Ask AudioStream: Down Sample Dilema

Hello Michael,

I'm very enthusiastic to see a site dedicated to computer audio. One thing would help immensely is to see an article clearly explaining what happens when file bit rate doesn't match either USB or Dac bitrate.

Take a file at 24/192hz, via USB 1.0 into a 96hz Dac? What happens? When are files upsampled? When are files downsampled? What is the file bitrate? What is the USB bitrate? What is the coming Dac bitrate? And what is the Dac output bitrate?

I've seen Dacs that say they take a 192hz input, but output only 96hz. Or upscale all input to 192. How does that effect sound? What happens at that point? Should we care? How do I play a downloaded 24/192 HDTracks file at full native resolution?

It's interesting that there don't seem to be any devices that clearly say, here is original file rate, here is the USB or SPDIF input rate, here is the native output and the upsampled output.

It's very time consuming, like guesswork, to assemble the whole chain to ensure 24/192 file, to high speed 24/192 USB 2.0 to 24/192 Dac input and finally 24/192 Dac output, AND know that that chain is actually being used in it's entirety without being downsampled at some point.

Finally, How do I know that what I'm listening to is native bitrate or upsampled/downsampled at some stage in the process?

Thank you for listening, and congratulations on the new site.

JX



Hi JX,

Great question!

I think the answer to your question is ultimately dependent on a bunch of specifics - which DAC, media player, platform...For example, running Pure Music with the Ayre QB-9, we can see the original file bit/sample rate in iTunes, the bit/sample rate that Pure Music is using to process the file, and on the face of the QB-9 the playback sample rate is displayed on the front panel. Obviously I've selected the easiest example ;-)

But I'd say for many cases this information is in the worst case not available or buried in software settings which are again dependent on platform (Mac/PC), media player settings and the DAC and the only way to know what's actually happening is to examine the output of the DAC. One point worth noting - Class 1 USB Audio is only capable of 24/96.

To complicate matters, as I'm sure you know some devices offer multiple digital inputs at varying bit/sample rates. Typically S/PDIF at 24/192, USB at 24/96 and wireless at best at 24/96. So the answer is not even device-specific rather input-specific. And again exactly how a file is processed by the OS is dependent on the OS version and how we've got it configured as well as what media player software we're using and how we've got it configured. And how a DAC handles incoming data is again not only device-specific but dependent on the media player and OS configuration.

I will give this more thought in terms of trying to come up with some generic rules but my initial take is the answer is too dependent on specifics. We can certainly talk about optimal generic setups for the various Operating Systems (there are plenty out there already and all of them suggest keeping the OS' hands off processing as much as possible), media player setup/optimization (most of the paid players have informative guides and offer upsample and NOS options) and finally the DAC of choice should provide the relevant information in terms of how it processes data and if it doesn't I'd steer clear.

In terms of upsampling and ultimate sound quality, I do not believe there are any hard-fast rules here either. Some systems may very well sound better playing files back in their native format while others may not.

Thanks for the thought-provoking question!

Michael Lavorgna
Editor, AudioStream



Thanks for the reply. I understand what you are saying. One last question I have is: Is there any point where loss of information occurs, and is there any point or condition one should avoid? For example, I understand it is matter of preference whether one likes an upsampling to 192 for example. But if you have a file at 192, and it's downsampled to 96, is that actual information loss?

btw, the pics from RMAF were great. It was interesting so see different combos of equipment that people were using and what gear is coming.

cheers,

jx



Hey JX,

Yes, when downsampling its safe to say that there is a loss of information and a resulting loss in sound quality. And in general, it's best to handle downsampling in your media player of choice or through file conversion otherwise your computer/OS will down sample to match your DAC's highest bit rate.

Thanks on the RMAF pics. I was of two minds on the room pics - show computer audio gear in context, or just focus on the computer audio gear. I think I'll do both for CES ;-)

Cheers and thanks for Asking AudioStream



If you have an Ask AudioStream question, just send me an email!

COMMENTS
Vincent Kars's picture

In case of USB the answer is simple.

A lot of USB DACs have a very simple receiver limited to 16/48.

The USB audio class 1 standard allows for 2 channel 24 bit / 96 kHz max.

This standard is implemented in OSX, Linux and Win.

From mid-2010 USB audio class 2 become available as a native driver in OSX and Linux

As it uses USB in High Speed mode you can do a 30 channel recording if needed.

All popular sample rates are supported in 24 bits up to 384 kHz.

For Win there are third party drivers.

You need of course a class 2 compliant DAC to use this.

A bit more on USB audio: http://thewelltemperedcomputer.com/KB/USB.html

 

By design down sampling removes information.

A 24 bit recording has a dynamic range of 144 dB, a 16 bit has 96 dB.

By down-sampling to 16 everything below 96 is lost.

The compensate for the artifacts, the 16 bit result must be dithered.

 

By design the highest possible frequency (the Nyquist) in PCM audio is half the sample rate.

If you go from 192 to 96 your Nyquist drops from 96 to 48 kHz.

In general up or down-sampling by a non natural number e.g. from 192 to 88 is even more complex as one has to interpolate.

Resampling software differs substantial in performance. Some create audible artifacts.

A bit more: http://thewelltemperedcomputer.com/KB/SRC.htm

Have a look at http://src.infinitewave.ca/ they compare a lot of re-samplers.

My personal preference is to play audio (within the limitations of the hardware) at its native sample rate.

This can be done by choosing the right media player.

Players like JRiver, MusicBee, Foobar allows you to choose drivers bypassing the Win mixer.

You get  a straight connection between the media player and the audio device by using drivers like WASAPI, ASIO, KS etc.

Michael Lavorgna's picture

And I appreciate your detailed response and your time spent sharing it.

For anyone reading this, you should also be reading Vincent's site - http://thewelltemperedcomputer.com

Cheers

deckeda's picture

Here's a page that purports to show what Channel D says Pure Vinyl/Pure Music can achieve with its conversions done in 64-bit. Don't ask me how that compares to the above.

http://www.channld.com/pure-vinyl_src.html

Channel D's Rob Robinson's stance, as I understand it, is that there is usable information on some LPs that a 96kHz sample rate cannot capture when digitizing. There's a video of him online, demoing at the 2010 RMAF, analyzing an Elvis LP with one of his software tools.

The finger snaps on "Fever" were spiking way up high past where a 96KHz sample rate could muster.

The philosophy, again as I understand it, is to:

A) fully capture the source first (as represented by 192kHz etc ...)

B) present it in a more practical way, if necessary, that preserves as much as possible. Meaning, if you captured at 96kHz you'd leave stuff on the table that would at least be "available" on a properly downsampled copy.

I don't claim to fully understand how throwing data away "cleanly" is any better than never having it to begin with, however part of a push for capturing higher, at say 192kHz is so that other tasks (RIAA eq, amplitude adjustments etc.) can be handled better higher up (i.e. while editing, just you would hopefully edit a video or image in hires and only downsample at the end.)

cheaptr's picture

hi,i have a dynaudio xeo wireless system

.appologies for being a dimwit, my question is if i use foobar to play 24bit files using the usb output on my dynaudio will the files be dithered as well as downsampled as the dynaudio xeo only accepts a signal up to 16/48.are the music files automatically both downsampled and dithered simutantanaly on playback. 

X