PONO

[LoopStationZebra]

Looks like all the copies have been taken down of that video. I can't find it anywhere...

[beats me]

Looks like all the copies have been taken down of that video. I can't find it anywhere...

Probably fell victim to YouTube’s new “flag for irrelevancy” feature.

[Davo]

Somebody correct me if I'm barking up the wrong tree here - apart from recording, isn't playback of more than 48khz 16 bit a total waste? Human hearing goes up to 20khz, most people can't hear anything after 17khz. 17khz x 2 = 34khz playback. So 48 just to be safe, though I don't think more than one out of a billion people can hear up to 24khz, if anyone.

24 bits? This mainly affects the noise floor and its relation to dynamics - the difference between the loudest and the softest sound. 16 bits is more than enough to accommodate the human ear.

And I've read that these very high sample rates sound better to some people purely because it sounds different, not better (subjective yes) than 44khz. These very high sample rates introduce harmonic distortion don't they, which colours the sound somewhat, giving people the impression it's higher quality.

What is the point of 192khz playback? Someone please inform me what I'm missing. The human ear can only hear so well...

I think you are confusing sample rates with audio frequencies.

If you want to make a digital representation of a sine wave at 17 KHz then you have to sample it at a number of points. If you want a good representation then you need more than one or two samples. Your error is in thinking that "17khz x 2 = 34 khz playback". Sampling at 34 khz will only give you two samples of the entire wave, so clearly a lot of information will be lost. If you take 10 samples during one complete wave cycle then you will need to sample at 170 khz, which should give a fairly good representation. It works te same way for playback. To re-create an accurate 17 khz sine wave you need more than one or two samples. I think thats how it works for straight sampled sound in a file like WAV, although encoding algorithms like mp3 manage to squeeze more information in. Please correct me if Im wrong.

[tedlogan]

Somebody correct me if I'm barking up the wrong tree here - apart from recording, isn't playback of more than 48khz 16 bit a total waste? Human hearing goes up to 20khz, most people can't hear anything after 17khz. 17khz x 2 = 34khz playback. So 48 just to be safe, though I don't think more than one out of a billion people can hear up to 24khz, if anyone.

24 bits? This mainly affects the noise floor and its relation to dynamics - the difference between the loudest and the softest sound. 16 bits is more than enough to accommodate the human ear.

And I've read that these very high sample rates sound better to some people purely because it sounds different, not better (subjective yes) than 44khz. These very high sample rates introduce harmonic distortion don't they, which colours the sound somewhat, giving people the impression it's higher quality.

What is the point of 192khz playback? Someone please inform me what I'm missing. The human ear can only hear so well...

I think you are confusing sample rates with audio frequencies.

If you want to make a digital representation of a sine wave at 17 KHz then you have to sample it at a number of points. If you want a good representation then you need more than one or two samples. Your error is in thinking that "17khz x 2 = 34 khz playback". Sampling at 34 khz will only give you two samples of the entire wave, so clearly a lot of information will be lost. If you take 10 samples during one complete wave cycle then you will need to sample at 170 khz, which should give a fairly good representation. It works te same way for playback. To re-create an accurate 17 khz sine wave you need more than one or two samples. I think thats how it works for straight sampled sound in a file like WAV, although encoding algorithms like mp3 manage to squeeze more information in. Please correct me if Im wrong.

Yup, I might very well be mixing things up. I'm no expert, merely going by articles I've read, like some linked a bit further back.

[stringtapper]

Somebody correct me if I'm barking up the wrong tree here - apart from recording, isn't playback of more than 48khz 16 bit a total waste? Human hearing goes up to 20khz, most people can't hear anything after 17khz. 17khz x 2 = 34khz playback. So 48 just to be safe, though I don't think more than one out of a billion people can hear up to 24khz, if anyone.

24 bits? This mainly affects the noise floor and its relation to dynamics - the difference between the loudest and the softest sound. 16 bits is more than enough to accommodate the human ear.

And I've read that these very high sample rates sound better to some people purely because it sounds different, not better (subjective yes) than 44khz. These very high sample rates introduce harmonic distortion don't they, which colours the sound somewhat, giving people the impression it's higher quality.

