Amarra

[machinehead]

Is there a way to play a file with a player like iTunes and capture the data output, say from the USB port and

measure the exact size or check sum? I guess the size could be the same and have bits changed, but a check sum should

verify the file is exactly the same. Is this process possible?

 

Are there such tools that can measure a digital data stream and analyze it?

 

This is the only way I can think, that you could verify that bits have not been changed.

 

[audiozorro]

Thanks cfmsp for giving more insight into soundBlade. I have heard of iZotope before and it seems to rate very highly among those who seem to know.

 

I haven’t heard of Wave Editor before, but since they offer free demo software, I would be remiss if I didn’t try it out.

 

 

[weiss2496]

Ok - shame on me, sorry Apple... I had the iTunes built in EQ switched on. If that EQ is switched off the spectrum looks rather more like a 24 bit bit transparent playback. (Whether it is truly transparent would have to be checked)

 

You can check the picture here:

http://www.weiss-highend.com/amarra/ITUNES50.jpg

 

And with a 20 dB attenuation (iTunes fader) here:

http://www.weiss-highend.com/amarra/ITUNES51.jpg

This shows that the fader is not dithered, BTW, as there are quantization distortion components creeping up.

 

Don't use that EQ unless you do not care and/or like its sound...

 

Regards,

Daniel

 

 

[HarryHWombat]

... for the mea culpa! Have you tested iTunes with the output set to 16/44 against 24/44? I know I bang on about this but I found something on the Lavry site that states explicitly that 16/44 media will not be bit perfect unless midi output is set to 24/44.

 

I'll post a link when I get back home.

 

[MahlerFreak]

Ah, that's better

 

Have you tried measuring the jitter spectrum from both? My guess is, they'd be essentially identical through a Firewire interface like yours?

 

 

 

 

[weiss2496]

Have you tested iTunes with the output set to 16/44 against 24/44?

 

No, with our Firewire interface it is not possible to set the wordlength. It is at 32 bits for transfer via Firewire independent of the actual wordlength of the file.

 

Daniel

 

 

[weiss2496]

Have you tried measuring the jitter spectrum from both?

 

No, did not. Do you mean in case there is a async SRC going on?

I measure the digital output, hence jitter is irrelevant (provided there is no async SRC).

 

For our Firewire DAC the software used to play back does not play any role when it comes to jitter (again provided there is no async SRC in the path) because the DAC is the master clock and the computer is slaved to it.

 

Daniel

 

 

[PeterSt]

That clears up things.

 

Daniel, before you're accused of presenting misinformation : I see you nicely adjusted the before PDF, but you forgot to adjust the text(s). There is no broader skirt anymore. The pictures are, say, the same now.

Even the title of that PDF may not be right anymore. But that's up to you (or the Sonic guys).

:-)

 

Keep up the good work,

Peter

 

[PeterSt]

Don't use that EQ unless you do not care and/or like its sound...

 

or if you like broad skirts. haha

 

[bite-the-byte]

....can you kindly share your technical point on this statement you provided:

 

"For our Firewire DAC the software used to play back does not play any role when it comes to jitter (again provided there is no async SRC in the path) because the DAC is the master clock and the computer is slaved to it."

 

I'm also thinking (but I'm less skilled than you) can't exist jitter due to SW routines of a player like iTunes, Jitter resides in the transmission. I hear to many comments about hipotectical more or less jitter thanks to one sw player vs. another., which I tend to think is a nonsense. Please Daniel shine your light over us....

 

Paolo

 

P.S.

Don't loose much time on the forum, BTW, as I ordered one new DAC1mkII and still waiting you to build it :-)

 

 

 

[HarryHWombat]

Over on the Lavry forum - hope posting the link is ok.

 

There is also an aspect of the conversion from the internal "32 bit floating-point" math to of Core Audio to the "fixed point" output that means if you do not set the output word length to "24 bit," a 16 bit (CD) file will not be reproduced with "bit accuracy." This means that the lower level bits of the signal may be changed, so the quality may suffer as a result.

 

http://www.lavryengineering.com/lavry_forum/viewtopic.php?f=1&t=690

 

@Daniel

 

No, with our Firewire interface it is not possible to set the wordlength. It is at 32 bits for transfer via Firewire independent of the actual wordlength of the file.

