Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Accuracy FLAC decoder (Read 29192 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Accuracy FLAC decoder

Reply #50 –
Quote from: greynol on
Even if it couldn't pass DTS, that still wouldn't explain alleged differences in bass response between flac and Monkey's Audio.


Of course not, one of them could be adding it to the DTS signal while compressing it ... uh ... *cough*

Accuracy FLAC decoder

Reply #51 –
Quote from: MMime on
just ABX your FLAC and APE files first (don't record your sound card output, just ABX the files). If you cannot hear the difference, this means that 1) the difference you hear is cause by confirmation bias OR 2) that there is a problem with your original software.

Or 3) that like any other tool, ABX test has some limitations. It's amazing, is not it? Although I have long understood this.

I wonder more thing: my problem is unique or somebody after you install Winamp does the same as "me" - report about 24bit in 16bit files?

Accuracy FLAC decoder

Reply #52 –
Quote from: Bublic on
I wonder more thing: my problem is unique or somebody after you install Winamp does the same as "me" - report about 24bit in 16bit files?


Your problem is uniquely imaginary.

Accuracy FLAC decoder

Reply #53 –
saratoga
You topics throughout wrote one message, many times. If they are removed, nothing will change. I have no doubt that getting a video screen with a running Winamp, you have not once write that all the problems have been invented.

P.S. Translator into English by the way, too, thinks I'm a Jedi Yoda. Please all treat with understanding to my spelling.

Accuracy FLAC decoder

Reply #54 –
Quote from: MMime on
Quote from: Neuron on
How does Windows require dithering for both FLAC and APE? You mean it requires dithering for everything?


From Win Vista on, yes, without ambiguity: everything is converted to float 32 by Windows and is only converted back to what the DAC likes right before the DAC. From what I read, dithering is applied in all cases, even for 24bit. In earlier versions, with kwin, things do not seem that clear however kwin arguably screwed things up in a number of setups (hence the audio stack rewrite).


I have not seen Windows dither 24-bit at all.  I've seen it do it to 16-bit audio.  At least with WASAPI as loopback in the place of using Stereo mix for cards that don't have it.  Unless that is caused by drivers or the fact I was using Windows 7 at one time.

Unless you referring to MME or directsound where I have had issues with -90 dB SNR with SPDIF in when I was expecting nothing but digital silence.

Accuracy FLAC decoder

Reply #55 –
Quote from: Chibisteven on
I have not seen Windows dither 24-bit at all.


First of all, Microsoft uses 'dither' as a synonyme for 'conversion to a lower bit-depth' in pretty much all its documentation. If you think this is inaccurate, you can contact their customer service to get a definite answer.

Second, dither is said not to be necessary for 24 bit audio. But that does not mean that there shouldn't be any: dithering when downsampling is always better.

Third, if your ears (an mine, and anybody's) cannot hear the difference anyway, then either way, that's all good.

Accuracy FLAC decoder

Reply #56 –
Quote from: MMime on
Quote from: Chibisteven on
I have not seen Windows dither 24-bit at all.


First of all, Microsoft uses 'dither' as a synonyme for 'conversion to a lower bit-depth' in pretty much all its documentation. If you think this is inaccurate, you can contact their customer service to get a definite answer.

Second, dither is said not to be necessary for 24 bit audio. But that does not mean that there shouldn't be any: dithering when downsampling is always better.

Third, if your ears (an mine, and anybody's) cannot hear the difference anyway, then either way, that's all good.



It's nothing to do with audibility or if I can hear a difference but rather lossless transfer from certain applications that don't have the ability to dump their data to a wav file to begin with to either work on or analyze but use common and supported rates.

And the concern that external digital devices are having their output needlessly altered in the signal processing chain before being recorded.

Accuracy FLAC decoder

Reply #57 –
Quote from: MMime on
First of all, Microsoft uses 'dither' as a synonyme for 'conversion to a lower bit-depth' in pretty much all its documentation. If you think this is inaccurate, you can contact their customer service to get a definite answer.

So here is a reply from a Microsoft representative:
https://social.msdn.microsoft.com/Forums/wi...api-shared-mode

==============================================
I can only find the word "dither" two times in your link.
Quote
Each sample is justified to the most significant bit, so it is easy to add additional precision on the bottom of a sample, or dither to a lower precision. The container size can be manipulated independently without implying any change in the data precision. A 24-bit stream in 32-bit containers (for efficient processing) can safely be transferred into 24-bit containers (for efficient storage space) without any data loss.

