24/192 Music Downloads and why they make no sense - Page 5

Nope.

Maybe sometimes, but not this time.



The quantization error is never removed.


More generally stated, quantizing an analog signal necessarily created qauntization error.


Only if done naively.


Exactly, but it does not remove the quantization distortion.


As a rule all real world program material sources have enough built-in random noise to decorrelate the quantization distoriton related to 16 bit digital.

But, it is the low pass reconstruction filter that actually removes the purported stairstep effect from the audiophile myth that Amir recited. This stairstep is related to the sampling frequency, and is outside the bandwidth of the digital signal.

The quantization error that dither decorrelates is in-band, and is therefore not affected by the low pass reconstruction filter. Two different things.

Amir's problem is that he continues to confuse the two.

But why downgrade a file in any way if bandwidth and disk space isn't an issue any more these days?

Continuing on with the digital image analogy... I myself got heavily into digital photography quite a few years ago in its early days. Back then with dial-up modems and hard drives of only tens of megabytes, file size was a factor. Having watched storage space and broadband explode over the years, those file size restraints have become largely irrelevant.

Yes an image with level 8 JPEG compression looks good and you would think that nothing is wrong with it. But when we started working with 16 bit RAW files - those moderately compressed JPEG's obviously had degradation. I even got to the point where I could see the difference between photoshop's compression level 12 to 11 JPEG. The slightest bit of compression resulted in a reduction to the dynamic range of the light in an image. They looked ever so slightly less 'real' and more like just a photo.

There's a scene in Baraka that demonstrates this quite nicely if you can compare the DVD against the Blu-ray. I'm not talking about the obvious resolution difference, but the difference higher compression has on the quality of light. There's a particular scene near the beginning of Baraka where some people are sweeping the paving stones around a temple in the golden early morning light. At least on the Blu-ray you could tell it was early morning light. (as a photographer you notice early morning light is different to late evening sunset light as things like the air is usually cooler and denser and with more moisture and less dust particles in it. It just has a different 'feel') But on the DVD a lot of the 'feel' of the light is lost and you could not tell if it was early morning or late evening light. The feel of the light is lost.

So having watched the increased quality of digital imagery thanks to increases in bandwidth and storage space... why not the same for music..???

No you didn't. You recited an audiophile myth.

You said:


Let's review a few of the properties of the steps:

1. Per the audiophile myth that you recited Amir, the steps are square waves.

2. Per the audiophile myth you recited Amir,, the steps have a frequency equal to the sampling frequency (in your examples, the sampling frequency seems to be 44.1 KHz)

What you seem to have either never learned or forgotten Amir, is the fact that a 44.1 KHz square wave has well-known harmonic components. They are a fundamental of 44.1 KHz and odd harmonics (third, fifth, seventh, etc.) of 44.1 KHz. IOW the square wave steps from the audiophile myth would be represented on a FFT plot by spikes at 44.1 KHz, 132.3 KHz, 220.5 KHz, etc.


The fact is Amir that the graphic you reference, in addition to being seriously flawed because its not what the author claims it to be (which is already agreed upon by many of us but apparently not noticeable by you), we can see from the labeling of the horizontal axis that the frequency range of this analysis ends at 19 KHz!

Now Amir, how would the above plot ending at 19 KHz show whether or not the stairsteps are removed, given that the stairsteps would be represented by spikes at 44.1 KHz, 132.3 KHz, 220.5 KHz?

So there we go - yet another rookie mistake of yours, Amir. You are trying prove you're right using a graphic that cannot contain relevant evidence because it ends at 19 KHz, and the evidence is at 44.1 KHz, 132.3 KHz and 220.5 KHz, etc.!

Amir, this is only compounded by the other error in your reference, being that it is not of 16 bit truncation, but actually of 8 bit truncation, which you also either can't grasp, or refuse to admit to!

In short Amir, your evidence is a very nice looking plot suffering only in that it is totally and utterly invalid vis-a-vis both its X and Y axis.

