XXHighEnd

[PeterSt]

Now, letting alone the rough changes, where will that put ABX testing ? Nobody hears a difference, but I do.

and

That’s exactly where ABX is about, it protects the individual against its own bias.

 

Although I understand your line, I don't understand it in the context (my line). If nobody hears a difference, the test is useless and besides that, the result is false (because there *is* a difference).

 

?

 

 

[tfarney]

Peter said:

 

"Try it. Take two different setups of any kind and of which you can expect differences really. Could be a tube amp vs. a SS. Now, everybody will agree that the first you might recognize as a difference will be about detail. So, you have e.g. Amp A and Amp B, and at listening to Amp B after Amp A, you're sure you hear a bell you did not hear from Amp A. Mind you, in this case it is the assumption that indeed Amp B shows the bell better than Amp A. But now try to *not* hear the bell from Amp A. Unless the differences are really rough, you can't avoid the bell anymore. You know it is there, and you will hear it.

This is just one example and I can give many others."

 

If you're listening for bells, or the quality of the reproduction of bells, the test is already invalidated. An ABX test is infinitely simpler and more useful in a case like this. It only determines if a difference can be heard, if the listeners can differentiate A from B with X as a control. If a difference is found, the quality of that difference is up to the ears and taste of audiophiles.

 

"I am fairly sure that if one doesn't hear differences, this is not about a low resolution system, but merely about not knowing what to listen for."

 

-- if the listener knows what to listen for, the test is invalidated. It is far too easy for us to hear what we expect to hear. It happens every day.

 

"I did not run into someone who could do what I accidentally learned to do, but I also did not run into anyone who couldn't do it after pointing it out."

 

-- Even more problematic if you're pointing out "what to listen for." This is the very heart of psychological bias.

 

As for the methodology of the testing through which you heard differences others did not, I apologize, I'm struggling with your English again. But from what I can understand, it doesn't look like testing, it certainly doesn't look blind. It just looks like listening.

 

I'm not trying to insult you, Peter. I'm sure you have good ears. But no one is safe from hearing what they want to hear, expect to hear, need to hear. Not me. Not you. Not the most experienced mastering engineers. Blind testing must be blind to be valid. It must include the X to have any results worth pointing to. The switching must be done by computer or by someone not participating in the test. It must NOT ask for quality judgments or discuss what is being listened for or heard. And it must be properly documented. Or it is not a test, it is just a bunch of folks listening to some equipment and talking about what they like. Nothing wrong with that. It's just not testing and it yields no valid data.

 

Tim

 

 

 

[tfarney]

"Now, letting alone the rough

 

Now, letting alone the rough changes, where will that put ABX testing ? Nobody hears a difference, but I do.

and

That’s exactly where ABX is about, it protects the individual against its own bias.

 

Although I understand your line, I don't understand it in the context (my line). If nobody hears a difference, the test is useless and besides that, the result is false (because there *is* a difference)."

 

If nobody hears a difference, the result is not false, the result is that there is no AUDIBLE difference. Very useful, I would think, in testing things that claim an audible difference that can only be measured by listening.

 

Tim

 

[PeterSt]

Yes Tim yes Tim yes Tim.

 

And again I apologize for my english.

 

Yes Tim. You're right Tim.

 

[BEEMB]

 

We've probably moved on since but I have been in touch with Amir Majidimehr who was ex vice president of the digital media division at Microsoft until he left last year. I was interested in his thoughts on the audio engines of XP and Vista and asked a few questions. As I say, some may not be totally relevant to the discussion.

 

Question 1. Amir, could you provide an insight into the audio fidelity under Vista and XP ? It’s common to bypass the K-Mixer in XP when outputting audio but my understanding is that the Vista audio engine is improved. WASAPI allows us to bypass the Vista engine. Is it required ?

 

Answer 1. Under XP, the situation is pretty bad for the most part. K-mixer (the engine which handles mixing and processing audio) has pretty poor algorithms for resampling audio and common operations such as changing volume. The former comes into play if you say, have some audio at 44Khz (CD) but your sound card only supports 48Khz. In this case, you get a lot of artifacts (e.g. the sound gets brighter and harsher).

