Yes, @JonesySA MEMS I dont have any personal experience attempting to use for high quality audio, although I know it has been attempted. It would not be my first choice, I would stick to crystal oscillators, otherwise I would be more inclined to go for something like a programmable oscillator/PLL . Certainly in microphones and sensors, but I think at the levels we expect from audio DACs generally, I would personally avoid it still.
Maybe better to say the ASRC will attenuate jitter from the I2S clock?
The less I2S jitter, the more likely a lower stable setting is possible for the ESS register setting 'DPLL_Bandwidth.' ESS recommends to set that to the minimum stable value. A number of people including myself have reported improved SQ with lower values.
Yes, that's true.
Being a DPLL it will have much better rejection of HF jitter than LF jitter. So I'd recommend a canned oscillator with a lowish noise power supply, doesn't need to be expensive, Elvee's denoiser comes to mind for example.
Just using a separate canned clock will have much better performance than a XTAL working with the oscillator circuit inside a noisy DSP... because the power supply noise will be much lower even from a cheap regulator than inside the DSP, and noise from the immediate environment is also much lower inside the canned oscillator than on the DSP chip.
ES9023 has no register settings btw.
Being a DPLL it will have much better rejection of HF jitter than LF jitter. So I'd recommend a canned oscillator with a lowish noise power supply, doesn't need to be expensive, Elvee's denoiser comes to mind for example.
Just using a separate canned clock will have much better performance than a XTAL working with the oscillator circuit inside a noisy DSP... because the power supply noise will be much lower even from a cheap regulator than inside the DSP, and noise from the immediate environment is also much lower inside the canned oscillator than on the DSP chip.
ES9023 has no register settings btw.
Is the attenuation as a result of the upsampling or is there some other mechanism? I was unable to locate an explanation of the 'DPLL_Bandwith'; how is the lower value set - via clock frequency in synchronous operation?
In the interim I will connect via i2s using the add-on board that comes with the a31 to generate clocks. My understanding however is that there is only one master clock in i2s protocol yet the es9023 already has an oscillator onboard generating this for the DAC. I assume I simply do not connect MCLK between them and the DAC will operate independently and upsample the signal?
If I disabled the DACs clock and connect MCLK then I would feed the sources 11.2mhz (which seems like bad idea).
Probably a silly question but my go-to i2s DAC is the pcm5102 and I find the ESS datasheet really limited.
I thought mems outperformed traditional oscillators? Why are they less preferred?
> Is the attenuation as a result of the upsampling or is there some other mechanism?
It's done by the Asynchronous Sample Rate converter inside the chip. It's a different system than oversampling.
> I was unable to locate an explanation of the 'DPLL_Bandwith'; how is the lower value set - via clock frequency in synchronous operation?
You can't set it on ES9023 because it has no I2C ports to tweak its settings.
> I assume I simply do not connect MCLK between them and the DAC will operate independently and upsample the signal?
Correct
> If I disabled the DACs clock and connect MCLK then I would feed the sources 11.2mhz (which seems like bad idea).
If you disable ES9023 MCLK it will not function. It works in asynchronous mode (due to the ASRC) so its MCLK does not need to be synchronized with the MCLK that generates your I2S signal.
> I find the ESS datasheet really limited.
Seems to be a tradition with ESS 😀
It's done by the Asynchronous Sample Rate converter inside the chip. It's a different system than oversampling.
> I was unable to locate an explanation of the 'DPLL_Bandwith'; how is the lower value set - via clock frequency in synchronous operation?
You can't set it on ES9023 because it has no I2C ports to tweak its settings.
> I assume I simply do not connect MCLK between them and the DAC will operate independently and upsample the signal?
Correct
> If I disabled the DACs clock and connect MCLK then I would feed the sources 11.2mhz (which seems like bad idea).
If you disable ES9023 MCLK it will not function. It works in asynchronous mode (due to the ASRC) so its MCLK does not need to be synchronized with the MCLK that generates your I2S signal.
> I find the ESS datasheet really limited.
Seems to be a tradition with ESS 😀
Close-in phase noise tends to be worse than for a good crystal clock. Its something that matters for some applications such as radar. Whether or not it matters for audio is something people still argue about.
i was just going thought the datasheet for Linkplay A28 on Parts express and observed pin # 37 mentioned as reference clock, has anyone probed it's purpose?
REF_CLK0 sounds like it might be MCLK output from the board. A scope could be used to check on that. Assuming that's what it is, since they refer to it as a 'REF' (i.e. for reference) makes me think it should probably be buffered if used in subsequent circuitry as an MCLK signal.
Hi @Markw4
Unfortunately do not have access to a Scope right now but maybe over the weekend, I might be able to access one all provided my health improves.
But tried a crude test when playing music to the Linkplay A28, I connected my multimeter to the WS = 44.10kHZ, BCLK = 1.4112MHz, and the Data out Pin was 0.300 ~ 0.282MHz was the reading.
Attachments
Hi @Markw4
Well that ref clock pin #37 didn’t work, I will see by giving it a pull up later in the weekend.
I currently tried a 50MHz txco connected as seen in the image and I have audio
infant the linkplay device announces “connected to your Wi-Fi” 😬.
During airplay the volume changes are sluggish. I feel the bass is a bit low but that might be the ESS9023 dac may be.
It is Arylic’s DAC board.
Well that ref clock pin #37 didn’t work, I will see by giving it a pull up later in the weekend.
I currently tried a 50MHz txco connected as seen in the image and I have audio
infant the linkplay device announces “connected to your Wi-Fi” 😬.
During airplay the volume changes are sluggish. I feel the bass is a bit low but that might be the ESS9023 dac may be.
It is Arylic’s DAC board.
Attachments
MEMS has the advantage of being almost immune to physical vibration. Physical vibration is probable a bigger problem in a traditional oscillator than close in noise - which will be hard to preserve in a LF noisy environment - like reality 🙂
It would be interesting to test a really good MEMS osc. - they have come a long way since 10 years...
SiTime MEMS oscillators had some ambitions when I contacted them 5 years ago.
//
It would be interesting to test a really good MEMS osc. - they have come a long way since 10 years...
SiTime MEMS oscillators had some ambitions when I contacted them 5 years ago.
//
