Some more crap from HDTracks?

[Manelus]

This is the spectrum from Shostakovich Symphony n.8 Pentatone, downloaded from HDTracks:

 

ALL Pentatones I have purchased, supposed to be 88.2 show no freq. above 22kHz.

The same applies to mDG ones from HDTracks.

 

Not bad sound, but definetly no 88.2 (If I have understood well).

I hope they are at least 24bits (Is there a way to check it?)

 

[wgscott]

It might actually be ok, because you don't see any hard cut-off like you do in the "Lush Life" example I posted awhile ago. Better that than the gain cranked to saturation. If you renormalize the plot, my guess is you will see signal reaching higher in frequency.

 

I've just started to download Beethoven's 5th & 7th from HDtracks, 24bit/88.2 kHz. My first concern was reading the liner notes, which says it is a 24bit/96kHz SACD, which is what that Lush Life one was. So why 88.2 kHz instead of 96kHz?

 

Anyway, here is what the first movement of the 5th looks like:

 

 

It looks like it extends slightly beyond 22kHz, but above that is noise. I'll take what you have over this any day.

 

[jhwalker]

Maybe I misunderstand what's going on here, but *should* there be any content beyond 22kHz? There are no orchestral notes that go anywhere near that high, and even overtones thin out *very* quickly. I'd be surprised if there was any significant content above about ~16kHz on *any* recording, regardless of the sampling frequency. I thought high sampling frequency was really about achieving a smoother wave function for sounds, not recording sounds so high no bat can hear them?

 

[bdiament]

Hi jhwalker,

 

"...Maybe I misunderstand what's going on here, but *should* there be any content beyond 22kHz? There are no orchestral notes that go anywhere near that high, and even overtones thin out *very* quickly. I'd be surprised if there was any significant content above about ~16kHz on *any* recording, regardless of the sampling frequency. I thought high sampling frequency was really about achieving a smoother wave function for sounds, not recording sounds so high no bat can hear them?..."

 

There are no orchestral fundamental notes up that high but there are plenty of orchestral harmonics that reach well beyond 20 kHz. For example, a muted trumpet can put out healthy amounts of 90 kHz!

 

I don't know about the recording that this thread is about, so I can't comment. But in view of some other threads along these lines, I'd say that generally speaking, as far as I'm concerned, a 22(.05) kHz limit means a recording was originally made at 44 kHz. (I don't care what anyone claims about deliberately adding filters to high res recordings; to me that is, to put it nicely, simply b.s.) It might be a fine sounding recording but then it is a fine sounding 44.1k recording, nothing more.

 

In the high res recordings I make, there is clearly content, low in level though it might be, well up toward the Nyquist frequency. My last recording, "Equinox" was made at 24/192 and shows content well up to 90+ kHz with significant harmonics well above 60 kHz.

 

Removing this content, even by taking the sample rate down to 96 kHz, which will still preserve information well above 40 kHz, results in audible losses. Clearly, I cannot hear 40 kHz sine waves (or even 39 kHz ;-}) but that doesn't mean there is no contribution down within the audible range.

 

KEF did research many years ago which showed that in order to preserve proper timing information, the required bandwidth must (according to their study) extend 5x beyond the "important" frequencies.

 

If a vendor isn't clearly saying their so-called "high res" files are not upsampled Redbook (when this is the case), from my perspective, they're being less than honest. However one feels about upsampled 44.1, it isn't at all the same as true high res.

 

As always, I suggest no one simply take my word for this. Try it yourself with a good converter: make a 2x (88.2 or 96k) or 4x (176.4 or 192k) recording and then record the same thing at 44.1k and upsample the latter with the finest SRC algorithm you can find. Then compare the two and draw your own conclusions.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[jhwalker]

@Barry -

 

Thank you so much for the detailed explanation. I really appreciate learning more about this stuff and more-experienced folks taking the time to educate.

 

In a well-engineered, 192/24 recording, would you expect to *see* the content above 20kHz on a soundwave chart, though? IOW, I believe you when you say it's there, but are the charts (examples provided above) a sophisticated enough tool to make a conclusion one way or another whether or not there is substantial high-frequency content? If not, what would prove it one way or another?

 

[wgscott]

For a 96kHz sampled track, for example, you often can see very clear signal all the way up to 48kHz. There are some examples posted in the other threads where this is discussed.

 

This is how it looks when things have gone wrong:

 

 

This is how it looks when it is working right:

 

 

and, even cleaner:

 

 

[Akapod]

To my (pretty limited) understanding, microphones are designed to be flat only through 20KHz. Obviously, manipulation of the recording could create tones above 20 KHz, but that doesn't seem the same.

 

[bdiament]

Hi jhwalker,

 

"...In a well-engineered, 192/24 recording, would you expect to *see* the content above 20kHz on a soundwave chart, though? IOW, I believe you when you say it's there, but are the charts (examples provided above) a sophisticated enough tool to make a conclusion one way or another whether or not there is substantial high-frequency content? If not, what would prove it one way or another?..."

 

Even something like the free Audacity will show this in a rudimentary way, as can be seen by the many screenshots folks have posted.

