24-bit/192kHz is pointless?

[DavidJPettifor]

I've been looking into 24-bit/96kHz - 24bit/192kHz sampling rates and came across Dan Lavrys comments (http://www.monoandstereo.com/2008/06/interview-with-dan-lavry-of-lavry.html). He appears to suggest that the "optimal sample rate" might be something between 60 and 70 kHz. It would also appear that the human ear can not hear anything far beyond 30kHz.

 

Does this mean that any music that is say, above 24-bit/96kHz is pointless? Discuss.

 

--

djp

 

[The Computer Audiophile]

Hi David - Thanks for the link. Dan Lavry certainly has a lot of knowledge in this area so I really can't counter anything he said in the interview. I do know that there are many people in high end manufacturing and music production that disagree with his statements in this interview.

 

Other than that I really can't say much. I do have 16/44.1, 24/96, and 24/176.4 versions of a single album. At least I can give the resolutions a listen.aCcording to Dan Lavry the 24/94 version should sound best of the three I have because it's closest to 70kHz.

 

Interesting.

 

[cs]

Actually, I think that even 21 bits and 70kHz is overkill.

 

Unless you are a very small child, or a cat, most people can't hear much above ~16kHz, so on that basis 44.1kHz is just fine. But what you really need is a DAC with an output filter that cuts off at 16kHz. None of them do of course, because of the industry's obsession with specmanship.

 

A dynamic range of 21 bits (128dB) is probably also over the top. Here's a little known fact (in audiophile circles, anyway) :- The instantaneous dynamic range of the human ear is typically ~30dB. Yes, really !

 

The other 90dB or so is achieved with signal processing in the brain, and is mostly an AGC effect. It is this property which is exploited in lossy codecs in the form of their psycho-acoustic masking. So 24 bits is somewhat excessive too.

 

I think a perfectly adequate spec would be 18 bits and 50kHz.

 

Chris.

 

[crion]

Higher sampling rates is not about wider frequency response but better impulse response with fewer pre-echos and ringing.

 

This is what Meridian is delivering with their new apodizing filters. The improved filtering (which works to mitigate/eliminate pre-echos and ringing from 44.1kHz/48kHz sources) is performed within the upsampling stage of the new 808.2 CD-Player and the DSP7200 Speakers. The new filtering/upsampling technology will be able to be uploaded to present speakers like the DSP8000 as well.

 

So assuming the impulse response gets better with an increase of the sample rate then technically 24/192kHz should be superior in that area as well, but if we can detect it, I don't know, I haven't done any blind listening tests to verify it.

 

Can anyone else detect in a blind listening test any discernable difference between a clip in 24/96kHz vs 24/192kHz? (20 rounds, 80%+ hit rate)

 

[cs]

Re :- "The improved filtering (which works to mitigate/eliminate pre-echos and ringing from 44.1kHz/48kHz sources) is performed within the upsampling stage of the new 808.2 CD-Player and the DSP7200 Speakers."

 

I think that the Meridian marketing machine has got itself confused there !

 

An apodizing (or windowing) filter must, by definition, start to roll-off within the passband. So in the context of audio, they can only really be used at the higher sample rates.

 

At 44.1 or 48kHz, there isn't enough guard-band between the wanted band and the closest alias to fit a useful windowing function in. True, it is a good argument for higher sample rates such as 96, 176.4 or 192kHz, but isn't a viable technique at 44.1 or 48kHz.

 

Chris.

 

[crion]

Hello Chris,

What if you performed the apodizing filter in the upsampling stage then? Where you do have access to 88.2kHz or 96kHz resolution from a 44.1kHz/48kHz source.

 

This is my understanding of where the apodizing filters are being used.

 

 

 

 

 

[tfarney]

I've got to say that I find 16/44.1 to be not only fine, but beautiful. Of course I'm 57 years old, played live electric music for a couple of decades and am probably incapable of hearing anything about 12 khz. There are a lot of things that golden-eared audiophiles and engineers seem to hear that get past me. I can't hear high sample rate mp3s vs lossless. To be honest, sometimes I can't hear 128kbps. What I can hear, every time, is flat, compressed or grossly boosted mastering. A 128kbps file of a great master will sound infinitely better than a lossless file that is over-compressed or noise-reduced or severely boosted in the bass and/or treble every time.

