Rk3308--8 channels changed to dual channels + recording gain

Change to dual channel

modify dts file

Related path ：

• Y:\hxy\RK3308\sdk-1.5\kernel\arch\arm64\boot\dts\rockchip
• Y:\hxy\RK3308\sdk-1.5\kernel\Documentation\devicetree\bindings\sound
• Y:\hxy\RK3308\sdk-1.5\kernel\sound\soc\rockchip

Passing instructions grep -rin "rockchip,rk3308-multi-dais" Find related files

Related documents ：

• rk3308-evb-dmic-pdm-v13.dts
• rockchip,multidais.txt
• rockchip_multi_dais.c

Modify the following sections ：

	pdm_i2s_dais: pdm-i2s-dais {
    
		status = "okay";
		compatible = "rockchip,rk3308-multi-dais", "rockchip,multi-dais";
		dais = <&pdm_8ch>, <&i2s_8ch_2>;
-	    capture,channel-mapping = <6 2>;
+		capture,channel-mapping = <2 0>;	
		playback,channel-mapping = <0 2>;
		bitclock-inversion = <1 0>;
	};

hold <6 2> Change to <2 0>

In limine SDK When I first got it , Is maintaining 8 When compiling firmware during channel configuration , use arecord Instructions cannot be recorded ( After recording, it is found that the data capacity is very small ), And the reason is dts There's one in the file vccio The port is not configured properly ：

&io_domains {
    
	status = "okay";

	vccio0-supply = <&vcc_io>;
	vccio1-supply = <&vcc_io>;
-	vccio2-supply = <&vcc_io>;
+	vccio2-supply = <&vcc_1v8>;
	vccio3-supply = <&vcc_1v8>;
	vccio4-supply = <&vcc_1v8>;
	vccio5-supply = <&vcc_io>;
};

from RK_EVB_RK3308_DDR3P116SD4_V13_20181029.pdf You can see each... In the file vccio The role of ：

understand asound.conf The role of

route ：Y:\hxy\RK3308\sdk-1.5\buildroot\board\rockchip\rk3308\fs-overlay\etc

When I first looked for the path, I found the following 4 A path ：

1. Y:\hxy\RK3308\sdk-1.5\buildroot\output\rockchip_rk3308_release\target\etc
2. Y:\hxy\RK3308\sdk-1.5\buildroot\output\rockchip_rk3308_64_dueros\target\etc
3. Y:\hxy\RK3308\sdk-1.5\buildroot\board\rockchip\rk3308\fs-overlay\etc
4. Y:\hxy\RK3308\sdk-1.5\buildroot\board\minnowboard\fs-overlay-graphical\etc

Related documents ：asound.conf and Asound_Configuration_CN.pdf

Recording gain

Gain patch pack patch

route ：Y:\hxy\RK3308\sdk-1.5\buildroot\package\alsa-lib

file ：0005-add-gain-for-audio-input.patch, There are patches , This is actually out of date , It is required to be changed RK default pcm.c It has been changed , If according to patch If you change the method of the package , The following new ：

1. 1171 Yes static void add_gain_rk_16(void* pIn,int length, int channels)
2. 1205 Yes static void add_gain_rk_32(void* pIn,int length, int channels)
3. 2096 Yes add_gain_rk_16((void*)buffer, tmp * pcm->frame_bits / pcm->sample_bits, pcm->channels); function
4. 2103 Yes add_gain_rk_32((void*)buffer, tmp * pcm->frame_bits / pcm->sample_bits, pcm->channels); function

patch Format parsing ： Reference link 01 Reference link 02

patch Some contents of the package are parsed as follows ：

diff --git a/src/pcm/pcm.c b/src/pcm/pcm.c	
index 9c90afb..281fbe3 100644
--- a/src/pcm/pcm.c		## The original document ,a Is the directory corresponding to the original file 
+++ b/src/pcm/pcm.c		## Revised document ,b Is the corresponding directory , This file can be found in the lower directory 

Here is @@- Number of old file start lines , Number of old lines + Number of new file start lines , Number of new rows @@; Old files from 825 Yes void DCBDoing Start , common 6 That's ok ; The new document is also from 825 OK, let's start , common 23 That's ok ：

@@ -825,6 +825,23 @@ void DCBDoing(void* pIn, int length, int channels)
     }
 }
 #endif
+void add_gain(void* pIn, int length, int channels)
+{
    
+    int i = 0;
+    int16_t * pInBuf  =  (int16_t *)pIn;
+    char *alsa_gain_level = getenv("ALSA_LIB_ADD_GAIN");
+    if(alsa_gain_level) {
    
+        int level = atoi(alsa_gain_level);
+        //printf("gain level:%d\n",level);
+        for(i = 0; i < length; i ++ ) {
    
