry here for some interesting reading on analog vs digital - it might make you go "hmmm".

IMHO, YMMV,

shinksma
Reply
Reply
post #65 of 95
3/24/08 at 12:23pm
Quote:
Originally Posted by shinksma View Post

I'm not sure what the maximum dynamic range of vinyl is, but I'm pretty sure it doesn't exceed 96dB.

iirc it is 78db, and hirez PCM 114db
Reply
Reply
post #66 of 95
3/24/08 at 2:17pm
Thread Starter 
Quote:
Originally Posted by shinksma View Post


So to see overly-compressed DVD-A encodes means the record company just re-sampled the overly compressed master used for a botched CD release, which didn't need to be over-compressed in the first place.

Actually, some of these DVDA stereo masters have no equivalent counterpart in CD -- in other words, the existing CD versions are either less, or more, compressed than the DVD-A version. So it's not necessarily just a case of transcoding a CD master.

E.g.,

- the current Queen 'Night at the Opera' CD is less compressed than the Queen DVDA stereo version

- the current Yes 'Fragile' CD is signficantly MORE compressed than the DVDA. If you have the current remaster of Fragile on the Rhino label, rip 'Roundabout' and take a look at it in a waveform viewer. It's incredible, and not in a good way.

- the CD side of the Talking Heads 'Speaking in Tongues' Dualdisc is a little bit more compressed than the DVD-A -- probably not audibly.
Reply
Reply
post #67 of 95
3/24/08 at 4:25pm
Quote:
Originally Posted by SiriuslyCold View Post

iirc it is 78db, and hirez PCM 114db

Well with vinyl, it's dependent on the formulation. Virgin Keysor Century, at it's best, was probably about -65-70 (with RIAA EQ) (it's not like there's a specific "number", there are way too many variables -- but of course even the slightest imperfection will give you a "tick", but the bulk of the noise floor on vinyl is from the rumble of the playback mechanism itself, and the smoothness of the disc surface -- usually dependent on whether or not an adhesive had been used on the back of the stamper.

HiResPCM (I'm assuming you meant 24-bit) is actually -144.
Reply
Reply
post #68 of 95
3/24/08 at 7:09pm
thanks for the correction!

I did mean 24b PCM - assuming we all agree 16b or 20b PCM is not hirez
Reply
Reply
post #69 of 95
3/24/08 at 9:15pm
Thread Starter 
Quote:
Originally Posted by SiriuslyCold View Post

thanks for the correction!

I did mean 24b PCM - assuming we all agree 16b or 20b PCM is not hirez

Both are 'hi rez' compared to vinyl. That was one of the selling points of CD way back when, remember? 'For the first time, listeners at home can experience the full dynamic range of the master recording!' Funny how that all worked out.
Reply
Reply
post #70 of 95
3/24/08 at 9:23pm
now you are going to get people arguing there's no such thing as "resolution" when it comes to analog and any discussion of it is irrelevant

and, just because people muck up the delivery it doesn't mean the medium is flawed
Reply
Reply
post #71 of 95
3/24/08 at 9:25pm
Thread Starter 
Babylon Sisters, Gaucho (Steely Dan). 96/24 DVDA mastered by Bob Ludwig



Overture, Tommy (The Who) 96/24 DVDA mastered by Jon Astley
Reply
Reply
post #72 of 95
3/24/08 at 9:27pm
can you zoom into that bit between 3:00 and 3:40, lets see it in higher resolution (sic)

and before anyone starts again, this is high resolution 24b / 96kHz (or 48khz) we are looking at, right? not a 16/44 stereo bit
Reply
Reply
post #73 of 95
3/24/08 at 9:39pm
Quote:
Originally Posted by krabapple View Post

Babylon Sisters, Gaucho (Steely Dan). 96/24 DVDA mastered by Bob Ludwig

Well, Bob is the best. He has to be to get people to come all the way up to Portland, Maine; when I went up there to wire it for the network, I almost froze to death (it was -18F in 1994). But obviously, this is an effort to make a very "loud" performance, using not only compression but dynamic clipping (simply chopping off peaks past a certain point). That's the only way you can get them that flat.