Since dithering in 24-32-24 is lossy, when it says "without any data loss" then it means dithering can be avoided.

Another paragraph with the word "dither":
Quote
An entity can change wValidBitsPerSample as it processes the data. For example, an application would know that a stream with wValidBitsPerSample = 24 must be dithered to 16 bits if the output driver indicated that it supported wValidBitsPerSample = 16 only.

Other than these two paragraphs I could not find the word "dither". None of these paragraphs say dithering must be applied in 24-bit.

But it really doesn't matter. Even if you can find some real Windows source codes (instead of pseudo codes or essays) to prove that it always performs dithering in 24-bit, it doesn't automatically mean it will always happen in real life, just like Bill Gates did not expect a BSOD when someone plugged a scanner in a Windows PC during a live demo session.

You can download my test files in this thread, use a wave editor to examine if there is any 24-bit dithering or not. I was not even using WASAPI exclusive mode, just DS. Examine the short silence part at the beginning of the DS file to see if it is pure silence or dither.
https://www.hydrogenaud.io/forums/index.php...st&p=854321


Accuracy FLAC decoder

Reply #58 –
Quote from: Chibisteven on
It's nothing to do with audibility or if I can hear a difference but rather lossless transfer from certain applications that don't have the ability to dump their data to a wav file to begin with to either work on or analyze but use common and supported rates.

If you need that in a "bit-perfect" fashion not only for your ears, but also for your eyes, then there might be some better ways like a virtual sound card in exclusive mode to make sure that neither Windows, nor the drivers, nor the hardware changes that. Even better, there might be some way to hijack Windows pipeline for a specific application by force-loading a DDL that redirects the stream to a file instead of Windows audio stack (for people familiar with Linux, this could work the same as proxychains but for audio). There might also be other ways to get the orginal sound as opposed to capturing it.

What I mean is either you consider the playback transparent and good enough (with all the dithering Windows engineers wanted) or you want to be absolutely sure that the output is bit-perfect (in which case there are better ways than trusting your OS audio stack and your sound card manufacturer).

Quote from: Chibisteven on
And the concern that external digital devices are having their output needlessly altered in the signal processing chain before being recorded.

Do you realize that in this specific case, you are talking about 1-2bit of "alteration" (which is needed in every but one case to obtain a more "genuine" sound) in a 24bit signal as opposed to the 16bit that your ear can hear?

Accuracy FLAC decoder

Reply #59 –
Quote from: bennetng on
Quote from: MMime on
First of all, Microsoft uses 'dither' as a synonyme for 'conversion to a lower bit-depth' in pretty much all its documentation. If you think this is inaccurate, you can contact their customer service to get a definite answer.

So here is a reply from a Microsoft representative:
https://social.msdn.microsoft.com/Forums/wi...api-shared-mode


Ok, then here we are:
Quote
The audio limiter APO will also kick in if it detects sample values that are outside of [a specific] range [...]

The audio engine works in floating-point samples internally and there may be dithering in the path.

You can avoid both of these by using WASAPI exclusive mode instead of WASAPI shared mode.


So even though you explicitly specified that you were using 24bit and disabled everything you could, he told you that there may or may not be dithering. His doubts can be caused by a number of reasons that only he can give. However, he didn't tell you something along the line "Don't worry: The audio limiter [...blah blah...], but Windows does not use dithering for 24bit sample so just stay in that specific range and you should get bit-perfect audio."

So at any rate, he did not rule out dithering for 24 bit samples, quite the contrary: this seems to be used at least in some cases (and "by default" if I extrapolate onto the comment about exclusive mode).

As far as my link is concerned, the quantity is not important: 'dither to a lower precision' is sufficient to say that 'dither' has be used in place of 'converted'. That's why I said MS used 'dither' as a synonym of 'conversion to a lower bit-depth' (they did that in some other places that I cannot find anymore, but well, your representative at least confirmed that dithering of 24bit audio is used).

Quote from: bennetng on
Quote
Each sample is justified to the most significant bit, so it is easy to add additional precision on the bottom of a sample, or dither to a lower precision. The container size can be manipulated independently without implying any change in the data precision. A 24-bit stream in 32-bit containers (for efficient processing) can safely be transferred into 24-bit containers (for efficient storage space) without any data loss.


