It’s true that by creating larger differences in the gain between bands you’ll see larger artifacts. Sort of a “with great power comes great responsibility” type thing. The artifacts are also greater when the bands are narrower, which is why you’re seeing it more prevalently in the lower frequency bands. The reason running two in series doesn’t show these artifacts is that you’re effectively doubling the length of the filter. That’s a trade off that increases CPU usage and increases latency.
FWIW, I assume you’re not often making 24 dB adjustments in adjacent frequency bands?
Dan, good to know about the filter length idea and the limits. You need to draw the line somewhere. In mastering it now dawns on me to double down for a final print if there are any bass boosts like at around 10db to 12db, by my research inducing very small ripple less that +/- 1 db.
So not always making 24db adjustments in music making or mastering, just in a music playback system I’m developing that implements the Equal Loudness Contour. Two plugs in series is not a drawback at all in that example.
BTW does one instance use one calculation no matter the bands count?