A/B and loudness
Two chains are resident at once so that switching is instant. The loudness is matched from the coefficients, never by listening to what is playing.
The two slots
- Slot A is the same thing the playback screen applies. Changing it in either place changes the same slot.
- Slot B belongs to the comparison. A design can be loaded into it without being heard, so the switch is instant when you do switch.
- Bypass is always available as a third position, and it is the 0 dB reference the matching works from.
- Switching crossfades over 5 to 10 ms and carries the convolution history across, so the new chain is correct from the first sample rather than after its filter fills.
- It is not blind. Which side you are hearing is always on screen, because the point is to move between your design and what it was before, not to test whether you can guess.
How the loudness is matched
correction [dB] = -( level of the chain, combined from its linear stages )
combined: gain, biquad, dc_blocker
counted as 0: every nonlinear stage, every spatial and imaging stage
limit: ±12 dBDrive is already compensated inside the node.
A nonlinear node holds its level as its drive rises, so it is counted as 0 dB here. If raising drive raises loudness, that is a defect rather than something the matching should paper over.
Spatial stages are not compensated.
Level changes caused by imaging and spatial nodes are outside the combination, so they are yours to manage with a gain node.
It is matched with a sine at the reference level.
Real music has a higher crest factor, so the match is close rather than exact.
At the limit, it says so.
If a slot needs more than ±12 dB, the correction is clipped and the screen states that the two are not fully matched. The fix is the gain staging inside the graph.
One nonlinear side changes more than tone.
If only one slot contains a nonlinear node, the output limiter is armed for the session, so the comparison includes its effect. The screen points this out.