|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
In Reply to: no ... you are wrong - reread my article more carefully posted by Christine Tham on June 14, 2004 at 02:13:16:
-6.66dB is a higher peak amplitude than -7.26dB
and -89.1dB is a lower minimum amplitude than -83.67-89.1+6.66 = -82.44 dvda
-83.67+7.26 = -76.41 sacd = less dynamic range.'relative dynamics' and dynamic range are proportional to each other.
Looking at the waveform isn't helpinf a lot as these graphs are rough estimates and not precise at the presented scale.
You can't draw any solid conclusions from these graphs.***PS - And oh, by the way, on the recordings done on the DVD-2200, the CD shows the greatest difference between maximum and minimum RMS - far greater than either SACD or DVD-A. Using your "logic" - CD has the best dynamic range :-) that shows you how much ultrasonic noise (even on DVD-A) influences the results. ***
The ultrasonic noise only affects the dsd stats as it is inheritent with the dsd format.
The tape noise is affecting all three...
The differences you point out are minute and could well be caused by the added hf noise.
Follow Ups:
*** Looking at the waveform isn't helpinf a lot as these graphs are rough estimates and not precise at the presented scale. ***Yes it is - I have the ability to zoom in to any precision in Cool Edit - far more precise than looking at statistics.
You are still confusing relative dynamics with dynamic range. They are NOT proportional to each other. relative dynamics is the difference in dB between any two points in a waveform. Dynamic range is the difference between the loudest signal and the noise floor.
Instead of arguing the point, why don't you do your own experiments and prove me wrong? All it takes is US$30 for the SACD and DVD-A, a US$100 soundcard and a universal player. You can download a trial edition of Adobe Audition, or Sound Forge but as I recall you already have Cubase.
Remember my assertion - any relatively "loud" point in the SACD recording is comparitively louder than the DVD-A in comparison to any relatively "soft" point in any part of the entire recording. Just don't select a soft point less than -60dB or so or you'll run into DSD ultrasonic noise.
The differences in relative dynamics between any two points can be up to 1-1.5dB - audibly significant.
Remember from digital sampling theory what i've empirically measured is actually possible. Remember that the analog reconstruction can have peaks higher than the highest digital sample. So it follows that different sampling rates can give rise to different reconstructions, which differ in relative dynamics.
One point of note: I am *not* saying the SACD recording is "better" than DVD-A, merely that it exhibits higher relative dynamics. It could well be that the DVD-A recording is accurately reflecting the "true" dynamics and the SACD is "exaggerating" dynamics. For example Dolby Digital exaggerates transients - that's why Dolby Digital always sound "punchy" in comparison to PCM.
***Dolby Digital exaggerates transients - that's why Dolby Digital always sound "punchy" in comparison to PCM.***Is this a documented / well known feature of Dolby Digital or is it just a hypothesis?
***Yes it is - I have the ability to zoom in to any precision in Cool Edit - far more precise than looking at statistics.***With cubase the wave representation in generated into a seperate file for performance reasons. This is only a crude wav representation where samples are simply 'decimated'. If you have 1 sec of wav on your screen then you have only 90..120 pixels to represent 96000 samples.
I'm not shure about cooledit or adobe audition. I did play with AA a couple of months ago but didn't look into this.
In any case the vertical resolution you can view is limited by your vertical pixel resolution on your screen.
***You are still confusing relative dynamics with dynamic range. They are NOT proportional to each other. relative dynamics is the difference in dB between any two points in a waveform. Dynamic range is the difference between the loudest signal and the noise floor.***
It really is of no consequense the difference between the loudest and the softest sample is proportional to the difference between any sample.
***Instead of arguing the point, why don't you do your own experiments and prove me wrong? All it takes is US$30 for the SACD and DVD-A, a US$100 soundcard and a universal player. You can download a trial edition of Adobe Audition, or Sound Forge but as I recall you already have Cubase.***
Perhaps I do, but I'm not that interrested in purchasing any more Diana Krall discs.