 

Unfortunately, we have no control over what really matters. Still, a beautifully recorded, mastered and transferred Redbook cd is a wonderful thing. Good enough that it is losing ground to much lower resolution downloads. Good enough that I think we can expect hi-res formats to remain a very rare specialty item, at least until there is a way to quickly download them.

 

Tim

 

[cs]

Hi Crion,

 

I think the problem here is that you need the 'near brick wall' filter to do the upsampling anyway, so the damage is already done.

 

It could be that they are using an apodizing filter which does encroach on the 20kHz passband somewhat, thus achieving the desired raised cosine (or whatever) shape. It would be interesting to get hold of Bob Stuart's AES paper on the subject to find out the details.

 

Chris.

 

[greatdane]

I'm not versed in the subject, but from my listening, I couldn't tell the difference between the 192/24 stereo tracks on a dvd-a to a standard redbook cd.

 

My DAC is only 44/16, but I'm thrilled with the sound I hear from it. I may be a purist, but I don't like the thought of upsampling standard resolution. I've yet to hear an electronic high bitrate track natively though, so that might change my mind.

 

[cs]

Re: "My DAC is only 44/16, but I'm thrilled with the sound I hear from it. I may be a purist, but I don't like the thought of upsampling standard resolution."

 

I tend to agree.

 

The current trend to use upsampling is a bit of a red herring, IMO. Some of the (allegedly) best sounding DACs are non-oversampling. And from a purely engineering perspective, DACs get less linear the faster they have to operate.

 

Chris.

 

[crion]

I would have to agree about upsampling in general. I usually prefer upsampling off from many producers products. Some produce distorsion even. Some produce "exciting" sound with upsampling but that is not how good upsampling should sound. It should simply be better SQ with no discernable side-effects. The best upsampler might even be the one producing "boring" sound, but when evaluating, try picture what DAC sounds truest, visit live performances, gain experience from all instruments sound and judge for yourself.

 

One upsampler frequently recommended in Computer Audio circles is the SRC (Secret Rabbit Code) plugin to foobar2000. In actuality with serious side-effects. In comparison with my Meridian G68 upsampler it added high frequency distorsion and stripped bass. SRC sounded more "exciting" at first, but when you really started to think about instruments, voices and performances it was just another subpar upsampler with added distorsion and altering of the soundstage (phase problems?).

 

Bladelius Gondul MKIII is another example of a high-end CD-player with excellent detail retrieval and soundstage ruined by the upsampling routine (384kHz), luckily you can turn the upsampling off on this product as well.

 

There are more examples likes these. So yes, I don't quite get why so many companies put upsampling in their products and default them to ON when they don't implement them correctly. Don't they listen?

 

 

[greatdane]

I associate upsampling with the use of equalizers...changing the actual sound to a desired sound. I know it's not quite the same, but to me, similar principle.

 

[The Computer Audiophile]

Hi guys - Good discussion here. Something seems fishy though. Most manufacturers I talk to employ some kind of sample rate changes in their DAC, but very few include a bypass feature. Almost all of them say the compromises of not changing the sampling rate are greater than changing it. Plus they could easily include a bypass if desired. This leads me to believe there is some oversimplification going on in this conversation. Not pointing fingers of course, I just want to raise the possibility. There are obviously some smart contributions in this thread :-)

 

Just out of curiosity, what do you guys think of video upscaling? Certainly not the same thing. Jus thought I'd pose the question :-)

 

[Ashley James]

I'm sure CS is correct in what he says because we've brickwall filtered at 16kHz and tested a load of golden ears, then removed the filter and reverted to our normal sample rate of 24/96 and no one has been able to tell in multiple switchings.

 

Those of you who've read other postings I've made on the bandwidth of music, the actual frequency response of a seriously good loudspeaker system and think the comments fair, will probably accept this.

