Dynamic Range, Audacity graphs, and High Res Question

[bleedink]

Ok this is a really sorta general question but I get a bit confused with some of the topics that take an audacity graph and debate the merits of what is contained therein. I recently saw a few versions of the new Wilco album and it seems that the 24/96 has the same dynamic range as the CD. To my ears the higher res mix sounds slightly fuller and all that but not much difference. Now the LP has a higher dynamic range (10 vs 7). It tends to sound better. However theoretically at least the LP should have less of a dynamic range right? So I guess my question is when do these graphs and dynamic ranges tell a story of the sound quality and when don't they? For example, some of the higher res files people have complained about sound better to my ears than the cd. However, if you look at the graphs and what not they seem to conclude that the cd will sound better based on these dynamics and brick walls and such. I guess my question is when do these things not tell the whole story? For example is it possible to have a better sounding recording thats seemingly brickwalled? Is it possible that the high res will still sound better despite not having the fuller dynamic range of its cd counterpart? I ask this again as it seems some people are upset about upsampling and such. I've heard some examples that clearly have to be upsampled because the state of recording at the time was such that 50k samples per second (like fleetwood mac used on Tusk) was considered overkill at the time. However, to my ears they sound better than the CD. Obviously my mileage may vary as they say. But I was hoping someone would chime in and help me understand these concepts. I've tried to follow along on some of the graph discussions but frankly am lost in the complexity. Can someone bring the discussion down a few notches for the three brain cells I have left to digest? Sorry to limit you to the three but it appears these are the only ones still working. Thanks!

 

[PeterSt]

Ok, I think this *is* complicated.

 

But let me first say that this honesty looks way better to me that listening to graphs ...

 

About the dynamic range thing : all programs for it will to this differently. I too have my own ways. This is a matter of averaging, do or do not take into account lead ins/outs etc., so maybe it doesn't tell much when different media are compared. But honestly I don't know. It depends too much on the "calculation" in the first place.

 

About the brick wall (shown) filtering ...

At least I don't say this will be a bad thing. Listening to higher frequencies ? I must laugh about it. Those who claim it sounds better (and include the graph) most often include graphs of which I "state" it shows a wrong thing. In the mean time everybody agrees it looks like the good thing.

I gave up.

 

We can well say you are just right on your listening (!) experiences.

But it depends on so many things ...

 

Take my own example : a DAC which doesn't touch the sound when it is about hires (it shouldn't) and I claim about all of the hires sounds (!) wrong. Otoh, since the DAC doesn't touch the sound, redbook *needs* that, and thus in software something has to be done about it (think upsampling).

Therefore no comparison is possible between what you perceive and what I do.

 

In the mean time each DAC behaves differently when it comes to filtering. This matters with the Hires (it should leave it alone) but it even more matters with redbook (it should do the best job there).

 

So, how to compare ?

Skip the graphs. They are hard to interpret, and the only thing I can tell is that I 90% of times don't agree with what others say about it.

 

What is this worth ?

Listen !

 

Peter

 

[Miska]

Now the LP has a higher dynamic range (10 vs 7). It tends to sound better. However theoretically at least the LP should have less of a dynamic range right?

 

It is one thing how much some media has dynamic range and another thing is how much of it is used.

 

Most of today's popular music is so compressed, that it would sound the same with eight bits or even less. When the level variation is within 3 dB range (like some local FM stations here), it's almost like noise. When there are no silent passages between all the time clipping noise, it is really hard to notice that the noise floor is at -60 dB instead of -100 dB.

 

So, with current loudness wars, even most of RedBook resolution is wasted, it would be capable of so much more.

 

 

[bleedink]

Thanks for getting back. So is there no relation to say dynamic range and sound quality or are you saying that the averages don't take into account transient ranges that may trump the average that the program calculates? Im not sure I understand your stance about all high res sounding wrong. Heck I consider some LPs to be high res with some of the other material around. Anyway, I definitely tend to trust my ears though I am not a strict traditionalist in terms of audio consumption. I tend to like the spaciousness of surround recordings and have found the interpretations of albums like say for instance the new DSOTM mix to be very illuminating and light years in terms of sound quality from what I am used to hearing, though nearly every version of that album sounds pretty good. I'm really interested in hearing/seeing some graphs/music that bucks the trend...say the lesser dynamic high res versions sounds better than say the cd version of said album even though its dynamic range is much better and has supersonic audacity spectrographs...nearly every high res recording I've heard have one thing in common for me and that is that their bass seems to always be more present. I can't think of too many recordings I've heard where I prefer the bass of the cd version to the DVD/SACD. Perhaps not so with vinyl but I guess that would be another discussion.

