[nicolo]

I think business ethics is a huge part of why Schiit has been so successful. I can tell you that Schiit buying American parts whenever possibe, treating their employees well, not bad mouthing other companies, letting their products stand on their own merit, and investing in the community are big reasons why I buy from them.

I just don't think evacant is explaining himself well. Punctuation and grammar would help a lot as I can't really follow his train of thought.

 
There's no such thing as business ethics. People have ethics, not organizations like private/public/for profit/non profit etc. Schiit's founders are ethical because they treat their employees fairly and they communicate transparently, not because they buy American parts. That's a personal choice made by the founders in part to help local vendors, but it also results in them getting possibly faster and better access to both raw and finished components. Please don't conflate the two.
 
As for evacant, there are far more appropriate forums than this one to talk about labor practices. This is after all a hobbyist's forum.

 

[rtaylor76]

I think we're getting of tangent here. Let's not argue about semantics.

 

[earnmyturns]

I'm idle this holiday weekend and away from all my nice audio gear, so I'm wasting time thinking about all of these PCM/DSD/MQA wars. I feel that there's a lot of confusion around terms like "lossless" and "bit perfect". Any bounded bit rate digital encoding of an arbitrary analog signal has to be lossy, in the sense that any exact representation of an arbitrary signal needs an infinite number of arbitrary precision coefficients (a simple Fourier expansion shows that). Any digital encoding has a "bit budget" that is bounded below by the mathematics of Nyquist and Shannon, and represents an approximation of the source signal. But of course audio signals are not arbitrary, they are constrained by the physics (and psychophysics) of voices, instruments, microphones, speakers, and hearing. In particular, our hearing is band-limited, and grows more so as we get older. So far, no news for you. But from these statements, it follows that the choice of digital encoding must be governed by criteria that take into account the relevant physical and psychophysical constraints. Perceptual coding systems attempt to do that by using their bit budget to match measured properties of human hearing. But all such measurements of human perception are subject to error, for two main reasons. First, humans are highly variable in their physical attributes and neurological endowment. Second, it is very hard to do fully controlled double-blind psychophysical experiments. As a result, perceptual encoders may do well on some mythical average population and average source materials, but not so well on the tails of those distributions (listeners or source materials). That's why many here prefer simple encoding techniques that are less bit-economical but rely only on the gross band-limitation (~22kHz) of the hearing system and the Nyquist theorem. That's what Redbook was designed for. But there are further complications. The digitally encoded samples do not specify a unique reconstruction by a DAC. Those same samples can be interpolated (even exactly, as with the Schiit multibit DACs) in many different ways by different DACs. In theory those variations should not matter because they only affect frequencies above the Nyquist limit, but in reality, DAC circuit nonlinearities may alias high-frequency energy into the audible spectrum. And finally to MQA. There's no reason in theory why one could not have a more bit efficient digital encoder than FLAC-ed PCM or DSD, it's not as if either of these are any close to the Shannon/Nyquist limit for the source materials we listen to. But getting closer to that limit seems to require making more assumptions about signal sources and targets that just a 22kHz band limit (there's some beautiful algorithmic complexity theory that shows that getting real close to the Shannon limit is computationally very hard, hence the need for simplifying assumptions). To the extent that MQA relies on such simplifying assumptions, it's just another perceptual coder, although maybe better than earlier ones. But their obsession with secrecy makes it very hard for anyone like me who actually has a bit of understanding of the issues to suspend disbelief. In contrast, whatever the sillinesses in the PCM vs DSD wars, the encodings are open, so we can judge ADCs and DACs by a combination of measurable and perceived attributes where we can factor out a well-understood encoding. Finally, what may go on in converting digital masters to MQA is a total mystery, since, as pointed out above, a digital representation of a signal is necessarily underspecified. That is, any post-encoding DSP necessarily makes assumptions about the original analog source. To the extent that those assumptions go beyond simple band limitation, we get into the ream of psychophysics with its inevitable uncertainties.

 

[rtaylor76]

You must have a doctorate.
 
I think I understand....
 
PCM has limitations
Our hearing has limitations
Even Schiit DACs interpolate the code
Double Blind studies are hard
MQA is a mystery
 
And I mostly agree.
 
I think you have come to the conclusions that most of us have. It is NOT the format. Well, lossless formats. It has more to do with the decoding of that signal.
 
Sure, there are just as crazy limitations on encoding as there are decoding. I see the forums and talk on the latest converters (ADCs), mics, microphone preamps, cables, clocking, etc. Even huge discussions on mixing in the box vs mixing with a console. Lots of talk about reel to reel tape saturation and compression and whether or not the plug-ins sound as good as the original vintage gear or this new clone Neve preamp sounds just like a real one for half the price.
 