 

CS knows a great deal more about this and I believe he knows how we do our DACs, but I'll explain anyway and he'll correct me if I'm wrong. Jitter is a political issue if not an actual one and sample rate conversion ensures we don't have any. We upsample to 96kHz because I understand that it makes for the best filter and the DAC is 24/192 but limited to 24/96 by the Optical digital lead that everyone prefers because it's more versatile.

 

Mart always says that in these sorts of designs, you're trading speed for accuracy, that everything is a compromise and that most of the differences you hear have nothing to do with the technology and everything to do with mistakes in implementation.

 

As I've said before, unless recordings and productions are complex ones and made throughout with 24 masters and hardware, there is unlikely to be an audible difference IMO.

 

[cs]

Hi Chris,

 

Yes, you are right that the majority of DAC chips these days are based on the Sigma-Delta principle, which always contain internal oversampling (brick wall) FIR filters. Typically, these increase the sample rate to 8x the input rate, which is normally enough to allow the Sigma-Delta modulator to work properly.

 

But many DAC box and CD player manufacturers often also use sample rate converter (ASRC) chips in front of the DAC chips to increase the sample rate by a non-integer factor to the typically trendy 192kHz. ASRC chips effectively also contain a brick wall interpolation filter function, so you now have two such filters in series.

 

There is some evidence that the resultant excessive time dispersion (due to the two filters) is audible and undesirable.

 

However, the plus side of the ASRC chips is that they contain narrowband digital PLLs, which greatly reduce the clock jitter fed to the DAC.

 

My opinion on this is that a better solution would be to dispense with the ASRC, but use reclocking to remove the jitter, then you only have the one brick wall filter (in the DAC itself). If you are using SPDIF, a good way to do this is to use one of Wolfson's SPDIF receiver chips. Unlike every other SPDIF chip out there, the Wolfson chip really does suppress jitter.

 

Chris.

 

[mpmct]

Great thread. It's fun to learn.

Don't stop now guys!

 

[The Computer Audiophile]

mpmct - I agree 100%. We've all seen this exact topic discussed on other sites, but it quickly turns into something nobody can learn from. I really appreciate the quality of responses here and the respect given by those in disagreement. It's a very enjoyable thread.

 

Thank you very much guys!

 

[Daphne]

I believe it is true, that with the higher sampling rates mentioned, signal information is captured above what is generally considered to be the human hearing range. From what I understand there is a good reason.

 

In the recording process there is distortion from the microphone right through the analog to digital conversion. There are many causes from filters to the type of timing crystal used. Then the digital signal is converted back to analog which can sometimes suffer from flaws on the CD to poor circuit design on a sound board. All containing artifacts which can be easily measured and heard, from the recording method used to the dithering of the signal.

 

So, sampling the waveform at higher frequencies and allowing for a greater number of bits per sample allows noise and distortion to be reduced. The DAC has the ability to over-sample, and sometimes in combination with up-sampling, which will greatly reduce distortion, mainly from jitter. From what I recall (I could be wrong), most of this distortion occurs in the higher range which we are most irritated by.

 

I realize this is a simple description. There are many factors that come into play with sampling at higher frequencies and greater bit rates. It has been a subject of debate for some time. In the pursuit of improving the quality of digital music reproduction there is a good deal of technical research, but there are still large gaps. Of course, we are discussing the audio industry, which means these gaps must be filled with subjective opinions. If this was the field of mathematics, any mathematician making like assumptions would be laughed out of the profession.

 

We have instruments that can accurately measure sound beyond what is claimed we can hear. However, sound waves are processed by the brain in two parts: the sound we can hear, and the sound waves we can feel, and no one knows exactly how the brain processes that information, especially within the ultrasonic range. So technicians can set up their instruments and measure sound until hell freezes over, but they cannot explain why music brings us so much pleasure. It seems there is an objective and subjective side to this science. Almost like coming up with an answer to the paradox - is Hell endothermic or exothermic?

