Benchmark DAC1 now available with USB

[5Kurt]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif 5Kurt, From our experiences, native drivers are bit transparent for 16-bit only (DAC1 USB not included...it is transparent at 24-bit). This is dangerous even when playing 16-bit files because volume adjustments in software, etc result in 24-bit words due to division. On the contrary, we rarely achieved bit-transparency with devices using custom drivers. But they were capable of 24-bit. The custom-driver devices, though not transparent, did not result in significant distortion. But, at the same time, all the bits are not making it through ok. The ideal solution would be a 24-bit native solution...hence our design goal.

Elias, thank you very much for the answers.

I got one last question. How is Vista compared to WinXp sonically with DAC-1. Is it worth the upgrade or shall we stick with XP?

 

[EliasGwinn]

The AP does not output USB. We play a "bit-test" audio file using a media player (Foobar, for example) through the stock USB port on the computer. Then we convert the I2S output from the USB chip (TAS1020B) to SPDIF and send that back to the AP. The AP compares the SPDIF signal to the bit-sequence it uses to create the "bit-test" audio file. It reports any errors - that is, descrepancies in digital data - to assure that it is the same data, bit-for-bit.

 

[EliasGwinn]

Quote:

Originally Posted by 5Kurt /img/forum/go_quote.gif Elias, thank you very much for the answers. I got one last question. How is Vista compared to WinXp sonically with DAC-1. Is it worth the upgrade or shall we stick with XP?

Vista handles native audio totally different then XP, but not necessarily better. Where XP will pass the audio at the original sample-rate (or the highest presented to kmixer), Vista has a setting to choose the output sample-rate. It then converts everything to that sample-rate. The sample-rate conversion is very well designed, however, and although it is not bit-transparent, it causes very little distortion.

To answer your question about upgrading, I don't think its necessary for the sake of audio, but things may change in the near future, as fixes and updates come out.

I should add that all the Vista testing we did was on Vista Beta 2, so there may be things different in the current release then what we had.

 

[audioengr]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif The AP does not output USB. We play a "bit-test" audio file using a media player (Foobar, for example) through the stock USB port on the computer. Then we convert the I2S output from the USB chip (TAS1020B) to SPDIF and send that back to the AP. The AP compares the SPDIF signal to the bit-sequence it uses to create the "bit-test" audio file. It reports any errors - that is, descrepancies in digital data - to assure that it is the same data, bit-for-bit.

That's funny. I see how you made the measurement and it makes sense. I also use the TAS1020 in my products and I definitely hear a degrading difference when using the Windows drivers on XP rather than a custom driver. In fact, I also hear a difference when I bypass KMIXER. The driver makes a bigger difference in my case though.

Steve N.

 

[audioengr]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif Vista handles native audio totally different then XP, but not necessarily better. Where XP will pass the audio at the original sample-rate (or the highest presented to kmixer), Vista has a setting to choose the output sample-rate. It then converts everything to that sample-rate. The sample-rate conversion is very well designed, however, and although it is not bit-transparent, it causes very little distortion. To answer your question about upgrading, I don't think its necessary for the sake of audio, but things may change in the near future, as fixes and updates come out. I should add that all the testing we did was on Vista Beta 2, so there may be things different in the current release then what we had.

You say that ALL the testing you did was on Vista. Does this include the bit-comparison testing of the driver?

Did you not do the same bit comparison using XP?

Steve N.

 

[thomaspf]

Quote:

That's funny. I see how you made the measurement and it makes sense. I also use the TAS1020 in my products and I definitely hear a degrading difference when using the Windows drivers on XP rather than a custom driver. In fact, I also hear a difference when I bypass KMIXER. The driver makes a bigger difference in my case though. Steve N.

Agreed!

Unlesss you use kernel streaming to bypass kmixer on XP the bit stream is modified.

If you use the directsound output on foobar you will go through kmixer if you select direct kernel streaming either directly or via an ASIO shim you will bypass it. On Vista the mixer has much better quality but it still changes the stream even if you have selected the matching sampling rate. On Vista you can use exclusive mode in addition to kernel streaming to bypass the mixer but I am unaware of any player app outside the pro music space.

I don't see how a TAS1020 could be tweaked to work any other way with the USBaudio.sys driver.

Cheers

Thomas

 

[EliasGwinn]

Quote:

Originally Posted by audioengr /img/forum/go_quote.gif You say that ALL the testing you did was on Vista. Does this include the bit-comparison testing of the driver? Did you not do the same bit comparison using XP? Steve N.

Ahh...thank you for pointing out that mis-statement. What I meant to say is we did all Vista testing on Vista Beta 2. We did all of the same tests on all of the major operating systems:

-Mac OSX
-Windows Vista (Beta 2)
-Windows XP (SP2)
-Windows 2000 (SP4)

Hope that answers your question!