***Remember my assertion - any relatively "loud" point in the SACD recording is comparitively louder than the DVD-A in comparison to any relatively "soft" point in any part of the entire recording. Just don't select a soft point less than -60dB or so or you'll run into DSD ultrasonic noise.
The differences in relative dynamics between any two points can be up to 1-1.5dB - audibly significant.***
Which might be caused by the dsd noise.
Remember that it is possible to have the higher sample + dsd (peak)noise and the lower sample - dsd (peak)noise.
Remember that a 1.5dB difference at the lower end of the scale only represents a minute voltage change.You can't this measure with a linear rular.
***Remember from digital sampling theory what i've empirically measured is actually possible. Remember that the analog reconstruction can have peaks higher than the highest digital sample. So it follows that different sampling rates can give rise to different reconstructions, which differ in relative dynamics.***
The sample rate isn't of any consequence for the reconstruction method. Only the parameters, such as the filters corner frequency , changes.
***One point of note: I am *not* saying the SACD recording is "better" than DVD-A, merely that it exhibits higher relative dynamics. It could well be that the DVD-A recording is accurately reflecting the "true" dynamics and the SACD is "exaggerating" dynamics. For example Dolby Digital exaggerates transients - that's why Dolby Digital always sound "punchy" in comparison to PCM.
I see too much variables that prevent any sound conclusion about dynamics. Measuring beteen pcm samples is indeed very dubious because as you indicated, the reconstructed waves amplitude differs from the samples amplitude.
*** In any case the vertical resolution you can view is limited by your vertical pixel resolution on your screen. ***Cool Edit allows you to zoom in down to individual samples - it does not rely on a low resolution representation of the waveform.
Cool Edit also allows to you zoom in vertically (as much as you want).
The only thing wrong with Cool Edit is that it interpolates a straight line between samples (but so does every other editing tool i know of). As per the AES paper, this is wrong.
What it should really be doing is drawing the reconstructed waveform (after removing frequencies above Nyquist) between samples. However, this takes way too much computational power. If it really did that, then more people will realise the reconstructed waveform can have peaks exceeding the highest digital sample.
*** It really is of no consequense the difference between the loudest and the softest sample is proportional to the difference between any sample. ***
Your statement highlights your ignorance and the reason why you confuse relative dynamics with dynamic range.
*** The sample rate isn't of any consequence for the reconstruction method. Only the parameters, such as the filters corner frequency , changes. ***
Instead of you parading your ignorance again, why don't you actually read that AES paper I referenced?
*** Measuring beteen pcm samples is indeed very dubious because as you indicated, the reconstructed waves amplitude differs from the samples amplitude. ***
Again, read that AES paper. Who knows, you might actually learn something for a change.
Cubase can show the interpolated wave form between samples.You're 'conclusion' has been exposed as wrong.
Calling me ignorant doesn't change that you made a dodgy conclusion on flimsy evidence.
.
But this paper is a sideline in this discussion.I do recognize that the resulting signal is different from the actual sample value due to the reconstruction filter.
This really means that you simply can't tell much by comparing a graph of a wav that indicates sample values.
That's one off the reasons why I can't dismiss the minimum RMS power and maximum RMS power values. IMO these values ar more telling.
PS. The differences really are minute and could creep in anywhere in the transfer mastering stage and also your soundcard and the used software.
.
Found also another article illustrating the problem.It may be the cause why your samples are clipping.
It can be caused by the dac running out of 'headroom' because the oversampled data from the DF is outside the da converter range or the data is clipped in the DF because the headroom in the DF itself isn't sufficent. (I was referring to the latter)
The analog circuits behind the dac are less likely to cause this trouble. But again this depends on sufficent power rail voltage for the analog circuit. I have seen players with power rails at +/- 8 Volts which leaves not much headroom...
It's possible the clipped waves can be caused in your player.