You must remember, the client (Steely Dan/et. al) are old school; the same would have been done to all of their music in the past. Listening to the tracks completely uncompressed would probably sound "weakish". This ain't classical. This is exactly what Bob gets paid the big bucks for.
Reply
Reply
post #74 of 95
3/24/08 at 10:26pm
Thread Starter 
Quote:
Originally Posted by SiriuslyCold View Post

can you zoom into that bit between 3:00 and 3:40, lets see it in higher resolution (sic)

and before anyone starts again, this is high resolution 24b / 96kHz (or 48khz) we are looking at, right? not a 16/44 stereo bit


96/24 = 96kHz SR, 24-bit resolution, of the delivery format.

Consider four broad categories of digital mastering re: dynamic range, from analog sources:

1) no substantial futzing with the full DR of the source, so peak levels vary a lot across the length of the track even at low magnification, with only one or a few actually reaching the highest track peak value. (e.g. the Billy Cobham , Neil Young and Deep Purple tracks). I would call THIS 'old school'.

2) dynamic range reduced, much less apparent peak level variance (i.e., a great many more peaks samples at or near the highest value, at low magnification), but still nothing clipping. The Talking Heads track is an example. Some of these tracks can even LOOK like they're clipping, in a low magnification view, but zoom in shows the 'clipping' is really a cluster of single peaks at or near the track maximum, not 'flattops'. Again, 'apparent' clipping at low magnification does NOT necessarily mean actual clipping. Further analysis is often required.

3) like #2, but with actual clipping -- 'flat topped' peaks. But there's only a few of them, or the ones that are there don;t last more than a few samples, so they're not likely to make as bad an audible impression.

4) like #2, but clipping is frequent, and often 10 or more samples in length when it does occur, and thus potentially much more audible than the others

Gaucho belongs to #3, a sort of 'walking the fine line' mastering -- Ludwig's a pro's pro, so this is consummately done for what it's trying to do, but why was it felt necessary to do at all for an 'audiophile' release?

Two increasing magnifications of a few seconds around 3.04 min. That plateau in the right channel is ~24 samples wide, btw.






IMO, the client being 'old school' - and famously anal about the sound of their releases -- would lead one to think that Steely Dan would NOT go in for loudness war mastering. But they seem to have changed with the times...the mastering of 'Two Against Nature' and 'Everthing Must Go' is even more 'market ready'.
Reply
Reply
post #75 of 95
3/24/08 at 11:27pm
thanks for the expansions - both imagery and explanation
Reply
Reply
post #76 of 95
3/24/08 at 11:43pm
Quote:
Originally Posted by SiriuslyCold View Post

this is high resolution 24b / 96kHz (or 48khz) we are looking at, right? not a 16/44 stereo bit

I'm not starting -- but understand this: Look at the waveforrms shots above. Specifically
The db value scale (the vertical axis) shows 3 values:
-3
-6
-12
then jumps to "infinite"

Every sound that you're seeing here -- at least clearly -- is above about -30db. So everything you're seeing here is taking place in the least significant 7 or so bits -- that is, of the 24 bits that this recording was mastered at, 17 bits of every 24-bit sample are representing silence and random noise. Imagine 17 bits being used for the detail in the "infinite" line, and the remaining 7 spreading out to the top and bottom. That's resolution.

Since 16 bit will take you down to -96, obviously what you're looking at would look -- and in fact would be identical -- in either 16 or 24 bit resolution. 24-bit would take you past -96 (silence) down to 48db even quieter -- that is, down to -144.

None of these resolution levels has anything to do with the spectacular amount of peak clipping that has been done to this recording, hence my earlier comments about not confusing resolution with dynamic range comression or clipping.