Unfortunately, this situation is far more common than it would seem because Microsoft’s minimum requirement of a sound card was to support only one frequency and that was 48Khz!!! Since much of downloaded or ripped CD music comes in sample rate of 44.1 Khz, you were constantly experiencing the poor upconversion by K-mixer on XP.

The effect of low quality sample rate is audio getting brighter, and losing ambiance. You didn’t need to be a super audiophile to hear the K-mixer artifacts in this situation.

Volume control effects were harder to experience. One way to tell was to play some really high fidelity music with good headphones and then adjust the volume control slightly lower and higher. See if the tone changes. If it gets brighter/harsher then you are hearing the “quantization noise” from K-mixer from XP. Note that this much more subtle than the resampling problem above.

 

Under Vista the situation is far, far better. Indeed, I would not bother with exclusive mode to get better fidelity. Vista has a completely revamped audio processing pipeline. It uses high precision floating point for internal representation of audio samples (and hence, doesn’t lose accuracy as it processes audio) and has brand new, high performance algorithms for resampling and volume control/mixing. Its noise floor is 144db down from what I recall.

 

Now people feel that anything that changes audio is bad but that is not the case. Yes, Vista does add “dither” to audio samples as it processes them. People look up definition of “dither” and see that it is “noise” and immediately get worried that something bad has happened to their audio. But dither is your friend in case of digital processing. It helps to remove distortion and convert it to benign noise and is a necessary component of signal processing. Any time you change the volume control for example, you MUST add dither or you get distortion. Ditto if you use speakers and have base processing to send the lows to your sub. So you have never experienced “bit exact” processing if you are using a home theater processor or AVR. Not using dither and passing the bits through would have been the worry here not the other way around! Look up dither in Wikipedia for some visual demonstrations of how helpful it is in processing digital samples.

So being bit-exact simply is not possible nor good. Bit-exact was important in XP to get past the nasty processing there. But with that part fixed, other than feeling good you are not getting anything out of being bit-exact.

 

Question 2. Having read through some of the thread over at AVS Forums it’s clear a lot of engineering has gone into Vista’s audio engine. But, given the above, what is the point of WASAPI / Exclusive mode if it does not enhance your audio.?

 

Answer 2. There are other reasons for having an exclusive API. The main reason is for music recording applications where people want to have lower latency (delay) than the standard audio pipeline allows. This is where ASIO originally came from. Another reason is if an application handles all processing and doesn’t want the OS to do more.

 

Question 3. In trying to achieve the best audio playback I can from my PC I’ve been comparing the output between different software. I was using iTunes streaming to an Airport Express, iTunes converts my uncompressed audio to ALAC on the fly – so in my opinion Media Monkey, using Vista’s own engine, ASIO and waveout sound different - better than iTunes via Airport Express. There is an audio player called XXHighEnd which utilizes WASAPI - and I believe the developer feels that ASIO is not worth bothering with on Vista. I’ve been comparing XXHighEnd to Media Monkey; it’s a very minimal program that loads your audio into RAM before playback with hardly any load on the rest of your system. I believe it has a more laid back sound and is just so easy to listen to – I honestly believe that I am not just hearing what I wish to hear. So could I ask your opinion on the following.. Can other applications impact your audio ? For example, processor instructions from other software can interrupt the playback and to what extent? Can audio quality be improved with less load on your system?

 

Answer 3. Windows is not a real-time system (nor is MacOS, Linux, etc). This means that at the extreme, the system can fall behind the sound card causing it to starve for data. Result is an audio glitch or pop or long delay where nothing plays.

One of the main reasons for audio glitches is running out of memory. When you are short of RAM, the operating system will suspend applications and attempt to move things around in memory and during which time, you may have a buffer underrun in the sound card and cause these artifacts. So yes, having too many applications active at once can cause problems (having them just sit there is not as bad as the operating system will probably page them out once and then leave them alone).

