Audio quality and Vista Options

[uart]

First up let me say that I'm still running Windows XP and I have no first hand experinece with Vista.

I've noticed that some people making claims that "audio" is much better in Vista, though what specifically they are referring to I don't know. The more outragous claims go something like "with Vista your old onboard sound will sound as good or better then the best PCI soundcard". Of course I think it's a load of bull, but what do the changes to audio in Vista really amount to?

[EagleScout1998]

I think that is a load of crap as well. But Vista does require significantly more computer power than XP. So it is possible that the people that are using Vista have better hardware (even onboard hardware) than the average XP user.

[benski]

The only thing that might have been improved is latency, which would only affect users of Pro Audio software like Pro Tools or Cubase. Supposedly, the new audio architecture allows for all sound cards to have low latency, instead of only sound cards with ASIO drivers (typically only expensive professional sound cards).

However, pro audio software developers aren't exactly lining up to use the new Vista APIs, as ASIO has been serving them well for almost a decade.

There's certainly nothing new that would improve sound quality except for maybe a new resampling algorithm.

[Gabriel]

Resampling is also vastly improved. A side effect is that it also allows per application volume setting.
However, many gamers are complaining because several processing tasks are now handled by the CPU instead of the sound card DSP.

[dantesan]

They could also maybe be referring to the fact that Vista does "auto-room-adjustments."

(Analyzes the room in realtime and makes adjustments for frequency bumps)

I am not sure whether this is on by default though. And, I would assume you would need a mic. (I'm not quite sure how they would do this without pink noise either) Hey, I'm just the messenger ya know.

[spockep]

They could also maybe be referring to the fact that Vista does "auto-room-adjustments."

(Analyzes the room in realtime and makes adjustments for frequency bumps)

I am not sure whether this is on by default though. And, I would assume you would need a mic. (I'm not quite sure how they would do this without pink noise either) Hey, I'm just the messenger ya know.

If this is true, how does Vista analyze the room you are in? I mean even top notch audiophile equipment doesn't do this great feat.

[tgoose]

They could also maybe be referring to the fact that Vista does "auto-room-adjustments."

(Analyzes the room in realtime and makes adjustments for frequency bumps)

I am not sure whether this is on by default though. And, I would assume you would need a mic. (I'm not quite sure how they would do this without pink noise either) Hey, I'm just the messenger ya know.

If this is true, how does Vista analyze the room you are in? I mean even top notch audiophile equipment doesn't do this great feat.

From what I've read (and I haven't used Vista since an early beta so I have no first hand experience,) this is a one-time thing, a kind of wizard to be run, which I guess uses pink noise or similar. I may be wrong on this, though.

Audiophile equipment doesn't do it because it doesn't have a microphone or any processor to do it with, and it's also probably pretty pointless for consumer equipment...

[spoon]

>Audiophile equipment doesn't do it because it doesn't have a microphone or any processor to do it with

Mid to high range denons do, they have a microphone supplied with their av amps to setup the amp to match the room.

[Gabriel]

I have a (a bit old) Denon equalizer which is bundled with a microphone in order to be able to create a flat frequency response, so as Spoon pointed, audiophile equipment can do it since years.

(yes, this is pink noise that is used)

[dantesan]

Yeah, frequency analysis of rooms is much cheaper than it used to be. A good mic, pink noise generator and an average preamp with a firewire interface can do miracles these days. Sure beats breaking out my Rane RTA (Real Time Analyzer) rack unit and doing it that way. Real Time Analyzers until recently (last 5 years) were the sole domain of pro audio gear. Now it's just plug-and-play. (I'm glad to hear it's working it's way into consumer audio receivers now) Hell, I thought it was awesome when they started incorporating speaker timings into consumer units. It's going to be an exciting future.

Anyway, from what I've read this is Microsoft's way of trying to convince audio people to dump their Macs, because you can now do "pro audio" on Windows. (Basically true. I run pro tools on a PC in my home studio) We use Macs at the larger studio I work for though. Pro Tools on both platforms is essentially identical. Of course, you don't get as many driver conflict/problems on a Mac usually.

This post has been edited by dantesan: Jun 11 2007, 00:20

[Woodinville]

They could also maybe be referring to the fact that Vista does "auto-room-adjustments."

(Analyzes the room in realtime and makes adjustments for frequency bumps)

I am not sure whether this is on by default though. And, I would assume you would need a mic. (I'm not quite sure how they would do this without pink noise either) Hey, I'm just the messenger ya know.

They use a signal somewhat like a chirp. Pink noise is spectacularly unsuited to such work.

They could also maybe be referring to the fact that Vista does "auto-room-adjustments."

(Analyzes the room in realtime and makes adjustments for frequency bumps)

I am not sure whether this is on by default though. And, I would assume you would need a mic. (I'm not quite sure how they would do this without pink noise either) Hey, I'm just the messenger ya know.

If this is true, how does Vista analyze the room you are in? I mean even top notch audiophile equipment doesn't do this great feat.

(if it's accepted) there will be an AES convention preprint on this in the fall AES.

