I need to use the demos with the internal microphone (M1) and speaker (SPK1) on the DM365 EVM, instead of the MIC input and LINE_OUT output.
I couldn't find this subject in the documentation.
I'm using a DM365 on a DM36X EVM REV E by Spectrum Digital.
Thanks.
Hello!
Let me to introduce the patch that adds VoiceCodec into the DVSDK4.2. If you have DM365EVM, please test it. I never knew how to do patches, and doing it the first time in my life.
I used the command 'diff' to create this patch.
Download the file by clicking here <7444.Add-VoiceCodec.patch.zip>
Unpack the archive. You will see a file 'Add-VoiceCodec.patch'.
Linux sources are located in $DVSDK/psp/linux-2.6.32.17-psp03.01.01.39
Make a backup copy of this catalog!
For example, copy the contents of the catalog
'linux-2.6.32.17-psp03.01.01.39'
into
'backup_linux-2.6.32.17-psp03.01.01.39'.
This patch modifies the contents of some files only within the Linux sources catalog.
If something goes wrong, simply delete 'linux-2.6.32.17-psp03.01.01.39', and then rename 'backup_linux-2.6.32.17-psp03.01.01.39' back to 'linux-2.6.32.17-psp03 .01.01.39 ', and you'll bring everything back as it was.
WARNING! This patch assumes that you have a new installation of DVSDK4.2 and you did not made any changes to these files: 'dm365.c', 'dm365.h', 'mux.h' and 'board-dm365-evm.c'. Otherwise, this patch may or may not apply.
Make a copy of the file '/etc/asound.state' from targetfs (be careful, from targetfs, and not from the host).
Put the file 'Add-VoiceCodec.patch' in catalog 'linux-2.6.32.17-psp03.01.01.39'.
Using the terminal go to that catalog and run:
$ patch -p1 -i Add-VoiceCodec.patch
In normal output should look as follows:
patching file arch/arm/mach-davinci/board-dm365-evm.c
patching file arch/arm/mach-davinci/dm365.c
patching file arch/arm/mach-davinci/include/mach/dm365.h
patching file arch/arm/mach-davinci/include/mach/mux.h
patching file drivers/mfd/davinci_voicecodec.c
patching file drivers/mfd/Kconfig
patching file drivers/mfd/Makefile
patching file include/linux/mfd/davinci_voicecodec.h
patching file sound/soc/codecs/cq93vc.c
patching file sound/soc/codecs/cq93vc.h
patching file sound/soc/codecs/Kconfig
patching file sound/soc/codecs/Makefile
patching file sound/soc/davinci/davinci-evm.c
patching file sound/soc/davinci/davinci-vcif.c
patching file sound/soc/davinci/davinci-vcif.h
patching file sound/soc/davinci/Kconfig
patching file sound/soc/davinci/Makefile
If the output looks different and there are errors, sorry. Restore the original catalog 'linux-2.6.32.17-psp03.01.01.39' (see above).
Device Drivers --->
...
<*> Sound card support --->
...
<*> Advanced Linux Sound Architecture --->
...
<*> ALSA for SoC audio support --->
--- ALSA for SoC audio support
<*> SoC Audio for the TI DAVINCI chip
<*> SoC Audio support for DaVinci DM6446, DM355 or DM365 EVM
DM365 codec select (Audio Codec - AIC3101) --->
< > Build all ASoC CODEC drivers
In paragraph 'DM365 codec select' you can switch the codec.
( ) Audio Codec - AIC3101
(X) Voice Codec — CQ93VC
You will return to the previous point and see that there is a new opportunity to choose the type of VoiceCodec as 8kHz or 16kHz.
Put a checkmark in the paragraph 'Show clocks at Linux startup'.
