I am deeply philosophically (as well as somewhat economically) opposed to the use of 2-way interfaces, including USB and I2S, as a means of audio interconnection.
First, it must be said that AES/SPDIF are perfectly fine. Under the normally specified limits (is it 30M for coax, 200M for AES ?) these can be shown to produce typically 200 pS jitter (my complex system does that), and that is 50 times below demonstrated audible limits in pure sine waves, and 100 times demonstrated audible limits in best case music on headphones only (famous AES publication).
In my system a typical input goes through 4 digital interconnections, 3-4 of them AES and the other SPDIF. I could add a few more AES layers, and it would still be 200 pS.
I banished Toslink only a couple years ago (and I had actual glass Toslink cables) because Toslink transmission is measureably worse, above 500 pS, but not because it actually makes a difference. I tried an unbalanced coax into my dac, which I was afraid might cause ground loops, but it didn't anyway, so I went wired all the way, wired is demonstrably better, and no sacrifice in this case. Complex systems like mine (which uses some prosumer equipment) generally work best with AES, thought, technically, there's a small loss in bandwidth that comes with the signal isolation, but much less so than toslink. But coax seemed fine with the Audio GD Dac 19 (I had coax muting issues with the Master 7 Singularity, though I haven't retested since enabling the PLL), and the Onkyo and Denon DACs I have used.
Anyway, this transmission jitter, produced by a dumb detector, is nothing in the final analysis. Any decent receiving system using PLL or Asynchronous reception can in principal reduce effective jitter to zero, so whatever jitter ultimately appears in the output could have been avoided (personally I prefer PLL which I think of as good enough, and it uniquely preserves the original data, which somehow I think is important, I will admit though I'm not an expert on what Asynchronous can do). Most jitter that we have seen has been highly avoidable with AES and SPDIF. And down below 200 pS, it wouldn't matter anyway if the DAC manufacturer did nothing about it. But certainly high frequency jitter can be reduced to zero, the only frequency variation being the comparatively slow changes of the source clock, which may have to be accomodated, but that very low frequency (much less than 1 Hz) jitter is even less audible, within the typical bounds.
[There is acutally a matter of trust here. We trust that the source of the transmission has a very stable clock, since in the end that is what is driving things, the other end has to follow in the long run, and it's best not to be snapped around. In the beginning CD players and the like may have been among the most expensive things in an audio system, and have the best clocks. Nowadays, it's typically to have inexpensive front ends, and have all the expense in the DAC system, so the DAC manufacturer would like to have it their way in the clocking...I think that may be possible with with USB and I2S...but personally I'm fine with trusting the front end, I always attend to that, in my AES/SPDIF practice. Or I trust another one of them... Rolling sources is my fun, you see. You who use cheap sources typically use USB or I2S now anyway, so keep on with your mega expensive dac clocks and so on, and never mind me, though I suspect that any non-super-cheap device is going to be OK, have long term clock drift at essentially undetectable levels anyway, so good for those who don't feel the need to bother with the fancy back end anyway.]
I2S itself is a two wire interface when computer systems have been going one wire. I2S was designed for doing communication within one box. It requires setting up a master-slave relationship, which in an interconnect system really requires software negotiation. Now you're talking proprietary software of all kinds, drivers, stuff you will endless have to buy and to mess with. And USB has already proven itself to be that way manifoldly.
AES/SPDIF has been plug and play since the beginning, it is an open system and easy to understand, and philosophically there's something I like about the free flowing nature of it. Easily divided, endlessly rerouted, distributed, switched, split, (not combined). That's the kind of flexibility anyone who is messing with a complex system wants.
But as always, the industry would prefer to sell top to bottom systems, wipe out all that complexity, and the drive to systems like USB and especially I2S exactly does that. Getting to any kind of distribution beyond 1 to 1 with I2S...isn't that pretty difficult?
In my view, that's back to being the victim of industry, rather than the beneficiary of it.