What is the point of 192khz playback? Someone please inform me what I'm missing. The human ear can only hear so well...

I think you are confusing sample rates with audio frequencies.

If you want to make a digital representation of a sine wave at 17 KHz then you have to sample it at a number of points. If you want a good representation then you need more than one or two samples. Your error is in thinking that "17khz x 2 = 34 khz playback". Sampling at 34 khz will only give you two samples of the entire wave, so clearly a lot of information will be lost. If you take 10 samples during one complete wave cycle then you will need to sample at 170 khz, which should give a fairly good representation. It works te same way for playback. To re-create an accurate 17 khz sine wave you need more than one or two samples. I think thats how it works for straight sampled sound in a file like WAV, although encoding algorithms like mp3 manage to squeeze more information in. Please correct me if Im wrong.

I think you are wrong actually. You're confusing frequency bandwidth with the time domain. A sample rate of 96kHz means that the AD converter is taking 96,000 samples of the sound per second. Therefore how many samples of a wave you get is dependent on a combination of the sample rate, the frequency of the sampled wave, and length of time that you're sampling.

What tedlogan was talking about with the 17kHz/34kHz business is what's called the Nyquist Theorem. Basically it means that in order to accurately represent one cycle of wave of frequency X the sampling rate must equal at least twice frequency X.

So at a sample rate of 48kHz the highest theoretical frequency that can be accurately represented is 24kHz. I say "theoretical" because real-life AD converters have to have analog filters in order to cut off at that highest frequency and avoid fold over (aka aliasing) and analog filters are going to have some kind of curve to them, however steep, and may not align exactly at that frequency. Most likely manufacturers will shoot for just under to eliminate the possibility of aliasing.

[Davo]

Thanks for this Stringtapper. It is indeed true in theory that in order to be able to reconstruct a signal the sampling frequency must be at least twice the frequency of the signal being sampled. As you point out though,this is a theoretical minimum and a number of engineering difficulties mean that this theoretical minimum cannot be achieved, which is presumably why music is recorded at such high sampling rates. Tedlogans post did not refer to Nyquists theorem so I gave what I thought was a useful illustration of what is happening when sound is sampled. It seems a little harsh to say that I was "wrong" as you put it. It seems correct (in practice) to say that " If you want a good representation then you need more than one or two samples.", and that information is lost if the sampling frequency is too low.

It sounded like Tedlogan was equating sound frequencies with sampling frequencies but if not then I was barking down from the wrong tree. However, if such high sampling rates are needed to overcome the limitations imposed by sound equipment for digitizing audio then doesn't this suggest that exactly the same limitations arise in the reconstruction of the original sound? This is presumably the rationale behind the Pono idea, which takes us back to the question that Tedlogan was posing.

[stringtapper]

Well interpolation is always happening in when sampling audio digitally. Interpolation between samples. So the interpolation that's happening between the two samples at that highest frequency is happening in the same span as the interpolation between any two samples of any other frequency. But I will have to study up on how accuracy is affected and even what "accuracy" means in terms of the upper threshold of the Nyquist Theorem.

The question of high sample rates, and particularly about 192kHz, is related to how much of the frequency spectrum is really needed for representing sounds in ways that we can actually hear. The limit, or minimum sample rate is not theoretical but very real as aliasing will indeed occur if there isn't a low pass filter implemented in the converter around that limit. All manufacturers implement these kinds of filters in order to avoid aliasing though.

Taking into account the frequency response of the microphone or direct input used, the threshold of human hearing, and the frequency response of the speakers used to reproduce the sound is the deal. That's why people are questioning whether a sample rate of 192kHz, yielding a maximum bandwidth of 96kHz is really necessary when we can't even hear it, let alone capture it or reproduce it with speakers.

I'm still reading up in depth about this but it's certainly all very interesting.

[beats me]

I think it will sell better if he changes the name to Algorithms by Young.

[stringtapper]

What gets under my skin about this is the blatant marketing speak about "putting the soul back into music."

In fact the more I read into the 192kHz debate the whole thing seems to be a clash of marketing vs. science. And marketing speak obviously wins as my new MOTU 828x does 192kHz and I'm sure many of the other consumer level interfaces do as well.

