i2s clock problem

[nikkov]

Hi,

I try working with i2s driver on last armbian (5.99 kernel 4.19.84) with nanopi neo and found trouble with I2S clock - it has very excessive jitter (see attached screen from my oscilloscope). I use small dts overlay for enable i2s and mainline i2s code without changes and with changes from codekipper, nanopi neo and neo 2 with same result, but old armbian image 5.65 (4.14.17) working without problem.

I also found a mention of a similar problem on forum volumio

 

sun8i-h3-i2s0.dts

/dts-v1/;
/plugin/;

/ {
	compatible = "allwinner,sun8i-h3";

 	fragment@0 {
 		target-path = "/";
 		__overlay__ {
			pcm5102a: pcm5102a {
			#sound-dai-cells = <0>;
			compatible = "ti,pcm5102a";
			pcm510x,format = "i2s";
			};
 		};
 	};

	fragment@1 {
		target = <&i2s0>;
		__overlay__ {
			status = "okay";
			pinctrl-0 = <&i2s0_pins>;
			sound-dai = <&pcm5102a>;
			pinctrl-names = "default";
		};
	};

	fragment@2 {
		target-path = "/";
		__overlay__ {
			sound_i2s {
				compatible = "simple-audio-card";
				simple-audio-card,name = "I2S-master";
				simple-audio-card,mclk-fs = <256>;
				simple-audio-card,format = "i2s";
		                status = "okay";

				simple-audio-card,cpu {
					sound-dai = <&i2s0>;
				};

				simple-audio-card,codec {
					sound-dai = <&pcm5102a>;
				};
			};
		};
	};
};

 

[nikkov]

I made some investigation and found that unstable phase is a consequence of sigma-delta modulation for audio PLL from this commit

Sigma-delta modulation make more exact frequency but it leads variable phase of clocks. This means that with the kernel> 4.15 some codecs will not work, as they will not be able to synchronize with the clocks.

My PCM5102A may work only with 48k, but not higher

[HEAudio]

Any fix for that?

I installed a rockpi S a week ago with new kernel with no problems.

[Christos]

On 11/22/2019 at 5:20 AM, nikkov said:

I made some investigation and found that unstable phase is a consequence of sigma-delta modulation for audio PLL from this commit

Sigma-delta modulation make more exact frequency but it leads variable phase of clocks. This means that with the kernel> 4.15 some codecs will not work, as they will not be able to synchronize with the clocks.

My PCM5102A may work only with 48k, but not higher

 

Hi @nikkov

 

Although quite more than a year old thread but it looks that I also stumbled on that problem too.

Terrible phase noise of the produced bitclock in all samplerates tested (48/96/192KHz). Attached the relevant oscilloscope screenshot.

It seems that this bitclock phase instability, jitter, creates problems in codec/DAC/ADC parts that have internal PLL and the outcome in my case is a bit of audible noise during play.

Tested the same board/hardware with a RPi and it works ok, but with OPiPC and latest Armbian with kernel 5.10.8 is having this audible output noise problem.

 

Did you manage to have a fix or possibly reverted the mentioned patch succesfully ?

(https://github.com/megous/linux/commit/417a54f606c4ccf75023da0d165febe1a30e2764#diff-874a1d4d688766992a150cf3b81a4d8c)

 

Christos