96Khz vs 192Khz

[paligap]

I hate to do this without a link/document, however, IIRC, and article in Sound & Vision mag by David Ranada pointed out a deficiency in the DSD digital system used in the SACD system. All I remember is an article about a problem with DSD, I don't recall specifics, if anyone can cite this info in a link, that would be a big help (S&V isn't in worldradiohistory.com yet).

I may go thru my copies of S&V to find this, it could be useful in the QQ library.

(edit) Found this:
https://sdg-master.com/lesestoff/attachment.pdf(cites S&V 2001-10, will check soon)


Kirk Bayne

I have lots of SACDs, but I got off the DSD train some time ago. All my players and receivers can decode it or throughput it or whatever the verb is, but if I recall correctly, something about the way it interacted with the Audyssey in my receivers led me to just output everything with PCM. Maybe those with high-end separates or those who eschew room correction are able to derive something from DSD that I can't. Oh, and 96Khz is plenty good enough for me, although I do have some content in 192Khz.

 

[MidiMagic]

There is no such thing as "pixel dust of digitzation", making terms up does not make them exist.

And for your observation to be true CD would have to have a similar or lower dynamic range to LP when CD actually has a vastly greater dynamic range. Any really quiet sound on LP will be easily captured on CD without clipping the loudest parts of the LP.

Actually, that is a term used for digitizing photos. When detail in an analog photo is finer than the fineness of digitization, the detail that is lost changes into pixel dust. This was usually a product of low resolution and low bit depth, so we don't hear the term much today

One expects the CD to be a different mastering. Any number of differences could exist that are simply choices, not some inherent 'pixel dust of digitization'. This is why I proposed digitizing your LP -- to demonstrate this to you.

I'm not sure I understand you.

You ran a TT *preamp* (aka 'phono stage') output to an ADC, as I suggested? And when you did, the 'subtle sounds' on the LP disappeared unless you recorded the signal 'near clipping'?*

Or do you mean something else?

(*recording peaks near 0dB is not a bad thing to do , btw, as long as you are careful to not actually clip**)
(**which means you are utilizing nearly all the available dynamic range of the digital format ...so for any low level signal to 'disappear' from the LP when doing this , makes no sense, unless the ADC method was extremely faulty. The noise floor of an LP is already so far above that of any proper digital recording and playback technology that a competent ADC will faithfully capture everything from the LP's 'silence' to its loudest moment)

Can you name the artist/album in question?

Two Albums:

Hot August Night - Neil Diamond (MCA)
Live - Steppenwolf (Dunhill)

I have both LPs and CDs of what purport to be the same album. I noticed when I got each CD that it was different from the LP in the subtle environment clues.

Later, I was experimenting with recording level on a Philips CD recorder I had just bought. I wanted to find the proper range of use for the VU meters on the recorder. The LP records were the sources (I had no way to copy a CD while varying level at that time). The ADC is inside the recorder.

I made several tracks of the same song at progressively lower levels. I chose the LP track for dynamic range.

The thing I noticed was that one of the lower level cuts sounded very much like the audio on the CD album. I first noticed it on Hot August Night (Crunchy Granola Suite), and then tried it on the Steppenwolf album (Monster: - the Spirit).

 

[par4ken]

To completely lose the quietest signal *of an LP* he would have to set his recording level very erroneously low.



16 bits is entirely adequate for the purpose of capturing the entire dynamic range of an LP.

Nobody said that you completely lose the quietest signal of an LP but for quality sound you still need the recording system to be better than the source. Recording with 32 bit (float) sounds noticeably better than using 16 bit. The difference is similar to the difference between MP3's and uncompressed PCM. The Hi-rez files can be converted to 16-bit that will still sound very good. MidiMagic is describing the subjective effect of his recordings, I know that Ssully completely rejects subjective sound evaluations.

I record at 192Khz as well because my sound card is capable of it. I believe that the declick processing works better at that higher sample rate. The files can then be downsampled to a more reasonable rate without any noticeable degradation.

 

[barfle]

Is this a good time to ask if you think SACD is better than DVDA?

I have a few albums in both formats. To me, the big difference is probably in the mastering. There is also the navigation menu on the DVD.

 

[barfle]

Some of these arguments remind me of some of the political arguments I hear on TV.

Or religion.

 

[admsh]

No, let me get more popcorn first.

Really didn't expect we'd actually need more popcorn. I was not prepared.

 

[Mr. Afternoon]

Really didn't expect we'd actually need more popcorn. I was not prepared.

I got you covered. Take some!

 

[jimfisheye]

There are a few audio interface (usually also stand alone recorders) that use two 24 bit ADC stages and stitch the result together into a 32 bit floating point data stream. The idea is to eliminate the need to gain stage the input for field recording. The goal isn't to preserve a skewed larger than life dynamic range with that! It's to give the opportunity to frame the dynamics after the fact and produce from distortion free recordings.