[tedlogan]

Some might find this video quite informative. I enjoyed watching it:

http://www.youtube.com/watch?v=cIQ9IXSUzuM

[Davo]

Filters are used to cut out anything above the "audible range" in order to set the frequency bandwidth, but those higher frequencies (like the harmonics shown in the oscilloscope) appear to be integral to the original sound. As air vibrations those higher frequencies are still air movements so they will reinforce or subtract from the audible range frequencies. They can be recorded digitally by using a higher sampling frequency. In the digital to analog playback process the samples are used to create electrical signals to drive amplifiers and loudspeakers. The encoded higher frequencies are converted to electrical higher frequencies as part of the analog signal construction. Any higher frequency electrical signal components will sum with the lower (audible) frequency components as part of the DAC proces. This is how recording frequencies above the audible range and using them in the DAC process might more closely recreate the original recorded sound. As far as I know there is no theorem that gives an upper limit for a sampling rate above which no further useful information can be encoded, but please correct me if i'm wrong.

[stringtapper]

Filters are used to cut out anything above the "audible range" in order to set the frequency bandwidth, but those higher frequencies (like the harmonics shown in the oscilloscope) appear to be integral to the original sound. As air vibrations those higher frequencies are still air movements so they will reinforce or subtract from the audible range frequencies. They can be recorded digitally by using a higher sampling frequency. In the digital to analog playback process the samples are used to create electrical signals to drive amplifiers and loudspeakers. The encoded higher frequencies are converted to electrical higher frequencies as part of the analog signal construction. Any higher frequency electrical signal components will sum with the lower (audible) frequency components as part of the DAC proces. This is how recording frequencies above the audible range and using them in the DAC process might more closely recreate the original recorded sound. As far as I know there is no theorem that gives an upper limit for a sampling rate above which no further useful information can be encoded, but please correct me if i'm wrong.

Well the whole problem here is how to define "useful" in the context that you're using it. From the research I've been doing this week it seems like it's still up in the air as to what impact ultrasonic frequencies may be having on audible frequencies in a way that could make a 96kHz file sound "better" or more "accurate" than a 48kHz file when the fact of the matter is that the frequency range that is gained in the 96kHz file (roughly 24kHz—48kHz) is not in the range of human hearing.

One thing that is known is intermodulation distortion can occur between ultrasonic and audible frequencies and Lavry claims that these effects are negative because the resultant distortion is never harmonic in nature. According to Lavry this effect gets worse as the frequency bandwidth is increased, which is one of his main arguments against using 192kHz or anything above 96kHz.

[re:dream]

Are we back in the sampling rate debate ? I thought this had been settled

[stringtapper]

Are we back in the sampling rate debate ? I thought this had been settled

When?

And what was the verdict?

[Forge.]

Technically, higher sample rates can be less accurate, a lot of people are starting to think around 60kHz might be ideal. I just had a broken hand set this morning though, so there's no way I'm typing it all out. You'll just have to Google it:

Dan Lavry

I was about to mention him too. 192 can actually be worse, it's certainly not worth the bandwidth and hard drive space.

There's no f##ing way Neil Young at his age could hear well enough for any of this to matter to him as much as he seems to think it does. I'm 37 and I've already lost most above 15k

And as the point has been made - the fact that it's portable makes the entire idea stupid. Are you going to really get the benefits sitting on a bus or in basically any noisy environment at all that you are likely to be in when you'd want a portable device?

I have to admit I haven't looked into it at all, but I am a big audiophile sceptic and have very little time for the sheer amount of bullshit in this industry. The Comb filtering you get from turning your head slightly when you're sitting on the sofa has a bigger effect on listening than any of the ultra subtle effects that he is trying to fix with this.

My most cynical self thinks it's purely an exercise to try and resell his catalogue

But ironically I was just whistling "Harvest moon" when I was walking the dog so I guess he wins.

Anyone into audio who hasn't already watched it needs to watch the AES Audio myths workshop: https://www.youtube.com/watch?v=BYTlN6wjcvQ - I bet anything that Ethan Winer dude has loads to say about what a crock it is.