 

[Miska]

say the lesser dynamic high res versions sounds better than say the cd version

 

Dynamic range (difference between loudest and quietest passages) is one thing. Then other things like transient attack reproduction accuracy (function of true bandwidth) is another thing.

 

Graphs are good for checking if you are getting authentic hires. How it sounds like depends so much on all kinds of mastering details in addition to the source format.

 

bass seems to always be more present

 

Sounds like modern frequency-banded compression - more "on your face"?

 

The old analog compressors are not as capable on the eq'd compression of modern digital algorithms.

 

Another thing is that there's no way to have completely fair way of comparing two resolutions. It depends so much on how the particular digital -> analog path behaves at each bit depth and sampling rate. In almost all cases for example the DAC's analog stage is the same for all sampling rates.

 

 

[bleedink]

"Ok, I think this *is* complicated."

 

Well said!

 

I guess this always comes down to what your ears like. But it is good to hear some of my subjective observations are not completely independent. By the way isn't the dynamic range of like SACDs and such something like 120db? I guess a dynamic range of 7 doesn't seem so high when you take that into account! Sort of unbelievable that one gets better dynamics on vinyl to this day for the most part. I understand that the source has a lot to do with it as well, but still! Are there any popular recordings that are incredibly dynamic? I take it there aren't too many rock recordings out there with a 20db dynamic range is that correct? Oh and thanks to all of you for keeping this well within the range of the capabilities of my three remaining brain cells!

 

[esldude]

Look at the attached Audacity file screenshot. At 96 khz I generated pink noise. In spectrum view it has content up above 40 khz. The second half of this tone has been brickwall filtered at 10 khz. All the blank space above 10 khz tells you there is no content up there. If this were seen on a 96/24 hires music file you would know either the original content had no higher frequency content or it has been filtered. If you saw response to 20-22khz and blank space above that you might figure the original 44.1 sampled CD file had simply been upsampled. If that space occurred above 24 khz you might think studio files at 48 khz had been upsampled. In either case you probably don't have a hires source. Now brickwall filters don't much have anything to do with dynamic range as they are a separate issue.

 

Also Fleetwood Mac's Tusk was recorded in 1979 and almost surely was recorded to analog tape. So a hires digital version from the master tape should be very good. But it might not show any real content above 30-40 khz as the tape would possibly not have response beyond that. It might not even have any real content that high. Dynamic range or more specifically signal to noise would likely not be even 16 bit either. The tape noise would have caused a noise floor you wouldn't get below even if you recorded the master tape in 24 bit.

 

[PeterSt]

I am glad you see something in that picture !

But I don't.

 

If you saw response to 20-22khz and blank space above that you might figure the original 44.1 sampled CD file had simply been upsampled.

 

That is your idea about it. Not mine.

Not of any importance, but at least this tells how I can look different at these graphs, with obviously different conclusions also.

 

So a hires digital version from the master tape should be very good. But it might not show any real content above 30-40 khz as the tape would possibly not have response beyond that.

 

Change that into 20-25Khz and you will interpret your next graphing experiment better.

 

I look like nit picking, but again, this is how the interpretations start off wrongly.

An of course I can be wrong, why not.

 

 

[PeterSt]

[although it may look like it, I am not talking to Miska]

 

Yes it is complex, and yes, again there is more going on than we normally might think of;

 

Dynamic range (difference between loudest and quietest passages) is one thing. Then other things like transient attack reproduction accuracy (function of true bandwidth) is another thing.

 

The first sentence tells that the dynamic range figures don't say much for an absolute measure. A measured high DR only tells that the distance between the lowest signal and the highsest is "high". But a low DR does not tell that the recording is poor(ly mastered). It could be, but it also could be that no soft passages are in there.

(and again, think about how this is measured, like 0.1sec of silence could incur for a high DR reported -> wrong of course).

 

The second sentence from above quote indeed is about another thing, as long as we understand that it is 100% related;

So, how to create a "long" steep transient when the DR isn't there in the first place ? It is like Miska told, "it is a function of the bandwidth" - and don't underestimate the merits of this ...

 

... for maybe more poor sound with higher bandwidth ...