So is someone actually using such an interface for digitizing their vinyl? Or is it more a fundamental misunderstanding and still recording from 24 bit converters in a standard audio interface? Just padding with zeros by recording 24 bit fixed into a 32 bit container. The broadcast industry products are pretty specific (and expensive!) and unmentioned above. My guess is the latter.

If you decide that the bottom 8 bits in a 24 bit capture should be your digital resolution floor as it were, that leaves 16 bits for program dynamics. That's 96db with the absolute quietest signal still having 8 bits or resolution. (Less than 8 bits resolution is pretty grainy and noise and not really meaningful. You can't very well compare a dynamic range down to 1 bit to the usable dynamic range above the noise floor in an analog system.) So 24 bits gives you true 96db dynamics. That's why we like it and it's the standard now. I'm sorry but if anyone thinks music needs to literally be able to cause permanent hearing loss and/or speaker destruction from larger impulses than that... Enjoy your shotgun symphony or whatever the heck you're listening to!

 

[par4ken]

The beauty of using 32 bit (float) is if you inadvertently let your record level exceed zero it likely won't clip. On occasion I've made a recording with my levels set a bit hot, after processing when the signal was normalised the level was actually lowered by that process. Now the recording can be saved as a 24 bit file without any clipping. Audio programs, like Audition and iZotope run their processing at 32 bit for that very reason to prevent degradation and reduce the chance of clipping.

 

[jimfisheye]

The beauty of using 32 bit (float) is if you inadvertently let your record level exceed zero it likely won't clip. On occasion I've made a recording with my levels set a bit hot, after processing when the signal was normalised the level was actually lowered by that process. Now the recording can be saved as a 24 bit file without any clipping. Audio programs, like Audition and iZotope run their processing at 32 bit for that very reason to prevent degradation and reduce the chance of clipping.

Only if you have an analog input stage that can handle the dynamic range and not actually clip AND you have an actual 32 bit capture from dual ADC stages. (Like some of the broadcast industry aimed field recorders with dual ADC stages.) If you record from 24 bit converters from a standard audio interface, clipping is still clipping and can't be undone. Modern mix engines are actually 64 bit floating point. (For about the last 15 years.) That means you can push overs through the mixing board with no clipping. You only need to respect the output being sent to your 24 bit fixed point DAC. (The real point of the resolution in the mix engine is to preserve the full 24 bit resolution of the original source elements throughout the mix no matter what attenuation you do.)

This stuff is all published spec and matter of fact. Please lets not telephone game using dual ADC stages or after the fact (ie after the recording stage) digital mixing with 64 bit fp mix engines as some magical new way to record or undo digital clipping!

I hope it goes without saying that recording a clipped signal from a 24 bit converter into a 32 bit floating point container does NOT allow for undoing the clipping after the fact!

 

[ssully]

Nobody said that you completely lose the quietest signal of an LP

It's what MidiMagic seems to maybe be reporting.

I write 'seems', 'maybe' because I still can't quite decipher what he is claiming to experience.

A difference between an LP and some CD version of it is trivial to explain -- it could come from a variety of production decisions

But he 'seems' to be reporting that he is literally losing low level signal (to the extent it exists within the limits of LP S/N) when he digitizes the LP output. If that's what he's actually claiming, it's hard to countenance if the digitization process is being done well.

but for quality sound you still need the recording system to be better than the source. Recording with 32 bit (float) sounds noticeably better than using 16 bit. The difference is similar to the difference between MP3's and uncompressed PCM.

Sorry, both of those sentences are untrue unless you're recording very badly -- although the second sentence has unwitting truth to it,in that MP3 vs PCM difference is far less audibly notable than audiophiles claim, given good encoding -- even more so 16 vs 32. So yes, the difference is 'similar' i.e. hard to hear assuming it's audible at all) .

It may be that MidiMagic's recording gear is actually *worse* than his TT/preamp in terms of noise, dynamic range, low-level detail , call it what you will. But that would be very bad gear.

The Hi-rez files can be converted to 16-bit that will still sound very good. MidiMagic is describing the subjective effect of his recordings, I know that Ssully completely rejects subjective sound evaluations.

He's not talking about converting hi rez files. He's (I think) talking about ADC -- digitizing an analog source.

I 'reject' explanations that aren't founded in good evidence.

I record at 192Khz as well because my sound card is capable of it. I believe that the declick processing works better at that higher sample rate. The files can then be downsampled to a more reasonable rate without any noticeable degradation.

Your 'belief' could do with objective testing. It might prove correct, it might not. (Something I observe generally for claims made by recording engineers. Subjectivism and unwariness of the flaws of sighted comparisons run strong in that group.)

But what MidiMagic is claiming is not obviously related to sample rates.

 

[par4ken]

Recording with 32-bit (float) requires only a 24 bit sound card. It allows recording levels above (digital) zero. I have yet to have a signal clip while recording, even though the signal went above 0 dBFS. It does use twice the disc space but once saved as 24 bit you can delete the original and get that space back. I'm using a Digigram VX822HR soundcard.

Obviously you can't take a clipped 24 bit signal and remove the clipping after the fact!