They are different. Yet every time I talked about quantization error in the context of 24 bits vs 16, you complained with the bit regarding reconstruction filter:


Clearly you had the two topics mixed or you wouldn't be giving one answer in the context of the other.


Per above, the confusion is all yours. I quoted what started the argument which only had to do with dither and noise shaping. I am glad we have resolved the confusion of you fighting a battle that didn't need fighting .

That is audiophile myth.

In fact, a properly made digital recording does not have worse distortion in the very quiet passages.

A digital channel actually adds no distortion. None. Nada. Zero.

Interesting. Every time I talk about reconstruction filters, you keep rattling on about dither.


Only because you kept talking about dithering when repeating the audiophile myth about stairsteps.

The bandwidth that is generally used with music recording has been adequate to capture its full detail and dynamic range for about 30 years. As far as 24/96 goes, relatively inexpensive hardware has been able to capture and record it for over a decade. We've known for over a decade that it makes no audible difference. Bandwidth is far from being the limiting parameter in the sound quality of recordings.

Correct me if I'm wrong, but only naive amateurs buy cameras based on their megapixels, right? Audio sample rates and megapixels are actually pretty comparable.

No audiophile myth. Only a confusion on your part where you think that is what I said vs what I really said. Some quotes to see if we can deal with it since you are being stubborn otherwise :

From WBF Thread:

"That is not true [having more samples helps] as a matter of math Jack . As long as we have twice as many samples as we have bandwidth, we can fully represent the signal. More samples are redundant and don't contribute to proper reconstruction."


Answer from Ethan:
"Thank you for busting that oft-repeated myth Amir. Since that myth has been debunked almost daily in the various forums I visit, I don't understand why people keep repeating it.

--Ethan"

What's more, the topic had nothing to do with the step-wise representation of samples in *time*. But rather, what happens when you use fixed voltage steps to represent the amplitude of the waveform. That distortion, despite what the rest of your post says, will not just generate noise above sampling frequency. The author's measurements show it to be in-band. My simulation show it to be in-band. As does Ethan's. You continue to confuse the two topics even though you say you know they are separate.

Amir, thanks for proving that while I can copy what you said in context any number of times, you can always deny what you said one time more.

You made me go back to look up what started it again and realized maybe I gave you too much credit for thinking I said something else Bolding mine:


Is it your position that having fixed bit depth is of no consequence in digital systems?

Irrelevant to the current discussion,

How could exploring what you meant by the "myth" be irrelevant to the current discussion?

I ask again, is the fact that samples have "fixed bit depth" cause for concern?

Would that be the masters recorded in 16 bit or 24 bit? Do do you mean a track being released on a CD in 16 bit is enough? Or do you mean a track recorded in 16 bit is enough?




That is correct. More pixels crammed onto x amount of space means smaller pixels. Smaller pixels cannot collect as many photons as a larger pixel so they result in a lower signal to noise ratio. i.e. they produce more noise because of their physically smaller size.

I don't see how increasing the sample frequency or bit depth has a negative effect like that of increasing pixel numbers though???

Amir, the myth relates to the stairstepping, and its removal, about which you have been shown to be totally wrong about in a number of ways, and which you still seem to be denial about. This other issue is just another one of your red herrings. As soon as you admit your errors in this area, we may talk about other issues.

16 bits is an overkill format, when it comes to recording music. It has about 20 dB more dynamic range than the signals that come out of microphones in recording studios and live venues.

Audiophiles don't get this, and never see the evidence that shows this to be the case. I'm a professional recordist who does both live recording and also studio work. So, I get my nose rubbed in this evidence every time I do a job.


That, too.


Tracks recorded with 16 bits are also enough.


I never said that using too many bits or too many samples causes technical problems for audio. It is just a waste. For short tracks such as are made for most studio work, the extra volume of data doesn't cause much difficulty. But, for live events running for the better part of an hour or even hours, they just waste time.

Amir, the customer is educated by someone, and that someone is often a person whose basic understanding of audio is highly flawed, even at the basic engineering level. Audiophile myths come from someplace, and they often trace back to the sales chain.


