Are frequency sweeps an accurate measure of SPL?

[Skol]

Dear REW community,

A hopefully simple question! Are frequency sweeps and accurate way to match the level of left/right stereo speakers?

To explain: I usually calibrate my speaker levels using pink noise and the SPL meter in REW (together with a handheld SPL meter as a double-check), but I always find SPL meters to be a bit ‘jittery’ and imprecise (even when C-weighted slow).

Recently, when reviewing some REW measurements of my room, I noticed that the left-side speaker frequency response was consistently 0.5dB lower than the right-side speaker response (when viewing the All SPL graph). This was especially clear when viewing both plots at 1/1 or 1/3 smoothing). So I decided to correct it by tweaking the left speaker output level and then running frequency sweeps to compare it against the right speaker response until they matched.

Are there any potential flaws in this method? It seems like an obvious and very easy solution to me, especially when making minor adjustments that would otherwise be difficult to gauge on an SPL meter. Unless I’m missing something?

Thanks in advance for any clarity

PS: loving the recent REW updates! Great work John.

 

[John Mulcahy]

That's a good approach, though you could do much the same more quickly using Pink periodic noise and the RTA, with the RTA window set to Rectangular and no averaging. You will also find the SPL meter readings a lot more stable if you use periodic noise rather than random pink noise. Using the view of the response lets you make an informed choice about where the responses should coincide, rather than relying on a single summed figure over a bandwidth which is what the SPL meter is providing.

 

[Skol]

Thanks John! Good to know it's a viable method for matching speaker output.

Also interesting to note your tip for using periodic noise and the RTA tool (and/or SPL meter), especially the point about "Using the view of the response lets you make an informed choice about where the responses should coincide". That's partly why I found using frequency sweep plots so useful, but your method would be much quicker as you say.

Thanks again