Redbook 44.1 kHz standard: theoretically sufficient timbral resolution?

[audiopro92]

But this completely neglects phase information above the range of human hearing. Air motion tweeters and super-tweeters that go up to 40kHz and higher are not there only to reproduce the harmonics of audio frequency information.

awesome response!

 

[Ron Resnick]

awesome response!

Thank you kindly for re-posting this. I was just about to post something similar, as this is my standard answer to assertions that our audio systems do not need to reproduce anything above the basic range of human audible hearing.

The four air motion tweeters in each of my ribbon panels literally start at 18kHz.

 

[sparkie]

While there will always be some who think otherwise, like most science-minded people I do not entertain the notion that somehow we can perceive frequencies above 20 kHz. Our hearing has upper limits, just like we cannot perceive any colors in the UV region, above the for us visible spectrum (unlike bees, for example, who can see stunning colors and color shapes in flowers where we simply cannot). Accordingly, missing high frequency content above 20 kHz would not form an argument against the Redbook CD 44.1 kHz standard. In this context is also worth noting that in the chain for vinyl playback ultrasonic content is greatly attenuated in order not to destroy cutting equipment, among others; this makes for a rather thin technical argument that vinyl is significantly superior to CD in this regard.

Of course, a greater bandwidth can always be advocated for purposes of avoiding artifacts in the audio band, yet this issue is frequently tackled by upsampling of 44.1 kHz digital to much higher frequencies (like 4 x 44.1 kHz = 176.4 kHz in my own DAC) and shallow filtering from there, so that for example phase issues in the audio band, such as introduced by the infamous 'brickwall' filtering, are avoided. I do find noteworthy the argument put forth by Chris Montgomery in his article "24/192 Music Downloads...and why they make no sense",

http://xiph.org/~xiphmont/demo/neil-young.html

that ultrasonic content is actually harmful because with a lot of audio equipment it leads to intermodulation distortion products in the audio band (and this can be tested with the files that he posts).

The argument that the sampling rate of 44.1 kHz is insufficient is mostly based on the stairstep model, which is the result of a fundamental misunderstanding of the application of the Nyquist theorem in digital that lead to standards around that frequency cut-off (44.1 or 48 kHz). That the stairstep model is false has been convincingly argued, among others, by Chris Montgomery in the above cited article.

Even more strongly, his video,

https://www.xiph.org/video/vid2.shtml

demonstrates that for any sinewave signal the 44.1 kHz bandwidth is sufficient, as it shows on an oscilloscope that even from a 20 kHz sinewave signal the analog waveform is reconstructed perfectly by a DAC, without any stairsteps (watch the first 9 minutes of the video, it will be stunning for those who are no familiar with this).

It has been suggested that any complex waveform can be synthesized from sinewaves, and thus according to the Nyquist theorem any music signal up to 20 kHz frequency can be perfectly represented by 44.1 kHz digital. Yet the following article by Chris Tham, "Exploring Digital Audio Myths and Reality Part 1" argues otherwise:

http://www.audioholics.com/audio-technologies/exploring-digital-audio-myths-and-reality-part-1

While Chris Tham concurs with Chris Montgomery's argument that the idea that digital is 'discontinuous' is a myth and that the stairstep model is bogus, he also points out that square waves and sawtooth waves cannot be represented accurately by 44.1 kHz digital, see figures 7, 8 and 9 in the article.

He concludes:

"Some could argue that we don't listen to sawtooths or square waves, therefore Figures 6-8 are not significant. But we do - some musical instruments have harmonic characteristics very similar to sawtooth waves. And pop/rock music often contain music generated by synthesizers - sawtooth and square waves are fundamental building blocks for digitally synthesized music."

Indeed, it has been suggested that the waveforms of trumpet sound are similar to sawtooth waves, and the similarity, the asymmetry in the waveform, is seen in the graphs in section 1.6.1 in the following article:

http://www.feilding.net/sfuad/musi3012-01/html/lectures/005_sound_IV.htm

It is also noteworthy that in digital synthesis trumpet sound often appears to be emulated using sawtooth waves. Yet strangely enough, I do not find notable weakness in 16/44.1 digital when it comes to reproduction of trumpet sound. I find the medium very convincing in this area. On the other hand, compared to top-level analog it shows, to my ears, some weakness in reproduction of violin or saxophone sound, sounds that are very rich in overtones. Having said that, I have not yet heard the very best playback in digital; here is the witness of someone who owns the Berkeley Reference DAC, next to very advanced vinyl playback:

http://audioshark.org/dac-reviews-9...e-alpha-dac-review-6331-page2.html#post106059