Another way you can get glitches of this sort if starting large programs which may block disk access needed to read the next chunk of data. Let’s say you decided to defragment your disc while playing music in your favorite music player. The extreme disk traffic may cause the player to be blocked while it is trying to read your music from the hard disk. Result is once again the same with audio being interrupted. In this case, having a dedicated drive for audio can eliminate conflicts.

So yes, having a dedicated system with plenty of RAM can be helpful in making sure the system plays everything fine. Note that in this context, we are just talking about something working or not. We are NOT talking about its fidelity. See more in #4 below.

 

Question 4. If your audio can be interrupted – can this introduce/cause jitter – or can that not happen at the data end ?

 

Answer 4. Per above, audio interruption has nothing to do with jitter or fidelity. The problems described only affect reliability of the system not quality.

Quality is however impacted by many other things, some of which are very hard to analyze. Jitter is a good example where audio samples are played earlier or later than they should. Sound cards have a clock that determines when an audio sample is output. That clock timing can be impacted by many things form power supply noise to what program is running on the CPU! You could envision that just moving your mouse may change the system behavior and cause jitter.

How much degradation occurs is very hard to say and depends on many factors beyond the PC itself. Most people are not capable of hearing modest or even extreme amount of jitter distortion. Can you hear differences between audio cables? If not, then chances are you won’t notice dither distortion either (not that the distortion is the same but they fall in the same marginal category). Even dedicated high-end systems can suffer from jitter distortion.

Unfortunately there is no way to really determine if you are suffering from jitter unless you have a reference that doesn’t. While there is equipment which measures jitter, impacting it in a PC can be very difficult..

 

In general though, anytime you can reduce the variations in the system, you can potentially improve the fidelity. The ear is far more sensitive to variations in fidelity than absolute levels. If there is a glitch that occurs once every five seconds, you will hear it much more than if the same glitch occurred 10000 times per second. So simplify your system. Get it to have a predictable load and activity when it plays music. Use best quality power supplies and components. And you are off on the path to reduce the last bit of distortion in your PC!

 

 

 

 

M.

 

[PeterSt]

Hey Matt, thanks !

A bit of a pitty that you asked Q3 and Q4 wrongly (it is not about interruptions !) but luckily Amir came back on the matter by himself.

 

If I may say so, his outlay on Vista's processing and dither opposed to WASAPI and the dither stuff ... b.s.

He takes for granted that

 

We resample, must do that because we want to hear our email coming during music playback, and use digital volumes.

 

Leave those out of the equation, and he just adds dither which can't be avoided.

Besides that, he resamples ALWAYS. From 44.1 to 96 and back when you play 44.1.

 

He did not change since that AVS thread ...

Did it occur to you that he just does not want to go the WASAPI direction ? OF COURSE NOT, HOW TO HEAR YOUR EMAIL COMING IN THEN ?

 

hahaha

Peter

 

 

[BEEMB]

 

Yeah, as I say I asked some of the questions prior to further reading on this forum and learning more myself.

 

However, I thought it could be useful to hear from somene involved in the programming side and I've asked him to come along and consider joining the forum.

 

I think comments outside of his thoughts on WASAPI are interesting though and mirror some comments made earlier within this thread. He lists a number of items that could affect audio but also questions whether you can hear or not.

 

That angle is kind of similar to Roseval's website; all these items may affect playback but only you can decide whether you hear them or not.

 

And that's just the thing; I'm totally convinced I have something better in XXHighEnd - that was the reason I created the thread. I wanted others to try it and report their findings. It generated a lot of discussion.

 

At the end of the day, only your ears can decide for you.

And I know there'll be one or two people saying that listening with your ears is wrong. But that's what I'm going to do. And I believe XX sounds better than iTunes converting my audio to ALAC and streaming it to my Airport Express, that I'm sure of.

 

Matt.

 

[BEEMB]

 

Peter,

 

Right now I'm listening to Lazy Eye by the Silversun Pickups... I've just compared Mediamonkey to XX.

XX Wins.

 

But, with Mediamonkey i use ASIO ... XX is using WASAPI. Am I simply hearing the difference between ASIO and WASAPI ??

