Questions on the Audibility of High-Q Peaks and Dips in Headphones

Just a random query that I can't easily fit into the shoutbox and that I've been curious about for a while.

Been following Keith Howard's (HiFiNews, but also offers independent measurements I believe) measurements for a while and they all seem to be using 1/3d octave smoothing in their final reports as overlayed on raw FR. This is something some others whose works I've been eyeing lately do, though there are exceptions like Brent Butterworth. I understand this might be fine for speakers because of how people might not be able to perceive very high-Q dips due to how the human auditory system perceives stimuli and tends to smooth things out, but I've not been able to find any literature explicitly stating the same for headphones.

Could just be I don't know how to use Google well. Do let me know if I'm being a dummy, thanks.

This is NOT an argument to provide smoothed FR responses by the way, I personally go 1/48 max if at all when generating squiggles, just that I think this is something people might occasionally overlook when interpreting graphs, at least with regard to FR without spectral decay to ratify the findings there.

Found some interesting old threads here on SBAF addressing DF/LDF curves and others (which were fun) but the question is more how much of what's accepted as permissible in speaker systems applies to headphone systems, considering the increased proximity of the transducer to the ears. In my case, I do personally hear narrower peaks as being less audible than broad ones (TH-X00 Buster Sword v MDR-Z1R fencing foil in the 10~11kHz region) and not really being able to catch any high-Q dips even when listening for them, but I readily admit not having heard nearly as many headphones as loads of other people and being a relatively inexperienced listener.

Found a paper of interest on AES supporting that peaks are more audible than valleys (http://www.aes.org/e-lib/browse.cfm?elib=10291 ), which I think everyone knows at this point, plus one on how timbral qualities can still be influenced by relatively narrow resonances (http://www.aes.org/e-lib/browse.cfm?elib=10291) but both studies based performance on loudspeakers so not sure how much of what's there applies to headphones.

I'm also aware that there's a profound difference between listening for test tones/headphones performance on sine sweeps and listening to music, the latter being exponentially more "chaotic", but what do I know I used to capture FR measurements on my phone using pink noise

 

I'm not aware of any scientific reasons for it, but I've definitely experienced narrow spikes. Back when I used to do some recording and mixing, sweeping high-q peaks was a surefire way to identify tonal problems with fizz, etch, and flub in distorted guitars... with amp sims especially there are problem frequencies up towards 6-8khz, 3.5khz, and down in the 400-800hz range. The only way to really identify them a lot of times is to sweep them with really tight EQ boosts. You can very audibly hear any spikes there that way. Whereas if you're making normal blind cuts, you can't find it.

Just sort of a way of identifying stuff that's already coming out, but is normally still hard to fully distinguish, by creating resonant peaks. It's stuff that's there, affecting the sound, and often causing different instruments to clash. The bane of double-tracking distorted guitars because of how the 'bad' frequencies resonate between the two tracks.

If you're accurate enough, you can even keep the cut tight enough to remove only those nasty-sounding spikes and still retain the overall tonal profile. Most people wouldn't hear it as being "cut" in that range, even though there are narrow sections of frequencies that have actually been pushed orders of magnitude lower. So yeah, cuts like that are definitely less perceptible.

Another adage from EQing is "boost wide" because it tends to sound more natural and less obtrusive. If we're to assume there's anything to this, finer peaks may not always register as a strong, jarring resonance, but rather just kinda wind-up sounding tonally 'wrong' even if not outright offensive.

Yet another thing worth considering when we're talking q-value and sweeping. A para eq is not linear. The distance between 100-200hz is about the same as the distance from 5-10khz. BIG audible difference between the two. As you go up in frequency, you need tighter Q values to retain the same precision. If you start with a moderate q value down at 10 hz and sweep up, you'll be affecting an increasingly wider range of frequencies as you move up. So what looks like a big bump on the lows may not mean that much, while a small peak on the highs could be affecting many frequencies. The scaling on the plot probably counts for at least as much in that regard.

 

@robot zombie Thanks for the input!

I don't actually listen for FR irregularities via sweeps (cuz... that sounds both painful and boring), but it was interesting to hear how weird resonances can actually be pushed down into near-oblivion without the overall sonic "colour" being affected. Definitely seems to support the idea that dips are much harder to detect for whatever reason, though I wonder at why that might be the case still.

Eh, it makes for easier listening, which counts for a lot as far as I'm concerned

When I was just getting around to learning about the MiniDSP EARS, how to use em etc, I noticed REW had more smoothing settings than I thought strictly necessary. Finding the option "psychoacoustic smoothing" and subsequently reading up on why the heck people might even consider using this is a big part of why I made this thread, just genuinely curious at how human hearing seems to be able to "listen past" big ol FR jaggies in 2ch and nearfield but how the same is intolerable on headphones.