- Daniel

 

So with Firewire does one bypass CoreAudio in OS X or does Core Audio "upsample" to 32 bits which one would hope it would just do by padding with zeros.

 

[Mr.C]

Now, unless my vision is messed up, the graphs of Amarra and iTunes are nigh identical making the difference people are hearing either from a different source than these graphs show, or a placebo effect. I guess my question is, what else could it be?

 

[The Computer Audiophile]

Hi Guys - I've been following this thread closely and enjoying every bit of it. It's really cool to have people from all over collaborating on this and trying to get to the bottom of something.

 

Peter - Let me address some things in your post.

 

"In the end it looks like not many people are interested in iTunes being bit perfect or not on the Mac. Try Google. Nothing but posts from CA ..."

 

There are actually a ton of people not only interested in the Mac being bit perfect, but depending on the Mac being bit perfect. The fact that all the posts are from CA hopefully says more about CA than a possible lack of interest from people. CA is the place to be for this stuff :~)

 

"I don't want to rant, and sure don't want to offend, but possibly Chris doesn't perform the test reliably. I recall one time at least that Chris reported something to be bit perfect (it could have been SongBird) while in a next post it was proved that it always upsamples, or whatever it was."

 

No offense taken at all. I'm constantly looking for better ways of doing things and I would hate to spread misinformation to the world. That wold hurt everyone. The Songbird example you mention is somewhat misleading. Songbird is bit perfect to the best of my knowledge on a Mac. In fact I tested it two weeks ago from 16/44.1 to 24/192. That said, my test for bit perfection is and has always been the HDCD indicator on my Alpha DAC. This is certainly not infallible but I think it's a pretty reliable method of determining bit perfection as we currently know it (the whole definition of bit perfection may seem objective and solid but the last few weeks have brought up some interesting conversations offline). In my review of the Berkeley Audio Design Alpha DAC I explained my reasoning behind using the HDCD indicator as a test for bit perfection. Here is what I wrote:

 

"If the HDCD indicator illuminates on the Alpha DAC, the data is uncorrupted. Theoretically, it is possible to alter HDCD data using specialized software while not touching the LSB, but all of the typical mechanisms that might alter data in a computer environment such as level shifting, dither, SRC’s, etc. will definitely affect the LSB. And, if the LSB is altered the HDCD code is lost. So, as a practical tool, presence of the HDCD light indicates no alteration of the data file."

 

Again, this is not a bullet proof method of testing, but I have yet to see it invalidated. It would be great if those who don't agree with this testing method could provide information about its inaccuracy so we could move on to a better method. I'm all for improving computer based playback and the better tests we have the better things will get.

 

 

 

[Mr.C]

From what Peter said before, Vista automatically switches to a bit perfect mode when it detects an HDCD signal but switches back for the other formats. This makes the test tricky because it is bit perfect and then switches when the test is over. The question is whether this happens with Mac OSX and core audio as well.

 

[The Computer Audiophile]

Hi HarryHWombat - I just tested a 16/44.1 HDCD track played back with iTunes on my Mac with AudioMidi set at 16/44.1. The results were bit perfect as far as I can tell. The HDCD indicator did illuminated. See my post above for my reasoning behind using the HDCD method as a bit perfect test.

 

[MahlerFreak]

Paolo, there was a thread or article on this site from Amir M, the architect of the Vista audio subsystem, that addressed this issue among others. I can't find it now, it may have been deleted?

 

In any case, the gist of Amir's comments, and those of Peter and others, is that jitter is a phenomenon at least partly related to digital clock instability, or EM noise inside the computing environment, or both. And it is certainly theoretically possible for different software programs which have different processor load patterns, memory access patterns, etc. to create different EM noise spectra inside the computing environment, and/or alter the stability of the clock by feeding noise back through the power supply line. The actual cause/effect relationships would be very complicated, however, and not simple to design around. Peter can maybe comment more on this.

 

The more the master clock is isolated from PS noise, and EM radiation, the less likely any SW changes could impact clocking of the bitstream. A device communicating via Firewire, or other asynchronous protocols, has its own master clock that is in no way tied to the computing environment, hence Daniel's comment.