However your example where the clipped signal level is below a similar signal at a higher level but without clipping contradicts this.What you can do to gain more insight is to record the redbooks spdif out and look if there are signs of clipping similar to the captured wav from the redbook. Look for evidence of clipping below the maximum signal level. If it doesn't clip in the digital output signal then the clipping is caused by your player.
This is a design problem and not a flaw in a format.
(PCM is at an advantage because overflow can be dealt with by having sufficent headroom in processing bits. DSD can't so that's probably why the max out at -6dB)Note that the problem is more likely to occur where high frequencies are at hot levels.
A hot testdisc (cd and dvda) would be a great tool to check a player/dac for this trouble before actual purchase.
If it shows clipping it's just a bad player.It shows once more that any conclusions about dynamic range of this particular dsd and pcm transfer are on thin ice.
It's likely that the difference is caused in your player.Frank
*** However your example where the clipped signal level is below a similar signal at a higher level but without clipping contradicts this. ***Actually, it is also likely that the clipped signal level is at 0dB FS and signals at a higher level are samples of the reconstructed analog waveform at points where it exceeds 0dB FS. remember we are taking a recording of a reconstructed analog waveform, it will be highly unlikely we are sampling at exactly the same points as the original digital recording.
*** PCM is at an advantage because overflow can be dealt with by having sufficent headroom in processing bits. DSD can't so that's probably why the max out at -6dB ***
Your statement is contradicted by those of ted smith and graemme previously on the asylum. given that these people have been involved in designing and building audio equipment, or have direct experience recording and mastering in DSD, i choose to believe them over you.
***Actually, it is also likely that the clipped signal level is at 0dB FS and signals at a higher level are samples of the reconstructed analog waveform at points where it exceeds 0dB FS. remember we are taking a recording of a reconstructed analog waveform, it will be highly unlikely we are sampling at exactly the same points as the original digital recording.***This can be the case. But the wave you showed has **many** samples showing clipping.
The troubles the papers refer to with reconstructed signals between sample points can only occur across a few samples.***Your statement is contradicted by those of ted smith and graemme previously on the asylum. given that these people have been involved in designing and building audio equipment, or have direct experience recording and mastering in DSD, i choose to believe them over you.
***You need to approach this logically. Whatever your guru's have posted about clipping in **recordings** isn't of consequence in this discussion.
With 1 bit at 64*fs you can't turn on (or off) more than that number of bits. So it's ***impossible*** in a pure 1 bit environment to get past 0dB. You simply run out of bits. This is true for recording and processing.
With pcm recording there isn't a need to keep the levels below -6dB.
Maxing out at near 0dB with 24bits isn't creating any problems as long as there are more bits available in the processing environment.
You simply don't need to waist dynamic range during recording.To prevent the problem described in the paper it's only necessairy to keep the maximum output for the production master below -6dB.
In the production stage you can still have the full resolution with pcm.Also this 6dB processing headroom is a big 'waist' of storage space.
With 1 bit is 'costing' you half the number of possible bit patterns.With 24bit pcm 6dB headroom is only costing 1/24th of the available bitspace.
"With 1 bit at 64*fs you can't turn on (or off) more than that number of bits. So it's ***impossible*** in a pure 1 bit environment to get past 0dB. You simply run out of bits. This is true for recording and processing."
I know of no system that can't be overloaded at some point. If you have such a system that can't be overloaded than I suggest you get a patent attorney, and a publicist.
d.b.
This wouldn't get me a fortune.It's very simple once you understand the difference between the recording stage and the processing stage in the production chain.
Create a recording using all the available bits, lets say 24.
Proces the recording in a DAW and apply 12dB gain.
If the processing is done in an 32bit fp environment you are still far from overloading your system.There are already 64bit fp systems becoming available.
This is virtually unlimited headroom in a processing environment.
sorry frank, i don't believe you are right and i continue to choose to believe ted and graemme know what they are talking about and you don't.
I'm right on this one.
yes, of course you are. how wrong of i to have even doubted you. :-)
nt
This post is made possible by the generous support of people like you and our sponsors: