Auto aidl

[galambosbalint004]

Can you give the aidl slider back? It ruins the sound, giving too much bass, when headphones need some boost.

[Pittvandewitt]

The slider was for buffer size. If your device has the AIDL audio HAL implemented, buffer size would use the exact same value that is used by default now anyway.
If I were to revert the change back, it would only be a placebo.
I'm not saying the sound you heard is the same as you heard before. Devices running the AIDL audio HAL but without aidl mode activated, suffered from a sound where the frequency response graph no longer matched the actual change in sound. This was problematic and it probably explains the sound you were experiencing.
As you may understand, I'm not planning on reverting the bug fix.

[noseph01-max]

@Pittvandewitt Thanks for explaining the graph mapping, that actually makes a lot of sense. I spent some time troubleshooting on my S25 Ultra (Android 16 / One UI 8.5) to see why the low end is acting up under the new automatic AIDL framework, and I think I've got a pretty good lead for you.

It definitely looks like a digital pipeline bottleneck rather than normal volume clipping. Here is what I found after digging into it:

• It's not standard clipping: Even if I back off the attenuator and channel balance to carve out over 20dB of headroom and throw an aggressive limiter on, the distortion stays exactly the same. It's a harsh, metallic grinding sound, not square-wave clipping.
• The 30Hz Sine Wave Test: If I play a pure, continuous 30Hz sine wave, it sounds perfectly clean with Wavelet toggled off. The absolute second I turn Wavelet on, the tone instantly breaks up into that grinding noise. That rules out any kind of physical hardware or speaker driver failure on my end.
• The Bluetooth Offload impact: If I go into Developer Options and turn off "Bluetooth A2DP hardware offload," the grinding noise drops drastically (though there is still a layer left on the deepest sub-bass).
• Other settings: I verified that Dolby Atmos, Legacy Mode, and the built-in Samsung EQs were completely disabled during all of this, so there are no other processing layers fighting for the stream.

Since changing the audio routing topology (toggling A2DP offload) heavily changes the intensity of the noise, but dropping the volume headroom does absolutely nothing, it really feels like a buffer starvation or timing sync issue tied directly to the new AIDL audio HAL.

Forcing the system CPU to handle software decoding seems to relieve a ton of the processing pressure, but the new framework's low-latency blocks still seem to be starving Wavelet's engine of the time it needs to calculate heavy low-end EQ before the hardware demands the next chunk of data. Hopefully this helps pinpoint where the pipeline is choking!