Since dithering in 24-32-24 is lossy, when it says "without any data loss" then it means dithering can be avoided

Oh dear... I am not saying it cannot. That was absolutely not what I tried to imply anywhere. I am saying that:
  • MS doc (and now your representative) implies that 24bit dithering is implemented.
  • MS doc (and now your representative) is vague about when it is used, but right or wrong, it is enough to reasonably suppose that it is used by default, especially after applying Windows audio filters (that's what they should do anyway).
  • It is most certainly used when the sound card is used in "shared mode" (from your representative comment).
  • It's been brought up only as a rhetorical argument: if the OP prefers APE to FLAC because (he said) he can hear the effect of that dithering, then what would he think if Windows dithered both files?
  • There is no reason why dithering should not occur on a playback pipeline: in the usual case, it keeps the downsampled sound closer to the original and in your case (24 bit -> 32 bit -> 24 bit supposedly without signal processing) it just doesn't change the perceived quality of the sound.


By saying "dithering can be avoided", you acknowledge then that dithering is very likely occurring by default when converting from 32bit to 24bit and therefore triggered a discussion about what you think I said as opposed to what I actually said, even after my trying to put that topic to rest.

Quote
But it really doesn't matter.


Precisely

Quote
You can download my test files in this thread, use a wave editor to examine if there is any 24-bit dithering or not. I was not even using WASAPI exclusive mode, just DS. Examine the short silence part at the beginning of the DS file to see if it is pure silence or dither.
https://www.hydrogenaud.io/forums/index.php...st&p=854321

Most DAC can't do real 24bit anyway and most of them detect silence and zeroes out. But even without that, if the amplitude of dithering is <1bit (this is the usual case with rectangular dithering), you wouldn't alter the silenced parts of your stream.  If I want to see the effect of dithering on your file, I need to align your signals (if there is no phase shift between the two, it should be pretty easy), then adjust the gain exactly to minimize the differences, and finally draw the distribution of the differences.

The question is: has that already been done?

Accuracy FLAC decoder

Reply #60 –
Quote from: Bublic on
Or 3) that like any other tool, ABX test has some limitations. It's amazing, is not it? Although I have long understood this.



What are the limitations of ABX which affect you in this situation?

Or are you just dismissing a serious point?
Creature of habit.

Accuracy FLAC decoder

Reply #61 –
The tone of your previous replies sound like you insisted there must be dithering in 24-bit therefore I made a reply. Now you acknowledged that it is not a must, then I am okay with it. I am not interested to discuss dithering is correct or not in some specific cases, I just interested to know if dither is present or not in the signal path when using different modes.

Quote from: MMime on
Most DAC can't do real 24bit anyway and most of them detect silence and zeroes out. But even without that, if the amplitude of dithering is <1bit (this is the usual case with rectangular dithering), you wouldn't alter the silenced parts of your stream.  If I want to see the effect of dithering on your file, I need to align your signals (if there is no phase shift between the two, it should be pretty easy), then adjust the gain exactly to minimize the differences, and finally draw the distribution of the differences.

The question is: has that already been done?


Can these two videos fulfil your needs? The first one is an ancient Roland USB audio interface sold at around 2001, another one is my motherboard's (Asrock B85m Pro4, Realtek ALC892) SPDIF output. The recording interface is Creative X-Fi Titanium HD.

https://youtu.be/TuEXZBmZlpM
https://youtu.be/HBwuspTZyvw

EDIT: Just want to clarify that person is not my Microsoft representative, the question is not asked by me, and it is dated in 2010. It is just a quick google search.

Accuracy FLAC decoder

Reply #62 –
Quote from: Soap on
Quote from: Bublic on
Or 3) that like any other tool, ABX test has some limitations. It's amazing, is not it? Although I have long understood this.



What are the limitations of ABX which affect you in this situation?

Or are you just dismissing a serious point?

For this situation, the determining factor is time. For example, to understand the difference in the sounds A and B, someone takes 2 seconds, and someone else for 2 months. You have agreed to conduct a rigorous and complex ABX test at least one week without being distracted by something extraneous? I think not. Of course all deadlines approached, but the essence of the difficulty (with little measurable difference at A and B), is this.