The screen should look like this:
--- ALSA for SoC audio support
<*> SoC Audio for the TI DAVINCI chip
<*> SoC Audio support for DaVinci DM6446, DM355 or DM365 EVM
DM365 codec select (Voice Codec - CQ93VC) --->
Voice Codec Sampling Frequency (Fs=16kHz) --->
[*] Show clocks at Linux startup (Debug feature)
< > Build all ASoC CODEC drivers
At the beginning of the Linux boot messages you see this:
*** VOICECODEC CONFIGURATION is 16 kHz
*** Board Clocks:
* PERI_CLKCTL=0x243F027C
* CLOCKOUT0EN is 0 = en
* CLOCKOUT1EN is 0 = en
* CLOCKOUT2EN is 1 = dis
* DIV1 is 15 = 16
* DIV2 is 4 = 5
* DIV3 is 63 = 64
* HDVICPCLKS is 1 = PLLC2SYSCLK2
* DDRCLKS is 0 = PLLC1SYSCLK7
* KEYSCLKS is 0 = RTCXI (MXI)
* ARMCLKS is 1 = PLLC2SYSCLK2
* PRTCCLKS is 0 = RTCXI (OSC)
* CLKIN = 24000000 Hz
* PLL1PreDiv is 7 = 8
* PLL1Mul is 81 = 162
* PLL1PostDiv is 0 = 1 -- en
* PLL1 CLK = 486000000 Hz -- en
* PLL1DIV1 is 0 = 1 -- en
* PLL1DIV2 is 1 = 2 -- en
* PLL1DIV3 is 1 = 2 -- en
* PLL1DIV4 is 3 = 4 -- en
* PLL1DIV5 is 1 = 2 -- en
* PLL1DIV6 is 17 = 18 -- en
* PLL1DIV7 is 0 = 1 -- en
* PLL1DIV8 is 3 = 4 -- en
* PLL1DIV9 is 1 = 2 -- en
* PLL2PreDiv is 7 = 8
* PLL2Mul is 99 = 198
* PLL2PostDiv is 0 = 1 -- en
* PLL2 CLK = 594000000 Hz -- en
* PLL2DIV1 is 0 = 1 -- en
* PLL2DIV2 is 1 = 2 -- en
* PLL2DIV3 is 0 = 1 -- en
* PLL2DIV4 is 28 = 29 -- en
* PLL2DIV5 is 7 = 8 -- en
* ARM CORE CLK = 297000000 Hz
* HDVICP CLK = 297000000 Hz
* DDR CLK supply for PHY = 486000000 Hz ; Real DDR clk = 243000000 Hz
* VoiceCodec CLK = 4096551 Hz ; Sample Freq = 16002 Hz
and this
asoc: CQ93VC <-> davinci-vcif mapping ok
ALSA device list:
#0: DaVinci DM365 EVM (CQ93VC)
Remove the old (AIC) configuration
root@dm365-evm:~# rm /etc/asound.state
Set the microphone sensitivity (from 0 to 3)
root@dm365-evm:~# amixer sset 'PGA' 2
Set the speaker volume (from 0 to 63 or from 0% to 100%)
root@dm365-evm:~# amixer sset 'Mono DAC' 50%
root@dm365-evm:~# alsactl store EVM
SPK1 Improvements
Unfortunately, the DM365EVM has a bad speaker (SPK1). You'll hear a very quiet sound from it. I can recommend (for your own risk) you to unsolder and unmount this speaker, and you can mount and solder the 3.5mm headphone jack instead the speaker. See photo.
My headphones have a resistance of 32 Ohms, and I made matching this resistance. Keep in mind that VoiceCodec has one channel, and you have to do a mono headset. Also, I recommend making a LP filter, for example, to cut off all frequencies above 3 kHz. You can carefully solder necessary parts to the back of the board. See photo.
Much easier to hear the sound through headphones!
As a result, I hear a loud sound in the headphones if I set 50% of the volume.
(amixer sset 'Mono DAC' 50%)
MIC1 Improvements
Much worse is the case with the microphone! I see that the problem apparently is that the impedance of the microphone is not match with the impedance of VoiceCodec. As a result, the microphone has a very weak sensitivity. You will have bent to it, literally screaming right into it.
Again, I can recommend to solder a normal and high quality electret microphone instead that there is. See photo.
You must connect a microphone through the BUFFER! It is necessary to match the output and input impedances.
As a result, I can record a loud sound using a value of 2 (range from 0 to 3) in the settings of the microphone.