What I might suggest, krabapple, is zoom in so that -96 is out near the edge, and see what is contained between -96 and -144; that would be the material that is different between 16-bit and 24 bit. (PS: It'll probably just be dithered noise. If you specifically zoom in on the fade out at the end of the song, you can see if the fade was done pre-dither or post-dither.) If you see where the noise floor actually lives, you can see how many bits are actually being used for audio vs. how many for random noise.
Reply
Reply
post #77 of 95
3/24/08 at 11:59pm
Quote:
Originally Posted by krabapple View Post

IMO, the client being 'old school' - and famously anal about the sound of their releases -- would lead one to think that Steely Dan would NOT go in for loudness war mastering. But they seem to have changed with the times...the mastering of 'Two Against Nature' and 'Everthing Must Go' is even more 'market ready'.

In mastering something as high-profile as this, the ultimate trust is put into Bob Ludwig and his spectacular track record, of having mastered more moneymakers than anyone else in history. See http://www.gatewaymastering.com/. A lot of clients don't even bother flying up, they just send the master up to Maine, and what comes out, comes out. They get a ref back for listening, but I can't imagine anyone complaining. Besides, flying into Portland, Maine can be scary. :-)
Reply
Reply
post #78 of 95
3/25/08 at 7:17am
Thread Starter 
Quote:
Originally Posted by PenteoSurround View Post

What I might suggest, krabapple, is zoom in so that -96 is out near the edge, and see what is contained between -96 and -144; that would be the material that is different between 16-bit and 24 bit. (PS: It'll probably just be dithered noise. If you specifically zoom in on the fade out at the end of the song, you can see if the fade was done pre-dither or post-dither.) If you see where the noise floor actually lives, you can see how many bits are actually being used for audio vs. how many for random noise.


Good idea. I'll get to this later. Then we can talk about how many home listening environments exist where a noise floor near -144 versus -96 dB, would make a difference

Btw, on the basis of Babylon Sisters, I wouldn't say the Gaucho DVDA stereo mastering has a 'spectacular amount of peak clipping' by any means...in the CD realm I've seen much, much more aggressive examples. As tracks with clipping go, this is somewhat conservative. The overall number of clipped peaks is low (though some are wide, like the one I showed). I'll run an analysis later that can generate the actual number of 'flat top' samples in the track.

Btw, I'm going to get my hands on the SACD version and see how THAT was mastered by comparison...DSD spec discourages clipping.
Reply
Reply
post #79 of 95
3/25/08 at 2:56pm
Quote:
Originally Posted by PenteoSurround View Post

of the 24 bits that this recording was mastered at, 17 bits of every 24-bit sample are representing silence and random noise. Imagine 17 bits being used for the detail in the "infinite" line, and the remaining 7 spreading out to the top and bottom. That's resolution.

I might re-phase this bit as:

Quote:


of the 24 bits that this recording was mastered at, 17 bits of every 24-bit sample are represented visually identically to silence and random noise in these screencaps. Imagine 17 bits being used for the detail in the "infinite" line, and the remaining 7 spreading out to the top and bottom. That's resolution.

Otherwise it sounds like there actually might be 17 bits of silence or random noise.

If we actually listened to a 7-bit encode, I hope we'd be able to tell the difference, though.

shinksma
Reply
Reply
post #80 of 95
3/25/08 at 3:18pm
huh, I thought Bobby L just threw on some Waves mastering plugz and was done with the deal... Guess those concrete-based speakers are good for sumthin.

For the sake of making sure we're seeing what's on the discs, can you step through the process for capturing the 6ch files please? Especially for SACD, the process, I'd think, would involve taking the analogue outputs from a player and capturing that. For DVD-A you could demux the audio to specific wav streams, and see the "raw" signal, but if you're capturing analgoue output, re-recording back into you DAW, then it's at least possible some of the banging atop the headroom could be a capture issue, not a source issue.
Reply
Reply
post #81 of 95
3/25/08 at 5:05pm
Quote:
Originally Posted by shinksma View Post

I might re-phase this bit as:



Otherwise it sounds like there actually might be 17 bits of silence or random noise.