 

Something like SpecraFoo will show this in much more elaborate ways but as can be easily seen, Audacity does the basics well enough.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[bdiament]

Hi Akapod,

 

"...To my (pretty limited) understanding, microphones are designed to be flat only through 20KHz. Obviously, manipulation of the recording could create tones above 20 KHz, but that doesn't seem the same..."

 

Actually, I'm not aware of any microphone design that would be deliberately bandlimited; its just that many have diaphragm resonances resulting in mid-treble peaks, followed by a roll off above that.

 

There is no need for (nor do I know anyone that would or any device that would perform "manipulation of the recording" to "create tones above 20 kHz". The very high frequency information exists in the real instruments and properly mic'd, will be captured in a wide band recording.

 

My Earthworks mics have frequency responses that look more like a power amps; they're "flat" to 40 kHz. The same company makes mics that are flat to 50 kHz. (With this in mind, consider the fact that the very high frequencies evident on some of these basic scans with Audacity, of real high res program material, are at a significant enough level to be recorded even while the mics are already rolling off. In another thread, someone put up an Audacity scan from a track on my latest recording, "Equinox", showing audio well up toward 90k with significant information just under 70k - this is not surprising with the percussion and trumpet.)

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[Akapod]

Appreciate the info.

 

[jhwalker]

Thanks, wgscott and Barry - really appreciate your taking the time to walk us (me) through it

 

[Russell_L]

"I've just started to download Beethoven's 5th & 7th from HDtracks, 24bit/88.2 kHz. My first concern was reading the liner notes, which says it is a 24bit/96kHz SACD, which is what that Lush Life one was. So why 88.2 kHz instead of 96kHz?"

 

I've read that all (or most of?) the 88.2 kHz files on HDtracks are done by Bruce Brown out of Puget Sound Studios in Seattle, directly from SACDs sent to him by HDtracks. The 96 kHz files on HDtracks come from other sources, I believe (like directly from the record label). With respect to the Kleiber Beethoven, I'm presuming that the original analog tapes were converted to 96/24 by DG and then converted to DSD for the SACD, which was then converted to 88.2/24 for HDtracks. Would have been nice to have the 96/24 file instead of this double-converted one! (Actually, the 96/24 tracks are available on DG's DVD-Audio release, but this disc is long out of print and it might be difficult to extract those tracks in any case...)

 

BTW, how does this recording sound? I've got the SACD and of course I could rip it, but only the CD layer at 44.1.

 

Russell

 

[wgscott]

I am ashamed to say I haven't yet listened. Hopefully that will change today.

 

[Manelus]

He we have an example of outstanding recording (Mozart Divertimento trio K.563 from BIS Records downloaded from eClassical).

It sounds great, and the spectrum is clean. I wish more Record labels - Online stores would follow that quality path.

 

 

 

[wgscott]

Could you play with the gain on your first example, just to humor me? (I had to do it from Preferences, although there might be a better way.)

 

[Manelus]

I'm not at home. I'll do tonight (Europe), and post the resuts.

I never "played" with gain, which gain do you suggest for this one?

 

 

[eggers]

I've just started to download Beethoven's 5th & 7th from HDtracks, 24bit/88.2 kHz. My first concern was reading the liner notes, which says it is a 24bit/96kHz SACD, which is what that Lush Life one was. So why 88.2 kHz instead of 96kHz?

 

SACD is DSD which is 1 bit 64F (44.1) = 2.8224 MHz. That can be converted to approximately 24 bit @ 88 kHz. Going to 96 kHz is mathematically pointless (as far I understand it) as you only get noise in the extra frequency range.

 

In your image:

http://www.computeraudiophile.com/files/Screen%20shot%202011-02-10%20at%207.44.22%20PM.png

 

The noise increasing in amplitude as you go up in frequency is not from the recording, but an effect of the DSD format.

 

[wgscott]

On OS X it can be tweaked in the Preferences for Audacity:

 

 

[Manelus]

As requested ,

Gain 40:

 

 

And Gain 60:

 

 

(Same track as first Post)

 

[wgscott]

Well, at least it is bricked at 23kHz instead of 22.05.

 

 

 

[1audio]

Most microphones used for recording do not have the extended response of the Earthworks. The Earthworks mikes have a 1/4 capsule with a very extended neutral response. They are similar in response to a 1/4" B&K or Microtech Gefell measurement microphone. Most popular recording mikes have a 1" or larger diaphragm. With the large diaphragm the response falls off pretty quickly above resonance (usually around 18 KHz) and may rise up again and fall again at higher frequencies, if the associated electronics will pass the frequencies. This is a physics issue with the wavelength of the sound, angle of incidence and the diameter of the diaphragm. Even the external screen has a large influence on the sound quality.

 

There is considered to be a lot of art to picking mikes, mostly about voicing the sound with the character of the mike. Many popular microphones are known to have a "good" quality of specific instruments and not good on others. It should be obvious where this could be heading. Of course the voicing may not work well with a different speaker.

 

I would be surprised to see much extension on a typical commercial recording, everything is going against it. Even the console may have "emi" filters to prevent distortions from RF and other HF noise. This is considered good engineering practice, don't pass information that is not needed, like all that above 20 KHz. The higher bit rate may not have any increased information to capture.

 

High resolution recording will take time to mature. The same was true for HDTV. The first few years were all upscaled SDTV and it really showed.