The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Launch of Pono Studio Headphones
Old 16th April 2014
  #3151
Quote:
Originally Posted by chrisj View Post
There was that one guy in the background of one of the Pono videos not smiling. He listened to it, and he said 'I don't get what's such a big deal'.

They took him out and killed him. Now all the rest of the listeners are suitably impressed!
OK. You win. heh
Old 16th April 2014
  #3152
Gear Head
 
bandpass's Avatar
 

Quote:
Originally Posted by theblue1 View Post
Um...

I think, for most of us in audio, the most familiar use of 1 bit sigma-delta is DSD/SACD.

Delta-sigma modulation - Wikipedia, the free encyclopedia
I'm sure I read they'd moved away from 1-bit DSD converters to multi-stage with the first stage with around 5 levels. But maybe that's just moving the problem to the later stage?
Old 16th April 2014
  #3153
j_j
Lives for gear
Quote:
Originally Posted by chrisj View Post
I could see another FIR-type filter being perceptually better (I'll accept as a given that it's analytically better) but I'm not sure it can really be called 'far from optimum' given that the optimized averaging has a function at lower frequencies, and these are likely the only ones we'll ever hear. To sacrifice effectiveness at this averaging in order to get better performance above human hearing seems not automatically better.
A moving average has a huge amount of high frequency loss at the intended maximum frequency. A proper FIR does the same averaging, has frequency response as good as you design, and is how to do this particular kind of processing.

You can use several kinds of FIR. You can use equiripple constant delay, you can move some of the zeros inside the unit circle and make the filter apodizing without going to IIR. You can even use cascaded FIR and IIR. Filter design is an old, old skill, first one I used at Bell Labs in 1976, and it is pretty well understood, really.
Quote:

Are you the JJ who did the 'top ten mistakes' list? I absolutely love #5, 'giving people a pan pot' (complete with pointing out the absence of time delay information, and that mono compatibility's not exactly relevant anymore).
Yes, that's me.
Quote:

I see the converter paper's advocating 19 bits word length or better. Sounds good to me

I do NOT see the paper arguing against a moving average filter versus a more high-order and well-behaved filter. If anything, they raise the time domain issue themselves. I'd need a reason to care about things like 'the moving average rejects the stop band poorly' given that our ears reject that stop band too.
I could write the paper if you want, but looking at the frequency response of such a filter is all one needs to do, it's very far from rocket science.

You will find 4 plots in this note. One is the moving average frequency and phase response. One is the frequency response from a 16 tab constant delay filter. One is the frequency response for a 16 tap minimum phase filter, and the last is the two impulse responses for the two filters. Note that the amplitude response of the two filters (constant delay and minimum phase) are absolutely identical modulo a small gain error that I failed to correct.

Note that the 16 tap filter removes a LOT more high frequencies, which means it removes MORE NOISE. I repeat MORE NOISE, i.e. it does a better job than the moving average filter can ever do.
Attached Thumbnails
Launch of Pono-movavg.jpg   Launch of Pono-16cdelay.jpg   Launch of Pono-fltrs.jpg   Launch of Pono-16ap.jpg  
Old 16th April 2014
  #3154
j_j
Lives for gear
Quote:
Originally Posted by nuthinupmysleeve View Post
The thing the anti-science types CONTINUALLY miss is, if there is ANYTHING that proves the 16/44 carrier as inadequate, science will happy change to match it.

Is 16/44 adequate? So far! Do I personally think it's the end state of audio delivery? Not sure... probably not. In the meantime, you don't choose a delivery system by magical thinking and "wishing".. you change it based on sound science.
All that's needed is solid, hard evidence.

Solid evidence never EVER includes sighted tests, i.e. tests where one knows the signal under test.

Solid evidence requires a PROPER test, of apples to apples, not oranges to guavas.

All it takes is real, cold, hard evidence.
Old 16th April 2014
  #3155
j_j
Lives for gear
Quote:
Originally Posted by Sounds Great View Post
At least? So in addition you are suggesting that the digitization can some how improve it?




