MicroSD Memory Card Transport Project

Synchronous I2S clocking scheme for ES9018

In his new BBS article, Chiaki gave more detailed explanation for his new clocking scheme for ES9018. I'd like to post the content here in English with my own words.

After I tried the "synchronous MCLK" method with Buffalo II, I have lost my will to restore the conventional scheme. The reason why is that music from the new one is far better than that from the conventional one.

In the "synchronous MCLK" method, SDTrans generates 90.3168 MHz(44.1 kHz x 2048) or 98.3040 MHz(48 kHz x 2048) clock signal as MCLK and the MCLK is directly connected to XI pin of ES9018 chip. This means we do use no oscillator on the Buffalo II board.
This clocking scheme is very common for such ordinary DAC chips as WM8741, PCM1795, AK4399. The only difference is that ES9018 requires a very high frequency for enabling OSF mode on fs 352.8 kHz or 384 kHz.

In the new clocking scheme of SDTrans, the 90.3168 or 98.3040 MHz clock is obtained by multiplying (4 x) the original output of each NDK oscillator of 22.5792 MHz or 24.576 MHz. The multiplier is implemented applying a DPLL circuit on FPGA.
Chiaki will adopt the combination of 90.3168 and 98.3040 MHz crystal oscillators and a divider in the near future instead of the DPLL-based multiplier. The divider approach may attain less phase noises for MCLK.
Anyway, in spite of the use of DPLL on SDTrans, I am fully satisfied with the resulting music on ES9018.

Bunpei
 
In his new BBS article, Chiaki gave more detailed explanation for his new clocking scheme for ES9018. I'd like to post the content here in English with my own words.

After I tried the "synchronous MCLK" method with Buffalo II, I have lost my will to restore the conventional scheme. The reason why is that music from the new one is far better than that from the conventional one.

In the "synchronous MCLK" method, SDTrans generates 90.3168 MHz(44.1 kHz x 2048) or 98.3040 MHz(48 kHz x 2048) clock signal as MCLK and the MCLK is directly connected to XI pin of ES9018 chip. This means we do use no oscillator on the Buffalo II board.
This clocking scheme is very common for such ordinary DAC chips as WM8741, PCM1795, AK4399. The only difference is that ES9018 requires a very high frequency for enabling OSF mode on fs 352.8 kHz or 384 kHz.

In the new clocking scheme of SDTrans, the 90.3168 or 98.3040 MHz clock is obtained by multiplying (4 x) the original output of each NDK oscillator of 22.5792 MHz or 24.576 MHz. The multiplier is implemented applying a DPLL circuit on FPGA.
Chiaki will adopt the combination of 90.3168 and 98.3040 MHz crystal oscillators and a divider in the near future instead of the DPLL-based multiplier. The divider approach may attain less phase noises for MCLK.
Anyway, in spite of the use of DPLL on SDTrans, I am fully satisfied with the resulting music on ES9018.

Bunpei

Good news Bunpei!

I wrote a long time ago that I used clocks with the 90.3168 and 98.3040 MHz frequencies (a huge upgrade compared to 80 or 100MHz clocks).
Maybe that have been one of the reasons I have both had good audio quality and no problems with playing 352.8k and 384k from SDTrans or USB to I2S units.
I am using 50 to 60 cm cables for the I2S without any degradations..
But I am using a different implementation of the I2S inputs than the Buffalos ;)
 
I wrote a long time ago that I used clocks with the 90.3168 and 98.3040 MHz frequencies (a huge upgrade compared to 80 or 100MHz clocks).
Maybe that have been one of the reasons I have both had good audio quality and no problems with playing 352.8k and 384k from SDTrans or USB to I2S units.

Strictly speaking, I think it's better for you to return the clock signal from the ES9018 dac side to the SDTrans side by dividing by 4 so that the configuration must stay in "synchronous". In other words, "one master clock in a system".

Anyway, I think it's worth exploring a synchronous configuration on ES9018 more.
Why don't you apply synchronous MCLK of your SDTrans192 Rev. 2.1 or ExaUSB to ES9018 masterclock input? As far as OSF mode is OFF, it may run very well. It might be better than asynchronous configuration.
In your case, you can control both physical wiring around oscillators on ES9018 dac board and register setting on ES9018 as you intend to do.
 
