Oscillators and their effects on sounds
I report my experiences on oscillator replacements on my SDTrans192.
SDTrans192 uses two crystal oscillators, 24.5760 MHz and 22.5792 MHz as master clocks for FPGA to generate I2S signals.
As the asynchronous FIFO buffer circuit for I2S generation designed by Chiaki must be very simple, quality of the master clock seems to be reflected to output sound quality very directly.
(I2S signals are completely synchronizing to one of the master clocks.)
1. Our standard oscillator components are;
Fox Electronics
Xpresso series FXO-HC735 Standard HCMOS Oscillators +-50PPM
FXO-HC735 | Information for FXO-HC735 Series from Fox Electronic
XPRESSO | XPRESSO Crystal Oscillator | Oscillator- Fox Electronics
It's a vanilla. Not so bad, however, we can expect more improvement
by replacing these oscillators.
2. I tried proprietary bipolar output crystal oscillators;
They were designed and produced by
a Japanese audio equipment vendor, FIDELIX
’´’áƒWƒbƒ^�[�…�»”*�UŠí(ƒNƒ�ƒbƒN)‚Ì�»�ì
(This page is only written in Japanese.)
Resulting sounds are less harsh, maintaining a good resolution and musically sounding.
3. The second replacement is;
CMOS output Crystal Clock Oscillator NZ2520SD
manufactured by a Japanese leading crystal maker, Nihon Dempa Kogyo Co., Ltd.
NZ2520SD(OA / AV)/Crystal Clock Oscillators/NDK
http://www.ndk.com/images/products/catalog/c_NZ2520SD_NSA3449E_e.pdf
The model has a remarkable low phase noise profile. According to their measurement result sheet, Phase Noise [dBc/Hz] 26MHz 3.3V for 5 samples
1Hz: max -76, min -81
10Hz: max -108, min -111
100Hz: max -136, min -138
1kHz: max -151, min -152
10kHz: max -156, min -157
The component is also available for personal users by way of a Japanese online web store, chip1stop. Unfortunately, they have no web page in English. However, you can contact to them through their web page.
chip1stop - One Stop Shopping Procurement Service for Electronic Components -
The price I bought was 1,500 JPY / piece ( 18 USD/ piece ).
My impression for their sounds is "high resolution tones, bass is enhanced accurately." I think this one has very high cost performance and its price
is very reasonable.
4. The last experience is of Mr. Ishida's Rb-based Oscillators;
Frequency Electronics Inc.
Rubidium Frequency Standard FE-5680A
Its technical manual says;
Phase Noise (fo=10 MHz)
@ 10 Hz: -100 dBc
@ 100Hz: -125 dBc
@ 1000 Hz: -145 dBc
I was very impressed by its high resolution tones when I listened to them for the first time. As I have not compared 3. and 4. at the same time,
I can't comment on comparison of 3. and 4.
Bunpei
I report my experiences on oscillator replacements on my SDTrans192.
SDTrans192 uses two crystal oscillators, 24.5760 MHz and 22.5792 MHz as master clocks for FPGA to generate I2S signals.
As the asynchronous FIFO buffer circuit for I2S generation designed by Chiaki must be very simple, quality of the master clock seems to be reflected to output sound quality very directly.
(I2S signals are completely synchronizing to one of the master clocks.)
1. Our standard oscillator components are;
Fox Electronics
Xpresso series FXO-HC735 Standard HCMOS Oscillators +-50PPM
FXO-HC735 | Information for FXO-HC735 Series from Fox Electronic
XPRESSO | XPRESSO Crystal Oscillator | Oscillator- Fox Electronics
It's a vanilla. Not so bad, however, we can expect more improvement
by replacing these oscillators.
2. I tried proprietary bipolar output crystal oscillators;
They were designed and produced by
a Japanese audio equipment vendor, FIDELIX
’´’áƒWƒbƒ^�[�…�»”*�UŠí(ƒNƒ�ƒbƒN)‚Ì�»�ì
(This page is only written in Japanese.)
Resulting sounds are less harsh, maintaining a good resolution and musically sounding.
3. The second replacement is;
CMOS output Crystal Clock Oscillator NZ2520SD
manufactured by a Japanese leading crystal maker, Nihon Dempa Kogyo Co., Ltd.