 

In other words :

One of the greater dangers with a higher dynamic range (the amplitude bandwith over which all the music spreads) is that the implied higher steepness of that can not be coped with by your chain (and as we know, one part not being able to follow is enough to destroy your music experience).

 

While one may perceive this as logic, it gets more complicated when we again incorporate that filtering, which will smear the transients (ahead of the happening, after it, or both). And :

 

While again this may come across as logic, it is *now* where things start to be really fuzzy, because the filter will respond on the transient itself, hence will smear the transient when it is there, and will not do that when it is not there. Or IOW :

 

The distortion coming from the ringing of a filter will be a function of that bandwidth available; the lower the bandwidth, the lower *that* distortion.

And no, I dit not say at all that thus the lower available bandwidth (lower DR in the recording) is to be preferred. But I will try to spend a small next post on this one.

 

Peter

 

[PeterSt]

[with apologizes I have to use my own stuff for the example]

 

So now think further;

 

What I do with my DAC is "not allowing for" any of those transients to be smeared. And please, the debate whether this is a good thing or not is not for this thread; it is about the idea of what happens further (and yes, now I was addressing Miska :-).

 

So imagine, there's this recording with a good dynamic range in there. This especially means that it is not compressed (the too high peaks for the digital domain involved (but just think redbook) are sustained, which especially means that the flow of the music will be at a fairly low level (could easily be 30dB under max) which *thus* allows the transients to stick about above that level (-> transients can be peaks of 30dB).

 

This is huge.

With the NOS1 this really happens. But now what ?

It theoretically implies that 1 sample transients can occur; not in really-real-life, but since synths can do anything, it may. And otherwise think a few more samples. Anyway, no speaker will be able to cope with this, right ? Not even your amps will. But in my case, the NOS1 sure does (measured).

 

I can tell you, that this encourages for problems. It went too far. Over the top.

 

We can go as far as thinking that where the amplifier is able to follow (doable, and mine does !) an *additional* probem will be created in the loudspeakers.

The high transients now encourage for distortion there ...

 

The debate could hop over to "better cables", knowing that better cables do nothing much more than fooling you because they filter. This filter is not a good thing because it is out of your control, and will never match what is exactly to filter, except for some smearing will work out for the better (mind you, in my proposed situation).

 

It is here where the lower dynamic range of a recording comes handy. It will sound better, because it can't imply the distortions described.

Is it a good thing ? no, of course not. The real solution lays in improving the speaker (once the amp can follow).

 

Again think further how a hires recording may sound better in your system, knowing that you won't be using my DAC but that other one with nice official filtering which smears. Don't right away think that this "smear" is a bad thing, because the filtering officially is a good thing. However, now suddenly it *is* about smeared redbook against not smeared hires at that level. All now becomes apples and oranges because the hires is subject to all kinds of other "smears", but think wrong appliances (like downmixes which nearly all DVD-A's are, that 100% guaranteed implying (close to 12dB) compression). But :

 

When *I* listen to redbook, and assumed that my filtering is OK only because I claim so (allow me that for this time), the redbook sustained all the transients, while the hires is as poor as it was delivered to me. Nothing to do about it anymore. Gone = gone.

 

And the latter is all in the realm of the now possibly distorting speaker ...

 

See ? this is how we all will perceive audio different; not because of our brain, subjective listening, moods, name it - but because it *is* different. Physically.

 

Peter

 

 

[Miska]

By the way isn't the dynamic range of like SACDs and such something like 120db?

 

SACD technically can have about that, or actually a bit more with recent encoders. But would any real world SACD disc have that kind of dynamic range in the music material is another question.

 

Are there any popular recordings that are incredibly dynamic?

 

I have number of SACDs with classical music that have everything from audience whispering to strongest passages of Mahler or Shostakovich symphonies. That's still not 120 dB.

 

Let's keep dynamic range of a media; distance of noise floor and loudest possible sound, and dynamic range of a recording; difference between quietest and loudest passage, separate.

 

The way these are commonly determined are different and shouldn't be directly compared.

 

 

[Miska]

for instance the new DSOTM mix to be very illuminating and light years in terms of sound quality from what I am used to hearing

 

I don't have the new version, but have they changed the relative loudness of the "hidden message"?

 

I have the first CD release (?, Capitol label) and the SACD version. And at least the RedBook layer of the SACD is compressed compared to that older CD release and the "hidden message" is now louder and much less "hidden".