From the post:

"A little backround, IMHO current digital state of the art still misses the best of analog on two marks: 1) reproduction of accurate timbres of instruments especially ones with significant high frequency overtones, 2) reproduction of the soundstage and spacing of instruments within that soundstage. [...] It was immediately apparent that the BADA Reference was doing something on issue 1 that I hadn't previously heard from Digital. In its own way, the BADA was as much a breakthrough on issue 1 as the Light Harmonic was on issue 2. There is a certain rightness about the best recordings through the BADA Reference. It flatters some of the most difficult to reproduce instruments like pianos and massed strings. Best of all, it didn't matter if it was Redbook or a HD source. The Redbook performance is striking. The BADA makes the format wars conversation almost silly."

So here it is argued that the timbral performance of Redbook digital matches the one of hi-rez; I have read somewhat similar testimonies about that DAC elsewhere.

So what is going on? Is the 44.1 kHz standard indeed theoretically, on a technical level, insufficient when it comes to proper timbral resolution of just acoustic instruments (disregarding odd non-sine waveforms from synthesizers), even if some people suggest that it does not matter in practice?

And what about the technical argument that ultrasonic frequency content from hi-rez digital can be harmful to sound reproduction in practice, because a lot of the downstream equipment cannot handle the bandwidth and causes intermodulation distortion products? (You can test this in your own equipment with the files Chris Montgomery posts in his above cited article.)

Isn't the digital nerve racking for some people?




I spent some time, experimenting recording bands and artists. Even though I would get into that part of it full time. But after using my 'audio test generating subjects' I'm more decline to sell that all off after I give the next guy after I hot rod them into perfection. Once I recorded them they went on their way, what I found out is they thought they took advantage of me, however, they were just my organic signal generators. The only cool thing was using different mics and all of that in the process and internalizing how three dimensional sound waves are translated into the wire.




The digital clock signal has its little issues mucking up the signal on the way in and out due to the digital clock's harmonics bleeding into the band. So the engineers placed filters in that switch in and out at those sample frequencies. There are other frequency filtering afterwards that the other sample rates would use. All of these do add into the phenomenon called insertion loss. Which is a drop of signal power when a circuit like a filter is inserted. The perception of the higher sampling rates is the perception of change in signal to noise when the sampling rate changes, because of adding or subtracting those filters. I learned the 'point instant' instead of a 'stair step model' in regards to sampling. But quite frankly,its a mood point. Because that is not where the loss truly happens. It happens right before that when the signal begins its trip into a media change.






(continued on next post)

 

[sparkie]

I used different fancy boxes, and i/o and all kinds of these "recording interfaces" that consists of a signal path of microphone preamp, a line buffer into a a/d conversion chip. The strata of what you buy is what you pay for comes seemed like it it was true, but later I found, that is just on the surface. Those recording interface manufacturers purposely kept their schematics off the market. For I guess of being afraid of someone noticing, they are pretty much all the same. Another interesting point I cam across is they don't even create their own circuitry past what the chip manufacturers have in their "sample circuit" of their device. Going back further in school times and remembering a lecture in class about data sheets and their example circuits. I remember the professor saying, "The example circuit found in data sheets are simple examples to demonstrate how the device works. If you are truly designing something for a purpose, you should only use this as a reference guideline instead of the exact way it supposed to build for all applications". Remembering that, and seeing that yes indeed they didn't do anything but assemble the example circuit in their computer interface boxes over and over again. All those fancy mic preamps and all that going into a generic circuit not designed specifically for the task, but was designed to demonstrate the converter chip. I wonder why what I recorded didn't sound correct like too much of the wrong bass, or lack of mid punch or even the airy highs like I would hear on a 'mastered track' . After some time recording people at different levels, and notice differences in signal, it dawned on me that it has this insertion loss, and other parts compounded this. I stepped back and thought about how the older recording devices were made like the reel to reel so many admired more and even came up with emulators to add its not so good factors in to attempt to make it sound like a tape machine. So , instead of refining the interface, they market dsp to try to fix something it can't because it was lost at the time of conversion/capture. So I went and modified an ADAT interface first, I removed all of the generic circuit and replaced certain things to fix what was recorded. The mic preamp in front, when I jumpered it out of the unit into a power monitor for me to listen sounded pretty bad compared to the professional mic preamps that I had in a rack. Tinny, and the bass out of balanced. Interesting that their line in didn't bypass this stage, but attenuated the signal so it would go through all this. So I decided to bypass it. After that, I found that the unit recorded better than what I could buy. I found out later only a few high dollar converters were set up like this and they are still not available in the music stores online or physically. You had to contact them, then someone would quote you through a distributor they dealt with. I noticed that a lot of the professional recording gear of the past has low frequency roll offs built into them. So on another interface, I replaced the capacitor circuit on the microphone preamp with one of those transformers from a famous mic preamp, and lowered the capacitance values on their coupling caps in that unit so I can achieve that 80 hz roll off I measured on my reel to reel tape professional machine. The results in both modded units were spectacular and someone said "did you record this on your reel to reel?" Needless to say, he was shocked when I told him.
(continued on the next post)