[BigFNDeal]

They could also maybe be referring to the fact that Vista does "auto-room-adjustments."

(Analyzes the room in realtime and makes adjustments for frequency bumps)

I am not sure whether this is on by default though. And, I would assume you would need a mic. (I'm not quite sure how they would do this without pink noise either) Hey, I'm just the messenger ya know.

I will investigate this when I get home from work tonight. There has to be a way to turn it off if it exists. This would use clock cycles very unecessarily imho.

[spoon]

In Vista it is a one time setup, it does not run all the time.

[tarsier]

I've noticed that some people making claims that "audio" is much better in Vista, though what specifically they are referring to I don't know.

I can think of one area that might be of interest to the HA people. They went to a 32 bit floating point "audio engine" in Vista. So any DSP that might be done by Vista is now done using 32 bit floating point math.

Of course, who wants any extra DSP done? If you want to bypass any Vista DSP, then your audio app should use the new WaveRT mode which will let it pipe data directly to the soundcard.

[Woodinville]

I will investigate this when I get home from work tonight. There has to be a way to turn it off if it exists. This would use clock cycles very unecessarily imho.

It is off until you profile a room and turn it on. In order to do a profile you need an omni or cardioid (no close-talking mikes), or even better an instrument mike (settings in UI for both).

Once you turn it on it uses a few cycles when you're doing output. It's not enormously expensive, actually. There are filter designs, and there are filter designs ...

[Hancoque]

Resampling is also vastly improved. A side effect is that it also allows per application volume setting.

The major improvement I see in this is that you can now set a "shared mode" sample rate. This makes it possible to bypass the internal resampler of the audio hardware even when using DirectSound output. I used the famous udial.wav to compare Vista's resampling quality with that of foobar2000's PPHS resampler and I couldn't hear a difference. There was clearly no siren sound in either playback method, but only a very high pitched one like some CRT TVs produce when you stand at the right (or wrong ) spot. I was using an Edirol UA-1EX set to 48 KHz.

I can think of one area that might be of interest to the HA people. They went to a 32 bit floating point "audio engine" in Vista. So any DSP that might be done by Vista is now done using 32 bit floating point math.

Combined with the above this is really something that could make kernel streaming or ASIO output superfluous for normal listening. Just set the native sample rate of your audio adapter in the shared mode configuration screen and you're done. The big advantage is that you can still hear the sound of other applications and even change their volumes in a central mixer application.

This post has been edited by Hancoque: Jun 16 2007, 15:38

[BlueWing]

Much has been written about Audio Processing in Vista, and it is very promising. Haven't tried room correction yet, as I don't have a mic.

[wyup]

Hello,
I have just installed Vista dual boot with XP and from the first moment I got the impression that it sounds better than XP, more detailed highs and bass, and somewhat better dynamics.
I have a M-Audio Audiophile 2496 RCA cabled to a 50 watt NAD amplifier and stereo PSB monitors. I have tuned Vista for reduced memory and CPU consumption by only running essential services and software.

I know my claims come just by personal perception, and aren't supported or proved with ABX, but I have noticed a significant improvement and wanted to contribute to this post. I run updated drivers on both XP/Vista, and play music on foobar2000, WMP and MPClassic.

[sld]

more detailed highs and bass, and somewhat better dynamics.

This forum appreciates your enthusiastic voluntarism very much, but seriously, phrases like that trigger off ToS #8. People demand statistical evidence for off-hand audiophile-speak.

I know my claims come just by personal perception, and aren't supported or proved with ABX

Then surely you know that not many will take you seriously, yes?

If audio is played with the best respective settings on Vista and XP, there will be no distinguishable difference for identical files.

Effects like Creative's 24-bit Crystalizer are not counted.

This post has been edited by sld: Mar 18 2008, 19:28

[Woodinville]

I think that is a load of crap as well. But Vista does require significantly more computer power than XP. So it is possible that the people that are using Vista have better hardware (even onboard hardware) than the average XP user.

Since this still seems to be running...

In Vista, "kmixer" does not exist. System processing is done in 32 bit float, and digitized to the rendering resolution with an anti-clipping limiter at the end of the chain. EQ, loudness EQ, room correction, bass management, speaker fill, headphone virtualization are all done in float.

In addition, a high-quality resampler has been added for streaming audio (still a fast resampler for hihg-channel-count dsound, which may have to do 100 channels at once), including dsound (low channel count, i.e. 1 or 2 stereo, 5.1 or 7.1 streams) that is good well beyond 16 bits.

The downmix algorithm for channel folddown when nothing else is called for has been modified so that you will no longer have channel dropouts.

So, in fact, Vista audio quality is, as long as it stays in real time, considerably better in quality.

In addition, Vista has exclusive mode, so you can get exact audio in-audio out from the engine if you want that.

If audio is played with the best respective settings on Vista and XP, there will be no distinguishable difference for identical files.

Ok, please show YOUR evidence for that.

This post has been edited by Woodinville: Mar 18 2008, 19:31

[sld]

When the ABX test is set up to test for lack of differences, sure, I'll do it.