 

That older CD is much closer to the LP I have.

 

 

[PeterSt]

I wondered before, and Miska's post about the theoretical 120dB of SACD makes me think of this again. A bit controversial maybe, and possibly with wrong perception :

 

So, the 24 bit recording allows for a higher dynamic range, were it about digital of course.

Regularly I talk about the dynamic range we can perceive during real life listening, that not being more than 70dB ever (which is my own empirical (explicit) finding).

 

This 70dB (or whatever it exactly is), is taken into account by the recording engineers (I recall a small discussion about this between me and Barry Diament somewhere). So :

 

Assuming an available DR of 120dB (just because it was mentioned in the before posts), it theoretically allows for no compression at all, because, well, it is not needed. Not because music will never exceed the 120dB but merely because we don't need more than that 70dB anyway.

 

But it is not working like that ...

 

When the 120dB really would be applied, it will be the most easy to indeed incur for recorded music with way more than this 70dB. Not 120dB immediately, but think 100dB or a little more (a rockband's drumkit will produce over 110dB fairly easy).

 

This means that the uncompressed (!) recording will play at such a low level (the rock drummer grabbing some brushes at last) they are inaudible. Now what. So, you crank up the volume so they *are* audible. But then he takes his sticks again, and there go your windows. Not because it is 110dB but because you had to tune the brushes into, say, 50dB so they are audible, and now the original 110dB smash will be in the area of 50+110dB (this is wrong math, but the idea is that it will be more than 110dB).

 

So, this violates heavily my 70dB of dynamic range we or our windows will be able to take (mind you, which is 70dB of dynamic range, and not SPL; this latter can be something like 120dB indeed).

 

While the above is one means of reasoning why a recording won't be constructed like this, the recording engineer has a few other reasons. They are in the noise area, and not loosing bits unnecessary. Well, never minding this too much - what it comes down to is that compression will be there always. To some senseable extend. So, highest peaks may be sqashed 20dB or so, and now all plays nicely in that 70dB range, in the mean time the SNR being quite okay.

 

To emphasize what I mean : I would not allow my recordings to contain a DR of 120dB. Or I'd never hear the soft passages anyway, and since I wouldn't know about what comes (and I don't like Mahler :-) my windows may go out.

 

So here too, some more is going on, and it might be unexpected ...

 

[Miska]

Not because it is 110dB but because you had to tune the brushes into, say, 50dB so they are audible, and now the original 110dB smash will be in the area of 50+110dB (this is wrong math, but the idea is that it will be more than 110dB).

 

Peter, as you said the math goes wrong here. Let's play with some numbers. If the drummers maximum is 110 dB, and if the brushes are 50 dB, then you have 110 - 50 = 60 dB dynamic range in use for the music material.

 

Let's assume also that it was recorded in fairly quiet environment with 30 dB background noise level. Now you have 110 - 30 = 90 dB dynamic range in use.

 

If you play at those volumes at home, which would have the same 30 dB background noise level, you would still have 16-bit digital noise 6 dB below your home's background noise. If you turn down the volume and play only 90 dB at the loudest point, digital noise floor is lower (26 dB below the background) and the brushes are at 30 dB background noise level. If you used only 16-bit digital volume control, then the digital noise would stay at that 6 dB below background.

 

 

[PeterSt]

Ok, let's see. And btw, this is not exactly about the bit depth -> DR argument. It is more the windows breaking argument ...

 

The real life rock drummer plays at ~110dB SPL.

We suppose his brushes play at 50dB (I never tested exactly *that*, to I assume it, and it is about the idea).

Also it is assumed that my playback means allows for a dynamic range of 120dB.

 

I put on the album, and the stupid rock drummer starts with his brushes. I perceive them at 50dB of SPL and I don't like it much. Too soft. So I put up the volume by 10dB. Now his brushes play at 60dB. Of course this is 10dB more loud than reality (50dB assumed).

Next the drummer digs up his sticks and starts to play at his loudest level, which is 110dB reality.

Obviously my speakers now will produce 120dB of SPL, assumed they can do it.

 

This is nothing I would like much. :-)

It is my own fault, because I put the volume so loud, but how could I know ...

 

What I am saying is that the distance between the 50dB of the brushes and the 110dB of the sticks -which is 60dB- is squeezed during the recording into ... make up something. Could be 20dB which seems fairly normal to me.