That is up to science to figure out, but it is pretty obvious to a lot of people with ears.
So, rather than provide falsifiable evidence, you simply insult people?

You're going to succeed in nothing more than getting people angry until you provide some evidence.

The way it works is: You make the claim, you provide the evidence.

I await your evidence. Falsifiable, reproducible evidence.
Old 16th April 2014
  #3156
j_j
Lives for gear
Quote:
Originally Posted by Sounds Great View Post
What about DSD? It is superior to any PCM format, by a lot.
Really, again, you make a claim but you provide no evidence.

Just provide that testable, falsifiable evidence, and talk somebody into running the actual research (they can hire me to do it if they like) and I'll tell you why.
Old 16th April 2014
  #3157
j_j
Lives for gear
Quote:
Originally Posted by bandpass View Post
That's describing a problem with (now obsolete) single-bit ADCs, not DSD per se. AFAIK, modern DSD is adequate (but inefficient compared to PCM) for transparency—I wouldn't expect a modern DSD recorder inserted into CD or Pono playback path to audibly degrade it.
You do all realize that almost every DAC on the market is using a DSD-like technique to put out the 20-20K signal, right?

The only conceptual difference is in the input.
Old 16th April 2014
  #3158
j_j
Lives for gear
Quote:
Originally Posted by chrisj View Post
-you can't really bury a tone inside the digital noise floor in the same way that you can hear an analog tone alongside analog noisefloor, because the quantization tends to screw it up.
Wrong, it works exactly the same way, assuming both noises have the same spectrum and level. Of course, in the real world, digital noise floors are a lot lower.

Of course, if you don't dither, you didn't quantize properly. If you don't do it right, it's not right. d'oh.
Old 16th April 2014
  #3159
j_j
Lives for gear
Quote:
Originally Posted by bogosort View Post
How "far" you can hear under the noise is a function of time (or, equivalently, samples). If you have one sample whose amplitude is below the noise, then you will hear only noise (this should be obvious: noise will have every amplitude randomly, so it is impossible to determine whether any particular amplitude is signal or noise).
Not quite. As to "hear" as opposed to measure, your ear, if you will recall, analyzes things via frequency analysis that is about 50Hz wide at low frequencies and about 1/4 octave wide at higher frequencies.

What that means, for any given frequency, is that a lot of the noise that is not included in that band simply does not count for purposes of detecting a given tone.

For instance, if we have white noise at power '1' that goes from DC to 20kHz,
in the band around 1kHz the noise will be some 20.6dB lower.

It is pretty well established that the masking threshold for noise masking a tone is about 5.5 dB, so you can detect a tone at 1kHz under that noise at about -26.1dB relative to the TOTAL noise level.

Of course, as pointed out elsewhere, filtering is a form of averaging. Mathematics continues to work as expected.

Bear in mind that in order to distinguish frequencies 173Hz apart, the filter length must be at least 1/173 of a second long, or about 5.7 milliseconds, MINIMUM for 1/e rejection and in practice will be much longer. So you're taking an average, via filtering, in your ear, over at leat 5.7 milliseconds, and in reality it's about 20 milliseconds.

Hope that helped somebody.
Old 16th April 2014
  #3160
Gear Head
 
bandpass's Avatar
 

Quote:
Originally Posted by j_j View Post
You do all realize that almost every DAC on the market is using a DSD-like technique to put out the 20-20K signal, right?
Yes, but they still use more than two levels, so Lip****z & Vanderkooy doesn't apply.

The question was do current techniques for generating DSD still suffer from the problems highlighted in Lip****z & Vanderkooy's paper.
Old 16th April 2014
  #3161
Gear Guru
 
Sounds Great's Avatar
 

Quote:
Originally Posted by j_j View Post
So, rather than provide falsifiable evidence, you simply insult people?

You're going to succeed in nothing more than getting people angry until you provide some evidence.

The way it works is: You make the claim, you provide the evidence.