If we actually listened to a 7-bit encode, I hope we'd be able to tell the difference, though.

shinksma


No, that's correct. About 12-14 bits is silence or random noise, even on the best recording; all the active audio is taking place in the least significant 12-16 or so bits; that's why the difference between a 16-bit and a 24-bit recording is barely audible; they're both higher than the usual noise floor, and are numerically identical with the exception of the most significant bits, which are just carrying noise, which is usually dithered to about -88, unless the recording was 24-bit digital from the beginning. Basicaly, bits 17-24 are just too darn quiet to hear (they are all below -96).

You can create a faux 24-bit recording from a 16-bit by simply shifting all the bits right 8, and then padding all the most significant bits with noise. It really doesn't matter, you can't hear it, and it helps with quantization error. 20-bit storage started in about 1993 for master storage since it can do -120db, and since 20 bits is an awkward size to work with register-wise, the industry settled on the daunting 24 bits, which even the best recording ever made could never approach the capabilities of (it exceeds the capabilities of the analog electronics in the preamps.)

12 bits still gives you an admirable 72db noise floor, exceeding the capabilities of all but the very best analog tape recordings.

And a 7-bit encode would sound just like any other, but would be noisy in the quiet parts. I can post some examples if you like. 7-bits still exceed, for example, the capabilities of FM stereo radio. (because of the subcarrier, FM Stereo can only do about a -50db floor. With an SCA it's more like -35. FM Mono is about 80db.
Reply
Reply
post #82 of 95
3/25/08 at 5:20pm
Quote:
Originally Posted by sharkshark View Post

For the sake of making sure we're seeing what's on the discs, can you step through the process for capturing the 6ch files please?

I would think they're about right; certainly it doesn't look that different than a contemporary CD mastering job.
Reply
Reply
post #83 of 95
3/25/08 at 10:03pm
Thread Starter 
Quote:
Originally Posted by sharkshark View Post

huh, I thought Bobby L just threw on some Waves mastering plugz and was done with the deal... Guess those concrete-based speakers are good for sumthin.

For the sake of making sure we're seeing what's on the discs, can you step through the process for capturing the 6ch files please? Especially for SACD, the process, I'd think, would involve taking the analogue outputs from a player and capturing that.

For SACD, indeed it would, and I've done SACD captures that way. I haven't shown any SACD waveforms on this thread (yet).

Quote:


For DVD-A you could demux the audio to specific wav streams, and see the "raw" signal, but if you're capturing analgoue output, re-recording back into you DAW, then it's at least possible some of the banging atop the headroom could be a capture issue, not a source issue.

I do know how to record analog input so as not to 'bang' the signal up against digital max, thanks.

But these aren't captures of analog output, they're rips. The ripping software is called DVDAExplorer.
Reply
Reply
post #84 of 95
3/25/08 at 10:08pm
Thread Starter 
Quote:
Originally Posted by PenteoSurround View Post

No, that's correct. About 12-14 bits is silence or random noise, even on the best recording; all the active audio is taking place in the least significant 12-16 or so bits; that's why the difference between a 16-bit and a 24-bit recording is barely audible; they're both higher than the usual noise floor, and are numerically identical with the exception of the most significant bits, which are just carrying noise, which is usually dithered to about -88, unless the recording was 24-bit digital from the beginning. Basicaly, bits 17-24 are just too darn quiet to hear (they are all below -96).


And it's why hearing that 24 vs 16 bit difference, when it can be heard at all, often requires listening to the fades or reverb tails, with the volume turned up to '11'.
Reply
Reply
post #85 of 95
3/26/08 at 12:59am
Quote:
Originally Posted by krabapple View Post

And it's why hearing that 24 vs 16 bit difference, when it can be heard at all, often requires listening to the fades or reverb tails, with the volume turned up to '11'.

Or 12.
Reply
Reply
post #86 of 95
3/26/08 at 7:28am
Quote:
Originally Posted by krabapple View Post

I do know how to record analog input so as not to 'bang' the signal up against digital max, thanks.


