Benchmark DAC1 now available with USB

[zheka]

Quote:

Originally Posted by puntloos /img/forum/go_quote.gif Many audiophiles would disagree

no doubts about this.

Quote:

Originally Posted by puntloos /img/forum/go_quote.gif That said, perfect reproduction would probably involve recreating the soundwaves that were present (i.e. 3D!) when the band was there, and this might be easier with more speakers. Right now though, multichannel is for fake effects. 'Boom!' from side speakers does not accurate sound make.

I think multi-channel digital audio is long since past fake sound effects stage.
well recorded 6 channel audio, even DTS/DDS, let alone uncompressed DVD-A, reproduced on good system is clearly superior to stereo , to my ears at least.

 

[slwiser]

Quote:

Originally Posted by zheka /img/forum/go_quote.gif AIX claim 5.1 at 24/96 as far as size limitations i think if MLP is used dual layer DVD (9G) can hold more that 2 hours of 6 channel 24/96 data

I guess I was all wrong there....but just maybe I was not wrong...read this from Dr. AIX;

http://www.audioasylum.com/scripts/t.pl?f=dvda&m=28265

Check out what he says in one of his posts, this one in particular:

http://www.audioasylum.com/forums/dv...ges/28266.html

It would appear that the words on the AIX website is open for some misunderstanding if what Dr. AIX is saying in this post is correct.

Note the liner page of an AIX disk says this for DVD-A. 5.1 "Stage" mix at 96 kHz/24 bits using MLP
and
Stereo Mix at 96 kHz/24 bis using PCM

The following is a discussion about MLP which appears to be a lossless format.

http://www.meridian-audio.com/p_mlp_mix.htm

This thread is a Benchmark DAC1 subject thread. I am sorry if I have push this to far off topic.

 

[EliasGwinn]

puntloos,

You asked why someone would use USB at all vs. a sound card with a digital output. This is a very legitimate question, as each has inherent pros and cons.

USB is obviously very convenient. Almost all computers (desktops and laptops) have an easily accessible USB port. It is true that it isolates the audio device from the high electro-magnetic (EM) field which exists inside of a computer. However, it doesn't necessarily cure the symptoms of this problem. As mentioned previously, the USB clock is just as prone to severe amounts of jitter. Often, this is intentional! Computer manufactures will add jitter to their clocks for the purpose of spreading the EM energy across a wider bandwidth (instead of a spike at a specific frequency) to measure better in FCC testing.

Some designers of USB audio devices alleviate this problem with asynchronous configurations, which makes the DAC clock the master, and the computer clock the slave. The problem with this is: when the computer is slave to an external clock, it will re-sample its data to the master. This re-sampling is often very poorly executed by the software, and results in severe distortion. The computer must be the master clock to ensure bit-transparency.

In our research and testing, we found the best configuration is isochronous/synchronous. This allows the data to be transmitted from the computer bit-transparently. This configuration, however, will have significant jitter. In the case of the DAC1, it is not affected by this jitter whatsoever, so this is a perfect solution all around. In fact, when we measure jitter-artifacts of the USB vs. the other digital inputs, there is no difference whatsoever. This is also evident in our listening room. All inputs perform identically.

Digital outputs on an internal sound card will probably be a little more subjected to jitter because of the environment, but otherwise should be about the same. In fact, an internal sound card is basically identical in operation with a few exceptions. A PCI card will have more direct access to data streams, depending on software configurations. Although USB is capable of handling data streams, it was not designed for it. In some specific systems with certain peripheral devices, drivers, etc, the USB performance may be compromised.

But, just like with any computer-related debate, it is entirely dependent on the specific system and configuration. There are very few absolutes with computer systems.

Thanks,
Elias

 

[puntloos]

Quote:

Originally Posted by zheka /img/forum/go_quote.gif I think multi-channel digital audio is long since past fake sound effects stage. well recorded 6 channel audio, even DTS/DDS, let alone uncompressed DVD-A, reproduced on good system is clearly superior to stereo , to my ears at least.

I do think this discussion is related to something we've previously touched upon: how do you separate the room from the speakers. Answer obviously is: you don't.

A very philosophical question then is: should a recording contain room characteristics? Would the ideal recording of a choir be a hypothetical 100 singers, each in their own soundproof booth singing into a single microphone, then mixing that together across a soundstage and let the playback room acoustics work themselves out?

Or should a recording try to 'force' the room it was recorded in into the room it is played back at by introducing (for example) echo that was in the recording room but not in the playback room, or trying to cancel out echo that is in the playback room but wasnt in the recording room.

If you want to 'transplant' a performance into your room, like the artist is standing there, then two speakers is all you need, and surround will just try to add distortedness into your room.

The only situation I can conceive of when 'surround' is useful is when you have some futuristic digital processor that first measures a room, and then tries to calculate what every speaker should be producing to in effect mimic the original performance (including acoustics) by adding or substracting accents.