Accuracy FLAC decoder

Reply #63 –
Quote from: bennetng on
The tone of your previous replies sound like you insisted there must be dithering in 24-bit therefore I made a reply. Now you acknowledged that it is not a must, then I am okay with it. I am not interested to discuss dithering is correct or not in some specific cases, I just interested to know if dither is present or not in the signal path when using different modes.

Concerning dithering itself for 24bit audio, I totally acknowledge that this is not strictly needed. The truth is that the noise coming from the studio and all the analog (recording) devices is (AFAIK) always higher than the +/-3 units random variations that dithering would introduce at most (and again, that wouldn't make any difference to our ears). Said otherwise, move your microphones by 1 inch at record time and you would get more variations than that.  Or said another way: loudspeakers are already struggling to playback 16bit audio with acceptable gains and phase, even with a perfect placement in a perfect room, so your 1-2bit variation around -130dB...  My point in this very specific regard is that dithering does not degrade the sound in any way (even at 24bit) and by a very large margin. So even if Windows does dither in his playback pipeline, no one should care at all.

Quote from: bennetng on
Can these two videos fulfil your needs? The first one is an ancient Roload USB audio interface sold at around 2001, another one is my motherboard's (Asrock B85m Pro4, Realtek ALC892) SPDIF output. The recording interface is Creative X-Fi Titanium HD.


That or, in the noiseless case, a typical rectangle dither is used... (as I said, the usual rectangle dither spans around +/-0.5bit, so there is no change for truly constant signal). That said, having no dither is far more likely (why would it change its dithering parameters in the first place?).

Accuracy FLAC decoder

Reply #64 –
Quote from: MMime on
That or, in the noiseless case, a typical rectangle dither is used... (as I said, the usual rectangle dither spans around +/-0.5bit, so there is no change for truly constant signal). That said, having no dither is far more likely (why would it change its dithering parameters in the first place?).

Please watch my video carefully. The dither I showed in the first video is from the original RMAA generated signal. The null test nulled everything to -inf means the whole DS-playback and MME-recording signal chain does not have any dither, and it is bit-perfect (if the limiter is not triggered).
Quote
Most DAC can't do real 24bit anyway and most of them detect silence and zeroes out

Since the null test also nulled the dither in the original file, that means your claim is false, at least among my equipment.

Accuracy FLAC decoder

Reply #65 –
Quote from: Bublic on
[...] to understand the difference in the sounds A and B, someone takes 2 seconds, and someone else for 2 months [...]
[...]with little measurable difference at A and B [...]


So, if I followed you correctly, you tell us that one of the limitations of ABX testing is that it could require a lot of time, I suppose in perfect listening conditions, if you focus only on this without any distraction, to get sufficiently used to each glitch of the samples to be finally able to tell difference?

Then you say that these conditions are rarely met and that's why ABX testing is limited. Is that correct?

Then three questions:

1) How much time and concentration did you allocate to spot the differences between the two formats?
2) You say that you can hear the differences in the bass, etc. Did you only try to spot them in an ABX test software, or that's just by listening, after carefully selecting the correct song in your player?
3) If you need that many favorable conditions (and you could add a planet conjunction at the top of that), then is that still relevant in your casual use?

EDIT: and as a complementary question... You are basically saying that the difference would be really really hard to spot, even by spending a few hours without distraction on them (I may remind the other readers that we are talking lossless here). Then you acknowledge that even you couldn't tell the difference if you were just given A or B (without being able to swap at will between them... for hours or months). Then what's the use?

Accuracy FLAC decoder

Reply #66 –
bennetng, please bear with me: I don't know everything nor can I understand everything as soon as I see/read/listen to it.

You indeed did exactly what I was talking about and proved that there is no dithering in your setup. \o/

Accuracy FLAC decoder

Reply #67 –
Quote from: Bublic on
Quote from: Soap on
Quote from: Bublic on
Or 3) that like any other tool, ABX test has some limitations. It's amazing, is not it? Although I have long understood this.



What are the limitations of ABX which affect you in this situation?

Or are you just dismissing a serious point?

For this situation, the determining factor is time. For example, to understand the difference in the sounds A and B, someone takes 2 seconds, and someone else for 2 months. You have agreed to conduct a rigorous and complex ABX test at least one week without being distracted by something extraneous? I think not. Of course all deadlines approached, but the essence of the difficulty (with little measurable difference at A and B), is this.


