Page 9 of 24

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 1:50 pm
by LowOrbit
[quote="Aleg

Mark

That sounds very promising.

Could you sort of separate the effect of the Raspberry and Buffalo DAC on sound quality?
Did you use the Buffalo before with another player?

Cheers

Aleg[/quote]

Hi Aleg/Chaps

I used the Buffalo with a WaveIO usb interface with my linear powered, passively cooled PC, when I was running either JPLAY, MQN or JLP. SO it was a straight substitution from PC/JLP to RPi/Volumio.

As per my previous note, the difference was significant and instantly heard. Dynamics, relaxed but detailed presentation, more like a very highend cd transport. I try to avoid hyperbole, but I am not going back to an intel pc for audio. I don't expect that the RPi is the endgame as there are compromises in the architecture, but Beaglebone Black (on order) looks promising for the longerterm.

I would not say the hassle was any worse than getting JLP running as a service.

Mark

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 1:53 pm
by LowOrbit
jkeny wrote:Don't mind my frustration & impatience, I'm just venting here rather than kicking the dog :)
It should be expected that there will be teething problems as the software is in beta & the hardware is relatively new.
Thanks Mark, for reporting from the front-line - did you try the linear supply yet - SQ improvement?
Hi John, Linux brings frustration out in funny ways. It has made me want to explode a penguin in the past.

I haven't got to the linear psu yet and am stuck overnight away from home so nothing doing until Friday at the earliest alas. I'll report asap!

Regards
Mark

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 2:34 pm
by Aleg
LowOrbit wrote:Hi Aleg/Chaps

I used the Buffalo with a WaveIO usb interface with my linear powered, passively cooled PC, when I was running either JPLAY, MQN or JLP. SO it was a straight substitution from PC/JLP to RPi/Volumio.

As per my previous note, the difference was significant and instantly heard. Dynamics, relaxed but detailed presentation, more like a very highend cd transport. I try to avoid hyperbole, but I am not going back to an intel pc for audio. I don't expect that the RPi is the endgame as there are compromises in the architecture, but Beaglebone Black (on order) looks promising for the longerterm.

I would not say the hassle was any worse than getting JLP running as a service.

Mark
Great

I have a BeagleBone Black on order as well, with a planned availability from 24th of February.
Also ordered a cheap US$20 I2S DAC from DIYINHK, just to be able to trial the I2S link.

Eager to get my hands on these two.

Cheers

Aleg

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 2:58 pm
by nige2000
is it one of these Aleg
ive been looking at them too

http://www.diyinhk.com/shop/audio-kits/ ... lator.html
http://www.diyinhk.com/shop/audio-kits/ ... lator.html

im a bit confused from what i read i think it may need a master clock
http://www.ebay.com/itm/330924147727

maybe ive the wrong end of the stick

diyinhk stuff takes a while to be delivered too

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 3:10 pm
by Aleg
nige2000 wrote:is it one of these Aleg
ive been looking at them too

http://www.diyinhk.com/shop/audio-kits/ ... lator.html
http://www.diyinhk.com/shop/audio-kits/ ... lator.html

im a bit confused from what i read i think it may need a master clock
http://www.ebay.com/itm/330924147727

maybe ive the wrong end of the stick

diyinhk stuff takes a while to be delivered too

Nigel

It is the second one http://www.diyinhk.com/shop/audio-kits/ ... lator.html

It is my first order from DIYINHK so have no experience yet, but they claim to ship within 2 days.

On reading the whole again I think it doesn't require an external clock because it is fed an MCLK from the I2S, but it say you can add an external clock
7) Extreme diyer can cut the external MCLK trace and uses the pre-allocated universal oscillator footprint space(CRYSTEK957/7050/5032/3225) to install a high quality 50M oscillator for asynchronous MCLK mode
Like to hear if you interpret this differently.

-------
Edit 1:
Looking at the pinout from the BeagleBone Black, I'm not sure yet if it provides an MCLK signal. They do have an oscilator on board, but have yet found a pinout for that signal.

Edit2: IIUC I2S requires 3 lines (not containing MCLK) but can contain an additional 2 lines (containg the MCLK)
The bus consists of at least three lines:
Bit clock line
Word clock line - also called word select (WS) or left right clock (LRCLK)
At least one multiplexed data line
It may also include the following lines:
Master clock (typical 256 x LRCLK)
A multiplexed data line for upload
Edit 3:
According to datasheet ESS9023 http://myl8test.files.wordpress.com/201 ... asheet.pdf
It requires a masterclock of >192 fs, so would need to be 36.864 MHz or better still 49.152 MHz for 192kHz support

Cheers

Aleg

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 4:40 pm
by jkeny
Guys,
I can tell you a bit about what's needed. The RPi doesn't output an MCLK signal so only DAC chips that have some other way of getting an MCLK signal will work.
The best bet would be to get one of these HiFiBerry DACs as it is already setup to work with the RPi http://www.hifiberry.com/dac. It uses the PCM5102 DAC chip - same as the one in the Ciunas but I provide MCLK to this chip synchronised to the other 3 I2S signals using one of two local oscillators. The other way of operating it is to allow it to generate it's own MCLK internally. The HiFiBerry board is setup to operate in this way.