(amixer sset 'PGA' 2)
Record WAV-file using a microphone (duration 10s)
root@dm365-evm:~# arecord -t wav -f S16_LE -r 16000 -c 1 -d 10 test16kHz.wav
Recording WAVE 'test16kHz.wav' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono
WAV-file playback
root@dm365-evm:~# aplay test16kHz.wav
Playing WAVE 'test16kHz.wav' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono
Soundtest
gst-launch -v audiotestsrc \
! 'audio/x-raw-int,width=16,depth =16,rate=16000,channels=1' \
! audioconvert \
! alsasink
Loopback
gst-launch -v alsasrc \
! 'audio/x-raw-int,width=16,depth=16,rate=16000,channels=1' \
! alsasink sync=false
Record AAC (encoder need 2 channels)
gst-launch -v alsasrc \
! 'audio/x-raw-int,width=16,depth=16,rate=16000,channels=2' \
! TIAudenc1 codecName=aaclcenc engineName=codecServer \
! filesink location=test.aac
Playback AAC (input file must be 16kHz, 2channels)
gst-launch -v filesrc location=test.aac \
! TIAuddec1 codecName=aacdec engineName=codecServer \
! alsasink sync=false
Important note: AAC-encoder and decoder operates with a sampling rate of 16kHz and above. You can not use 8 kHz mode of VoiceCodec to encode/decode AAC.
If you have just installed DVSDK, (you never rebuild it), then when you run 'encode-a test.aac', you can see 'failed to create audio device'.
Try to rebuild dvsdk-demos! ('make demos_clean', 'make demos', 'make demos_install')
./encode -a test.aac -p 16000 -t 10
The file has a duration of 5 seconds, and the sound is twice as fast (if you listen the file on the host).
Need something to fix in the code.
Open 'audio.c' from dvsdk-demos_xxx/dm365/encode.
Look for this one block within the code.
/* Use supplied params if any, otherwise use defaults */
params = envp->params ? envp->params : &defaultParams;
dynParams = envp->dynParams ? envp->dynParams : &defaultDynParams;
params->sampleRate = dynParams->sampleRate = envp->sampleRate;
params->bitRate = dynParams->bitRate = envp->soundBitRate;
Just add those two lines after this code.
params->channelMode = IAUDIO_1_0;
dynParams->channelMode = IAUDIO_1_0;
This is because the VoiceCodec is a mono-device.
Rebuild encode demo.
Try again.
./encode -a test.aac -p 16000 -t 10
It works!!!
Please test it ALL!
Help!
Something strange is happening with VoiceCodec ((((
In short, there are no problems if you capture sound only. But the strange artifact begins to appear after a while, if capture audio along with video together.
Try this
./encode -a test.aac -r 16000 -t 3600
All was good. No artifacts there for an hour recording!!! The same is true if you use 'arecord' or whatever when there is no video capture but only sound capture is.
But if you try this
./encode -a test.aac -r 16000 -v test.264 -y 2 -t 600
(In my case I use the PAL-cam through TVP5146 (composite), so I use the '-y 2' option.)
Quite possibly the sound at the end of AAC-file will be slightly different than in the beginning.
Please check it!
The strange effect is not immediate. It can start after 2 minutes of recording time, sometimes a little later, like after 7 or 9 minutes of recording time. Usually, this effect is seen if the record at least 15 minutes. I think that if the record at least half an hour, this effect is exactly appear.
And if this thing befallen, then when you run the program again, the effect is seen immediately. And it will not disappear until you reboot your board.
For no apparent reason, SUDDENLY there appears a high-frequency noise.
Here is WAV-files ... <0535.VoiceCodecEffect.zip>
I cut out the beginning and the end for clarity. You will hear a few seconds of sound, and a few seconds later you hear that suddenly appeared artifact. This file was converted to WAV from the AAC. In fact, the problem is not with AAC-encoding.
In the encode-demo file (audio.c), I inserted a piece of code that records audio to WAV. And if I record the sound directly into WAV (not using the AAC-codec), this effect appears in WAV-file.
Then I tried not to use DMAI for sound capture. I put a piece of code (to the 'audio.c' func 'audioThrFxn') that initializes and uses the ALSA natively (using the ALSA functions 'snd_pcm_open', 'snd_pcm_start' and 'snd_pcm_readi'). And it did not help. Effect was again in WAV-file. The problem is not DMAI.
The problem is not hardware like. Nothing strange there on the oscilloscope.
I tried to see, with software WAV-viewer, how this artifact looks.
And I found that each one sample from seventeenth is distorted.
WHAT IS IT?
This effect occurs only with VoiceCodec and does not appear if we use the AIC.
This effect appears only when there is to be audio-capture along with video-capture.
Any ideas? What are the
thoughts on this? What to do?
I think that VoiceCodec useless if
you can not get it to work properly with the video capture.