NZ2520SD(OA / AV)/Crystal Clock Oscillators/NDK
http://www.ndk.com/images/products/catalog/c_NZ2520SD_NSA3449E_e.pdf
The model has a remarkable low phase noise profile. According to their measurement result sheet, Phase Noise [dBc/Hz] 26MHz 3.3V for 5 samples
1Hz: max -76, min -81
10Hz: max -108, min -111
100Hz: max -136, min -138
1kHz: max -151, min -152
10kHz: max -156, min -157
The component is also available for personal users by way of a Japanese online web store, chip1stop. Unfortunately, they have no web page in English. However, you can contact to them through their web page.
chip1stop - One Stop Shopping Procurement Service for Electronic Components -
The price I bought was 1,500 JPY / piece ( 18 USD/ piece ).
My impression for their sounds is "high resolution tones, bass is enhanced accurately." I think this one has very high cost performance and its price
is very reasonable.
4. The last experience is of Mr. Ishida's Rb-based Oscillators;
Frequency Electronics Inc.
Rubidium Frequency Standard FE-5680A
Its technical manual says;
Phase Noise (fo=10 MHz)
@ 10 Hz: -100 dBc
@ 100Hz: -125 dBc
@ 1000 Hz: -145 dBc
I was very impressed by its high resolution tones when I listened to them for the first time. As I have not compared 3. and 4. at the same time,
I can't comment on comparison of 3. and 4.
Bunpei
They can accept orders for any frequencies in the specified range.
Its price is a standard price. However, their delivery period is three months in the worst case. Minimum quantity for one order is one piece!
As for the phase noise measurement technical data, I got it from a sales person of the company(The document is written in Japanese language).
Bunpei
We use C8051F361(Silicon Laboratories). Its clock rate is 24.5 MHz and running at quad speed mode. They say the actual performance is 100 MIPS. Chiaki's firmware is coded in assembly language so as to exploit the maximum performance.
We just use SPI. Chiaki found a "magic" method for reading and transferring data from SDHC memory card to FPGA(I2S signal generator).
Based on the two points explained above, Chiaki obtained enough data transfer rate for 352.8 kHz / 24 bit. However, to our regret, 384 kHz / 24 bit is just beyond the capability of SDTrans192.
Bunpei
Dear xrcd24bit,
I tried to send a diyAudio private message to you and failed to do that.
Your user account on the forum seems to reject a private message.
Please send me your ordinary e-mail. My address is
bunpei<at>ta2<dot>so-net<dot>ne<dot>jp
Best regards,
Bunpei
Problems on playing DXD audio files on Buffalo DACs
As some of you may already know it on Twisted Pear Audio - Buffalo II thread in this forum, playing DXD audio files ( 352.8 kHz / 24 bit PCM WAV files) on "Over sampling Mode" of ES9018 DAC chip by connecting standard Buffalo32s or Buffalo II DAC products to SDTrans192 through I2S interface has some problems, two type of noises and an unstable lock state.
If you replace an original 80 MHz processor clock with 100 MHz one, you can eliminate the problems practically and enjoy a quality of 352.8 kHz / 24 bit sources.
Though Chiaki and Bunpei are making efforts to find a route cause of the problems, we have not identified it and not get effective or possible solutions yet.
The two types of noises are;
A. Continuous white-noise like irregular noise at "STOP" state
(Bit Clock signal is output, but SDATA signal is kept low.)
ES9018 is loosing a lock.
This can be suppressed if you set certain register values for
"AUTOMUTE" setting.
B. A loud "click" noise at the timing of switching from or to 352.8 kHz / 24 bit
file.
You can't remove this noise easily by register setting.
When we get any new findings or corrective measures, we will report them on this thread.
Bunpei
As some of you may already know it on Twisted Pear Audio - Buffalo II thread in this forum, playing DXD audio files ( 352.8 kHz / 24 bit PCM WAV files) on "Over sampling Mode" of ES9018 DAC chip by connecting standard Buffalo32s or Buffalo II DAC products to SDTrans192 through I2S interface has some problems, two type of noises and an unstable lock state.
If you replace an original 80 MHz processor clock with 100 MHz one, you can eliminate the problems practically and enjoy a quality of 352.8 kHz / 24 bit sources.
Though Chiaki and Bunpei are making efforts to find a route cause of the problems, we have not identified it and not get effective or possible solutions yet.
The two types of noises are;
A. Continuous white-noise like irregular noise at "STOP" state
(Bit Clock signal is output, but SDATA signal is kept low.)
ES9018 is loosing a lock.
This can be suppressed if you set certain register values for
"AUTOMUTE" setting.
B. A loud "click" noise at the timing of switching from or to 352.8 kHz / 24 bit
file.
You can't remove this noise easily by register setting.
When we get any new findings or corrective measures, we will report them on this thread.