I think windows vista comes with something resembling this technology but I doubt it will measure up to audiophile quality (yet), not to mention that a true surround system in my mind would have to consist of 4 or 8 identical speakers of sublime quality, which is very very very expensive. Some weak '2 decent fronts, 2 crappy rears, and some hidden away subwoofer' setup is something for people who also like many blinking lights and knobs on their audio components as well.

As for a recording where it sounds like you are one of the musicians i.e. 'singers are behind you, the choir in front of you.. well.. thats not how I experience audio when I go to a concert.

 

[EliasGwinn]

As for the DAC1 being capable of handling DVD-A, it is absolutely capable. DVD-A is nothing more then normal PCM digital audio at high resolutions. The DAC1 can handle resolutions above 192 kHz.

The problem, as you all are noting, is many players do not actually stream the digital information at their true resolutions because of DRM.

We've been testing many consumer-level DVD-A players to determine which, if any, play at full-rez. The only ones we found that do so are no longer in production. The DRM police have managed to ruin all the great benefits DVD-A was supposed to offer.

AIX is one exception. (If I'm not mistaken), they are not including the copyright protection status-bit that causes the DVD player to engage sample-rate conversion.

Thanks,
Elias

 

[zheka]

@ puntloos
I agree that good quality multi-channel audio system is indeed very expensive. In addition it is more difficult to set up and probably allows for fewer compromises than stereo one. Plus there is shortage of well recorded multi-channel material. That’s why I am sticking with stereo for now, not because it sounds better but because it makes more sense.

BTW, the “futuristic sound processors” are here already; take a look at DEQX product lineup http://www.deqx.com/index.html for example

@ slwiser
I am not sure I see any contradictions between what Dr. AIX is saying in that post and the liner page notes.

Respectfully,

gene

 

[Crowbar]

Re: high resolution recordings. There are many; theck out classicrecs.com or various Chesky recordings, there are 24/96 on DVD-Video (compatible with any player), and 24/192 on DVD-A, without DRM.

Re: sound processors. Such sound processors are severely limited and cannot reproduce the sound because of how the recording was made in the first place.

There are only two ways to really reproduce proper sound at the ears. The first is ambisonics, where an array of microphones is facing outward from a single point for the recording; then the soundfield is computed with spherical harmonics, and reproduced by a spherical arrangement of speakers around the listener. The more speakers are used, the more accurate the reproduction. Downside is you need more speakers for better results, and the sweet spot is pretty small.
The second way is binaural recordings made with microphones in the ear canals of a dummy head, and played back by in-ear-canal headphones. Since the outer ear differs significantly between individuals and makes a big difference in positional perception, to get the full effect, it is necessary to convolve the signal with the specific listener's own HRTF (heat-related transfer function). These can be measured with an anechoic chamber, which is impractical, or a laser scan of the head can be made and then the HRTF computed by finite boundary method simulation. But even without that, good binaural recordings are far better than stereo. To play them over speakers, one needs to use DSP to compensate for the crosstalk (left speaker to right ear and the reverse), which is of only limited effectiveness.

 

[Crowbar]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif The problem with this is: when the computer is slave to an external clock, it will re-sample its data to the master.

I don't know about various versions of Windows, but Apple and Linux programmers have confirmed for me this is not the case with their systems. They will not resample the data in asynchronous mode. If Windows is indeed resampling the data in asynchronous mode, that is not in the spirit of the USB Audio Class standard specification.

 

[music_man]

does the dac1 sound any different if you feed it a 24/192khz signal than if you feed it a 16/44.1. signal. i can't tell if i can really hear a difference. maybe because it just converts everything into 110khz anyways?

music_man

 

[Wavelength]

Elias,

Quote:

Some designers of USB audio devices alleviate this problem with asynchronous configurations, which makes the DAC clock the master, and the computer clock the slave. The problem with this is: when the computer is slave to an external clock, it will re-sample its data to the master. This re-sampling is often very poorly executed by the software, and results in severe distortion. The computer must be the master clock to ensure bit-transparency.

This is not how ASYNC works at all. First off the upper layers have no idea what the SYNCIN rate is. They merly know that the DAC's enumeration indicates that it supports rates like 44.1, 48, 88.2, 96 with a bit depth of let's say 24 bits. Therefore there is NO resampling done anywhere in this mode. Well other than what the KMIXER does but let's for this argument let's assume it's bypassed.

The USB driver seeing that the enumeration indicates ASYNC mode realizes there is two pipes. The data pipe out to the dac and the SYNC IN pipe. The Sync IN pipe is like a flow control mechanism.

The computer will send down data at the rate indicated by the application layer. Let's say 44.1... again for those of you that don't know this if the data is 16 bits say redbook then that is padded to 24 bits and piped out. In the ASYNC controller the buffer is measured and if the buffer is fine the ASYNC device tells the computer that it is on track by sending a stream to the computer over the Sync IN pipe that the rate is good. If the steam is too fast then the Sync IN pipe will tell the computer to slow down and if too slow to speed up.