I got a clear picture of the artifact as shown!!!
In my previous post I published a WAV-files.
They were obtained by converting from AAC to WAV using ffmpeg.
Unfortunately, it seems that as a result of this transformation, the effect is distorted.
I suggest a more pure experiment!
If you are in directory '$DVSDK/dvsdk-demos_4_02_00_01/dm365/encode/' are the original sources, you can do the following:
Rename the file in this directory 'audio.c' in 'backup_audio' (without the 'c' extension).
Take here these sources <2211.audio-to-wav.zip>
Put in the catalog these three files:
* MyWavLib.h
* MyWavLib.c
* audio_for_wav.c
Just rebuild the demo ('make demos', 'make demos_install').
If you then want to return everything as it was, just delete 'MyWavLib.c', 'MyWavLib.h', 'audio_for_wav.c' and rename 'backup_audio' to 'audio.c'.
On the host computer, run the sound output
host $ gst-launch -v audiotestsrc ! alsasink
Connect your computer sound output to the input VoiceCodec with a cable. (we will not use the microphone.)
To record only sound (5 seconds), the demo should be run like this.
./encode -a dummy.aac -p 16000 -t 5
The parameter '-a dummy.aac' needed to run the audio stream. As a result, AAC-file is not created, and in the directory with the application will create a 'test.wav'.
Open the 'test.wav' using WAV-editor (for example, 'audacity').
Simply adjust the volume level of your sound card on your computer.
To do this, run the audio capture on several occasions, and eventually until you get to a wav-file was a normal sound level.
Finally, start the capture audio along with video capture for a long time (eg at least 15 minutes).
./encode -v test.264 -y 2 -a dummy.aac -p 16000 -t 900
Effect should appear.
Here WAV-file in which I left only the part where the effect appears 7536.VoiceCodecEffect_directWav.zip<>
PLEASE HELP!
It is very necessary to understand what was happening.
I apologize for the late reply.
to Michael:
Hi, Michael.
The general problem is mismatch of impedances.
Do not forget that the input resistance of VoiceCodec is 10k.
But, what is the output impedance of your MIC?
You need to do match of these values.
It is possible that the output resistance of the microphone is to low.
As I wrote earlier, I strongly recommend to place a buffer.
Of course, do not forget to make a design for amplification, offset and filtering.
See details and application notes for your mic.
In short, you need to make your custom mini-board with few parts.
You need to connect your mic to the input of this mini-board and connect outputs from the board to the inputs of VoiceCodec (use wires).
More specifically, you must connect the output wires to the place where was old demounted microphone (M1). The photo that I published earlier this mini-board was not visible.
Here is an example.
I hope that you will succeed and the sound level is will good. Please, write a short message in this topic that you got it.
When you get a good level of sound you can test shows whether you have "the effect of distorting the sound after 15 minutes of audio recording with a video" or not.
Please write about it. You will be doing this great help to me. Because I am suffering.
I could hear the sound recorded from VoiceCodec slightly better than from AIC (slightly less noise) for the Fs = 16kHz.
And it would be great, if the effect of distortion is not apear.
to YJ Liu:
Hi, YJ Liu.
Unfortunately now I have to use the AIC.
Unfortunately I do not have time to win this problem, because I need to write much more code for other aud&vid capture tasks.
This damned VoiceCodec was take a lot of time for me.
I could not devote time to it.
But I'm planning a couple of months to return to this issue.
I plan to try another board rev.F with TMS (Now I have DM365EVM rev.E with TMX DM365)
I plan to use newest PSP 03.21.00.04 (I want to note that this is not part of the DVSDK4)
I plan to use latest Linux kernel (AGAIN, this is not part of the DVSDK4)
I plan to examine VoiceCodec registers (suddenly something happens to them during the recording)
I plan to use anything )))) … (If no help, can I pray?)
When I bought DM365EVM board and downloads from the Internet DVSDK4, I thought that VoiceCodec will work.
I was not lucky one? What am I doing wrong?
to ALL:
Hey guys!
Do not be silent!
I'm the only one stupid?
Please, post your results (good or not).
If you have all the good you do not be lazy and write so concisely. "All is well, I have my VoiceCodec writes more than half an hour of audio along with video and effect does not occur." It's not a hard for you? (Please specify your DM365 revision).
You could help to stop my suffering.
ALL IN ADVANCE THANK YOU! I hope for your help!
