I have no idea if the SPL meter app on my phone is properly calibrated (the app company does specifically calibrate for my model of phone), but for my use cases it doesn't matter.
My loudest use case is playing orchestral or ensemble music at the same levels I would experience if I was playing with the group on stage. I do this for those occasions when I want to play along. I play tuba, which is a loud instrument, and when playing along I want to play at performance levels and not have to hold back to hear the ensemble in a realistic way.
In some ways, this is an unattainable use case, simply because recordings are made with "out-front" balance, and when I'm sitting in the ensemble, I'm right next to the bass trombone, or perhaps in front of the tympani, while the higher instruments are "over there". But if the recording I'm accompanying is loud enough, I can play at performance levels and still hear the ensemble clearly.
To get an idea of the requirements for this use case, I measured SPL (A weighted) in my current ensemble. I recognize that it does not really measure the peakiest part of peaks, but it does hold the peak value, and I'm only using it for comparison. Playing a generally loud work (general ensemble tutti with lots of percussion at the conclusions of the piece when we are playing at fortissimo--NOT when we are playing 1812 Overture and being accompanied by 105mm Howitzers), I measure peaks at 108-110 dBA SPL and a constant sound of 95-100 dBA. Those are fairly low (tympani and bass drum), so in absolute terms they were much louder. And they are percussive so I'm sure my meter is integrating the signal over too long a sample period to capture the true peak. Whatever. When I play a similar work on my system at home, I want peaks at 108 dB and a constant output at 95-100 dBA. I do not find that classical recordings measure with less dynamic range than the live group, but that may be because of where I sit in the group, or the rehearsal room, or a range of other possibilities. It's louder on the first row of the audience, simply because those patrons are closer to the high instruments than I am, and those instruments sit in the most sensitive range of our hearing. They are also designed to project their sound out front. The only instrument I don't want to side
behind is the French horn. Those folks on the front row of the audience are probably getting hammered harder than I am on the back row of the stage.
My listening position is 6-7 feet from the speakers, but it's a larger space, and I'm not always sitting at the LP. When I'm playing tuba, I'm sideways from the speakers, so the loudness of the speakers has to be greater than at the LP to fulfill my requirements.
I got similar average measurements with my brass quintet but with louder peaks. In a brass quintet, the high instruments are trumpets and there is no percussion, so the peaks are sustained notes by trumpets listened at close range. Usually, the members of a quintet sit to one side of the trumpet blast zone--it's even louder right in front of the bell.
So, my answer to the question is a sustained 100 dBA on average, with peaks of 108-115 dBA or more, playing big classical works or brass quintet, measured generally in the room. At that level, I can play along without holding back. When I put prog rock on at the same level, I start to worry about structural damage to my home because of the difference in mix. Obviously, I don't exercise this use case when my wife is on the same side of the county.
For routine listening, I play (dynamically recorded) prog rock at home at about that level, aiming the average output to be in the 95-100 dBA range, when I'm alone and ready to focus on listening. Peaks depend on the dynamics of the recording, but may be 20 dB higher, and I would prefer that they do not compress. For routine listening when others are at home, 80 dB on average is plenty. I set REW to record sweeps at 90 dBA SPL, using pink noise for the SPL calibration.
I do NOT watch movies at reference level, because my wife is watching them with me and she just doesn't like it that loud. That said, more than once the well-recorded rumble of thunder in the soundtrack has caused me to look out the window to check for incoming bad weather.
When I was using four Advents powered by two 125-wpc amps, I could reach that level, though with a little strain in the upper frequencies. The two Revel F12's, powered at the same 125 wpc (referenced to 8 ohms) is just as loud, reflecting their greater efficiency. I don't hear strain in the upper frequencies, but there is a bit of compression in the lows, though I think that's running into the limits of the amp as much as the speakers. My incoming Buckeye will deliver more than twice the power of my current amp. My current amp probably delivers 150 wpc into the nominal six ohms of the Revels, with 1.2 dB of headroom, whatever that means. The NC502MP will deliver more like 275 watts at the knee, but over 400 at 1% distortion, so it should give me another 3 dB, assuming the speakers can take it. Rick Sykora and Buckeye are providing clipping indicators for me for a reason
Rick "who listens at 70 dB tops for most listening when working and when others are at home" Denney