What you're referring to is the C-weighted SPL observed at the listening point as relayed through the speakers. The question that was asked was how many dB down from full-scale is the magnitude of the band-limited random noise (the 'test tones'); the two things are related, but not the same.

For example, you could have a 0 dBFS file (i.e. all bits hi) being played back through an audio system, but the SPL is a function of the volume control on the 990, the amplifier gain, and the speaker sensitivity. For example, a full-scale file could be played back at 45, 57, 78, 80...or whatever dB SPL, but the file itself remains full-scale.

What he's really asking is when the tones are generated, how many dB down from full-scale is the magnitude of that file. Remember that the value in dB relative to full scale is not a function of the sound pressure level achieved - it's a function of the file itself.

For example, if the magnitude of file "A" is full scale (0 dBFS) and the magnitude of file "B" is 70.7% of the value of file "A", then file "B" is - 3 dBFS. Likewise, if the magnitude of File "C" is 50% that of file "A", then we know that the magnitude of File "C" is - 6 dBFS. We can say this because we know that file "A" is full-scaled, that is, 0 dB.

That's why I suggested the measurement approach that I did, because comparing the magnitude of an unknown file (the 'test tones') to the magnitude of a known file (the full-scale file) for a given volume setting on the 990 would tell you how many dB down is the magnitude of the unknown file.

You could make a crude approximation of the value though by using a SLM and playing the full-scale file (for a given SAFE volume setting) noting that value, and then invoking the tones and noting the SPL observed for the band-limited noise. It's not as accurate as a measurement of the signal coming from the 990 (because the speaker-based approach relies upon speakers, and is influenced by the room, its acoustics and so on - and therefore where you measure affects what you measure; these caveats do not apply when the measurements are made driectly from the preamp's outputs).

Still, if you played the known full-scale file at a reasonable volume setting (maybe -40 or - 50 dB on the 990 to start), noted its SPL, then noted the SPL generated for the test tones (again, for the same volume setting on the 990), the DELTA in SPL would be a crude approximation of the scaling of the test tones relative to full scale (because speaker SPL is proportional to applied voltage, not power).

In other words, if the fullscale file, for a given volume setting on the 990 yielded 87 dB SPL, and the test tones yielded 58 dB(same volume setting on the 990 and the measurement mic in the same location used to measure the full-scale file's SPL), the you could say that the test tones are approximately 87 - 58 = 29 dB, or, 29 dB down from full-scale, or written another way, -29 dBFS.

Again, using the electrical signals (preamp outputs) is a more reliable method though as it eliminates all the influences of room modes, overall acoustics, speaker sensitivity / directivity / frequency response, and so on.