Strictly speaking, I think it's better for you to return the clock signal from the ES9018 dac side to the SDTrans side by dividing by 4 so that the configuration must stay in "synchronous". In other words, "one master clock in a system".

Anyway, I think it's worth exploring a synchronous configuration on ES9018 more.
Why don't you apply synchronous MCLK of your SDTrans192 Rev. 2.1 or ExaUSB to ES9018 masterclock input? As far as OSF mode is OFF, it may run very well. It might be better than asynchronous configuration.
In your case, you can control both physical wiring around oscillators on ES9018 dac board and register setting on ES9018 as you intend to do.

I do not think that matters much due to the clocks will be out of sync if you divide the MCLK by four (some hardware) and the length of the wires when it is two separate PCB boards...

As it is now in my system I use special built 0.5ps jitter clocks in my DAC, the SDTrans, the exaU2I and other USB -> I2S devices I use.
The clocks are also made with a improved duty cycle, ppm etc..
I expect a 5% improvement in the duty cycles and low jitter are more important for the end result than the divide and long delay length method...
 
As it is now in my system I use special built 0.5ps jitter clocks in my DAC, the SDTrans, the exaU2I and other USB -> I2S devices I use.

You have very good clock devices! Are they for 90 MHz range or 20 MHz range or both? How did you measure or confirm the value 0.5ps by using what kind of measuring instrument in your running environment?

The better clocks you have in transport devices, the better results you will have in "synchronous configuration" of I2S for ES9018.
 
Great SDtrans-Buffalo II "in-sync" configuration

When I'd got Buffalo II, SDtrans, SunRing PPS capacitor and so on, each time I felt sonic improvement.
I had enjoyed wav files playback very much with these, although still prefered vinyls.

Syncronous mode gives furthermore, and I 'm astonished at that this configuration shows some more
informations previously missed, and I think these more detailed informations help listener's imagination enough
to consider the sounds as "more 3D-ish" "Lifelike" with well-mastered recordings even when they are redbook 16bit/44.1khz.

Now I can enjoy digital playback as much as vinyls. I 'm disappointed with the sonics only when playing poorly mastered ones.


I thank Bunpei san, Chiaki san and Twisted Pear for sharing their tireless effort for digital audio.

Cheers
 
Syncronous mode

Interesting news!

How do you connect the SDTrans clock to the Buffalo2?
A photograph perhaps?



/S

Syncronous mode gives furthermore, and I 'm astonished at that this configuration shows some more
informations previously missed, and I think these more detailed informations help listener's imagination enough
to consider the sounds as "more 3D-ish" "Lifelike" with well-mastered recordings even when they are redbook 16bit/44.1khz.

Now I can enjoy digital playback as much as vinyls.
Cheers
 
>How do you connect the SDTrans clock to the Buffalo2?
>A photograph perhaps?

First of all, you must remove Crystek TCXO from Buffalo II board. You would be unable to use BII with other transports by doing this.

Then SDtrans' special EEPROM -Chiaki san kindly sent to me- for this purpose enables you to feed 90.3168mhz or 98.3040mhz to BII' s XI pin
from SDtrans' MCLK pin. You also need moving SMD "JP5" to "JP6"

Another way is using external master clock to feed both SDtrans and BII.

Sorry for no pictures because I don't have decent digital cam :eek:


Bunpei might explain better.


Regards.
 
To wktk_smile.

I appreciated your exact explanation very much!

This is a picture of my attempt.
(The AVCC regulator sub-board is detached in order to take this picture. The board must be set as usual.)
 

Attachments

  • SN3G0046.JPG
    SN3G0046.JPG
    52.1 KB · Views: 705
Chiaki and Bunpei would like to recommend ES9018 DAC chip users try "synchronous master clocking" scheme on the chip.

In order to try this, the first thing you need to do is removing (temporarily) an original clock device on your DAC board.