You can get lost chasing it all. The simple fact is that that THE most accepted digital format are redbook CDs. Up until recently, that was Tidals top tier - 44/16. And now all this MQA business, but I digress. That is why the marketing direction of Schiit is "DACs Made for the Music You Have, Not the Music You Have to Buy." 
 
And I agree that most of the stuff about MQA is hype, but they need it. It is borderline hyperbole. I admit that I was take aback at SACD when I first heard it. Logically, DSD makes sense. What I didn't know at the time, that I know now, is that it has more to do with the decoding process. That high frequency noise was able to be filtered out. Now, after hearing real multibit PCM decoding, I am just in sheer awe at the fidelity given the file size and other limitations as you mentioned (Nyquist/Shannon).

 

[earnmyturns]

  You must have a doctorate.
 
I think I understand....
 
PCM has limitations
Our hearing has limitations
Even Schiit DACs interpolate the code
Double Blind studies are hard
MQA is a mystery
 
And I mostly agree.
 
I think you have come to the conclusions that most of us have. It is NOT the format. Well, lossless formats. It has more to do with the decoding of that signal.

I do have a PhD, and training in EE, math, and CS, but clearly my teaching ability falls short, and all of this is mainly from memory, my information theory books being on the other coast at the moment. What I was trying to point out is that no digital encoding is truly "lossless," in the sense that the digitized signals of bounded bitrate have a maximum information content that in general will be less than the information content of an arbitrary analog signal. So, the real debate is not between "lossless" and "lossy," but rather between the kinds of assumptions and approximations that different encodings make. The losses in PCM (and, arguably, DSD) are relatively simple to characterize, In PCM, which I understand best, there are two sources of loss. First, the sampling rate (Nyquist-Shannon theorem) should be higher than twice the highest frequency represented in the original signal's power spectrum, otherwise there will be aliasing. The 44.1kHz rate thus assumes that there's no power in the original signal above 22.05kHz, which is the reason for the "brickwall" filters used on such sources. Second, the sample values (amplitude) are discretized: in the standard L(inear)PCM code, into fixed width bins encoded as signed binary numbers. That discretization is lossy too. So, to some extent, every encoding requires the DAC to guess the source signal. MQA is lossy, but so are DSD and MQA. Whether the guess is good or less good depends on a complex match between the characteristics of the source, of the code, or the DAC itself, of the analog stages, and of the listener's ear and brain. In other words, YMMV.

 

[mikoss]

I was happy Schiit clarified their position on MQA. In my opinion, it's purely a hype train. I stand with the opinion that most people don't care and will happily listen to their mastered for iTunes music in lossy format. The ones who care about quality either enjoy redbook, analog, or even hi-res without worrying about bandwidth or space. Personally, if I'm going to pay for a 24/192 file, why would I sacrifice some of it to lossy encoding just to save space? It's a crazy proposition.

As for streaming, MQA might stand a chance if it is noticeably better sounding on consumer grade products. Going out on a limb here and going to say 320/redbook/hi res is fairly indescribable in those areas. I'm also calling bs on the "Master Quality" title. It's a regular digital recording that's been compressed.

Give us actual nicely mastered material, that's all we want.

 

[rtaylor76]

  I do have a PhD, and training in EE, math, and CS, but clearly my teaching ability falls short, and all of this is mainly from memory, my information theory books being on the other coast at the moment. What I was trying to point out is that no digital encoding is truly "lossless," in the sense that the digitized signals of bounded bitrate have a maximum information content that in general will be less than the information content of an arbitrary analog signal.

 
Same could be said of any medium. ANY medium. Domains on tape, grains on film, pixels and sensors, which Nyquist filters also have to be used in digital imaging sensors, but anyway...
 
PCM is still about the closest to the original source as you can get. So I think there will be always issues to face. I think the trick is to try to get to those things that really muck things up, like the anti-alias filter that does not have much ringing, which Dave and Mike addressed in their mega-combo-burrito filter. 
 
And I think the term "lossy" is just an industry standard term in reference to file size compression - don't try to take it out of context, but yes, you are correct. I understand what you mean, no need to explain it further. It's like when I was trying to explain to my father in law the difference between what "Genetically Modified Organism" means as he was taking to mean a much larger definition than what it actually is, but I could not tell him any differently.
 
Sorry, not trying to change the subject, but I am just saying we all know even the data we are trying to decode is not perfect in the first place to their limitations and thus, "lossy". Got it.
 
And most of us past the age of 20 can't hear past 17k anyhow. Higher resolutions with more HF info can just give way to shallower slope filters with less ringing. This ringing is what really mucks things up. And the noise, and the blah blah.