I await your evidence. Falsifiable, reproducible evidence.
Not insulting at all. The ear part is just pointing out what actually matters. A lot of people can hear these differences, and apparently a lot of people can't, or haven't bothered to even try, because they have already made up their minds.

I am more than willing to take a properly set up test. So far I haven't been invited to one. I don't need the test. The difference between the same music on CD versus SACD is so strikingly different to me that at amazes me that someone could hear the two and NOT hear that difference. If someone wants to test me to attempt to prove me wrong, by all means, bring it on.

I haven't to this day been provided with proper documentation of any test that was done and the details of how the test was conducted. Just talk that "lots of people were tested and they couldn't tell the difference".
Old 16th April 2014
  #3162
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by j_j View Post
Note that the 16 tap filter removes a LOT more high frequencies, which means it removes MORE NOISE. I repeat MORE NOISE, i.e. it does a better job than the moving average filter can ever do.
Sounds good. Presumably it's still a moving-average-LIKE filter and still does that functionality, too?

I can't imagine that there would be a problem with amending the design to include the additional 12 taps. The important part is that there's an area roughly four taps that makes up the bulk of the filter behavior. There's no reason at all that there couldn't be some more taps following up and cleaning up the behavior.

I don't feel there should be any lower-level taps happening before the four (three and a bit?) primary taps, but after them seems fair game.
Old 16th April 2014
  #3163
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by j_j View Post
Wrong, it works exactly the same way, assuming both noises have the same spectrum and level. Of course, in the real world, digital noise floors are a lot lower.

Of course, if you don't dither, you didn't quantize properly. If you don't do it right, it's not right. d'oh.
mmm. I am not absolutely certain this is the case, but one thing I do know is the significance of dither in the equation. None of us argue that undithered noisefloor masks such information very effectively.

I did an ABX once, years ago, where I thought I could hear the difference between hi-res noise and one-bit noise. And indeed I could—when the blast of raw noise was truncated.

When I dithered the raw noise with TPDF dither ('more noise'!) the files became indistinguishable.

So, perhaps I am wrong. Of course, in saying this, you're also saying that the '120db' or '95db' noise floors associated with digital audio aren't the whole story by a long shot. Signal can lurk beneath there, possibly far beneath there. This is another argument for higher word length, or (as with the averaging) higher effective word length.
Old 16th April 2014
  #3164
Lives for gear
 
nuthinupmysleeve's Avatar
 

Quote:
Originally Posted by Sounds Great View Post
A lot of people can hear these differences, and apparently a lot of people can't, or haven't bothered to even try, because they have already made up their minds.
The only people so far who can hear the difference haven't bothered to try, or haven't bothered to try fairly.

Have you done an a/b/x test? Please do and report your results!

Quote:
I am more than willing to take a properly set up test. So far I haven't been invited to one.
I pointed you to free 96/24 NIN files you can download. You can proceed to simply downsample any one file to 44/16, then upsample it back to 96/24. Then download a free a/b/x tester app. Free. (Be sure and run your audio interface at 96/24.. if you can't do this then you cannot test).

Sorry but I don't think anyone is going to do the 10 minutes of work to set it up for you. If you have the intellectual curiosity, it's easy to test yourself.

If you don't know the NIN music, find a different file you are interested in. archive.org has more samples for you to check out, as do any of the services like hdtracks.

Quote:
I don't need the test. The difference between the same music on CD versus SACD is so strikingly different to me that at amazes me that someone could hear the two and NOT hear that difference.
I don't mean to offend you, but... this is just accepting your own ignorance. I don't understand your choice to remain ignorant or in denial of your own confirmation bias. I don't know why you choose to hold on to the belief, in spite of all evidence to the contrary, that it doesn't apply to you.

Now, that doesn't mean you won't hear the difference.. you might! But you can't KNOW until you do a blind level matched a/b/x test.