 

Daphne

 

[mpmct]

I read these two interviews at the site

referenced at the start of this thread, that as

far as I'm able to understand, reinforce the comments

Mr. Lavry made regarding "bunches of bits", and their

value, where/when ... or not ... ?

 

Very interested on takes on either of these interviews/comments,

from those of you who clearly have expertise/experience some

of us wish we had ...

 

Daniel Weiss (Weiss Labs)

 

Bob Katz (Digital Domain mastering)

 

 

[audiozorro]

Striving for superior formats is never pointless. Several people that have digitized LPs have noted significant sonic improvements with 24bit/192kHz conversions over 24bit/96kHz. SACD aficionados generally claim to hear significant sonic improvements of 2.8224Mbit/s over 24bit/96Hz DVD-As. LPs, SACDs, and DVD-As are generally considered to sound better than Redbook CDs at 16bit/44.1kHz. And of course Redbook CDs are generally acknowledged to sound superior to any of the lossy compression formats. Just imagine the day when the music industry finally has a format that is universally accepted by listeners as sonically superior to all of the above.

 

It may be pointless to argue with someone who hears no difference. There could be several factors at play including ones’ hearing, training or audio equipment. I doubt that I could tell the difference between a violinist playing a $10K violin from a $300K violin, but there are people who can.

 

Fortunately on this issue several recording studios, such as Reference Recordings, Linn Records, and 2L, have provided multiple file formats for us to hear the difference. Let your hearing and pocketbook be your guide.

 

I happen to believe there is no problem as long as the majority of people find that the higher resolutions sound superior or hear no discernible differences when comparing two different resolutions. If however a majority of listeners find that a higher resolution sounds worse, then I believe that the limit at that point of time has been pushed too far.

 

 

[DavidJPettifor]

Picking up on mpmct's comments regarding similar comments made by Bob Katz and Daniel Weiss from the same link, posted originally, these are the comments that directly relate to this thread. First, from Bob Katz;

 

MI: How are the 192 khz recordings? Does so much information bring us closer to the original recordings and are there more problems due to the increase of content?

 

BK: I have not had enough experience with 192 kHz to say. I like the results I'm getting at 96 K, and in my book I make a convincing argument that it is the converter design that counts far more than the sample rate. We have always known that a well-designed 44.1 kHz converter sounds much better than a mediocre 96 kHz model. And this has always been true. I believe that a good designer will be able to make a 96K converter that sounds as good as anything at a higher rate. But designers are getting lazy, and it is cheaper and easier to get a good sound at a higher rate because the filters are less complex and easier to design. There is nothing magic about the higher rates; it's not the higher frequencies that we're hearing, but rather, more linear performance from 20-20 kHz! Keep that in mind... We really should be labeling converters by their resolution, not by their sample rate.

 

... and then from Daniel Weiss;

 

MI: What sampling frequency is enough? Some say that even 44 khz 16 bit is good if mastered well, other say that 24bit and 96khz is more than enough, and again third camp say that only 192 khz (DSD) or even more would bring the quality of analogue or natural timber to life.

 

DW: For the human hearing 44.1/16 is enough. For technical reasons it is advantageous to go with higher sampling rates. 88.2 or 96 would be fine. Anything higher does not make much sense in my opinion. The advantages of higher sampling rates (higher than 44.1) are that the filters in the A/D and D/A Converters can be made less demanding and thus less colouring. For the processing of digital audio, which is necessary during e.g. the mastering process, it is advantageous to have high sampling rates as well.

 

Seems to substantiate Dan Lavry's comments and also pick up on points made by fellow posters here in this thread regarding high sampling rates during the mastering process.

 

--

djp

 

[wappinghigh]

Daphne is right.