The good thing about ASYNC mode is now you can use a really clean ultra low jitter master clock, instead of the jittery derived clock inside the part. Remember gang USB in it self has no jitter because there is no clock riding on top of the data like there is in SPDIF. **** BUT there is intrinsic jitter most of which happens inside the USB controller. Some is generated between the USB controller and the dac because of power supplies, poor grounding yadayadayada...

Anyways this external master clock can be very clean will then generate the BCLK, WCLK and clock out the data.

I released code base two for my products on wed this took a couple of months where I ripped out all the bs code from the reference design to include only out going stuff. I tried SYNC, Adaptive and Async and found that Async was the best at least for the TAS1020. I rewrote the Software PLL used in Adaptive as they where updating the MCLK ever 4ms and that made Adaptive actually work about the same as SYNC mode in regards to jitter.

In any mode you get bit perfect data. There is nothing special about Sync over any other mode. If the data is screwed up anywhere it will be between the application layer and the USB Driver.

In many cases though with Elias product the jitter will be some what removed in the upsampling section of the dac.

Since I don't use upsamplers ASYNC is the mode for me.

Thanks
Gordon

 

[Crowbar]

Thanks for clearing that up. That was my understanding from our email exchange, but after Elias' comment I started wondering whether there was some issue with the way Windows was handling things. My guess is he was referring to the kmixer resampling.

What I don't understand is why you went NOS. Async removes the jitter issue, but what does that have to do with oversampling?
Quote:

Since I don't use upsamplers ASYNC is the mode for me.

This statement seems to connect two unrelated things, NOS and asynch mode.

There's no reason to make a DAC NOS whether one has low or high jitter or uses asynch or other modes. There are two cases with NOS: either there's no analog filter, which can be OK within the DAC with a non-feedback high bandwidth analog stage, but then is almost guaranteed to cause TIM and intermodulation related problems in the power amplifier or headphone amplifier, or have an analog filter which since it can't be infinitely steep to filter an image beginning immediately above the band, will alias image energy into the audio band. And even if the whole system has extremely high bandwidth into the MHz (including the speakers) so that images are not a problem, you end up relying on the very imperfect lowpass filter of the human ear, which is a problem given the references posted here and elsewhere that ultrasonic energy can influence perception.

 

[EliasGwinn]

Puntloos,

On the question of: Why does the DAC1 re-sample to 110 kHz?

Here is why: it is the highest frequency to maintain the full oversampling of the D-A chip. EVERY D-to-A chip on the market cuts the oversampling rate in half to accommodate 192 kHz. This will also implement a different type of digital low-pass filtering which is inferior to the filter used at and below 110kHz.

This is also why most recording engineers don't use 192 kHz. The higher bandwidth seems appealing, but the stat-of-the-technology is such that 192 kHz conversion is actually inferior to 96 kHz.

Also, the DAC1's oversampling ASRC and resulting 110 kHz sample rate reproduces 96 kHz signals much more faithfully then a D-A converting the original 96 kHz signal. This is because the Nyquist frequency is on the slope of the filter (attenuated, but not completely). This is undesirable for two reasons. The first reason is the Nyquist frequency is not faithfully converted to analog (ie, the analog bandwidth of 96 kHz conversion is actually less then 48 kHz). With the DAC1, the full bandwidth of a 96 kHz signal can be faithfully reproduced. The second problem with 96 kHz conversion is the frequencies at and above Nyquist (48 kHz and up) are not completely attenuated, so some aliasing and imaging will occur. With the 110 kHz upsampling and conversion in the DAC1, the frequencies below 55 kHz are not in danger of being aliased.

Thanks,
Elias

 

[Crowbar]

If I remember correctly, the DAC1 uses AD1853. That has internal oversampling (called interpolation filter in the datasheet). Are you bypassing that when feeding the 110 kHz in, or does it get 4x or whatever by the DAC chip's internal filter?

 

[EliasGwinn]

Puntloos,

When you listed possible DAC's you might buy, you mentioned that the Bel Canto has the same UltraLock system as the DAC1. This is not true. They call their system UltraClock (which sounds strikingly similar to UltraLock...perhaps just coincidence

). More importantly, however, it is not the same circuitry. I am not familiar with the effectiveness of their jitter reduction technology, but I can assure you that the DAC1's jitter reduction technology is much more then a plug-and-play solution. Their are several design considerations that require meticulous engineering to achieve.

You also mentioned that it has a better input stage then the DAC1. Can you elaborate on this a little bit?

Also, please see my previous post concerning 192 kHz.

Thanks,
Elias

ps. I realize that the thread has moved on to new topics, but I don't want to leave any questions unanswered.

 

[EliasGwinn]

Quote:

Originally Posted by audioengr /img/forum/go_quote.gif As for the jitter in a stock DAC-1, this is addressed in their design by using asynchronous upsampling like many other DAC's. It certainly does reduce the jitter of the incoming stream, probably better than many DAC's. However, this technique does not totally eliminate jitter IME. When I remove the upsampling chip and replace it with an I2S interface driven by my FIFO reclocker, the audible jitter is noticable lower. Steve N.

Steve,

What method did you use to determine the jitter attenuation?

Thanks,
Elias