Quote:
I haven't to this day been provided with proper documentation of any test that was done and the details of how the test was conducted. Just talk that "lots of people were tested and they couldn't tell the difference".
The Emperor’s New Sampling Rate -- Are CDs Actually Good Enough?
Old 16th April 2014
  #3165
Gear Guru
 

Quote:
Originally Posted by Sounds Great View Post
A lot of people can hear these differences
Quote:
A lot of people claim they can hear these differences
fixed that ^^ for you

Quote:
haven't bothered to even try, because they have already made up their minds.
Quote:
people who refuse to test blind are the ones who haven't bothered to even try, because they have already made up their minds.
fixed that ^^ as well

Quote:
I am more than willing to take a properly set up test.
Quote:
I am more than willing to take a properly set up test, but since I haven't done so yet I am going to refrain from making claims about 'what I can hear' .
see, just a small qualifier and your remarks ^ now stand up on their own

Quote:
I don't need the test.
Quote:
I don't need the test because I am the one human being on the entire face of the earth who is immune to placebo and expectation bias.
I can't think of anything different to add to a statement as hubristic as: "I don't need the test"
Old 16th April 2014
  #3166
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by j_j View Post
Not quite. As to "hear" as opposed to measure, your ear, if you will recall, analyzes things via frequency analysis that is about 50Hz wide at low frequencies and about 1/4 octave wide at higher frequencies.

What that means, for any given frequency, is that a lot of the noise that is not included in that band simply does not count for purposes of detecting a given tone.
This is an excellent point, thank you.
Old 16th April 2014
  #3167
j_j
Lives for gear
Quote:
Originally Posted by Sounds Great View Post
The difference between the same music on CD versus SACD is so strikingly different to me that at amazes me that someone could hear the two and NOT hear that difference.
The problem is, here you make a claim. In order to support it, you need to show that the exact same signal was recorded on Redbook vs. SACD to start with, for instance.

You insist that this is "so strikingly different". If so, you should be able to whiz through even a crappy test. So it might be worth YOUR time to prove YOUR claim.

Be careful and include controls, though, please.
Old 16th April 2014
  #3168
j_j
Lives for gear
Quote:
Originally Posted by chrisj View Post
So, perhaps I am wrong. Of course, in saying this, you're also saying that the '120db' or '95db' noise floors associated with digital audio aren't the whole story by a long shot. Signal can lurk beneath there, possibly far beneath there. This is another argument for higher word length, or (as with the averaging) higher effective word length.
How is this another argument for longer word length, then? You don't need longer word length to get the 26dB below the flat noise floor at 1kHz, for instance. That just happens.
Old 16th April 2014
  #3169
j_j
Lives for gear
Quote:
Originally Posted by chrisj View Post
Sounds good. Presumably it's still a moving-average-LIKE filter and still does that functionality, too?
Any filter is doing a weighted average. That's what any linear filter does.

But the way to know how good the average is, and how useful it is for in-band signals is to figure out the frequency response. You did take time to look at the filter response characteristics for the moving average, and the 2 16 tap filters, did you not?

Quote:

I can't imagine that there would be a problem with amending the design to include the additional 12 taps.
It's 16 taps this time because that's what I chose.

Note that you can NOT extend the "moving average" filter beyond 4 taps without severely impacting the in-band frequency response, which is going to hurt a lot.

You CAN extend the filter designed in the frequency domain as much as you want, within practical limits, and it will continue to get better as it gets longer.

Quote:
The important part is that there's an area roughly four taps that makes up the bulk of the filter behavior. There's no reason at all that there couldn't be some more taps following up and cleaning up the behavior.
Um, no, for the moving average filter you can only have 4 taps. With a filter designed in the frequency domain, you can have many more, leading edge or trailing, or both.
Quote:

I don't feel there should be any lower-level taps happening before the four (three and a bit?) primary taps, but after them seems fair game.
You did look at the two FIR filter responses, didn't you? Didn't you notice that one of them has non-constant-delay behavior that actually has in-band phase shift and for which the largest taps were at the beginning?