 

Science would need to be employed here. If bioscience was to approach this topic, then the only way to resolve it would be through a double blind trial, including a placebo ( say the same track 24/192 vs 24/96 vs something silly: say highly compressed mp3). There are a lot of well know trials in pharmacology that demonstrate time and again the power of the placebo effect. I recall one that revealed that patients who were taking the branded ie (expensive) analgesic drug, rated their pain control as much better than the SAME (generic- much cheaper) unbranded drug. What's even more interesting is that if you then compare a third "tablet" (which is actually sugar- the placebo) and tell the patient it is the latest discovery (and most expensive) of all the medications, you'll find a group of people who will actually rate this as the best pain controller of all. Even though it is just sugar. In other words the power of the mind and the desire to "believe" in an improvement is actually often a powerful effect in it's own right. The power of the placebo. This is the basis of many so called "benefits" of alternative medicine. I suggest a read of several of Richard Dawkins works for a better understanding of this.

 

The same logic applied to this debate would suggest there may be a "placebo" effect if the listener was informed that they were listening to the "best" most expensive, "state of the art" digital encoding (even if they weren't). The power of the mind might release some hormones etc that may make the listening musical experience more pleasureable, even if the ear itself couldn't actually decifer the difference in frequency. It doesn't mean the listener is dishonest or "lying" etc. They might truely believe they notice a difference.

 

At the end of the day does it matter? The answer is only if there is deception in the marketing/sale of the drug (in this case music). If the consumer is making an informed choice, then it is probably of little concern. Other more practicle issues become important. eg availability of drug (music). How easy it is to swollow (play). Any side effects! etc.

 

I'd love to see audiology science really tackle this one. Make a great PHD project for a budding audiologist.

 

AB

 

[Ashley James]

David is on the nail!

 

Perfectly put and thanks too for Bob Katz and Daniel Weiss for bringing wisdom to this discussion.

 

In our opinion Compressed material is better than many realise too, but seems to suffer more from poorly implemented DACs and Amps, hence some having such strong feelings against it.

 

Let's not forget too that vinyl amounts to about 8 Bit resolution and Analogue tape to 11 Bit, I'm not terribly comfortable with vinyl but some Analogue tape can be very good indeed IMO.

 

Ashley

 

[mpmct]

"About high resolution

I do a fair amount of subjective testing in the audio field, in addition to my audio production and studio design work. Over the past six months, I’ve gotten caught up in a food fight on the Internet about the audibility of various audio processes." ...

 

http://www.moultonlabs.com/more/24_bits_can_you_hear/

 

( Article continues with the bulleted list at page bottom.

It's kind of hard to tell that these things are hyperlinks.

About high resolution

About blind testing

About hearing small differences

About audibility

About our hearing

About bit resolution

About bandwidth

About accuracy )

 

 

 

[audiozorro]

Let’s continue this discussion;

 

*

*MI: “What sampling frequency is enough? Some say that even 44 khz 16 bit is good if mastered well, other say that 24bit and 96khz is more than enough, and again third camp say that only 192 khz (DSD) or even more would bring the quality of analogue or natural timber to life.”

 

This confirms to me that the so called “Perfect Sound Forever” still leaves a lot to be desired. Personally I don’t need any double blind testing to convince me that I prefer the sonic qualities of SACDs and LPs over Redbook CD resolutions. I don’t believe that Reference Recordings is trying to fool me in believing that their HDCD and HRx files are sonically superior to their RDBK CDs.

 

BK: “I have not had enough experience with 192 kHz to say. I like the results I'm getting at 96 K, and in my book I make a convincing argument that it is the converter design that counts far more than the sample rate. We have always known that a well-designed 44.1 kHz converter sounds much better than a mediocre 96 kHz model. And this has always been true. I believe that a good designer will be able to make a 96K converter that sounds as good as anything at a higher rate.”

 

I could not agree more. Implementation is everything. Superior audio specs do not guarantee superior sound and long term audio satisfaction.

 

To me this is all the same tune, different dance. I remember the THD wars from over 30 years ago. I’ve heard all the arguments that all amps sound the same, all cables sound the same, etc. It’s fine if you believe that and it’s not wrong for someone else to disagree. I for one am glad that we now have so many good format choices. And I sincerely hope that companies like Reference Recordings, Linn Records and 2L can encourage enough others to perfect the Perfect Sound Forever.