 

Again the album starts playing, and again I turn up the volume until 60dB of SPL comes to me.

Again the drummer takes his sticks, and now he plays at 80dB through my speakers. Nice.

Not reality at all, but nice.

 

That the *reason* to record like this is avoiding being into the noise level, is something coincidental, but also convenient. See the "Nice".

 

Of course my example can be exaggerated, because nothing tells us that the brushes play at 50dB with an environmental (or molecule) noise of 30dB. It can play at 31dB obviously. This, while the maximum still can be 110dB. Nothing changes about the latter, while the former just means more soft playing by the drummer. Now the dynamic range implied is 79dB, but *still* I hear nothing and put up the volume so the softly playing drummer comes to me as 60dB again.

This time 60+79 = 139dB and with some nice (24 bit resolution system and) amps now my windows won't break but my speakers do.

 

I don't think this is stupid theory, but just practice. Unless we find a couple of albums in our collection which really exhibit this recorded dynamic range.

 

My whole point is :

I never saw a real difference in utilized dynamic range for 24 bits over 16 bits. Maybe this already starts the way *that* math works, knowing that conversion from 16 bits to 24 bits the most officially will create the same SPL for -0dBFS. It just oughts to be like that. This, while it is not necessary at all (we could just sustain the 32767 level) in order to next add on top of that (48dB !).

 

Looked at this from the latter angle, I'm the most sure that no recording engineer operates like this. Still it can be done at the cost of the additional resolution (now lacking). So, the other way around again, the additional headroom is used for resolution (a good reason btw), and the additional dynamic range (headroom) is not increased because of that.

Still it can be anyhthing in the middle (a little for resolution and a little for higher dynamic range), but *still* it is not done like that.

But *when* done, there go my windows.

 

In the end it is and remains related to the 70dB DR we can perceive (ok, determined by me). So, we set the volume such that we

a. hear something and

b. won't blow anything during the track.

 

Ad a.

In near 100% of the time we play much softer, so we miss all the soft little details anyway. This is not because we want to miss the details, but playing against window-breaking is not what we do most of the time. Well, I don't, although my wife usually tells me different.

 

The (stupid) moral remains : when no compression is applied we will only loose more soft details. With compression the soft details can be heard the same time as the loud smashes. The less "DR bandwidth" the better that works.

And the more lousy it sounds of course !

 

Peter

 

PS: I can recommend Turtle Records for the highest utilized dynamic range, measured by transient steepness (just redbook). For their recordings they used a special reel to reel machine of which I didn't quite get the details but to me it looked like a 60's device (which was still in use a few years ago). Maybe Chris knows the details, or made a photo of it a few weeks back).

 

 

 

[Miska]

I don't have anything else to add, than that some classical SACDs I have have uncompressed DSD layer while having compressed CD layer. By rough guesstimate this matches roughly range from 40 dB to 110 dB.

 

CD layer is compressed because it is becoming a bit too close to the resolution limits since the average level would be most of the time so much below 0 dBFS level. (Noise 20 dB below the signal is already audible hiss.)

 

When I'm listening to the DSD layer of those, I need to turn up the volume 20 - 30 dB higher than when listening Pink Floyd. And then the loudest peaks can "shake the room".

 

Just a few weeks ago I was at concert of Mahler's 6th at the brand new Helsinki Music Centre. And the dynamics in this acoustically very good space sounded absolutely tremendous. Try to reproduce something like that at home...

 

 

[Sri]

The only way to reproduce a concert hall at home is to invite friends who will cough at inappropriate times or my favorite beef start clapping at the end of Tristan when the curtain starts to come down but the orchestra hasn't stopped playing.

 

[wgb113]

All things equal Hi-Rez sounds better to me.

 

When you start tinkering with Dynamic Range it becomes more of a preference, though I tend to favor more dynamic range over higher resolution with lower dynamics.

 

A recent example is the new HiRez version of Nirvana's "Nevermind". I prefer the Mobile Fidelity CD version and even the original Geffen release on CD to the new HiRez download.

 

Something like McCartney's "Band On The Run" I appreciated the cleanup of the 25th Anniversary CD remaster over the original CD release but the unlimited HiRez download trounces both.

 

IMO we should be pushing for better dynamic range before or at least in conjunction with higher resolution. Most rock recordings with a DR rating of less than 7 simply sound like noise and are worse the better you playback system.