Do look at the plots, please. They aren't there for fun.
Old 17th April 2014
  #3170
Motown legend
 
Bob Olhsson's Avatar
 

Dither prevents distortion. It's not about masking anything.
Old 17th April 2014
  #3171
Gear Guru
 
Sounds Great's Avatar
 

Quote:
Originally Posted by nuthinupmysleeve View Post
H


I pointed you to free 96/24 NIN files you can download. You can proceed to simply downsample any one file to 44/16, then upsample it back to 96/24. Then download a free a/b/x tester app. Free. (Be sure and run your audio interface at 96/24.. if you can't do this then you cannot test).
services like hdtracks.
How can a do a test on my computer? I have no idea what the sound card could do and how IT would factor into any kind of output. And how does this pertain to DSD?

My argument isn't really about different PCM schemes. I have an Alesis Masterlink. I CAN hear the differences in higher rates/bits, but it still sounds like PCM. I can understand why most people would have trouble with the differences as they aren't that big. They are there though, and I have no doubt that people listening to the Pono are noticing details they aren't use to.

DSD is another story all together. It does not sound like PCM. It doesn't sound like anything, but what was recorded. I would love to set up your ABX with CD vs: SACD. But then you all say the test can't be trusted because we can't be sure that the masters used were the same. So the only option is to record the SACD onto CD and then compare the two I guess. Is my Masterlink an acceptable recording device to compare the two? Or when I can tell you which is which 100 out of 100 times, will you then tell me that the problem must be the converter in my Alesis?
Old 17th April 2014
  #3172
Lives for gear
 
nuthinupmysleeve's Avatar
 

Quote:
Originally Posted by Sounds Great View Post
How can a do a test on my computer? I have no idea what the sound card could do and how IT would factor into any kind of output. And how does this pertain to DSD?
Tell us what computer you have and if you have a separate audio interface. We will figure it out.

Quote:
My argument isn't really about different PCM schemes. I have an Alesis Masterlink. I CAN hear the differences in higher rates/bits, but it still sounds like PCM.
You can only test what you can test. Testing DSD is much more difficult, so do that next if you can reliably pick out 24/96 vs 16/44. That will require a much more complex setup.

For the record.. I think if you would have asked most of us if we thought could hear a difference, we all would have said yes. I thought I would be able to.

Quote:
DSD is another story all together. It does not sound like PCM. It doesn't sound like anything, but what was recorded. I would love to set up your ABX with CD vs: SACD. But then you all say the test can't be trusted because we can't be sure that the masters used were the same. So the only option is to record the SACD onto CD and then compare the two I guess. Is my Masterlink an acceptable recording device to compare the two? Or when I can tell you which is which 100 out of 100 times, will you then tell me that the problem must be the converter in my Alesis?
Honestly you are getting caught up in what others think. It doesn't matter what we think, it only matters what YOU think. If you can tell the difference in a/b/x testing, then don't worry about it. BUT it's important to be fair in your testing, level matching, no tells in your switching, etc.

To answer your question, if you were to record to your master link from a DSD stream via analog and then a/b/x between your DSD and the master link, that would be a decent though not perfect test. As long as you level match and you do so in a way there are no tells, I think it would be a worthwhile test!
Old 17th April 2014
  #3173
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by chrisj View Post
Sounds good. Presumably it's still a moving-average-LIKE filter and still does that functionality, too?
They're weighted, though, which is why they have superior frequency response. The moving average is just a bad idea for a LPF. Comparing the two weighted filters, the linear phase version has superior phase response (constant delay below Nyquist), and the minimum phase filter has the "prettier" time response (apodizing).