 

[Ashley James]

I would caution against any comparisons unless the levels are identical. This means playing a 1 kHz test tone through each player being tested and measuring the output with a DVM. Switching between each will need to be instantaneous too.

 

The reasons are that even if it sounds worse, people always prefer the loudest of a group. Equally you have no actual audio memory, only a tiny percentage of people have perfect pitch for example, so unless switching is instantaneous, you have no chance of remembering what you were listening to, so proper comparison is impossible. Gross differences yes, but caution is needed.

 

This is all well documented and understood and I've been humiliated by being the subject of experiments to prove this.

 

Ash

 

[BEEMB]

 

I'll retest to see if there are any noticeable change in the volume.

 

Right now I'd best get ready to head off out....

 

 

[PeterSt]

I am so sorry, because I guess it is my fault that you got literal on "interrupts" because I said this earlier in the thread :

 

also not about preserving the capacity, because indeed the capacity is far over sufficient (XX uses virtually no CPU at even 24/352.8). It is merely about interrupts (now think of the hdd access again) and how the CPU deals with its cache and copying of 2L cache, and more.

 

... and I recall you wanted to hear about "interrupts" ...

 

But I was referring to system interrupts here, which is just computer language I'm afraid;

When a task is ready with its job (think of a hdd which got what it had to get), it interrupts the processor by saying "hey, I'm ready here !". This, by no means, implies that audio playback is interrupted (glitches, ticks etc.) unless the setup is wrong (most often wrong drivers). Only when the load gets too high this happens in normal-setup-circumstances.

 

Again, so sorry.

Peter

 

 

 

[PeterSt]

But, with Mediamonkey i use ASIO ... XX is using WASAPI. Am I simply hearing the difference between ASIO and WASAPI ??

 

Try Foobar/WASAPI. Then you'll know.

Or Foobar/ASIO.

 

To my ears ASIO has a certain signature, and I don't like it much.

Comparing with ASIO by itself is dangerous because quite a few implementations of it exist. But I still hear the common denominator of it.

 

[PeterSt]

True. But there is no way levels *can* be different at bit perfect playback.

 

Levels will be perceived diffrently though because of e.g. harshness vs. no harshness. But that is another matter, and not to be equalized.

 

Peter

 

[Roseval]

But, with Mediamonkey i use ASIO ... XX is using WASAPI. Am I simply hearing the difference between ASIO and WASAPI ??

 

Drivers can have a sound influence.

Try MM+WASAPI

A nice candidate might be Foobar

You can compare it with XX using WASAPI and a large buffer (memory play)

 

[BobH]

Good effort, sir! And thanks to Amir for providing the answers.

 

I can actually see the point now, for Peter, in using WASAPI when you also consider his declared preference for NOS Dacs. There's little point in running a NOS Dac if the source has been up/over/down/back/sideways/ sampled en route! But that also means that great care has to be taken not to introduce additional noise, as there isn't going to be an opportunity to either remove and/or disguise it. Hence the obsession with the smallest of details. (Am I close, Peter?)

 

Most Dacs now employ oversampling as part of their audio chain. As was stated in the reply, if most people are not going to be able to hear the results of the processing that goes on in the PC, and it has been included, with good technical reasons as to why, in order to enhance their listening experience, then there seems little point in advising them to avoid it. Unless, of course, you can provide very good, equally technical, reasons why they should want to. In this scenario I think that an explanation of context and a few facts and figures is warranted.

 

A Vista PC running the user's media player of choice running into the dac of their choice - job done. Sounds good to me!

 

We all know the deserved reputation of XP's kmixer/volume double act and therefore why we would want to avoid them. If Vista's audio engine is as good as Amir says it is then the case for avoiding it is much less clear.

 

IMHO.

 

ps - The above statement in no way implies that I shall stop tweaking, btw. I does it 'cos I likes it!! :)

 

[mpmct]

 

FWIW ...

 

http://www.moultonlabs.com/more/little_bit_more_about_dither_and_bits/

 

 

And this is just for fun, but it might be worth considering ...

