I honestly have no idea as to the REAL answer for why Lower Sideband on 160, 80 and 40M but all others use Upper Sideband. The anomaly is 60M sandwiched in between 80 and 40M being USB and 30M no Sideband as it is CW only. So, the argument everything 40M and below is always LSB is invalid.
To stir up the pot there is a regular evening net at 7213 kHz +/- that operates on USB not LSB. So why do they use USB? The other regular 40M USB mode is FT-8. The net has a large participation, and the net control is in the mid-west with very strong signals heard here on the left coast.
However, that 40M USB net was put to great use the other evening as I tweaked the LSB USB frequencies on an Arduino Digital LO/BFO. One of the old-time tests in setting the BFO frequencies is to use the computer between your ears in the listening mode.
Close your eyes and switch sidebands with the dial set on NO station. The background sound that you hear when toggling between USB/LSB (supposedly when set correctly) should sound very close. Of course, this test is the real-world approach because your filter, especially if homebrew may not be exactly symmetrical and the sidebands may not be spaced exactly 3kHz apart.
Once I found the correct two BFO frequencies switching to USB for this net... it was spot on. The typical approach is to find the filter center frequency and then use frequencies 1.5 kHz above and below as a starting place.
Now in the old days when separate BFO crystals were used especially in commercial vacuum tube rigs the procedure (of which I am not fully certain of) involved measuring power output and the proper setting followed a formulaic approach of power reduction of a certain amount at some specific value of BFO frequency. You would typically see small trimmer caps on the BFO crystals to "net" them to the proper frequency.
One company did not use the trimmer cap on the BFO crystals was Collins Radio. I have several KWM-1's and if you change the 455 kHz Mechanical Filter you must also change the BFO crystal. Evidently the replacements were sold as a set.
The nominal bandwidth of the KWM-1 filter is 3 kHz. I happened to find a 2KHz wide same type plug-in filter as used in the 75A4. I thought this might sharpen things up a bit on the receive end. What I didn't consider was that the installed BFO crystal was tailored to be placed on the slope of a 3 kHz wide filter and changing the filter to 2kHz made that BFO frequency fall on the wrong part of the filter slope.
Were there a trimmer cap I might have been able to change the BFO. BUT likely NOT! If you added trimmers to crystals such as like in a VXO as you go up in the crystal frequencies (say 12MHz) a small C change can move the frequency a lot. At 455 kHz you would hardly see a wiggle in the frequency change... likely the reason Collins didn't include it.
USB/LSB convention is more or less like the assumption that two Bob's Big Boy hamburgers will always get the 300-pound Mary Jo into the back seat of the 57 VW Beetle. This was the typical process.
I also marvel that many commercial / industrial and military radios in the 2 to 12 MHz range were USB standard. While sitting in the comm shack at Chu Lai, imagine my disappointment when I found out the A/N PRC-47 only tuned USB, and it didn't cover 20M.
With certainty, I will receive several emails with an explanation of why the convention and again with certainty they will all be different and most likely not the real reason. A parallel paradox... driving on the right-hand side of the road and it might have something to do with shifting gears.
Them that know can make things go.
73's
Pete N6QW