I2s vs USB

[Blizzard]

Why would fiber>electrical conversion have more jitter than LVDS> SE i2s conversions?

 

I'm not even sure if it's possible to convert I2S to fibre then back to I2S. If so I don't know anyone doing it.

 

Since toslink spdif causes more jitter than coaxial, I would assume the same with raw I2S. LVDS is almost the same as the TMDS used for HDMI. There's a reason TMDS is used over copper rather than optical for HDMI.

 

The LVDS conversion is extremely low jitter causing. Especially if done inside an FPGA.

 

Anyways the new wave of cutting edge renderers will be built into the DAC's. So LVDS conversion won't be required. LVDS was really a solution for a problem in the past.

[Blizzard]

This is quite true.

 

 

I agree with Blizzard here.

 

 

Blizzard is right that ultimately the USB interface has to interface to the i2s input of the DAC chip. What is missed is that to get i2s there is still some conversion only it is being done internal to the "transport" - very few processors have a "native" or "raw" i2s output. If you utilise something like the Pink Faun card for example you are simply moving the conversion from the USB bus to the PCIe bus.

 

 

Well yes i2s is a (defacto) standard ... but external to a box it is no longer a standard.

 

 

Instead you have the jitter causing steps involved in getting a processor to output i2s (this is excepting dedicated processors such as SHARC which natively do i2s).

 

The i2s is part of the audio codec which is added to the ARM core (for example) albeit all part of the same silicon.

 

 

With respect thats supposition ... it could utilise an internal USB to i2s device in the same way the Aries and Aries LE use an internal USB to SPDIF converter.

 

The steps for getting the I2S out of a renderer are far less jitter causing than USB. Besides the best renderers send the I2S direct from the chips.

 

General purpose computers aren't involved with the best setups. If they are, they are just used as a data source and GUI.

[Audio_ELF]

The steps for getting the I2S out of a renderer are far less jitter causing than USB. Besides the best renderers send the I2S direct from the chips.

 

General purpose computers aren't involved with the best setups. If they are, they are just used as a data source and GUI.

Can you give an example of "the best renderers"?

 

I will give you that anything with a SHARC/Blackfin processor (and I would think other similar DSP) will output i2s directly, but anything with an ARM core (for example) will also have some kind of audio codec on the silicon and it is that which outputs i2s (afaik).

[Blizzard]

Can you give an example of "the best renderers"?

 

I will give you that anything with a SHARC/Blackfin processor (and I would think other similar DSP) will output i2s directly, but anything with an ARM core (for example) will also have some kind of audio codec on the silicon and it is that which outputs i2s (afaik).

 

2 great examples are the MSB network renderer, and the Ravenna interface in the Merging NADAC. Both of which don't require LVDS conversion for the I2S/DSD.

 

I'll also add that both of them are slaved to the DAC master clock. Another extremely important attribute.

 

 

Speaking of the Merging NADAC, there's finally a review out. Now all the Ravenna bashers can explain how they are getting such great sound out of an interface protocol not suitable for high end audio. I heard Chris with have one to review on Aug 31st as well.

 

http://positive-feedback.com/audio-discourse/impressions-the-merging-technology-nadac-mc-8-dsd-dac/

[Audio_ELF]

2 great examples are the MSB network renderer, and the Ravenna interface in the Merging NADAC. Both of which don't require LVDS conversion for the I2S/DSD.

Okay ... so lets take the MSB network renderer, do you know what processor it uses?

[Blizzard]

Okay ... so lets take the MSB network renderer, do you know what processor it uses?

 

Not sure what they have hidden in that box. Probably a combination of CPU and FPGA. Either that or Hybrid CPU/FPGA. But I imagine the FPGA end of things outputs the I2S direct in the purest way possible.

[4est]

You said fiber, not toslink. ST fiber is a more robust method, and likely doable.

I'm not even sure if it's possible to convert I2S to fibre then back to I2S. If so I don't know anyone doing it.

 

Since toslink spdif causes more jitter than coaxial, I would assume the same with raw I2S. LVDS is almost the same as the TMDS used for HDMI. There's a reason TMDS is used over copper rather than optical for HDMI.

 

The LVDS conversion is extremely low jitter causing. Especially if done inside an FPGA.

 

Anyways the new wave of cutting edge renderers will be built into the DAC's. So LVDS conversion won't be required. LVDS was really a solution for a problem in the past.

[4est]

These seem to be very operative words here.

But I imagine

[Blizzard]

You said fiber, not toslink. ST fiber is a more robust method, and likely doable.

 

Optical Toslink uses fibre. I can imagine that's what would be used for HDMI if it was better than TMDS/LVDS.

 

Anyways, no point in exploring this further when there's better ways to handle I2S, such as direct over PCB traces to the DAC chip.

 

For me when I find a best way to do things, I don't dwell on ways to make inferior methods a bit better. I focus on the best ways, and ways to make the best ways better.

