wmy-espressif
commented
1 year ago Hi @0x0fe , thanks for your question. I am a Bluetooth guy and I will check the Bluetooth stack for you first.
I would like to ask some questions on how to reproduce the issue, so that I can perform some local tests:
- You mentioned you use a simple A2DP firmware, did you use the a2dp_sink example provided with esp-idf, or you an example in esp-adf? Do you make some changes on the example? I would like to know how I can get a minimum application to reproduce.
- As far as I know, DOWDR2 has embedded PSRAM(2M), compared with D0WDQ6. You have mentioned you have performed multile tests including enable/disable PSRAM. Can you provide your sdkconfig file used to reproduce this issue, both for the cases that PSRAM is enabled and disabled?
0x0fe
xiongweichao
Answers checklist.
General issue report
So the context is a consumer product design which uses an external DAC (ES7148) and provides audio playback via different sources, MMC, wifi and bluetooth (A2DP), initial tests were conducted on ESP32D0WDQ6 and everything worked fine, the A2DP is smooth, no stuttering or anything. The A2DP firmware is based on ESP-IDF v4.4.4.
For final implementation we switched to D0WDR2 since we use both wifi and bluetooth and a relatively large application logic, additional PSRAM is required.
Sadly we found that the D0WDR2 cannot work normally for the A2DP, there is a massive stuttering, and the playback is totally abnormal. There is no playback issue when playing audio from MMC.
In order to confirm the problem we performed multiple test, at first changing compile time options such as Flash speed (40M and 80M) CPU speed (160M and 240M) PSRAM enabled/disabled, and even buffer sizes for I2S and A2DP, we found no difference at all, in all cases the stuttering was still present with the same magnitude on the D0WDR2 and no stuttering at all on D0WDQ6.
Then we decided to make comparative tests with devkit instead of the product hardware, we changed the DAC for one without master clock (MCLK) as i suspected it could be related to MCLK, so we used : -one devkit with D0WDQ6 + MAX98357A -one devkit with D0WDR2 + MAX98357A Identical pinmap, and identical firmware binary flashed in both.
We tested the same way, with an iphone as source and running a simple A2DP firmware (also tried different source such as PC): -D0WDQ6 : perfectly smooth, no stuttering at all -D0WDR2 : massive stuttering all over, unusable
With the log in verbose mode, there is absolutely no error or warning of any kind, both output exactly the same thing. So the problem lies somewhere deep and is not a simple buffer underrun or something (well unless it doesnt gets catched).
Sadly, we can now confirm the problem comes from the D0WDR2 chip itself and not the hardware, the DAC or the firmware. I have no idea yet of what causes this, nor what exactly changed in the die between D0WDQ6 and D0WDR2, both are rev 3 so i would have expected the D0WDR2 just be a D0WDQ6 with a PSRAM attached, but apparently it is not the case ,I2S block? bluetooth stack? who knows.
So please espressif have a look at that. Attached is a quick audio recording of the problem (from an iphone).
Dowdr2 a2dp stuttering.zip