DS and source clock


#1

A simple question: being the DS masterclock, is the source clock accuracy important or not?

Thanks


#2

In general no, the source clock accuracy isn’t terribly important. Ideally it should be within +/- 100ppm of nominal which is the Redbook standard and most everything is. But even going outside that isn’t really a problem.


#3

Thanks Ted.

My streamer is based on Raspberry Pi3 and an I2s over hdmi output module.

It works great, definitely the best source I’ve ever had, but I’m wondering if I can further improve it using a better clock, being the Raspberry clock not a very good one.


#4

If there is a crystal or oscillator module just for the audio, upgrading it might help. But I’m reasonably sure that they are just dividing down the system clock and upgrading the system clock won’t make much difference at all - accuracy isn’t very important and dividing a clock down will add enough jitter to swamp any clock improvements.


#5

This article may be worth reading: http://www.dimdim.gr/2014/12/the-rasberry-pi-audio-out-through-i2s/

Then in the comments they recommend this as an easy way to improve the Pi’s i2s http://support.hifiberry.com/hc/en-us/articles/205711451-DAC-Pro-connect-external-I2S-DACs


#6
vortecjr said This article may be worth reading: http://www.dimdim.gr/2014/12/the-rasberry-pi-audio-out-through-i2s/

Then in the comments they recommend this as an easy way to improve the Pi’s i2s http://support.hifiberry.com/hc/en-us/articles/205711451-DAC-Pro-connect-external-I2S-DACs

Thanks Jesus, for pointing! I have the Pi3/HifiBerry DAC+ Pro/Audio-GD I2S-to-HDMI package running now as a Roon Bridge. If it sounds better than my Pi2/Audio-GD module without the HifiBerry DAC? I have no idea. Too many variables for the moment. All I can say is it sounds great but that already did the Pi2/Audio-GD package with the piCorePlayer installed.

Thanks again!

Mario

pi3_hb_audiogd.jpg


#7
Ted Smith said In general no, the source clock accuracy isn't terribly important. Ideally it should be within +/- 100ppm of nominal which is the Redbook standard and most everything is. But even going outside that isn't really a problem.
Is the clock accuracy important with DSD128 ?

I read there is one school of thought that says the clock accuracy and jitters are the key to digital audio.


#8

I believe Ted was responding in the context of the Direct Stream DAC, which does not rely on the clock in the source (such as a transport) and does a very good job of eliminating any jitter from the source.


#9

stevem2 is correct. I was answering the original posters question.

Still there are three things that your question brings to mind:

As I tried to say above (and as the links which vortecjr explain) the amount of jitter Pi’s I2S implementation completely swamps any jitter/accuracy of it’s clock.

Clock frequency accuracy isn’t important in digital audio - it matters not a hoot if the clock is a little too fast or a little too slow. The phase noise of a clock matters quite a bit - i.e. how closely each cycle looks like the previous cycles. Still clocks that are well built tend to be more accurate as well as have less phase noise.

The higher the sample rate the more jitter matters. Somewhere between DSD128 and DSD256 higher clocks add more noise via jitter than they lower noise via the higher sample rate.


#10
Ted Smith said stevem2 is correct. I was answering the original posters question.

Still there are three things that your question brings to mind:

As I tried to say above (and as the links which vortecjr explain) the amount of jitter Pi’s I2S implementation completely swamps any jitter/accuracy of it’s clock.

Clock frequency accuracy isn’t important in digital audio - it matters not a hoot if the clock is a little too fast or a little too slow. The phase noise of a clock matters quite a bit - i.e. how closely each cycle looks like the previous cycles. Still clocks that are well built tend to be more accurate as well as have less phase noise.

The higher the sample rate the more jitter matters. Somewhere between DSD128 and DSD256 higher clocks add more noise via jitter than they lower noise via the higher sample rate.


I guess that is why you have maxed out the DS’ DSD at 128. I have read previous comments here that you are considering enabling higher DSD sample rates. Is that still on the cards?

Also, since the DS alsready converts all streams to DSD, is there much of a point or sonic benefits from using a software upsampler like HQ Player into the DS?


#11
yacheah said I guess that is why you have maxed out the DS' DSD at 128. I have read previous comments here that you are considering enabling higher DSD sample rates. Is that still on the cards?

Also, since the DS alsready converts all streams to DSD, is there much of a point or sonic benefits from using a software upsampler like HQ Player into the DS?