AES/SPDIF is good enough! Let's stick with that!
First, it must be said that AES/SPDIF are perfectly fine. Under the normally specified limits (is it 30M for coax, 200M for AES ?) these can be shown to produce typically 200 pS jitter (my complex system does that), and that is 50 times below demonstrated audible limits in pure sine waves, and 100 times demonstrated audible limits in best case music on headphones only (famous AES publication).
In my system a typical input goes through 4 digital interconnections, 3-4 of them AES and the other SPDIF. I could add a few more AES layers, and it would still be 200 pS.
I banished Toslink only a couple years ago (and I had actual glass Toslink cables) because Toslink transmission is measureably worse, above 500 pS, but not because it actually makes a difference. I tried an unbalanced coax into my dac, which I was afraid might cause ground loops, but it didn't anyway, so I went wired all the way, wired is demonstrably better, and no sacrifice in this case. Complex systems like mine (which uses some prosumer equipment) generally work best with AES, thought, technically, there's a small loss in bandwidth that comes with the signal isolation, but much less so than toslink. But coax seemed fine with the Audio GD Dac 19 (I had coax muting issues with the Master 7 Singularity, though I haven't retested since enabling the PLL), and the Onkyo and Denon DACs I have used.
Anyway, this transmission jitter, produced by a dumb detector, is nothing in the final analysis. Any decent receiving system using PLL or Asynchronous reception can in principal reduce effective jitter to zero, so whatever jitter ultimately appears in the output could have been avoided (personally I prefer PLL which I think of as good enough, and it uniquely preserves the original data, which somehow I think is important, I will admit though I'm not an expert on what Asynchronous can do). Most jitter that we have seen has been highly avoidable with AES and SPDIF. And down below 200 pS, it wouldn't matter anyway if the DAC manufacturer did nothing about it. But certainly high frequency jitter can be reduced to zero, the only frequency variation being the comparatively slow changes of the source clock, which may have to be accomodated, but that very low frequency (much less than 1 Hz) jitter is even less audible, within the typical bounds.
[There is acutally a matter of trust here. We trust that the source of the transmission has a very stable clock, since in the end that is what is driving things, the other end has to follow in the long run, and it's best not to be snapped around. In the beginning CD players and the like may have been among the most expensive things in an audio system, and have the best clocks. Nowadays, it's typically to have inexpensive front ends, and have all the expense in the DAC system, so the DAC manufacturer would like to have it their way in the clocking...I think that may be possible with with USB and I2S...but personally I'm fine with trusting the front end, I always attend to that, in my AES/SPDIF practice. Or I trust another one of them... Rolling sources is my fun, you see. You who use cheap sources typically use USB or I2S now anyway, so keep on with your mega expensive dac clocks and so on, and never mind me, though I suspect that any non-super-cheap device is going to be OK, have long term clock drift at essentially undetectable levels anyway, so good for those who don't feel the need to bother with the fancy back end anyway.]
I2S itself is a two wire interface when computer systems have been going one wire. I2S was designed for doing communication within one box. It requires setting up a master-slave relationship, which in an interconnect system really requires software negotiation. Now you're talking proprietary software of all kinds, drivers, stuff you will endless have to buy and to mess with. And USB has already proven itself to be that way manifoldly.
AES/SPDIF has been plug and play since the beginning, it is an open system and easy to understand, and philosophically there's something I like about the free flowing nature of it. Easily divided, endlessly rerouted, distributed, switched, split, (not combined). That's the kind of flexibility anyone who is messing with a complex system wants.
But as always, the industry would prefer to sell top to bottom systems, wipe out all that complexity, and the drive to systems like USB and especially I2S exactly does that. Getting to any kind of distribution beyond 1 to 1 with I2S...isn't that pretty difficult?
In my view, that's back to being the victim of industry, rather than the beneficiary of it.
AES/SPDIF is good enough! Let's stick with that!
No comments:
Post a Comment