 

[ssully]

The beauty of using 32 bit (float) is if you inadvertently let your record level exceed zero it likely won't clip. On occasion I've made a recording with my levels set a bit hot, after processing when the signal was normalised the level was actually lowered by that process.

Mmmmbut....it's still clipped. If it's actually clipped at input. You didn't 'undo' the clipping by downcoverting to 24 (with dither or not). That would defy the laws of physics (restoring information after it's been lost)

Hi bit recording allows for more headroom *and* you can apply massive processing without accumulating audible errors in the finished product. That's why it's done.

 

[par4ken]

But what MidiMagic is claiming is not obviously related to sample rates.

I agree with that.

But he 'seems' to be reporting that he is literally losing low level signal (to the extent it exists within the limits of LP S/N) when he digitizes the LP output. If that's what he's actually claiming, it's hard to countenance if the digitization process is being done well.

I think that he is reporting the loss of low level detail, not a complete signal loss.

Mp3's do have a gritty sound to them compared to uncompressed files.

 

[par4ken]

Mmmmbut....it's still clipped. If it's actually clipped at input. You didn't 'undo' the clipping by downcoverting to 24 (with dither or not). That would defy the laws of physics (restoring information after it's been lost)

Hi bit recording allows for more headroom *and* you can apply massive processing without accumulating audible errors in the finished product. That's why it's done.

It's not clipped!!!!! I don't know if the program sets up the VU meters differently (lowers the zero point) but I do not get any clipping while recording above zero.

 

[jimfisheye]

Recording with 32-bit (float) requires only a 24 bit sound card. It allows recording levels above (digital) zero.

This is bluntly false.
No manufacturer of these products makes any such claim. The published format spec for 24 bit PCM digital would have to be intentionally dishonest and have the entire industry play along with no leaks. All 3rd party tools (like even a basic bit depth meter) would have to be playing along as well. That's just absurd!

If you think this is still true and everyone else industry wide is lying, you could torpedo Sound Devices and their expensive-ass dual ADC stage 32 bit output interfaces/stand along recorders if you could prove an industry wide deception like this. There'd be an avalanche of lawsuits.

 

[ssully]

Actually, that is a term used for digitizing photos. When detail in an analog photo is finer than the fineness of digitization, the detail that is lost changes into pixel dust. This was usually a product of low resolution and low bit depth, so we don't hear the term much today



Two Albums:

Hot August Night - Neil Diamond (MCA)
Live - Steppenwolf (Dunhill)

I have both LPs and CDs of what purport to be the same album. I noticed when I got each CD that it was different from the LP in the subtle environment clues.

Later, I was experimenting with recording level on a Philips CD recorder I had just bought. I wanted to find the proper range of use for the VU meters on the recorder. The LP records were the sources (I had no way to copy a CD while varying level at that time). The ADC is inside the recorder.

I made several tracks of the same song at progressively lower levels. I chose the LP track for dynamic range.

The thing I noticed was that one of the lower level cuts sounded very much like the audio on the CD album. I first noticed it on Hot August Night (Crunchy Granola Suite), and then tried it on the Steppenwolf album (Monster: - the Spirit).

Still not clear on what you have done.

Please describe the TT-->preamp-->ADC chain. (include the connectors and the model # of the CD recorder)

Here's a youtube upload of Monster (source format unknown). At what time point do you detect something missing that you hear on your LP?

 

[ssully]

I agree with that.

I think that he is reporting the loss of low level detail, not a complete signal loss.

He reported this about LP vs CD versions:
The sounds I am missing are subtle sounds from a live audience. In the records I have, I hear people talking in the background (unintelligible speech) and other crowd noises. They are not there on the CD of the same live album.

He *seems* to be reporting the same difference when he digitizes the LP.

Mp3's do have a gritty sound to them compared to uncompressed files.

Another myth. Also, to simply refer to 'MP3' as if there were not a range of quality level options for encoding, is misleading.

 

[par4ken]

This is bluntly false.
No manufacturer of these products makes any such claim. The published format spec for 24 bit PCM digital would have to be intentionally dishonest and have the entire industry play along with no leaks. All 3rd party tools (like even a basic bit depth meter) would have to be playing along as well. That's just absurd!

The original signal can still be only 24 bit or even 16 bit. When processing is done more bits can be used to prevent the degradation of the sound and allow for massive headroom. That is why It is done. Audio programs use it internally.

Quoting DVDdoug from the Audacity forum "So if you record in floating-point, the software/drivers are making a conversion. The conversion between 16-bit or 24-bit to 32-bit floating-point, and back is lossless."

 

[jimfisheye]

The original signal can still be only 24 bit or even 16 bit. When processing is done more bits can be used to prevent the degradation of the sound and allow for massive headroom. That is why It is done. Audio programs use it internally.

That's what I was trying to explain earlier with the 64 bit floating point mix engine most modern DAWs use. You're finally getting it!

If that earlier comment where you literally suggested a clipped 24 bit recording could be unclipped if you recorded it to a 32 bit container was miscommunication, then fair enough. That was a mighty false statement there!