+            int curChannel = i % channels;
+            if(curChannel < channels - 2 ) {
    //audio loopback no need to add gain
+                (*pInBuf) <<= level;
+            }
+            pInBuf++;
+        }
+    }
+}

modify pcm.c

route ：Y:\hxy\RK3308\sdk-1.5\buildroot\output\rockchip_rk3308_release\build\alsa-lib-1.1.5\src\pcm

stay 8 When the sound track , The recorded sound is very low , After adding the recording gain patch , Enter commands on the console export ALSA_LIB_ADD_GAIN=4 To achieve recording gain ; When I just changed it to two channels , The above command input on the console is invalid , At this time, we have to be right pcm.c Make changes (notepad++ A quick way to jump to a line ：Ctrl+G)：

1.2090 That's ok ,snd_pcm_readi There is such a condition in the function ：

if(pcm->channels != 2 && tmp > 0){
    ...}

Change to the following ：

if(pcm->channels <= 2 && tmp > 0) {
    ...}

2. Add one more 2089 One of the lines printf The information is as follows ：

printf("dcb size=%d,tmp=%d,sam_bits=%d,framebits=%d,format=%d,channels=%d,\n",size,tmp,pcm->sample_bits,pcm->frame_bits,pcm->format,pcm->channels);

3. Comment out add_gain_rk_16 Functions and add_gain_rk_32 Judgment condition in function ：

if(curChannel < channels - 2 ) {
    ...}

Add the following statements respectively ：

printf("rk_16 channels:%d level:%d \n",channels,level);
printf("rk_32 channels:%d level:%d \n",channels,level);

1. After a change , Need to go to cd work/hxy/RK3308/sdk-1.5/buildroot/output/rockchip_rk3308_release/ perform make alsa-lib-rebuild command
2. I've tried in Y:\hxy\RK3308\sdk-1.5\buildroot\output\rockchip_rk3308_release\build\alsa-lib-1.1.5\src\pcm Proceed under make, But it didn't work

mic Hardware circuit diagram

Related documents ：EVB_CN.pdf
mic specifications ：MSM261D4030H1CPM

6+0+2 Microphone array scheme description

Circular microphone array , For different microphone positions , Different channel combinations can be realized , For example, the scheme in the following figure , Arranged equidistant on the circumference 6 A microphone

In this microphone array scheme , Circular array microphone can also be designed for voice wake-up function , So there is no need for a separate wake-up microphone . meanwhile , For better echo cancellation , On the left and right channel speaker network , It is provided by the master controller ADC Recover the audio signal at the output end , This mining method can be closest to the actual effect of loudspeaker playing .

therefore 6+0+2 Microphone array means ：6 A circular microphone (capture)、0 A wake-up Mike 、2 Road mining (playback) sound signal .

PDM Schematic diagram of microphone

In circuit , Every time 2 One number corresponds to one PDM_SDI The signal , Here's the picture , therefore 6 A wheat needs to be used 3 individual SDI The signal ：

Connecting seat

FPC：Flexible Printed Circuit, Flexible printed circuit
The microphone array daughter board passes through 30Pin_0.5mm Spaced FPC Line and RK3308_EVB Motherboard connection , Its interface is defined as follows ：

PDM The microphone corresponds to PCB The plate screen printing is : RK_EVB_DMIC-PDM_6C70_V11_20180427J

Console console test

sound recording arecord test

1. First use the following 3 An instruction ：

• export ALSA_LIB_ADD_GAIN=4
• arecord -Dhw:0,0 -c 2 -r 8000 -f S16_LE /data/test01.wav
• aplay /data/test01.wav

Use the following printf Out of log The information is as follows （ It will be commented out in the source code later ）：

rk_16 channels:2 level:3 
dcb size=1000,tmp=1000,sam_bits=16,framebits=32,format=2,channels=2,

Changed the file system type test

2. If the file system type is changed to ext2 Type words , It can lead to aplay Can't play audio ,arecord Can record （ Yes push Come out and audition ）

dts Test of parameter modification significance

problem 1：capture,channel-mapping = <2 0>; Does it mean that only 6 individual mic Medium 2 individual , Which two ？

answer ： Through physical tests , Touching on the wheat will make a harsh sound , So we can judge it is mic1 and mic2.

problem 2： It was used mic1 and mic2, Which is the actual recording taken mic The data of ？ Which is mic Which one is louder ？ Still together ？ The software here chooses it by itself ？

answer ： Leave through the sound source mic Play far and near , Listen to the difference between the left and right speakers , If it is synthetic , The sound from those two speakers should be the same , Not really , It is mic1 and mic2 Corresponding to the left and right channels of the horn .

problem 3： The manual says PDM Support 8 What does road recording mean ？ Because there is only 6 A wheat , There is only 6 The way is right , What are the other two routes for ？

answer ： It is possible that the other two routes are simulation mic, Or other input sources , Like headphones .