You sir, have flouted TOS #8. Where's the evidence for the discernibly better resampler in Vista? All the big numbers you throw around are useful for DSP work. For playback, a previous test conducted here has shown that most people's ears don't go beyond.... 13-bit?


And you don't seem to understand 100% accurate audio reproduction, at least, beyond the listener's ability to distinguish. If I claim that you will hear no difference between a FLAC, Wavpack and Vorbis q6 encode of the same musical file, do you want to call on me to perform a test too?

This post has been edited by sld: Mar 18 2008, 20:02

[krabapple]

Oh my. This could be um, interesting. <gets popcorn>

[Woodinville]

When the ABX test is set up to test for lack of differences, sure, I'll do it.

You sir, have flouted TOS #8. Where's the evidence for the discernibly better resampler in Vista? All the big numbers you throw around are useful for DSP work. For playback, a previous test conducted here has shown that most people's ears don't go beyond.... 13-bit?


And you don't seem to understand 100% accurate audio reproduction, at least, beyond the listener's ability to distinguish. If I claim that you will hear no difference between a FLAC, Wavpack and Vorbis q6 encode of the same musical file, do you want to call on me to perform a test too?

Yes, I'm calling you on all of your unsupported, false claims. You have made a testable claim without any evidence. Now produce your evidence.

You are the one insisting that there is "no distinguishable difference". Prove it. Run the ABX test yourself, and don't try to shift the burden. Prove your claim. You made it, not I.

Regarding the resampler, you can measure in the resampler in Vista yourself, as I have. No, I don't have the numbers at hand presently, I measured it rather before Vista was released. If you've any experience to speak of in PC audio, you'd know that there is a substantial problem with resampling in Kmixer, and that the presence of a high-quality resampler is the solution for that problem. This is not some "new bit of news" but something that people in the field have known for years. I'm not making any new assertion, I'm stating the common understanding, and a common understanding that I have personally tested and verified, BEFORE product release.

That would go for the bass management, loudness EQ, graphic EQ (only Vista version, please!), room correction, speaker fill, headphone virtualization, and so on, as well. And I'm quite certain that all of them run in 32 bit float for calculations, as opposed to the Kmixer version of (some of them) that ran in 16 bit signed integer, including filters, etc. You are aware of the need for computation headroom in filters, are you not?

So go do the measurements. We're not talking hypotheses, guesses, or anything beyond COLD HARD FACT here, so get with the program and go make the measurements. If I tell you numbers, you obviously won't believe a word I say. (but the difference between 120dB and 50db is pretty significant, I'd say)

Evidence is commonplace that denies your "13 bit" number. Try a 13 bit piano music recording. I don't even have to run that test (again), because I know what happens. Do it in a quiet room, though. I'll cheerfully agree that in a common office, 13 bits is sufficient, but that has nothing to do with human hearing to speak of, and everything to do with how noisy the modern world is. Some of us actually listen to music in a living room instead of a computer lab.

A 13 bit INSTANTANEOUS dynamic range is about right. (meaning difference between highest energy and lowest perceptable energy across frequency at one instant). I've said this in my own tutorials and talks. However, with recorded music, in a quiet space, played back under good conditions, its very easy to show that 16 bits is marginal, ACROSS TIME, i.e. from noise floor (inaudible or not) to peak level, and if the noise floor of the 16 bit system is at all peaky at low frequencies, i.e. not flat or high-pass-like, 16 bits is not enough in a quiet environment.

That result can trivially be seen from a look at Fletcher or Stevens' work on zero loudness curves, coupled with a knowledge of modern headphone sensitivity.

I am ignoring your last bit of uninformed ad-hominem nonsense. I would suggest that you do not repeat it again.

Oh my. This could be um, interesting. <gets popcorn>

Gimme a handfull of that popcorn, will you, I missed lunch and I'm in a bad mood.

[john33]

OK, just to throw a little more into this, I have a test system running 32 bit Vista Ultimate and, although I haven't and have no intention of ABXing this, the quality improvements in Vista so far as audio is concerned are sufficiently great that ABX is irrelevant. If you haven't tried this, your comments are pointless!

Don't get me wrong, I'm not a great advocate of Vista, although there are some things I do like about it, but in the audio reproduction area, great strides certainly have been made.

[Xenno]

I agree with SLD. Examine the playback flow with an M-Audio 2496 using MA drivers (that must bypass kmixer and route to the patchbay which can enable/disable MA's monitor mixer). You have format (ogg, mp3, etc)>codec (decompress to wav pcm?)>soundcard (DAC conversion)>amp. Yes ... gross simplification. I don't know how system sounds get into the audio stream with a 2496 setup with kmixer out of the loop. I also don't know if the 2496 or it's drivers do any resampling but I would think not (at least on the primary audio channel). So ... if I basically have it right, then how could Vista sound any better than XP, or FooBar better than WinAmp or Media Center for that matter.

No intention of testing Vista ... ever. What kind of setup do you have John?