Yet another Audiophile Myth!

There is no such thing as the low level distortion that you just claimed was inherent, Amir.

Amir, you clearly don't understand even your own posts, which show that dither eliminates (not masks, but eliminates) the low level distortion that might be there were dither or other quantization randomization strategies weren't used as a matter of course.



Yet another repeat of this Audiophile Myth!

I repeat Amir, the alleged distortion you are talking about simply does not exist.

Like this microphone?
http://www.neumann.com/?lang=en&id=c...s&cid=u87_data

I appreciate you not wanting to discuss the topic anymore . But as you say, we still have a disagreement as you confirm here so we should sort it out. After all, when it comes to math and science, there should only be one answer and not two. So let's review again what you said:


It is clear that you are saying that having fixed levels does not create distortion. Otherwise, why would it be a myth?

If it is a myth that the steps add distortion, how about this. Why do we have 16 bits in our CD format? Why not 1 bit at the same sampling rate? If the steps by magic go away as you have been saying, there should be nothing wrong with that. Who wants to raise their hand and agree with Arny that it matters not that we have steps in our amplitude?


Red herring? Other issues? I am only asking you what you meant by the myth.

As I said yesterday, my sense is that you confused time domain quantization with amplitude quantization. That is evidenced by the fact that you keep thinking the distortion is only created at multiples of sampling rate rather than in band. From your original post:


Yet simulation and measurements posted in this thread including that of Ethan's show that amplitude steps create distortions in the audible band, well below the sampling rate which is 2X higher than our highest frequency. Here they are again:



And



Clearly we have distortion in the audible range below half the sampling rate. And we see visible distortion created due to fixed amplitude steps below the resolution of our original signal. Since the first chart is generated using analog measurements post DAC, then we know that noting in our digital system removed them.

The "audiophile myth" as I stated in whatsbestforum is about fixed samples in time being problematic. That myth we can bust using sampling theorem. And the solution is the band limiting of our system.

There is NO myth related to fixed steps causing distortion. Mathematics this time *guarantee* that it will be there rather than the other way around. There is an entire science around quantifying it, and how to deal with it using dither.

Bottom line, you are dabbling in audio alchemy, imagining solutions to problems that mathematically can be shown to be false. If you still think this is a myth and that you can recover our signal using "reconstruction filter," you may want to note this line from the Wiki: http://en.wikipedia.org/wiki/Quantiz...nal_processing)

"Because quantization is a many-to-few mapping, it is an inherently non-linear and irreversible process (i.e., because the same output value is shared by multiple input values, it is impossible in general to recover the exact input value when given only the output value)."

Saying otherwise amounts to believing in unicorns and such. Who said only one camp is guilty of that?

As I said earlier, your arguments in this area is the best reason to deliver 24 bits to the consumer. If you can't figure out the science behind this, what hope there is for some creative type trying to produce music?

When did you get in the business of educating audiophiles?


Were you in sales? Just in this thread you said that time and frequency domain are not connected. Or that quantization levels are a myth when it comes to inducing distortions. You believed in audibility of ultrasnoic tones creating audible distortion and the argument that we are harmed by existence of wider bandwidth. And by implication, we know how to build an audio system that has 22 Khz audio response but incapable of doing so at 44 Khz.


Inherent? You are saying when I generated my tone using high resolution samples, it came with these harmonic distortion spikes?



I think you are going way past believing in unicorns and think Jurassic Park movie dinosaurs are real .


Ah, now you are saying my posts are accurate enough to be used as a proof point by you here? Yes, dither does that. That is what I said yet you kept arguing with me left and right. The point was that I can't rely on people like you knowing it and leaving the checkbox on in the sample conversion. Folks read that dither is noise and may think it is a bad thing to add to their recordings.


So a visual spectrum that shows them is a falsity? Seems like you are back to thinking frequency domain distortions don't exist in time domain. As I said, unicorns and such are alive in your world as much as audiophiles. Sad to say.....