I haven't ruled out higher sample rates, but it isn't an immediate goal.

It’s a question of taste where to upsample: The DS has a very good PCM upsampling chain and a very good sigma delta modulator. (FWIW: Having an external device or external software convert PCM to DSD isn’t the most direct route since the DS will still run that DSD thru it’s own sigma delta modulator.) Still there are many upsamplers out there, many with their own sounds, as with all things audio you need to choose for yourself the sound you like the best. Personally I don’t like many of the software upsamplers - to me they seem to slightly muffle PCM and smear the sound stage a little.


#12

Hi Ted,

What’s your thought’s on a clock with phase noise like this for DSD usage?

Purestream-Clock.png


#13

That looks pretty good. The close in phase noise is great and tho some other oscillators out there have a lower asymptote, below a certain level it just doesn’t matter for audio so I think that would be a fine clock. Then the question is: Can they keep that low of phase noise to and thru the last reclocker and/or digital switch?


#14
Ted Smith said That looks pretty good. The close in phase noise is great and tho some other oscillators out there have a lower asymptote, below a certain level it just doesn't matter for audio so I think that would be a fine clock. Then the question is: Can they keep that low of phase noise to and thru the last reclocker and/or digital switch?
It's going in a pure DSD 256 only DAC using the AKM AK4490 chip in DSD bypass mode. There's only 1 clock for both the XMOS USB chip,and the AK4490 with this DAC.

The XMOS chip, clock, and DAC are only a few inches apart to there’s an extremely short signal path between the clock and the DAC chip.

No reclockers, and no PLL’s. The XMOS chip will generate the DSD 256 bit clock and LRCLK by simply using the rising and falling edges of this clock. This allows the 11.2896Mhz clock for DSD 256 to be generated without any PLL’s or clock splitting.

The AK4490 will basically just be a DSD 256 conduit only using the SCF to assist with the low pass filtering. The rest of the DSD filtering will be done with a second order discrete filter set @120Khz.

This DAC is designed to be used in conjunction with HQplayer upsampling to DSD 256 with their 7th order modulator. The combination of modulator, SCF in AK4490, and discrete filter will ensure the noise floor is down at under 130dB@ 120Khz.

So similar concept to the DS, only with a high powered Intel I7 quad core based SRC/SDM engine rather than FPGA.

There will be phase noise testing done at the clock inputs on the AK4490.


#15
Holzohr said
vortecjr said This article may be worth reading: http://www.dimdim.gr/2014/12/the-rasberry-pi-audio-out-through-i2s/

Then in the comments they recommend this as an easy way to improve the Pi’s i2s http://support.hifiberry.com/hc/en-us/articles/205711451-DAC-Pro-connect-external-I2S-DACs

Thanks Jesus, for pointing! I have the Pi3/HifiBerry DAC+ Pro/Audio-GD I2S-to-HDMI package running now as a Roon Bridge. If it sounds better than my Pi2/Audio-GD module without the HifiBerry DAC? I have no idea. Too many variables for the moment. All I can say is it sounds great but that already did the Pi2/Audio-GD package with the piCorePlayer installed.

Thanks again!

Mario

Hi Mario,

So no improvement with the DAC+ Pro for DSD128? What was the point then since it just copies the I2S input from the Pi?

I looked at the Audio-Gd module specs and they should be enough to transmit DSD128. Even if a relatively long signal path (looking at the insides of an Audio-Gd DAC, a quite long cabel is used internally) so I really think the Pi is missing power…

I’ll be trying the I2S output of an Odroid C2 soon, I’ll let you know how that went!

Charles


#16
Linvincible said

Hi Mario,

So no improvement with the DAC+ Pro for DSD128? What was the point then since it just copies the I2S input from the Pi?

I looked at the Audio-Gd module specs and they should be enough to transmit DSD128. Even if a relatively long signal path (looking at the insides of an Audio-Gd DAC, a quite long cabel is used internally) so I really think the Pi is missing power…

I’ll be trying the I2S output of an Odroid C2 soon, I’ll let you know how that went!

Charles


Hi Charles,

the benefit of the DAC+ Pro is it makes the I2S connection available/usable on Roon. Without the HifiBerry DAC there was no sound via I2S.
Unfortunately, no DSD128 by this solution. Maybe a Pi “4” will bring success in this case. 105_gif

Pls keep us informed about the Odroid C2.

Mario