|
|
Hi Mr. Jacky Chiang, Could you provide your plan to implement it ? |
|
|
Hi Wooshin, We will evaluate and check the change and effort in SW architecture on Synaptics BSP and get back to you soon. |
|
|
Hi Jacky, In step 2, you don't need to copy data for speaker. App will out mic data to speaker. Thanks. |
|
|
Hi Wooshin, Do you have test app ready for our testing ? Thanks~ |
|
|
Hi Wooshin, We think that if the audio patch of AMP SOC is not used, it may have a latency of sound output from APP. If possible, we need test APP to observe latency of this function if comment 7230 is ready. Thanks. |
|
|
Hi, Sita I have no any test app, so you have to make record app using below code and check if mic data is recorded correctly. recorder = new MediaRecorder(); recorder.setAudioSource(MediaRecorder.AudioSource. VOICE_COMMUNICATION); Thanks. |
|
|
Hi, Sita I will explain again. There are various karaoke APP and some app will use audio record class for making karaoke score. So you have to make like below sequence. 1. check is it open "audio record class (MediaRecorder.AudioSource MIC type=VOICE_COMMUNICATION) from APP ? 2. If opened, send mic data to audio record class 3. If not opened, send mic data to SPEAKER or HDMI. If you have any question, please inform me. Thanks. |
|
|
Hi Wooshin, As our experience, if human's voice path is from "microphone->audio HAL->framework->app->framework->audio HAL->speaker", it will has big and unacceptable latency. For example, you will hear your voice > 300ms later after you speak. You will feel your voice is late and not sync with music. It is not acceptable for karaoke application. That's why we added audio patch(a fast path from "microphone->audio HAL->speaker") in Eagle and SWAN to lower the latency in current implementation. If karaoke app needs to record audio for score, we still think it need to output to speaker by audio patch not by karaoke app. Otherwise the latency is a big problem. |
|
|
Hi River, Inopiatech implemented it and did test. They report there is no delay to SKB and skb decide to use same method. @Sita Inopiatech share test APP. Please do test using attached APP. |
|
|
Hi Sita, You can download karaoke test APP. Please check it. https://drive.google.com/file/d/1xcQpk5SuSbgfd9lGv15wlL1K1ACkmFiF/view Thanks. |
|
|
Hi Wooshin, I will test the feature with this APP. Thanks for information. Thanks. |
|
|
Hi Wooshin, Test image path on alt server: /release_by_fxn/swan_test_image/Karaoke_Record/BFX-UA300_20210618_SD_v524r23_DelayTest. The following notes about test image: 1. The MIC setting function(DEVICE_SetMicOnOff) is not called in the new APP(app-release.apk, skb), so we hard code to enable MIC in default. Kumyoung APP is still working normally. 2. Please verify if the MIC setting function(DEVICE_SetMicOnOff) will be called in new APP or not with APP vendor. It will affect our schedule for feature implementation. 3. The voice from audioTrack of new APP sounds very quiet, so we think the volume of voice maybe need to be amplified in APP. We keep finding root cause for this behavior. 4. Other audio feature(including audio routing, NUGU, youtube, and BTV) have been not fully tested yet. We cannot ensure these feature is working normally. Thanks. |
|
|
Hi Sita, Thanks for image. I will test it and share the result. |
|
|
Hi Sita, SKB decide to go original way due delay. < Original Idea > 1. Please keep sending data to amp driver to output from speaker. 2. Copy data for speaker from step1 and send it to audio record class. 3. Android Audio Record Class get data using "MediaRecorder.AudioSource MIC type=VOICE_COMMUNICATION" 4. Some Apk will use it. (example, Karaoke app uses it to get score) I am sorry for confusing. Thanks. |
|
|
Hi Wooshin, Test image has been uploaded to ALT server, file path: /release_by_fxn/swan_test_image/Karaoke_Record/BFX-UA300_20210625_SD_v524r23_DelayTest Some notes for this test image: 1. Following #c7553 comment, we change some behavior in HAL layer. 2. Enable MIC in default for testing. Kumyoung APP is still working normally. 3. The delay time of the both APP is about 160~180 ms. 4. Other audio feature have been not fully tested yet. 5. Please confirm with APP provider vendor if MIC setting function(DEVICE_SetMicOnOff) will be called. We still need some time to change code. Thanks. |
|
|
Hi Sita, I have attached new test app. Please check it. Thanks. |
|
|
Hi Sita, SKB did check delay time, delay is 127ms. They are saying maximum delay should be under 100ms. Please check how to do reduce delay under 100ms. Thanks, |
|
|
Hi Wooshin, We would like to know how SKB checks the delay time(127ms) and what APP is used? Could you confirm it with SKB? In addition, we will check the delay time of audio stream in audio patch to find out how to reduce delay time. Thanks. |
|
|
Hi sita, I will check it with SKB. thanks. |
|
|
Hi Wooshin, Test image has been uploaded to ALT server, file path: /release_by_fxn/swan_test_image/Karaoke_Record/BFX-UA300_20210721_SD_v531r03 We found the a way to reduce delay time of MIC and It works. The latest delay time is about 95ms (maybe +/- 5ms). Please confirm the test image. Thanks. |
|
|
Hi Wooshin, I already commit the modification about this feature. Thanks. |
|
|
Hi Sita, I have done test for delay of karaoke. I feel there is too much delay time. Could you check it is committed correctly ? Thanks. |
|
|
please check my comment. |
|
|
Hi Sita, Do you have any update ? |
|
|
There are Karaoke mic delay results [Test Enviroment] FW: 10/1 Daily Build Karaoke app version: app-release-5th.apk DVT2: No any rework (ESMT Amp without A113X) DVT3: Reworked mic and ref delay issue (Wellang Amp with A113X) [Test Results] DVT2: 96ms, 4.649 ~ 4.735 DVT3: 160ms, 1.564 ~ 1.727 DVT3.jpg (251,940 bytes) |
|
|
Hi Wooshin, The I2S path for DVT2 is: VS550 ------ ESMT amplifier The I2S path for DVT3 is: VS550 ------IS2 Buffer IC ------ A113X ------ Wellang amplifier Obviously, the 60ms delay is increasing in DVT3 by A113x or Buffer IC. I guess DVT3 will be the main reason. Please also help to check if A113X cause the delay and can optimize to reduce delay. |
|
|
Hi Wooshin, From HW's information, the I2S Buffer IC only delay 4.8ns. So, please check if you can shrink the delay in A113X. |
|
|
Hi Wooshin, I measured the I2S data before and after A113X with version 0x30 in the attached picture. As you can see, the time difference is about 59.2ms. So, the extra delay is caused by A113X. A113X_ver0x30_I2S_delay.jpg (34,047 bytes) |
|
|
Hi River, Actually, i think that mic data only was increased because i couldn't feel delay of music from karaoke app. If there is delay in A113x side, it should be delayed both but it is not. |
|
|
Hi Wooshin, Any audio (music or microphone) will pass A113X and all these sound will has about 60ms delay in A113x. But when you are singing, you will listen to music that is just outputted by A113X and then your mouth starts to sing. So, you are based on the music that is already output from A113x (And the next music is already in A113X buffer and will output very soon). And your voice is just starting to enter microphone and need to walk through the 160ms path (mic->VS550->A113x->Wellang->Speaker). |
|
|
Hi River, I can understand your explanation of this issue but my feeling is not good. If I will get any comment from SKB, I will share that on here. |
|
|
The delay time is caused by A113X. So we close this issue. If you still have any problem, you can reopen it. |
- yk.jung ( (ALTech) Younkwang Jung ) developer