I've been arguing for years that people should aim to hear what the mixing and mastering engineers heard. But they obviously don't care about that. If they did, every audiophile would have extensive acoustic treatment. Arguing about infinitesimal amounts of distortion while ignoring the elephant in the room - peaks and nulls spanning 30+ dB - shows that most 'phooles have very poor priorities, and don't really care about high fidelity as much as they think they do.


When did I say that music quality should be downgraded? I might have said that 44/16 is adequate, but that's what I use to record in the first place. So do a lot of others.


What the hell is analog sound? 3 percent distortion of analog tape? 5 percent distortion and <20 dB crosstalk at best from vinyl?


Agreed. 128 kbps probably is good enough for some material, but it's definitely not good enough for everything. However, 128 kbps lossy compression is a far cry from uncompressed CD quality. So that's a red herring.


That is a lie. I was told not to post anywhere but my own forum section. The only reason my posts a few weeks ago in that other section weren't deleted is because the thread was about my own upcoming book! Steve even threatened to delete me from that thread until someone pointed out the hypocrisy of not allowing me to participate in a thread about my own book.

Amir, I don't take people on. I do call bullsh1t when I see it, but I always do so nicely, and I always explain why I think [whatever] is bullsh1t. It is always others who are rude and insulting to me first. If you really wanted your forum to be a place for intelligent and civil discourse, you'd have banned the others who make trouble and call me wrong without ever explaining what is right. This is the real issue. People like Mike Lavigne are ignorant and arrogant, and deep down they know they can't defend their position. I imagine you know this, and realize how ridiculous was Mike's list of "what digital is unable to capture." Yet he was allowed to state his nonsense, and I was not allowed to correct him. Truly pathetic.


Again with the red herrings. That's a specious argument because nobody listens that way.


Sure, but if one distortion is 100 times larger than another, which it is in this case, only someone ignorant about human perception would believe the smaller distortion matters.

--Ethan

This.

Correct me if I am wrong, but can't you make the CD sound exactly like the vinyl if one wanted?

Would you mind sharing the URL of your discussion on this topic between you and Mike? I am very curious about this.

Yes, lossy JPG compression is noticeable. But reducing "high res" audio to 16/44 is not. So there's another red herring. All this talk of fish is making me hungry!

--Ethan

Yes! I've made that very point many times. If someone prefers the degradation added by vinyl, they can record their turntable with clean digital, and never worry about the vinyl degrading over time with repeated playing. Same for the sound of analog tape. I've also suggested that audiophile labels could offer music recorded that way for those who prefer such distortion.

--Ethan

Where is this discussion ethan? you and mike. I must read this

This is a follow-up, and it starts by linking to the original:

http://www.whatsbestforum.com/showth...esolved-issues

--Ethan

Yes.

The microphone by itself only has about 100 dB dynamic range from its self-noise to the SPL that is used for specifying distortion.

But, when used to record music, its effective dynamic range is limited by acoustic noise floor of the room. While there are rooms with residual noise in the 20-30 dB SPL range, putting a few people into the room who move, have heatbeats and move their limbs to say play instruments pushes that up another 10-20 dB.

The manufacturer gives its SNR as being 80 dB or less.

But which tonearm/cartridge combination? I understand what you're saying and while the words above don't cause me any personal affront, I can see how the choice can cause vinyl aficionados raised blood pressure which is why I've rewritten the above choosing softer terms.

Hate to be hypercritical but it's 105 dB dynamic range per the link.
The heartbeats, arm movements, and breathing are all part of the performance.

The distortion of factual information is not necessary. The point you are trying to make is that 16 bit dynamic range is more than adequate for recording once all the noise floor is taken into account and not that the equipment isn't good enough. Frankly the vast majority of pop music doesn't come close to 16 bits dynamic range.

I know people love to win an argument, and obviously there a lot of back story among posters, but each side of the debate has some merit without both sides distorting facts, misrepresenting each others statements, and adding ad hominem attacks.