 

[PeterSt]

Paolo, there was a thread or article on this site from Amir M, the architect of the Vista audio subsystem, that addressed this issue among others. I can't find it now, it may have been deleted?

 

It was a quoted email conversation, IIRC performed/posted by BEEMB (Matt).

 

[PeterSt]

Chris, cool. I for one am the most satisfied how this turned out. Hard to believe maybe, but I am.

 

Now, what's next ?

 

Edit (PS) : You might try the HDCD light on Vista at using Direct Sound (from Foobar, whatever), and check whether that works. I suspect it just does not, because I actually never heard of HDCD in relation to this. DTS for sure turns all in bit perfect mode.

No hurries, but it would be good to know (and I don't have these "lights", otherwise I'd do it myself of course).

 

Peter

 

[The Computer Audiophile]

Found the thread containing Amir's comments. I just did a search for Amir and it was the only thread.

 

 

http://www.computeraudiophile.com/content/XXHighEnd#comment-13673

 

[weiss2496]

Jitter plays a role when digital data gets converted to analog or vice versa, i.e. in the A/D or D/A converters. Or in a asynchronous Sampling Rate Converter where the filter coefficients are calculated on the fly, depending on the ratio of the input to the output sampling rate, i.e. jitter has an influence there as it can change that ratio. These are the only three instances where jitter plays a role in digital audio quality.

So in a scenario where there is a digital audio source and a D/A converter connected to it, then, assuming the D/A converter clock is (virtually) jitter insensitive, it does not matter at all what kind of source that is. Can be CD, hard disk based, satellite transmission, DAB, solid state memory, DVD, PC based, MAC based, OSX, Windows - you name it. As long as the bits are read off the memory w/o errors it simply does not matter where the bits have been stored or at what rate fluctuation they are transmitted.

The point is to use a D/A converter which is built to be jitter insensitive. If that is the case and the bits fed to that D/A converter are not altered in any way, there simply is no sonic difference between all those sources.

 

So with Firewire does one bypass CoreAudio in OS X or does Core Audio "upsample" to 32 bits which one would hope it would just do by padding with zeros.

That I have to check with our Firewire driver. I would expect that the bits are padded with zeros. I will do a bit transparency test with our software / hardware.

Also plans are to do a firmware upgrade to our Firewire enabled units which allows to check for bit transparency by playing specific audio files which contain bit patterns which can be recognized and verified by the firmware.

 

Daniel

 

 

[HarryHWombat]

Excellent Daniel thx and also thx Chris for checking the 24/44 against 16/44 thing. TBH though I don't trust the method because I do not fully understand it I shall look into the method further .. it's good to learn!

 

[The Computer Audiophile]

"Also plans are to do a firmware upgrade to our Firewire enabled units which allows to check for bit transparency by playing specific audio files which contain bit patterns which can be recognized and verified by the firmware."

 

Very cool idea Daniel.

 

[machinehead]

Mr. Weiss,

 

Thanks for taking the time to post on our forum, I feel as though we are very privileged to have such consultation from a man of your knowledge and standing in the industry. Also thanks for entertaining our every question down to the minute detail, this is truly a great learning experience.

 

 

Jeff

 

[MahlerFreak]

I can personally attest that if you are a Weiss customer, you get this same level of responsiveness on your issues from Daniel Weiss personally. Best customer service I've ever had for any audio product, and I've had quite a few

 

 

 

 

[bite-the-byte]

Daniel, that's clarify finally, I missed a competent comment on that jitter speculation, and now I got it.

 

BTW, just finished over one hour of comparison between osx iTunes and Soundblade. You know my disappointing to find Peak LE clearly worse than iTunes? Well now I can tell SB is definetely better than iTunes. No small difference, the sound presentation changes a lot. Music flows with much more sense of breath, and "analog" like natural density, definetely involves more, image is rock solid fix for all instruments and voices, putting them in more clear 3D stage. Some edgeness in attack and decay for percussive sound (piano, drums etc..) loose the artificial hyper edging and gets much closer to real live event. Voices, sax and trumphet solos are so organic and breathing I felt emotion running on my skin... So if Amarra is that, it brings a real value. To join the ranking group: if iTunes 6 to 7, than SB 9

All my test at 16/44.1 via firewire.

 

Paolo