Could I hear a difference between the two with fc of 22.05 kHz? Haven't tried, but probably not.
Old 17th April 2014
  #3174
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by Bob Olhsson View Post
Dither prevents distortion. It's not about masking anything.
Amen.
Old 17th April 2014
  #3175
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by Sounds Great View Post
DSD is another story all together. It does not sound like PCM.
Are you referring to single-bit DSD? You do realize that it has flawed quantization (see Lip****z & Vanderkooy), yes? And if you're referring to multi-bit DSD, well that is PCM.
Old 17th April 2014
  #3176
Gear Maniac
 

Quote:
Perhaps a bit OT, but I second this with my whole heart. When I record, I try to look away from the computer as much as I can to focus on the music. I also avoid buying chintzy controller gizmos with all kinds of color lit buttons for this reason. The only thing is - with a name like this, shouldn't Airwindows run on (gasp...) Windows too?
Old 17th April 2014
  #3177
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by bogosort View Post
They're weighted, though, which is why they have superior frequency response. The moving average is just a bad idea for a LPF. Comparing the two weighted filters, the linear phase version has superior phase response (constant delay below Nyquist), and the minimum phase filter has the "prettier" time response (apodizing).
I think we probably want to go with time response over phase and perfection of the cutoff, what with the striking listener reactions from Pono's moving average and Meridian/Ayre/Esoteric's apodizing filters. The thing is, it's commonplace for the ritzy high ender types to be listening over pretty fancy tweeters, wire of adequate capacitance, amplifiers that will pass supersonics cleanly, and so on. The fact that they pay hundreds of times what these things are probably worth (and paint them with magic paint and sit magic rocks on them) is sort of irrelevant when the baseline for that stuff starts out pretty capable (if you omit other genres of high-endiness like expensive SET amps powering single-driver speakers, which are about freedom from certain artifacts at the expense of EVERYTHING else).

It suggests that time response might be paydirt for getting this stuff to sound better. It's just that a moving average is not the ONLY way to do good time response.

I'd really like to see a 1K squarewave off the Pono. I'd like to see what it does with it, and if it manages to pull off a 'round corners with no ripple at all' behavior. A 10K squarewave might be more relevant.

As for DSD, it might be a 'the very air is alive!' (yeah, with overshoots and a tizzy of relentless noise!) situation that makes it so 'obviously different' to experience. I honestly think Pono might have an edge over DSD for naturalness and ease of presentation, for that reason. We can gripe about the moving average filter (assuming they're stuck on that and won't budge) but it is still going to spit out less craziness above 30K than a DSD converter. The thing about DSD is extremely good resolution at very low frequencies, since the error all concentrates way up high and has to be filtered out. But Pono looks to be just as good, I think, and without the issues inherent in DSD.
Old 17th April 2014
  #3178
Gear Guru
 
Sounds Great's Avatar
 

Quote:
Originally Posted by chrisj View Post

As for DSD, it might be a 'the very air is alive!' (yeah, with overshoots and a tizzy of relentless noise!) situation that makes it so 'obviously different' to experience.
Is that your experience as well? Do you notice this "relentless noise" when listening to DSD?


Quote:
Originally Posted by chrisj View Post
I honestly think Pono might have an edge over DSD for naturalness and ease of presentation, for that reason.
Have you compared them?


Quote:
Originally Posted by chrisj View Post
We can gripe about the moving average filter (assuming they're stuck on that and won't budge) but it is still going to spit out less craziness above 30K than a DSD converter.
Why are you concerned with what is being spit out above 30k?

Quote:
Originally Posted by chrisj View Post
The thing about DSD is extremely good resolution at very low frequencies
You mean as in the bass sounds more like the actual bass? Finally we agree.
Old 17th April 2014
  #3179
Gear Guru
 
Sounds Great's Avatar
 

Quote:
Originally Posted by bogosort View Post
Are you referring to single-bit DSD? You do realize that it has flawed quantization (see Lip****z & Vanderkooy), yes? And if you're referring to multi-bit DSD, well that is PCM.
I don't know, in actually listening I haven't really noticed that flawed quantization. Do you notice it when listening to DSD?
Old 17th April 2014
  #3180
j_j
Lives for gear
Quote:
Originally Posted by Bob Olhsson View Post
Dither prevents distortion. It's not about masking anything.
Not sure what that's directed to. It's not, but it does create a flat noise floor (which is the goal), which does mask things too far below that noise floor, because it's noise.
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump