Are multi-driver IEMs meaningfully non-minimum-phase?
I got interested in this because I keep seeing the claim that you can’t EQ a cheaper IEM to sound like an expensive multi-driver IEM because crossovers make them non-minimum-phase.
Thankfully, this is all measurable, so we don't have to rely on guesswork or anecdotes.
So I calculated excess group delay from HypetheSonics impulse-response measurements for a few IEMs.
The result surprised me: most of the deviations are tiny. Even the larger ones are still well below the lowest mean audibility thresholds reported by Liski et al. (2021): about −0.56 ms / +0.64 ms under very revealing test conditions.
The reason minimum-phase behaviour matters is that, for a linear minimum-phase system, the magnitude response also determines the phase and transient response. So things like “driver speed” or “transient quality” are not separate hidden variables once the FR is matched.
That doesn’t mean FR contains literally everything. Distortion, channel matching and fit can still differ.
And even if someone could hear a 0.1–0.2 ms excess-delay difference, why assume it is a good thing?
If I EQ a Salnotes Zero to match the magnitude response of a more expensive IEM, and the Zero remains closer to minimum phase, then the remaining phase difference only proves that they are different — not that the extra delay is responsible for better “technicalities”.
So I’m increasingly skeptical of the argument that multi-driver phase behaviour is a major reason why EQ can’t reproduce the sound of another IEM.
Sources:
HypetheSonics Jdata database
Liski, Mäkivirta & Välimäki (2021), Audibility of Group-Delay Equalization, IEEE/ACM TASLP. DOI: 10.1109/TASLP.2021.3087969