 

http://www.moultonlabs.com/more/yeah_i_think_i_can_hear_it/

 

[glt]

Great links, thanks.

 

[PeterSt]

Bob,

 

At reading Amir's outlays on why Vista is better, this should be placed in the proper contexts. And I don't think they were, as I don't think Amir is as objective as he could be.

 

1. In XP it never has been much of a problem to avoid KMixer. So why compare things with KMixer in the chain ?

 

2. In Vista "KMixer" (called Audio Engine there) cannot be avoided. So better emphasize how good it is !

 

3. In Vista "Kmixer" is no good at all, because it is in the chain most often without any reason. And worse as with XP. In XP there's only unnecessary dither (nothing is resampled etc. but the LSB is dithered) while in Vista there's also unnecessary resampling (only when the input to the Audio Engine is 24/96 nothing happens, as per experience by many (I never checked that myself, but I believe it)).

 

Now, IMO it would be wrong to think that where Vista resamples so much better than XP, it is a good thing to do. It never is.

But there is more :

 

Where Amir highly depends on his internal 48 bit float whatever processing, he passes the ever so important part of things having to go back to 16 bits when your soundcard/DAC is. And at going back from his 48 bits to 16, you loose as much as everything. The result is as worse as it was in XP, although in XP it might have been worse because of the resampling itself (per many process steps) which also seem to have worked at the 16 bit level.

I am very clear about this : he doesn't understand, or prefers to have a second agenda.

 

Please note that when things really come down to where they should be (and it all has been in that long AVS thread, indeed me pestering Amir quite a lot), the only answer left for him is : go Vista/WASAPI Exclusive mode.

And by itself he is right.

 

What he was wrong about though (at that time, but to a large extend still) is that WASAPI was (and merely is) an unfinished product with lots of flaws and in the in fact undoable task of creating a player in that environment. Man, I was the first, and what struggles I have been through.

At this moment you can see it by looking at the other WASAPI players, and non of them working as flawlessly as XX does, but this is just because I was ahead one year. Translate this into a year of additional struggle if you want. And oh, even XX is not flawless at this time, because of the lacking 24 bit USB support I can't seem to work around.

 

In the end, I only wanted to make clear you shouldn't say this :

 

A Vista PC running the user's media player of choice running into the dac of their choice - job done. Sounds good to me!

 

because it won't be good at all.

 

My conclusion :

In XP there is no problem because you can avoid KMixer easily;

In Vista KMixer can't be avoided, and you suffer from Amir's ideas which are not good.

The solution would be WASAPI or ASIO, and since the latter also works in XP, why go Vista.

 

One thing has really changed :

 

With WASAPI one can create a sound engine right from the very bottom, which is what I did. Note that this is already not the case with ASIO, which always makes use of a standard "base program", similar to Kernel Streaming btw. And :

While Amir tells that Windows is not a real time environment, the latency of WASAPI is so good, that you might just as well approach all like being real time. For those who understand : I play 24/192 with 48 samples latency.

 

There's little point in running a NOS Dac if the source has been up/over/down/back/sideways/ sampled en route! But that also means that great care has to be taken not to introduce additional noise, as there isn't going to be an opportunity to either remove and/or disguise it. Hence the obsession with the smallest of details. (Am I close, Peter?)

 

You are so right Bob. Now add to this the link about dither from mpmct. As Amir, that link assumes digital volume to be applied, or anyway that digital volume impeeds for dither. Man, what are they wrong; when digital volume is applied with some knowledge and intelligence (I don't say the latter because of IQ claims, but because it is a tough job really), no dither is needed. So, XXHighEnd does not apply dither but does allow (its own only !) digital volume. Now guess what sounds better ...

I only want to say, one can have huge stories about the best implementations, but leave such baddy out is always better. And note that the digital volume of XX is not there to have a digital volume, but to allow the preamp to be out, which is another baddy (with proper impedance matching DACs).

 

So in the end it is all over the place (and was I finished talking :-), but it may take mpmct's second link to get it and perceive it.

For those who rather don't go that indeed tyering route, better convert to MP3 because it saves space !

