See my reply #2 above. The schematics you've seen with a 20k feedback resistor are for the non-inverting configuration, yours is inverting and the usual 1k/20k is replaced here with 10k / ~200k to keep the input impedance to a reasonable value. In my reply I explain how you can also modify the gain by changing these resistors.
As for the additional components shown in the non-inverting configuration of post 9, which I assume also apply to the inverting one:
- RSN, CSN are required for stability
- L, R is required if you have a potentially capacitive load, like a very long speaker cable
- CC improves clipping behaviour, but it actually reduces phase margin, which is compensated for by Cf, Rf2
Always a good idea to include them all, especially RSN, CSN, the others as required depending on your application.
Yes to RSN and CSN having the same values. CC, Cf and Rf2 would have to be scaled accordingly (22p, 5.1p, ~200k) but on second thought I'm not sure they're required in an inverting configuration, since CC would simply act as a low pass filter at the input (since In+ is connected to ground) and, as I said, Cf and Rf2 are used to compensate for CC, so you either install all three or none. Also, since these components are supposed to affect clipping behaviour and you have very efficient speakers, I very much doubt that you're getting anywhere close to clipping, unless you listen at insane levels...
I think the first thing to try, which is easy enough, would be to add RSN and CSN at the output, and then try to decrease the gain as I explained in post #2, e.g. decreasing the feedback resistor, since the extra input resistor doesn't seem to have worked for you, although I don't really know why, it's a perfectly valid approach too. Remember not to reduce Rf below 100k since a minimum gain of 10 is required for stability according to the datasheet.
I think the first thing to try, which is easy enough, would be to add RSN and CSN at the output, and then try to decrease the gain as I explained in post #2, e.g. decreasing the feedback resistor, since the extra input resistor doesn't seem to have worked for you, although I don't really know why, it's a perfectly valid approach too. Remember not to reduce Rf below 100k since a minimum gain of 10 is required for stability according to the datasheet.