 

Bill

 

[esldude]

Often small amounts of compression are perceived as greater dynamic range. A good many listeners prefer this as the music sounds a bit more energized that way. They will deny that is what they are hearing most of the time. Wider uncompressed music will sound a bit dead to many listeners. I think this is one of the attractions of tube amps coupled through transformers. Those often compress the sound as a side effect (among other things) as they are driven near their limits. LP's are usually one way or another compressed vs the master tape just to get the track cut into the vinyl.

 

Even so, darn few recordings come close to using up even the 96 dB of dynamic range available on the CD. Differences in recordings as always are most often just down to how they are mastered. Hires recordings may well be better due to different mastering as much as anything else.

 

[esldude]

"I am glad you see something in that picture !

But I don't."

 

It was meant to show what filtering looks like in Audacity. It does that pretty clearly. Can you not visually perceive the difference, I am pretty sure you can? I was saying nothing about what that sounded like, only what it looked like in a spectrum graph in Audacity.

 

"If you saw response to 20-22khz and blank space above that you might figure the original 44.1 sampled CD file had simply been upsampled.

 

That is your idea about it. Not mine.

Not of any importance, but at least this tells how I can look different at these graphs, with obviously different conclusions also."

 

Actually it is not my idea or yours. It simply is what would be shown in Audacity if something were brickwall filtered by being derived from a 44.1k redbook file and then passed off as hires by being resampled to 96 khz. As that was one of the OP's questions it seemed more pertinent than whatever you are talking about here.

 

 

 

[Jud]

The debate could hop over to "better cables", knowing that better cables do nothing much more than fooling you because they filter. This filter is not a good thing because it is out of your control, and will never match what is exactly to filter, except for some smearing will work out for the better (mind you, in my proposed situation).

 

The top cables from the manufacturer I favor are not "better" in the sense of having lots of Teflon dielectric (in fact they have no Teflon at all; they use a very loose mesh shield, which means the dielectric is almost exclusively air), specifically so that they do not have the dielectric "pumping" that results in smear. Though I have wondered occasionally what the timing is like for the dielectric pumping effect (absorption and release of electrical energy), and whether it occurs slowly enough to result in smearing or other audible effects for either digital or analog cables.

 

[PeterSt]

Please sort out the difference between "brick wall filtering" and "filtering". Next try to recognize the difference in e.g. Audacity. The "filtering" with hires of course. Start with DSD.

 

And might you think it can be seen by that way too low binned spectagram graph you put up, or the way too tiny FFT picture ...

No.

 

But it isn't about these sort of "mistakes". It is about the conclusion people seem to draw from such mistakes.

Besides, it's for the discussion only. Way on topic.

 

Regards,

Peter

 

[esldude]

"Please sort out the difference between "brick wall filtering" and "filtering". Next try to recognize the difference in e.g. Audacity. The "filtering" with hires of course. Start with DSD."

 

Well there is filtering and brickwall filtering is filtering done at extremely steep rates. Rather than sloping off gently like a first order filter it drops abruptly over a very narrow frequency range. Which is exactly the sort of filtering applied in the screenshot of Audacity I reported. Again the OP asked about interpreting graphs from Audacity. DSD wasn't mentioned though I fail to see what relevance it has in general in this discussion. I get the feeling you are having some discussion amongst yourself without us on this forum reading more than half of it.

 

If you have some better illustrations about the OP's question of interpreting Audacity results then how about putting some of them up so maybe it will be clear exactly what your point is.

 

Just for comparison I am attaching a screenshot where pink noise was done at 96 khz and the second half here was filtered at a 2nd order 12 dB per octave rate. Along with the other version with a brick walk filter.

 

[bleedink]

I first of all thank all of you for providing me with far more information than the three brain cells I have can actually process. It certainly is not as simple as I would have liked! But, I'm not sure that anyone has really provided me the examples I was looking for...perhaps they have and my brain just doesn't comprehend it. Mainly I am looking for instances where the graph would indicate a poor recording but in fact the recording sounds better than a version that "looks" like it should. Though I must thank you all for breaking down what some of these graphs are actually telling me. Im just trying to resolve this "cd version" is better than the high res version because of a graph when listening to the title indicates to my ears at least a deeper and more preserved original state. Because as a few others have chimed in I have never heard a high res recording where at least the bass was not significantly better each and every time, upsampled, brickwalled, whatever. Always sounds deeper, more engaging. So to my mind these graphs don't tell the whole story. Not that the high res is better every time but I've heard few that were not wholly better sounding than their cd counterparts and not just because the dynamics had been quashed out of it. This would be the effect of bandwidth as Miska stated right? Oh and just to be fair I read up about the Tusk recording. It was done digitally with a sample rate of around 50k. Again in 1979 this was considered overkill. The recording kit they used is apparently so good that some SACDs have been released directly from these masters. There was one pressing done that was made directly from the digital masters. An interesting aside: if 16/50 was considered overkill around 1980 I think we can safely assume that most digital masters around this time period are 16 bit at best. This basically confirms what many of us have thought...ie that early digital recordings that are marketed as high res have to be upsampled as the recording at that time in the digital realm simply did not allow for 24/96 or what have you.