 

[sparkie]

Then the mixing game happened in post. There was all kinds of fun putting it together, good and bad. without modifications I had to do all this crazy processing stuff to get something sounding nice. After I modified it, it was esier to put together and also didn't have to do a lot of processing.

Then there was mixdown. I found it interesting that inside the multi track software, it sounded awesome, when I went to mix it down or 'bounce to disk' it sounded low and duller when I wrote it on a disk and played it. I know I did everything right, what was the deal? Later, I did a project and sent it into a mastering guy after auditioning several ones and looking at what they were doing. Some would eq it, other slammed it into a limiter or both but one thing I notice about all of them is they shortened the dynamic range and analyzed the effects, they what I called integrating soft clipping. So the format of CD and their loudness war ended up who can soft clip the best. When I analyzed the raw masters, I find that they were causing +3 to+5 dbfs of signal to exist. The digital scale is -150 to 0 dbfs, how were they doing it (soft clipping) I also notice they never let the signal fall below -3 dbfs which means to me, they are doing that to overcome the D/A insertion loss of the filters.




So the professionals ruined the digital format. That is why there very few dynamics. But on the other hand, too dynamic rendered a softer signal after conversion. Because we fell below a signal to noise before its fully converted back to analog. Some people swear by higher sampling rates, but I found that all it was a signal relationship of the insertion loss I discussed earlier. Later, I found if I mix really low and then master it up digitally, I gained more space and what was more what I mixed. It sounded great, until I burned it to CD and it was lower, softer than when I played it on a windows player.




I further analyzed the D/A process electrically and how circuits are built and noticed a few flaws inherited by the digital process. The sad thing about it is that one could make a wonderful D/A conversion circuit. But the commercial material was damaged to be mastered to go through the inefficient systems that were built to reproduce them.

 

[Geoffkait]

Thank you kindly for re-posting this. I was just about to post something similar, as this is my standard answer to assertions that our audio systems do not need to reproduce anything above the basic range of human audible hearing.

The four air motion tweeters in each of my ribbon panels literally start at 18kHz.

I was once a dealer for Golden Sound’s Ultra Tweeters, a quantum mechanical device with a claimed frequency of 2 or 3 GHz.

 

[Rexp]

Then the mixing game happened in post. There was all kinds of fun putting it together, good and bad. without modifications I had to do all this crazy processing stuff to get something sounding nice. After I modified it, it was esier to put together and also didn't have to do a lot of processing.

Then there was mixdown. I found it interesting that inside the multi track software, it sounded awesome, when I went to mix it down or 'bounce to disk' it sounded low and duller when I wrote it on a disk and played it. I know I did everything right, what was the deal? Later, I did a project and sent it into a mastering guy after auditioning several ones and looking at what they were doing. Some would eq it, other slammed it into a limiter or both but one thing I notice about all of them is they shortened the dynamic range and analyzed the effects, they what I called integrating soft clipping. So the format of CD and their loudness war ended up who can soft clip the best. When I analyzed the raw masters, I find that they were causing +3 to+5 dbfs of signal to exist. The digital scale is -150 to 0 dbfs, how were they doing it (soft clipping) I also notice they never let the signal fall below -3 dbfs which means to me, they are doing that to overcome the D/A insertion loss of the filters.




So the professionals ruined the digital format. That is why there very few dynamics. But on the other hand, too dynamic rendered a softer signal after conversion. Because we fell below a signal to noise before its fully converted back to analog. Some people swear by higher sampling rates, but I found that all it was a signal relationship of the insertion loss I discussed earlier. Later, I found if I mix really low and then master it up digitally, I gained more space and what was more what I mixed. It sounded great, until I burned it to CD and it was lower, softer than when I played it on a windows player.




I further analyzed the D/A process electrically and how circuits are built and noticed a few flaws inherited by the digital process. The sad thing about it is that one could make a wonderful D/A conversion circuit. But the commercial material was damaged to be mastered to go through the inefficient systems that were built to reproduce them.