 

Peter

 

[tfarney]

"A Vista PC running the user's media player of choice running into the dac of their choice - job done."

 

I think that's very good news. It means that now, either PC or Mac users have a straight and simple path to excellent computer audio. That should help bring more music lovers into the new world of computer hifi instead of frightening them with the absurd expenses and/or daunting complications of tiny, subjective (if audible) tweaks stated in the usual super-sized audiophile hyperbole, and guaranteed to send the curious running back to their cdps.

 

Sometimes technology shatters ceilings and destroys old notions of price/performance. Some things really are easy and inexpensive, even very good things, even excellent things. Audio solutions don't always have to have all the attraction, joy and practicality beaten right out of them. They don't always have to be tweaked until they are exclusively expensive or absurdly complex. Really. Sometimes they actually can be understandable to, and within the reach of many, and really be as good as it gets.

 

This is one of those times. Take your OCD meds and try to enjoy it, guys.

 

Tim

 

[DavidJPettifor]

tfarney wrote:

"A Vista PC running the user's media player of choice running into the dac of their choice - job done."

 

I think that's very good news. It means that now, either PC or Mac users have a straight and simple path to excellent computer audio. That should help bring more music lovers into the new world of computer hifi instead of frightening them with the absurd expenses and/or daunting complications of tiny, subjective (if audible) tweaks stated in the usual super-sized audiophile hyperbole, and guaranteed to send the curious running back to their cdps.

 

Sometimes technology shatters ceilings and destroys old notions of price/performance. Some things really are easy and inexpensive, even very good things, even excellent things. Audio solutions don't always have to have all the attraction, joy and practicality beaten right out of them. They don't always have to be tweaked until they are exclusively expensive or absurdly complex. Really. Sometimes they actually can be understandable to, and within the reach of many, and really be as good as it gets.

 

This is one of those times. Take your OCD meds and try to enjoy it, guys.

 

Hooray! I think we're there.

 

--

djp

 

[tfarney]

...anyone takes offense at the OCD remark, have a couple of spares:

 

:)

 

Tim

 

[BEEMB]

 

Excellent ... now you can all move to the "Move Me" thread and talk about your favourite, most moving music. ;-)

 

(please - I like listening to other people's suggestions).

 

[audiozorro]

I would caution about making absolutes. What was stated is often true, that

 

“I would caution against any comparisons unless the levels are identical. This means playing a 1 kHz test tone through each player being tested and measuring the output with a DVM. Switching between each will need to be instantaneous too.

The reasons are that even if it sounds worse, people always prefer the loudest of a group. This is all well documented and understood.”

 

But it is not always true. A perfect example is the many passive preamps or TVCs that have a +6dB gain switch. Many, if not the majority of listeners find superior sonics with the gain switch off and the gain is primarily there for those who need or want more gain.

 

Despite the fact that the sound pressure meters and voltmeters will document the obvious, that the +6dB gain is louder, it is often not the preference.

 

 

 

[Ashley James]

I'm afraid Audiozorro that it is true and in the case of matching levels, it is certainly true because we're talking about fractions of a dB. Not 6dB! Which amounts to four times the power requirement!

 

The performance of a passive Preamp depends on the load that it sees from the amplifier it is connected to. Most modern amplifiers comply with EMC regulations and often have filters on the input that turn it into a tone control, therefore I wouldn't use one for hi fi. It's also possible that increasing gain by as much as 6dB could cause some clipping on transients and audible distortion.

 

To clarify. If you are comparing different audio components, the only safe way to do it is to match all the levels to within a fraction of a dB using a Digital Volt Meter and 1 kHz test tone. Switching between each should be instantaneous because you don't have a memory of what you've heard. This is very important for tests where differences are small because lots of experiments have been done that show a difference of 1/10dB can be enough to sway the test in favour of something that is less good.

 

Obviously where big and obvious differences are being compared, there isn't as much need for caution.

 

Worth reading: http://www.zaphaudio.com/evaluation.html

 

[PeterSt]

FYI : The 6dB switch of a TVC changes the frequency response of it. The higher frequencies will be attenuated.

 

Peter