Bunpei
I have no noise or lock problems playing 352.8k 24bit or 32bit (bit length converted for testing).
And I tested just now with a 80 MHz clock instead of the 100 MHz clock I normally uses - no problems..
On my ES9018 there are separate JFET regulators / capacitor multipliers (20F) for all voltages including clock and 1.2 volt voltages.
On the Buffalo DACs with ES9018 I expect the uP controller are running on the 3.3 volt for the ES9018.
The ES9018 chips are extremely sensitive to all kind of electrical noises and have intermittent lock problems even at 44.1k/16bit.
I used combined series (JFET) and shunt (JFET) regulators, but due to intermittent lock problems I redesigned without the shunt regulators and increased the capacitor values by a factor of 10.
After this rebuild I have not experienced the intermittent lock problems and the 352.8k are only tested on this last release..
I have now finished a new PCB design and will start evaluating this new version as soon as the Arduino software with IR remote control works with the 4*20 lines display and hopefully I2C communication with the SDTrans for displaying the name of the played song and album name etc.
The SDTrans will also be controlled from the DAC/preamp - play, stop, next melody, next directory etc..
My SDTrans units (I have four of them) are upgraded / modified and Bunpei and Chiakis version 2.1 have included some of my upgrades.
My SDTrans have less noise and jitter than standard, and my ES9018 DAC are much improved compared to a Buffalo DAC - maybe the combination are resulting in the noise free and no lock problem playing.
And I tested just now with a 80 MHz clock instead of the 100 MHz clock I normally uses - no problems..
On my ES9018 there are separate JFET regulators / capacitor multipliers (20F) for all voltages including clock and 1.2 volt voltages.
On the Buffalo DACs with ES9018 I expect the uP controller are running on the 3.3 volt for the ES9018.
The ES9018 chips are extremely sensitive to all kind of electrical noises and have intermittent lock problems even at 44.1k/16bit.
I used combined series (JFET) and shunt (JFET) regulators, but due to intermittent lock problems I redesigned without the shunt regulators and increased the capacitor values by a factor of 10.
After this rebuild I have not experienced the intermittent lock problems and the 352.8k are only tested on this last release..
I have now finished a new PCB design and will start evaluating this new version as soon as the Arduino software with IR remote control works with the 4*20 lines display and hopefully I2C communication with the SDTrans for displaying the name of the played song and album name etc.
The SDTrans will also be controlled from the DAC/preamp - play, stop, next melody, next directory etc..
My SDTrans units (I have four of them) are upgraded / modified and Bunpei and Chiakis version 2.1 have included some of my upgrades.
My SDTrans have less noise and jitter than standard, and my ES9018 DAC are much improved compared to a Buffalo DAC - maybe the combination are resulting in the noise free and no lock problem playing.
Last edited:
Dear RayCtech,
with great excitemen I read that you are working on a remote controlable SDTrans192! Sonically the SDTrans is outstanding, and adding a remote control is just what it needs to reach Nirvana
Will you make the remote control design available to the public (I am very interested)?
Regards,
with great excitemen I read that you are working on a remote controlable SDTrans192! Sonically the SDTrans is outstanding, and adding a remote control is just what it needs to reach Nirvana
Will you make the remote control design available to the public (I am very interested)?
Regards,
Last edited:
To add only a remote control you can use the TinyIR2 from Tauntek or other similar solutions..
TinyIR2 Learning IR remote control receiver
You can then use any remote wand you already have,
and connect gnd, +3.3 volt and four wires to the buttons.
As I have not tested the TinyIR2 with the SDTrans I can hook up one during the weekend and verify that it both works as intended and do not cause any noise or other problems...
TinyIR2 Learning IR remote control receiver
You can then use any remote wand you already have,
and connect gnd, +3.3 volt and four wires to the buttons.
As I have not tested the TinyIR2 with the SDTrans I can hook up one during the weekend and verify that it both works as intended and do not cause any noise or other problems...
Dear RayCtech,
with great excitemen I read that you are working on a remote controlable SDTrans192! Sonically the SDTrans is outstanding, and adding a remote control is just what it needs to reach Nirvana
Will you make the remote control design available to the public (I am very interested)?
Regards,
Last edited:
Remote control
Dear RayCTech,
thank you for the tip! I am kinda analphabetic when it come to remote control design solutions
Yes, please hook one up and test it out. Would be much appreciated! Maybe you could take a photo of the setup that shows how you wired it.