This is one of the most interesting posts I've read on WBF, thanks!
Are there recordings you've done that are available on Tidal etc?

 

[Geoffkait]

Then the mixing game happened in post. There was all kinds of fun putting it together, good and bad. without modifications I had to do all this crazy processing stuff to get something sounding nice. After I modified it, it was esier to put together and also didn't have to do a lot of processing.

Then there was mixdown. I found it interesting that inside the multi track software, it sounded awesome, when I went to mix it down or 'bounce to disk' it sounded low and duller when I wrote it on a disk and played it. I know I did everything right, what was the deal? Later, I did a project and sent it into a mastering guy after auditioning several ones and looking at what they were doing. Some would eq it, other slammed it into a limiter or both but one thing I notice about all of them is they shortened the dynamic range and analyzed the effects, they what I called integrating soft clipping. So the format of CD and their loudness war ended up who can soft clip the best. When I analyzed the raw masters, I find that they were causing +3 to+5 dbfs of signal to exist. The digital scale is -150 to 0 dbfs, how were they doing it (soft clipping) I also notice they never let the signal fall below -3 dbfs which means to me, they are doing that to overcome the D/A insertion loss of the filters.

So the professionals ruined the digital format. That is why there very few dynamics. But on the other hand, too dynamic rendered a softer signal after conversion. Because we fell below a signal to noise before its fully converted back to analog. Some people swear by higher sampling rates, but I found that all it was a signal relationship of the insertion loss I discussed earlier. Later, I found if I mix really low and then master it up digitally, I gained more space and what was more what I mixed. It sounded great, until I burned it to CD and it was lower, softer than when I played it on a windows player.

I further analyzed the D/A process electrically and how circuits are built and noticed a few flaws inherited by the digital process. The sad thing about it is that one could make a wonderful D/A conversion circuit. But the commercial material was damaged to be mastered to go through the inefficient systems that were built to reproduce them.

Obviously overly compressing the sound is bad for dynamics, but one assumes business was falling rapidly so they had to do something. However, there are a lot of other perhaps less obvious reasons - hiding in plain sight! - why digital playback is rarely satisfactory to me here in the future.

this is my shortlist and yes I know what you’re thinking, “but isn’t Reed Solomon error detection/correction and the CD laser servo feedback system supposed to take care of all that?”

Isolate the CD player/transport from seismic type vibration.

Isolate the DAC from seismic type vibration.

Reduce scattered laser light inside the CD transport that would otherwise get into the photodetector.

Stabilize and damp the CD.

Eliminate the electric static charge on the CD.

Eliminate electric static charge on objects in the room.

Improve the transparency of the CD clear layer. Polycarbonate is only 90% transparent to the CD laser.

 

[sparkie]

Obviously overly compressing the sound is bad for dynamics,

Compression is a necessary evil in recording and production. Without it everything would be thin and out of scale with each other when the instruments are recorded individually.

converters made or modified for CD mastering changes the depth of the mix and like always the relative distance between loud and soft get shorter, the louder the mix is.

Here is someone covering one of the first converters available outside of having the custom made ones or an engineer soft clipping into a high end converter several times. Notice in the video the mixes changing depth in exchange for loudness.

 

[Geoffkait]

I referring to aggressive over-compression, i.e., loudness wars, not to compression in general.

 

[sparkie]

I referring to aggressive over-compression, i.e., loudness wars, not to compression in general.

You can only compress a mix so much. But limiting which is a type of compression is the loudness war key. In the video above in post 129, the magic circuit that allows them to go above 0dbfs is a limiter.

 

[Geoffkait]

A lot of CDs that have been coming out over the past ten years or so are flatlined. Apparently if there is some limit to the compression they utilize they haven’t reached it yet. Ever take a look at the dynamic range database?

Album list - Dynamic Range DB

 

[sparkie]

A lot of CDs that have been coming out over the past ten years or so are flatlined. Apparently if there is some limit to the compression they utilize they haven’t reached it yet. Ever take a look at the dynamic range database?

Album list - Dynamic Range DB

There is limits to it depending on the frequency balance and relative power. But what really limits it is too short of a distance between load and soft. which sounds over cranked and distorted. The other extreme of having the distance great reveals a soft and weak recording. But normally an un-mastered track has 4-6 db differences between loud and soft. That normally gets mastered to around 2-3 db difference. But CD media is -3 to -5 difference which is less than 1db on the recorded material and signal +3db greater after the dac and about twice what 0dbfs can deliver.

 

[Geoffkait]

Shirley you don’t mean 4-6 dB difference between loud and soft.