[Blizzard]

These seem to be very operative words here.

 

Assuming MSB knows what their doing. Which I think is extremely likely.

 

Based on their track record and reputation, I can't see them going through the trouble of building this device to offer their clients an inferior method of listening to their DAC's.

 

Since they offer USB, I2S over LVDS, and direct I2S options for their DACs, they have no reason to be biased towards any method of doing things. So chances are if you ask them the best way, you'll get an honest answer.

[4est]

Why would you imagine that they would use toslink when ST optical exists AND is better? Here we go again speculation based on opinion... I am not suggesting that any of this is better, or worse even. I simply do not know. My objection is that you have read some spec sheets, and now claim to "know the best way" whilst pretending to understand all of the underlying engineering. I get that you have a product in the incubator and want some street cred, but the world is not as black and white as you make it to be.

Optical Toslink uses fibre. I can imagine that's what would be used for HDMI if it was better than TMDS/LVDS.

 

Anyways, no point in exploring this further when there's better ways to handle I2S, such as direct over PCB traces to the DAC chip.

 

For me when I find a best way to do things, I don't dwell on ways to make inferior methods a bit better. I focus on the best ways, and ways to make the best ways better.

[Blizzard]

Why would you imagine that they would use toslink when ST optical exists AND is better? Here we go again speculation based on opinion... I am not suggesting that any of this is better, or worse even. I simply do not know. My objection is that you have read some spec sheets, and now claim to "know the best way" whilst pretending to understand all of the underlying engineering. I get that you have a product in the incubator and want some street cred, but the world is not as black and white as you make it to be.

 

I'm not suggesting anyone use Toslink. I'm saying Toslink is optical fibre technology. But it's not better than copper coaxial SPDIF technology. Either way going any way besides I2S direct over short PCB traces is going backwards.

 

I'm not looking for any "street cred" I'm just clearing up misconceptions by stating facts that are well known among industry leaders.

[4est]

Look, we got on this because you made a suggestion that this was not doable, and I still disagree and feel ST optical might be similar to LVDS i2s. Presently, depending upon length, if I had the need I would be inclined to try running regular single ended i2s over BNC- especially if I were using a separate master clock.

I'm not suggesting anyone use Toslink. I'm saying Toslink is optical fibre technology. But it's not better than copper coaxial SPDIF technology. Either way going any way besides I2S direct over short PCB traces is going backwards.

 

I'm not looking for any "street cred" I'm just clearing up misconceptions by stating facts that are well known among industry leaders.

[jabbr]

The steps for getting the I2S out of a renderer are far less jitter causing than USB. Besides the best renderers send the I2S direct from the chips.

 

General purpose computers aren't involved with the best setups. If they are, they are just used as a data source and GUI.

 

Maybe in your brane. But it is widely known by the very highest regarded and universally acknowledged industry leaders that general purpose CPUs are in the very best setups.

[Blizzard]

Look, we got on this because you made a suggestion that this was not doable, and I still disagree and feel ST optical might be similar to LVDS i2s. Presently, depending upon length, if I had the need I would be inclined to try running regular single ended i2s over BNC- especially if I were using a separate master clock.

 

Well if you can come up with a interface for I2S transmission over ST optical that's superior to I2S direct over an inch or 2 of PCB traces, you may be on to something.

[Blizzard]

Maybe in your brane. But it is widely known by the very highest regarded and universally acknowledged industry leaders that general purpose CPUs are in the very best setups.

 

Yes for the GUI, and in cases like SRC/modulation to the source audio prior to the renderer you are right. But SRC/modulation on PC is really only beneficial to lower Rez formats based on my experience with HQ player over the last month.

[Blizzard]

Yes for the GUI, and in cases like SRC/modulation to the source audio prior to the renderer you are right. But SRC/modulation on PC is really only beneficial to lower Rez formats based on my experience with HQ player over the last month.

 

Personally I prefer mobile devices for GUI. Unless I want to breathe new life into redbook quality tracks. Then HQ player is good for that.

[jabbr]

Yes for the GUI, and in cases like SRC/modulation to the source audio prior to the renderer you are right. But SRC/modulation on PC is really only beneficial to lower Rez formats based on my experience with HQ player over the last month.

 

That was very tongue in cheek on my part. Yeah so HQPlayer is only good for everything except your DSD256 recordings. Niche product. PeterSt also prefers general CPU for use with NOS1a. What do you suggest for DSC1 or Lampizator? The demise of the x86 architecture has been imminent since the late 1970s. If those darn Intel engineers weren't so pesky, the world would be more architecturally pure.