I have a few remote controls from QLS Hifi that I used to use for the QA550
Regards,
Dear RayCTech,
thank you for the tip! I am kinda analphabetic when it come to remote control design solutions
Yes, please hook one up and test it out. Would be much appreciated! Maybe you could take a photo of the setup that shows how you wired it.
I have a few remote controls from QLS Hifi that I used to use for the QA550
An externally hosted image should be here but it was not working when we last tested it.
Regards,
Today Bunpei, Nakagawa (an engineer) and myself got together at Nakagawa's home in Tokyo to test the effects of different clocks used in the SDTrans192. We started with the original FOX clocks, and proceeded with the NDK clocks, as well as a clock built up by Nakagawa with bipolar transistors. The FOX clock set was quickly deemed to be the worst of the three alternatives with the sound being relatively diffuse, undefined, and lacking energy. The real contest was between the hand made Bipolar clocks (constructed on a PCB the size of a stamp). The Bipolar clocks made the sound warm and friendly. Quite listenable, but kind of soft, and still lacking in resolution and dynamics. In comparison the NDK clocks really cleared up the "view" and gave both definition (pinpointing of each image in the sound stage) and dynamics (both on a macro and micro level). We did feel that the Bipolar clocks also had a lot of merit (and we kept on going back and forth multiple times to exactly pinpoint the differences), but also since they are a work in progress, the conclusion was that the NDK clocks were the definite winners in this round. They really bring the SDTrans up to a higher level.
We used two alternative DACs for this listening: (1) A current generation BUF II with IVY 3 and standard, original Twisted Pear 80 MHz clock. (2) The second DAC used was the ESS Sabre 9018 Evaluation Kit with a Nakagawa-designed Bipolar 101MHz clock (that we find to be really excellent as a clock for the DAC).
The listening results of the 3 x alternative clock sets on the SDTrans192 was repeated with both DACs and identical ranking was found to be true with both DAC combinations. Each of the DACs had their own sonic merits, and we preferred the BUF II/IVY 3 in some respects, while the ESS Evaluation Kit with 101MHz clock was preferable in other areas (e.g. the BUF II/IVY 3 had a better defined, stronger bass, while the Evaluation Kit with the Bipolar 101MHz clock seemed to have a preferable upper range. We did not pick "a winner" here.
This time all of the source material was EAC ripped 44.1kHz files, and this test was not focused around high-sampled files.
We used two alternative DACs for this listening: (1) A current generation BUF II with IVY 3 and standard, original Twisted Pear 80 MHz clock. (2) The second DAC used was the ESS Sabre 9018 Evaluation Kit with a Nakagawa-designed Bipolar 101MHz clock (that we find to be really excellent as a clock for the DAC).
The listening results of the 3 x alternative clock sets on the SDTrans192 was repeated with both DACs and identical ranking was found to be true with both DAC combinations. Each of the DACs had their own sonic merits, and we preferred the BUF II/IVY 3 in some respects, while the ESS Evaluation Kit with 101MHz clock was preferable in other areas (e.g. the BUF II/IVY 3 had a better defined, stronger bass, while the Evaluation Kit with the Bipolar 101MHz clock seemed to have a preferable upper range. We did not pick "a winner" here.
This time all of the source material was EAC ripped 44.1kHz files, and this test was not focused around high-sampled files.
Last edited:
Dear elecon,
very interesting information! Thank you for reporting it here.
In post #143 Bunpei reports on listening qualities for a few clocks, too. Have you and or the team been able to correlate any clock specification to sonics?
Does the clock with lowest low freq phase noise sound the best?
BTW,
what clock freq did you use for the NDK clock?
Regards,
very interesting information! Thank you for reporting it here.
In post #143 Bunpei reports on listening qualities for a few clocks, too. Have you and or the team been able to correlate any clock specification to sonics?
Does the clock with lowest low freq phase noise sound the best?
BTW,
what clock freq did you use for the NDK clock?
Regards,
I will let Bunpei report regarding clock frequencies which are the two fixed frequencies used in the SDTrans192 as per its design by Chiaki. (I think you will find these elsewhere in the thread.) We did not look at the correlation between low freq phase noise vs. sound, and I am not aware of whether we have such data available. (Again I would prefer if Bunpei reply to such questions.)
I have removed both oscillators from the SDtransport192's PCB and now inject a 24.5760MHz clock signal (I upsample all files to 32bit/192kHz, best results in combination with my Buffalo) from my D-Labs Neutron star (the best clock I know of, NewClassD Neutron Star). I can confirm Bunpei's and elecon's findings - quite a deal more detail and transparency while at the same time the tonal balance and warmth has become much more "right" and life-like. Clock upgrade definitely recommended! 