The ES9023 board from DIYINHK will work with a mod (you will need to provide an external MCLK from a local oscillator), as it can work asynchronosly i.e MCLK does not have to be synchronised with the other I2S signals (All the ESS DAC chips, including the ES9023 has an ASRC in it which allows this - this is why Mark's Buffalo works with it). The only thing is that the ESS DAC chips sound a bit better when working synchronously i.e MCLK is in phase with the other I2S signals & you can't do this with the RPi. However, Mark's Buffalo would have worked with his WaveIO in this asynchronous way too (I presume) so his reports of the sound improvement with the RPi is encouraging.

So the RPi has limitations in what DAC chips it will work with using I2S i.e only those that don't need MCLK signal.
The Beaglebone Black (BBB) has got MCLK I2S output along with the other I2S signals, AFAIK.

It remains to be seen how clean/jitter free the MCLK signal from the BBB or any of these boards is. It certainly wouldn't be as clean as the clock from a local oscillator. Whether this affects the sound & by how much remains to be seen/heard.

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 5:17 pm
by Aleg
jkeny wrote:Guys,
I can tell you a bit about what's needed. The RPi doesn't output an MCLK signal so only DAC chips that have some other way of getting an MCLK signal will work.
The best bet would be to get one of these HiFiBerry DACs as it is already setup to work with the RPi http://www.hifiberry.com/dac. It uses the PCM5102 DAC chip - same as the one in the Ciunas but I provide MCLK to this chip synchronised to the other 3 I2S signals using one of two local oscillators. The other way of operating it is to allow it to generate it's own MCLK internally. The HiFiBerry board is setup to operate in this way.

The ES9023 board from DIYINHK will work with a mod (you will need to provide an external MCLK from a local oscillator), as it can work asynchronosly i.e MCLK does not have to be synchronised with the other I2S signals (All the ESS DAC chips, including the ES9023 has an ASRC in it which allows this - this is why Mark's Buffalo works with it). The only thing is that the ESS DAC chips sound a bit better when working synchronously i.e MCLK is in phase with the other I2S signals & you can't do this with the RPi. However, Mark's Buffalo would have worked with his WaveIO in this asynchronous way too (I presume) so his reports of the sound improvement with the RPi is encouraging.

So the RPi has limitations in what DAC chips it will work with using I2S i.e only those that don't need MCLK signal.
The Beaglebone Black (BBB) has got MCLK I2S output along with the other I2S signals, AFAIK.

It remains to be seen how clean/jitter free the MCLK signal from the BBB or any of these boards is. It certainly wouldn't be as clean as the clock from a local oscillator. Whether this affects the sound & by how much remains to be seen/heard.
John

Thanks for this clarification.

As I'm focussing on the BBB I might just be lucky to be able to grab an MCLK signal from that board.
I noticed the BBB has a 24.576MHz oscillator onboard, but could not yet identify the pinout for that signal.

It will never be my main DAC it is just for testing the I2Sconnection. If I would go for an I2S DAC I would go for higher specs and other/better IV-stages.
An external MCLK can always be added later if required for testing.

Cheers

Aleg

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 5:20 pm
by jkeny
Yes, Aleg
If there is no MCLK I2S signal the the ES9023 needs an external oscillator of 50Mhz or less (I think 50 MHz is max for ES9023 & 90 or 100MHz max for other ESS DAC chips). The oscillator speed doesn't need to be an integer multiple of fs so no need for 49.152 MHz - 50Mhz clock will do, in fact it may be better for it not to be an integer multiple?

Edit: Ah you posted concurrently.
Yes BBB is probably the simplest solution regarding MCLK but my jitter concerns might apply.
I'm not sure why it uses a 24.576Mhz - presumably it uses some internal PLLs to multiply this osc speed to the internal speed required by the ARM chip) - the 24.576MHz would only be suitable for one sample rate family - ideally you would need 2 audio clock sources for the I2S MCLK signal

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 5:43 pm
by Aleg
jkeny wrote:Yes, Aleg
If there is no MCLK I2S signal the the ES9023 needs an external oscillator of 50Mhz or less (I think 50 MHz is max for ES9023 & 90 or 100MHz max for other ESS DAC chips). The oscillator speed doesn't need to be an integer multiple of fs so no need for 49.152 MHz - 50Mhz clock will do, in fact it may be better for it not to be an integer multiple?

Edit: Ah you posted concurrently.
Yes BBB is probably the simplest solution regarding MCLK but my jitter concerns might apply.
I'm not sure why it uses a 24.576Mhz - presumably it uses some internal PLLs to multiply this osc speed to the internal speed required by the ARM chip) - the 24.576MHz would only be suitable for one sample rate family - ideally you would need 2 audio clock sources for the I2S MCLK signal
It uses 24.576Mhz because it is primarily present to create the HDMI audio signal.
With this oscillator frequency your limited to 48/96/192 range of sample frequencies according to the ES9023 datasheet.

I think the clock may be available on P9 - pin25 in mode 2.

Cheers

Aleg

Re: Tweaker's Rash

Posted: Wed Feb 19, 2014 7:37 pm
by jkeny
Aleg wrote: It uses 24.576Mhz because it is primarily present to create the HDMI audio signal.
With this oscillator frequency your limited to 48/96/192 range of sample frequencies according to the ES9023 datasheet.

I think the clock may be available on P9 - pin25 in mode 2.

Cheers

Aleg
Ah, yes forgot it had HDMI
Yes, that's the speed family supported by 24.576MHz (or a multiple of). 22.5792MHz (or a multiple of) is needed for 44.1/88.2/176/352 speed family
Sorry I can't help with the pinouts, ATM