 

The advantage I have is that I really don't care who "wins" as long as I get better sound. I am a consumer who would buy a card if a reasonable one existed which could output LVDS/I2S/DSD if it existed. Upcoming project is to build the DSC1 ... boards sitting on my desk for a few months now ... it takes DSD in (over "I2S" lines) which typically is from Amanero USB-I2S ... but could be from LVDS if that existed ... or could be from an inbuilt ARM board. If an ARM board existed that took fiberoptic Ethernet in (e.g. SFP) then I would already own it ... in any case there are options and no one right way of doing things.

 

In my experience the future is too hard to predict and I look for options that can be implemented now or in the near future.

[4est]

I do not think it is fair to deal with this in that context. If one only needs to travel an inch or two, there is no reason to go USB and this topic is moot.

Well if you can come up with a interface for I2S transmission over ST optical that's superior to I2S direct over an inch or 2 of PCB traces, you may be on to something.

[Jud]

That was very tongue in cheek on my part. Yeah so HQPlayer is only good for everything except your DSD256 recordings. Niche product. PeterSt also prefers general CPU for use with NOS1a. What do you suggest for DSC1 or Lampizator? The demise of the x86 architecture has been imminent since the late 1970s. If those darn Intel engineers weren't so pesky, the world would be more architecturally pure.

 

The advantage I have is that I really don't care who "wins" as long as I get better sound. I am a consumer who would buy a card if a reasonable one existed which could output LVDS/I2S/DSD if it existed. Upcoming project is to build the DSC1 ... boards sitting on my desk for a few months now ... it takes DSD in (over "I2S" lines) which typically is from Amanero USB-I2S ... but could be from LVDS if that existed ... or could be from an inbuilt ARM board. If an ARM board existed that took fiberoptic Ethernet in (e.g. SFP) then I would already own it ... in any case there are options and no one right way of doing things.

 

In my experience the future is too hard to predict and I look for options that can be implemented now or in the near future.

 

Get that DSC1 going already! Because you know as soon as you start Miska will bring out the DSC2, and that's the one I'm waiting for.

[Blizzard]

That was very tongue in cheek on my part. Yeah so HQPlayer is only good for everything except your DSD256 recordings. Niche product. PeterSt also prefers general CPU for use with NOS1a. What do you suggest for DSC1 or Lampizator? The demise of the x86 architecture has been imminent since the late 1970s. If those darn Intel engineers weren't so pesky, the world would be more architecturally pure.

 

The advantage I have is that I really don't care who "wins" as long as I get better sound. I am a consumer who would buy a card if a reasonable one existed which could output LVDS/I2S/DSD if it existed. Upcoming project is to build the DSC1 ... boards sitting on my desk for a few months now ... it takes DSD in (over "I2S" lines) which typically is from Amanero USB-I2S ... but could be from LVDS if that existed ... or could be from an inbuilt ARM board. If an ARM board existed that took fiberoptic Ethernet in (e.g. SFP) then I would already own it ... in any case there are options and no one right way of doing things.

 

In my experience the future is too hard to predict and I look for options that can be implemented now or in the near future.

 

My experience with HQ player is its great for 16/44 PCM. With 24/96 depends on the album, and anything higher in PCM or DSD I prefer bypassing the HQPlayer processing. But this is with my Resonessence Mirus. May be a different story with a lower end DAC.

 

Some of the upcoming Ethernet renderer boards would be a perfect companion for the DSC-1. LVDS wouldn't be required as you can interface direct via CMOS I2S, as well as slave DAC master clock to it.

[Blizzard]

I do not think it is fair to deal with this in that context. If one only needs to travel an inch or two, there is no reason to go USB and this topic is moot.

 

Exactly. This is why USB interface technology is pretty much Moot with today's technology.

 

Might as well talk about how to make CD transports better.

[4est]

If I understand it, it wouldn't take much to get balanced out using one "stereo" board per channel.

Get that DSC1 going already! Because you know as soon as you start Miska will bring out the DSC2, and that's the one I'm waiting for.

[4est]

You are acting as if this is all new, and everyone will be doing it that way next year. I doubt everyone will be ditching their expensive DACs next year because they do not have a built in streamer. Formats evolve slowly. There aren't even many LVDS items available to date, and people are still fixated on DSD.

Exactly. This is why USB interface technology is pretty much Moot with today's technology.

 

Might as well talk about how to make CD transports better.

[Blizzard]

You are acting as if this is all new, and everyone will be doing it that way next year. I doubt everyone will be ditching their expensive DACs next year because they do not have a built in streamer. Formats evolve slowly. There aren't even many LVDS items available to date, and people are still fixated on DSD.

 

Not saying everyone will instantly sell their USB based DAC's. What I'm saying is technology has evolved to the point where superior Ethernet based small form factor renderers are available.

 

Because of this they will be the #1 choice for DAC manufacturers who have the desire to build a competitive product.

 

There will be no need for the USB option when the Ethernet option is available.

 

General purpose computers will just be used to run GUI's or online upsamplers/modulators like hqplayer.

 

All of which take place prior to the chain of USB today as well as with the new Ethernet renderers.