Check out the section on "long term listening" in this article from nwavguy: http://nwavguy.blogspot.ca/2011/05/subject...ive-debate.html
In long-term listening (months), people failed to hear a small difference when there was one, but later they could identify the difference in an ABX test.

I'd agree with the suggestion of recording the audio out of your computer playing APE and FLAC, if you have the gear to do it.

Accuracy FLAC decoder

Reply #68 –
Quote from: Bublic on
For this situation, the determining factor is time. For example, to understand the difference in the sounds A and B, someone takes 2 seconds, and someone else for 2 months. You have agreed to conduct a rigorous and complex ABX test at least one week without being distracted by something extraneous? I think not. Of course all deadlines approached, but the essence of the difficulty (with little measurable difference at A and B), is this.


So no actual problem with ABX, just an unwillingness on your part to do one.
Creature of habit.
 

Accuracy FLAC decoder

Reply #69 –
MMime
First of all, I have no confidence that the output of KS foobar2000, when there manually selected 16-bit, will behave as well as MME, DS output Winamp on my configuration. One of the older versions of foobar 1.1.5 has taken root on the computer, it seems to be a reference for music, although installed and newer, portable versions too.
1-2.) This is a spontaneous, random process, it has in my case there is any strict boundaries in time. A desire to listen to this or that music, I listen to it for yourself, and note that the APE 16/44 tend to sound a bit more articulate, which is especially noticeable in the area of bass and midbass ??percussion instruments. It was not much at all the records, and usually those that were issued in 80s and then transferred to lossless. Most clearly this effect can be discerned in the style of rock music and related music styles, I suppose. I admit that at best in the world of musical equipment time "go faster."
3) Whatever it was, to some extent, ABX is a stress for the human psyche, ie altered state of consciousness. What is fraught with committing more errors than the state, which is uninterested in anything particular listener.
Quote from: MMime on
Then you acknowledge that even you couldn't tell the difference if you were just given A or B (without being able to swap at will between them... for hours or months). Then what's the use?

The meaning of ABX, you mean? Probably to recognize and organize absolutely obvious. But there are not so obvious, to study which ABX test, as I understand it, from the very beginning was not intended.

Quote from: eric.w on
I'd agree with the suggestion of recording the audio out of your computer playing APE and FLAC, if you have the gear to do it.

While I will leave it at that, because I can not afford to make any decent record.
Thank you for the link!

Accuracy FLAC decoder

Reply #70 –
Quote from: Bublic on
1-2.) This is a spontaneous, random process, it has in my case there is any strict boundaries in time. A desire to listen to this or that music, I listen to it for yourself, and note that the APE 16/44 tend to sound a bit more articulate, which is especially noticeable in the area of bass and midbass ??percussion instruments. It was not much at all the records, and usually those that were issued in 80s and then transferred to lossless. Most clearly this effect can be discerned in the style of rock music and related music styles, I suppose. I admit that at best in the world of musical equipment time "go faster."


You have absolutely, totally, and with great verbiage, failed to describe how errors in digital audio reproduction sound.

THIS is why nobody believes you are actually hearing anything.

It isn't that the possibility of errors do not exist, but what you are describing is hallucinatory delusional bunkum.  The fact they are irreproducible is even more damning.

And it isn't the language barrier, you're quite articulate and translating well.  You're just spouting nonsense.
Creature of habit.

Accuracy FLAC decoder

Reply #71 –
Soap
You should not boil the way. I was asked to clarify some of the details, and I did it. I did not intend to deceive anyone.

Accuracy FLAC decoder

Reply #72 –
Quote from: Bublic on
Soap
You should not boil the way. I was asked to clarify some of the details, and I did it. I did not intend to deceive anyone.


Not accusing you of attempting to deceive anyone else here.  I'm accusing you of having deceived yourself.

It's good you clarified the details, but (avoiding idioms is hard) I am solely pointing out that you have described, through your details, illusions.  Common illusions all people are susceptible to, thus why this forum REQUIRES objective evidence such as double-blind testing when making such claims.

When people have real, actual, legitimate, problems in their digital audio reproduction their problems do not sound, at all, as you describe your problems sounding.
Creature of habit.

Accuracy FLAC decoder

Reply #73 –
Now that is clear nothing will be clarified in a satisfying way i vote for the recycler for this whole thread.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Accuracy FLAC decoder

Reply #74 –
Removing the thread would be unwise. It is possible for example to make a mark in the title, "is not confirmed". If there is any need.