 

[asdf]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif Ahh...thank you for pointing out that mis-statement. What I meant to say is we did all Vista testing on Vista Beta 2. We did all of the same tests on all of the major operating systems: -Mac OSX -Windows Vista (Beta 2) -Windows XP (SP2) -Windows 2000 (SP4) Hope that answers your question!

. . . linux? Any testing on linux???? Anything????

 

[Jon L]

Quote:

Originally Posted by thomaspf /img/forum/go_quote.gif Agreed! On Vista the mixer has much better quality but it still changes the stream even if you have selected the matching sampling rate.

That's counterintuitive that if you set the Vista to 44.1kHz, it'll still "resample" to 44.1kHz, and I don't know why they didn't design it to just bypass Vista mixer in this scenario.

Same thing happens with SRC (Secret Rabbit Code) resampler in Foobar v.0.8.3. If you set it to 44.1kHz and activate it, it changes the 44.1kHz stream with audible difference. I don't know what to call it, but I kind of prefer Foobar with SRC "resampling" to 44.1kHz, which sounds a bit smoother and richer.

 

[EliasGwinn]

Quote:

Originally Posted by audioengr /img/forum/go_quote.gif That's funny. I see how you made the measurement and it makes sense. I also use the TAS1020 in my products and I definitely hear a degrading difference when using the Windows drivers on XP rather than a custom driver. In fact, I also hear a difference when I bypass KMIXER. The driver makes a bigger difference in my case though. Steve N.

Although I wouldn't dispute what your ears tell you, we found that digital audio data could be streamed bit-for-bit perfectly through kmixer when using usbaudio.sys. As mentioned before, this was proven using a digital audio bit stream with a known bit sequence, and it was found to be identical coming out as it was going in. Aside from the bit-test, we also measured the audio performance (Freq response, THD, IMD, etc) of this setup, and it was completely similar to the all other digital inputs of the DAC1 up to 96/24. The third test was listening. I am still conducting that test as we speak, as are several others here at Benchmark. With a CD transport feeding the Coax input, and the computer feeding the USB with the same music, no one has been able to differentiate the two inputs.

With that being said, there are several things that can affect your audio within a computer besides simply going through kmixer. However, we have found that when everything is configured properly, the digital audio can be streamed transparently through kmixer.

We will be publishing an online Wiki that will document everything we have found with various media players, OS's, etc.

Also, with all due respect, the Windows driver is only one aspect of the streaming technology. I am not familiar with your product, but the firmware for the TAS1020B is the main difference between our solution and others on the market. The majority of the development time of this product was spent programming the firmware for that chip. It is a completely custom solution designed to handle streaming USB audio gracefully.

 

[EliasGwinn]

Quote:

Originally Posted by asdf /img/forum/go_quote.gif . . . linux? Any testing on linux???? Anything????

Not yet. We don't have a linux machine set up, but I am considering doing that in the near future. I'll try to keep you all updated, but the new Benchmark Wiki should have an announcement when any new developments or tests happen.

 

[thomaspf]

While the ability to hear any differences between a bit perfect stream and a kmixer modified stream is one thing that is a bit different from the question of being bit perfect.

Deleted the rest to work through the follow-up

 

[The Monkey]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif I do most (98%) of my listening through studio monitors (JBL 6332, Tannoy Reveal 6, and K&H 4-ways). I hope my lack of headphone experience doesn't disqualify me from this forum .

Not at all. Come to the NYC Meet tomorrow and we'll get you sorted out.

 

[koob]

Quote:

Originally Posted by thomaspf /img/forum/go_quote.gif there is no question that kmixer modifies 16/44.1 PCM streams and there is nothing you can do about it short of bypassing kmixer.

1) Microsoft documentation clearly states that kmixer will pass a single audio stream through unaltered if the destination device supports the incoming sample rate.
2) Elias states that in Benchmark's testing, the stream recovered from the USB passthrough was bit-for-bit identical to the stream heading into the kmixer/usbaudio stack.

In light of these 2 facts, how do you subtantiate your claim that "there is no question that kmixer modifies" the stream?

Quote:

Originally Posted by thomaspf /img/forum/go_quote.gif Your test might not be perfect in that you don't catch these modifications but you can take my word for it that they are different.

Again, Benchmark's test compares the outgoing digital stream with the recovered digital stream and Elias states these were bit-for-bit identical. Are you suggesting the test is "not perfect" in that they can't tell the difference between a 1 and a 0 at a given stream position? Since the entire test is in the digital domain, it doesn't leave room for golden-ear testing/opinion. Are we supposed to take your word over direct binary comparison?

 

[thomaspf]

I have also spend a long time doing testing. Something does not add up, so we just need to get to the bottom of it but there is nothing to get emotional about.

Maybe I am wrong or the by pure accident the pseudo random test stream happens to get passed through unharmed. I am not doubting that the DAC1 is a great product and I always like computer focused audio devices or I would not be here. We will just have to find out.


Cheers

Thomas