 

[yay101]

I do have a PhD, and training in EE, math, and CS, but clearly my teaching ability falls short, and all of this is mainly from memory, my information theory books being on the other coast at the moment. What I was trying to point out is that no digital encoding is truly "lossless," in the sense that the digitized signals of bounded bitrate have a maximum information content that in general will be less than the information content of an arbitrary analog signal. So, the real debate is not between "lossless" and "lossy," but rather between the kinds of assumptions and approximations that different encodings make. The losses in PCM (and, arguably, DSD) are relatively simple to characterize, In PCM, which I understand best, there are two sources of loss. First, the sampling rate (Nyquist-Shannon theorem) should be higher than twice the highest frequency represented in the original signal's power spectrum, otherwise there will be aliasing. The 44.1kHz rate thus assumes that there's no power in the original signal above 22.05kHz, which is the reason for the "brickwall" filters used on such sources. Second, the sample values (amplitude) are discretized: in the standard L(inear)PCM code, into fixed width bins encoded as signed binary numbers. That discretization is lossy too. So, to some extent, every encoding requires the DAC to guess the source signal. MQA is lossy, but so are DSD and MQA. Whether the guess is good or less good depends on a complex match between the characteristics of the source, of the code, or the DAC itself, of the analog stages, and of the listener's ear and brain. In other words, YMMV.

I feel like you will appreciate this: https://www.meridian-audio.com/meridian-uploads/ara/coding2.pdf

I'd love to hear your thoughts if you would be willing to share, your maths experience goes well beyond my own.

 

[Billheiser]

well its important to me ,ive read quite a few forems on this sight but ive never come across one that questions the ethics of the companys that manufacture audio euipment or the places they invest in .id say its the decideing facture of everything i buy ,from food to phones to earphones ,but i know most people dont give a s---- ,well not until it comes round to effects them

Your opinion of how the world is messed up is important to you, but does not belong on this thread.

 

[Rudiger]

About audio transport. The wireless transport unit I have is the Nad Dac 2, I connect it to the Bifrost. It works by radio waves (It cause some interfere when walking near to it ...). But I'm pretty satisfied.
That said, I would prefer by far a WIFI solution. But I can not find an equivalent. Chromecast Audio tempts me a lot, but it must use compatible software etc. which is very restrictive. (I mainly use foobar 2000 from PC and various music websites).
If anyone knows a simple solution, or if Schiit wants to make one, I'm interested ...

 

 

[yay101]

About audio transport. The wireless transport unit I have is the Nad Dac 2, I connect it to the Bifrost. It works by radio waves (It cause some interfere when walking near to it ...). But I'm pretty satisfied.

That said, I would prefer by far a WIFI solution. But I can not find an equivalent. Chromecast Audio tempts me a lot, but it must use compatible software etc. which is very restrictive.

If anyone knows a simple solution, or if Schiit wants to make one, I'm interested ...

Arm based computers like the pi support:

Upnp.
Airplay.
Various streaming services.
Web radio.
Support for all formats including DSD.
Local file support.
Many front ends including web browsers and apps on all platforms.
You can upnp out from foobar to the pi to output the exact wav end result from foobar.

Worth considering.

 

[Rudiger]

Arm based computers like the pi support:

Upnp.
Airplay.
Various streaming services.
Web radio.
Support for all formats including DSD.
Local file support.
Many front ends including web browsers and apps on all platforms.
You can upnp out from foobar to the pi to output the exact wav end result from foobar.

Worth considering.

 
Thank you. And would it be possible to simply stream the music from my PC to the Pi (and to the bifrost) ? This is equivalent to stream from my PC to another PC if I understand correctly.
I do not want to get rid of my PC which is also connected to my stereo and that my wife uses, etc.

 

[yay101]

Thank you. And would it be possible to simply stream the music from my PC to the Pi (and to the bifrost) ? This is equivalent to stream from my PC to another PC if I understand correctly.
I do not want to get rid of my PC which is also connected to my stereo and that my wife uses, etc.

Indeed the upnp out plugin for foobar 2000 will send the signal from foobar to the pi as plain pcm. The other option is to share the music folder with the pi and perform the decoding on the pi.

 

[Rudiger]

Well, finally I think I'll go to the Chromecast audio (connected to the bifrost) and install an UPnP pluggin to foobar.

 

[earnmyturns]

I feel like you will appreciate this: https://www.meridian-audio.com/meridian-uploads/ara/coding2.pdf

I'd love to hear your thoughts if you would be willing to share, your maths experience goes well beyond my own.

Thanks, started to read it, got my guard up soon enough with the claim that dither achieves infinite quantization and temporal resolution, but I'll persist through the end.