The second step is to connect MCLK output pin of I2S connector of SDTrans to ES9018 XI pin.
There are two ways at this stage. The easier way is to provide 22.5792 or 24.576 MHz MCLK. In this case, you need to set "OSF =OFF" on ES9018 so as to maintain a stable lock on ES9018 from 44.1 to 384 kHz. You need neither EEPROM upgrade nor hardware jumper change. The users of SDTrans192 Rev. 2.1 can also try this.

The another way is to provide 90.3168 or 98.304MHz MCLK. In this case, you can set "OSF=ON", "DPLL BW=LOWEST" on ES9018 and have a stable lock from 44.1 to 384 kHz. The high frequency MCLK is generated in FPGA and you need to have Mark 17 EEPROM upgrade. You also need to moving the jumper SMD chip on JP6 position to JP5. SDTrans192 Rev. 2.1 is not compatible this setting. My opinion is the setting does not necessarily bring the best result. I rather favor "OSF=OFF" mode.

Apart from "synchronous clocking", you might be interested in 384 kHz/32 bit play. To achieve this, you need to have Mark 16 or Mark 17 EEPROM upgrade.

Please contact to Bunpei to get Mark 16 or Mark 17 EEPROM upgrade.
 
Last edited:
Effect of Cornell Dubilier Electronics (CDE) Film Capacitor

Some of you may have known very well the great effect of Cornell Dubilier's giant film capacitors.

I recently added the big film capacitor in the picture below to the +5V power line of my SDTrans384 and Buffalo II.
The effect is amazing. More power, strength in all ranges maintaining purity and resolutions.

In my case, 944 type, 100 microF 1200V DC.
The cap is made by Cornell Dubilier Electronics (CDE).
Digi-Key - 338-1921-ND (Manufacturer - 944U101K122ACM)
(The largest capacitance available is 220 microF/800V DC.)

I'd like to recommend you try one!
 

Attachments

  • SN3G0048.JPG
    SN3G0048.JPG
    47.9 KB · Views: 1,991
I am a bit surprised that you seem to use a standard laboratory power supply, but I am not at all surprised that you will get better sonics if you add a large PP cap to this supply. I would ague that lab PS are not at all suited for audio use.

I would never use a laboratory power supply for audio equipment. That will for sure not give optimal sonics from the digital or analog sections of audio gear, IMO.

Personally, I use a dedicated transformer for my Buffalo2, and two separate transformers for the analog IV sections (dual mono) PlacidBPs. All my Placids have modified rectifcation sections, and added decoupling polystyrene caps on all outputs and inputs.

And my SDTrans192 has its own linear regulator supplied from its own transformer.



Some of you may have known very well the great effect of Cornell Dubilier's giant film capacitors.

I recently added the big film capacitor in the picture below to the +5V power line of my SDTrans384 and Buffalo II.
The effect is amazing. More power, strength in all ranges maintaining purity and resolutions.

In my case, 944 type, 100 microF 1200V DC.
The cap is made by Cornell Dubilier Electronics (CDE).
Digi-Key - 338-1921-ND (Manufacturer - 944U101K122ACM)
(The largest capacitance available is 220 microF/800V DC.)

I'd like to recommend you try one!
 
Last edited:
I agree.
I have the tridents and they make a good sonic improvment with many improvements, for ex the sound stage and a more organic sound. For the price the Tridents are a bargain sonically.




Hi Bunpei,
I noticed that you do not have the trident regulators for your buffalo II. While that cap no doubt improves things (for $60 it had better!), I'd be willing to bet the tridents would be even more of a benefit.
 
Dear staccatiss and alazira,

Thank you very much for your comments and recommendation.

It's true that a versatile laboratory power supply is not one of the best power supplies for audio use. However, it's also true that some of them provide a certain quality level.
In this case with the giant film capacitor of CDE, even the usual laboratory power supply showed a very good performance. It was a surprising change for me. That's why I posted the previous recommendation.

I could not buy TPA Trident regulators because they had been sold out for a long while. Now, they are on sale again. I think it's time to buy.
I appreciated much your recommendation.

Bunpei
 
Chiaki is now suspending a distribution of his SDTrans384 board. Though he have no fixed plan for future as of now, he may restart his activity after autumn. He has some ideas for improvements. When some are proved to be effective enough, he will release a new revision.