Wasn't it mentioned on the blog that njm4556 are rated 70 mA each but can actually sink/source 100 mA of current if need be?
8x reduction in THD is really impressive!
That is right about the chip current source/sink and the same sort of situation exists for the O2 booster board. The 70mA figure is continuous - assuming the DIP8 package heat dissipation limits are not exceeded - while the 100mA is for short periods. But that is how music works out anyway of course, the whole "music power" issue where the peaks in music are often 2 or 3 times higher than the average.
I tend to be on the conservative side and use 60mA for the continous figure for those NJM4556A chips. He is pulling that number off a graph in the NJM4556A datasheet and not a table, so there is a bit of graph-reading wiggle room there.
With the booster board the LME49600 output chips are rated at 250mA continous and can actually do that from a heat sinking standpoint on the board. At the project's Google Drive link I have some photos of the board being tested to the limits at 250mA by powering a 8 ohm desktop bookshelf speaker, something it would never be expected to do normally. In those photos I have a thermocouple on the chips reading the case temperature.
BUT... the problem here is the voltage regulators in the O2 headamp which are not heatsinked. So although the booster board output chips could handle that much current steady state, the O2's voltage regulators would fry. In those O2 booster board maximum load test photos I have the O2's power supply bypassed and the booster board fed directly with a +/-15Vdc lab power supply for that reason.
So what it all means is that just like those NJM4556A chips's 100mA peaks, the O2 booster board can do 250mA peaks for short periods - musical peaks - but steady state is more around 180mA due to the O2's unheatsinked voltage regulators.
That THD reduction is as per the chip data sheets. THD is worse on the high frequency (20kHz) end of things, so that part of the chip THD graphs winds up as the limiting factor. Now whether that additional reduction in THD is actually audible - well that is for all of you guys with ears a lot better than mine to report one way or the other!