Here is more on Tusk:

"There is, however, one interesting conundrum: It was an early digital recording. The Soundstream digital 16 track recorder was one of the first to emerge in the late 70's, preceding the more widely used Tascam, Sony and 3M machines that accounted for most digital recordings made between 1980-85. If for no other reason, this recording has suffered near total rejection by analogue purists. Denounced by audiophiles as nowhere near as good as albums made using the prevailing analog tape systems of the day. Perhaps it's time this was re-assessed. The Soundstream machine was designed to record wide dynamic range classical music while its use in rock and pop was, up until then, virtually unknown. It had huge headroom designed into it and perhaps far more importantly, recorded at 50Khz/16bit, a noticeable increase in sampling quality over standard CD. It was even seen as astronomical overkill at the time. 50Khz was higher than the not-yet designed DAT tape system (48/20). The system also had a very hi-spec mixdown capability all running at full spec. This was an excellent machine built without compromise! Teldec, known for their early digital recordings of orchestral music, have even started to release some of their old Soundstream recordings on SACD, such is the quality of the originals. So what happened with 'Tusk'? The difference I suppose lies in the inflexibility of the Soundstream system; where it was 'rock' standard to use 24 track multi-channel with several mixdowns and send-returns ('wet'), this system needed to be run mostly 'dry' and was limited to 16 channels of pure digital. It wasn't possible to create an album that 'sounded like' Rumours. Who knows, perhaps it is because the production and the general presentation of the sound was rather 'flat' as a result, or perhaps it was because the wild 'glossiness' of the 'Rumours' album was entirely absent, but somehow Tusk has never got it's fair recognition in the world of audio."--from a blurb about this release

BTW this particular recording has a dynamic range of 13

 

@ESLDUDE I guess I see what you are saying but Im not sure what your graphs are showing me or what I was to get from it.

 

[bdiament]

Hi bleedink,

 

"...However theoretically at least the LP should have less of a dynamic range right?..."

 

In the case of CD (or 16-bit), I believe a lot of folks tend to confuse signal-to-noise ratio with dynamic range.

 

While one can, in theory record a very low level signal on CD, what often is not taken into account, yet is quite plainly audible, is that those low level signals will be significantly lower in resolution than those near the top of the dynamic range.

 

A good orchestral recording will have an average level ~20 dB lower than the maximum peak. If the maximum peak is represented on the CD (or other 16-bit recording) as something close to 0 dBFS, the average level material, at ~-20 is going to be encoded with somewhere between 12 and 13 bits of resolution. (16-bits only happens in the top 6 db... or 6.02 for those who are counting. ;-})

 

The lower level harmonics of most instruments, as well as the ambient sound of the recording locale itself will be still lower in level and hence be encoded with still lower resolution.

 

This is a large part of why many hear a bleaching of instrumental harmonics, a diminution of spatial cues and an overall coarsening of the sound with 16-bits.

 

The same recording on vinyl will not suffer any of this. Granted, any surface noise in a less than good pressing will be louder relative to the lower level material but the sound of the violin will not change the way it does with 16-bit and the sound of the auditorium will not be foreshortened as it is with 16-bit.

 

So ultimately, in my opinion, the idea that a 16-bit format has greater dynamic range than vinyl is something that is not borne out in the real world of the listening room.

 

There are other aspects too. One, actually in favor of 16-bit (sometimes) is that steep low frequency dynamic changes are better represented in the digital medium -- IF they are close to full scale, level-wise. On the other hand, vinyl seems to display more dynamic gradations, particularly in the middle and especially at the bottom of the dynamic range.

 

Of course, with a good 24-bit recording, none of the above applies.

 

All just my perspective, of course.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com