...I was not implying that you personally did not, just that it was a consideration to take when ripping SACD, one that some may not have thought about.

Bang aside, is there any other explanation possible for the clipping? Ie., could DVDexplorer be introducing artifacts in the demuxing/transcoding to WAV? Do you see the same clipping on the m/c waves? For fun and giggles, I'd like to see all 6 ch of the same track with the offending clips.

I have no doubt that there's lots of badly mastered discs, I'm just a little saddened that ANATO is one of them, I've personally always enjoyed my DVD-A, aparently to my ears' detriment.
Reply
Reply
post #87 of 95
3/26/08 at 8:38am
Thread Starter 
Why would the software introduce clipping on rips from some discs,but not others?
Not all of the DVDA stereo wavevforms I've shown, show evidence of major compression and clipping (e.g. Neil Young, Billy Cobham, Deep Purple). Yet they were all ripped by the same software, on the same drive.

I haven't looked at many multichannel waves, because I'm still figuring out how to archive those for streaming. As I showed, the Queen mix appears to be less compressed in multichannel than in stereo. But that's a remix, of course -- a completely new product, not a remaster.

Btw, the software is out there and it's free. Anyone here with a PC and a DVD drive can 'interrogate' their own DVDA collection.
Reply
Reply
post #88 of 95
3/30/08 at 10:23am
Quote:
Originally Posted by PenteoSurround View Post

Um.... the most significant bits are usually carrying just noise. It's the least significant that actually carry the sound. In a 16-bit recording (from an analog source with no noise reduction) the four most significant bits are noise, meaning that it's really about a 12-bit recording.

You've got your least/most-significant terminology totally backwards.

Stop and actually listen to what you're saying. What's the "most-significant" bit? The one with the largest impact! The one that if it changes swings the speaker by 50% of its travel. Not the ones below the threshold of hearing! You can't have noise or silence in the most-significant bits and information in the least-significant bits - you wouldn't hear the information!

In audio encoding, the simplest way to think of it is as binary fractions. Each sample ranges from -1 to 1.

If you've got a 16-bit signal, you've (roughly speaking) got 1 sign bit and 15 bits below the binary point: samples range from -0.111111111111111 to +0.111111111111111.

If you add extra bits, you get extra less-significant bits at the bottom. Those may well contain noise, or nothing, as you say. But that's fine, because they're the least-significant bits. Each new bit gives a finer control as it's 6dB quieter than the bit before. The most significant bits remain the ones closest to the binary point.

Because of your "most/least" reversal, some of your explanations come across as if you're saying that the extra 8 bits are more significant than the existing ones, but don't get used in case they blow the speakers up - a 24-bit sound could be 48dB louder than a 16-bit sound. But I'm sure you don't really believe it's implemented like that!
Reply
Reply
post #89 of 95
4/1/08 at 10:29am
Quote:
Originally Posted by KMO View Post

You've got your least/most-significant terminology totally backwards.


You know, it's funny. I've been making a living with digital audio for 23 years (analog before that) and rarely do I admit that someone has truly blown my mind.
I have made digital audio fly across the country and around the world, untouched by compression, live, more than anyone else on the planet.
But you're completely right. In my brain, I was thinking about it backwards.

It doesn't invalidate any of the dynamic range vs. resolution arguments. I've never really had to think about it that way before (usually I'm working with the sound of some saturated tape transfer) so in explaining it, I was talking off the top of my head.

I love it when I truly learn something significant. Feel free to "neener neener neener", everyone.
Reply
Reply
post #90 of 95
4/6/08 at 7:02pm
Krabapple--how come that Talking Heads song at the beginning of this thread looks soooo bad, but sounds sooooo good? I listened to it for the first time the other night with a friend of mine who has heard many MCH recordings here at my house. As I was thinking "That's the best one I have heard yet", he turned to me and said "That's the best sounding song you have ever played for me here."
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Surround Music Formats

AVS Top Picks

AVS › AVS Forum › Audio › Surround Music Formats › 'hi rez' DVD-A? Sometimes.