How much does frequency response actually matter?

This is what I was attempting to hint to - the rest is in the review, accessible to everyone.
How is it possible that 3. and 4. order distortions are significantly reduced?


The effect is that by taking out the oscillation source, the distortion effects it produces are also removed.
Nothing in real life is absolute - removing means reducing to a significant degree, not deleting.

You can view break up mode like a separate signal source that is related to the rest of the moving membrane but not in linear manner.
Yes the PEQ applies to fr response, but the fr-resp is not detached from distortion effects - and vice-versa.
That is why it can and will significantly alter timbre, perceived resolution and even soundstage performance.
This can not be explained by simply lowering energy at some region. It is predominantly the nonlinear distortions that mess up the fine things.
I know because I use these circuits in speakers.
It does not make regular ceramic membrane to a cvd diamond - but it does get quarter of the way there.

Take another driver that would not have the break up mode there and apply the same dip - you'd have hard time noticing any difference.
Now speculatively, add mega spike there, it would get annoying tonally but the soundstage, resolution etc stuff will be more or less the same - because you would mostly add quantity, not nonlinear distortion components. I should try this at some point and report back.

I agree wrt headphones or perhaps even widebander drivers - too much going on at once to hope a crude trick like a notch filter at some annoyance will be sum total helpful.
With well made speaker drivers the breakup crap is well out of filter bandwidth - you only need a bit extra help to clean it up. The point is that many people are mistaken in that the spike in tonality itself is the only problem, the other crap that comes along the ride is just as harmful or more.

 

Corrected or not - the notch filter targeted to breakup mode would do something.
He did not do the IMD with the notch filter.
You'd see some changes as with the H3 and H4 - where he did measure it.
The observable changes would be mostly around the break up mode zone not at 1 kHz.

 

I think much of the out of band peak that is still noticeable comes down to the phase response. When we use a notch filter to tame the peak the phase response also improves.

This is what I suspect happens with the ceramic tweeters.

Another example with my Hathor speakers: Shallow XO at 200Hz. Woofer breakup at 6kHz. It absolutely needs to be corrected! In this case a notch filter isn't the most elegant solution, but it also works.

@purr1n A notch filter and widebander can absolutely give good results, but it has to be meticulously measured and simulated. And the parts quality has to be sufficient of course. PIO caps. High quality resistors and coils.
At least it gives good results to my ears with the flat cone widebander. However breakup is quite high in frequency, roughly 8 and 14kHz.

This is a fantastic result. That M142A driver is something really special. M142T + M74B + M34B is something I am dreaming of building some day. Imagine a hard cone, wide dispersion speaker with a 97dB sensitivity! Not possible before.

 

Yeah, I remember that. For some drivers it makes sense to spend more on xover parts than the driver itself. In your case the coil is quite large and the driver isn't very resolving (I've heard them a couple times), so really no need.

For the widebander in Hathor the coil needs to be absolutely tiny and have virtually zero DCR to get the the Q high enough. Hence why I use cored coils not usually intended for audio. And the woofer is maybe my favorite in this size range (sounds very responsive) hence why spending money on the XO parts pays off. It's not stupid to spend 1k on the like 8 crossover parts. But I'm not sure I'd throw Duelund CAST PIO caps at a high Q notch filter. The treble already sounds very good to my ears.

Wasn't intentional. However for speakers I think it's somewhat different than for headphones, anyway.

Spoiler: Side argument: Dispersion

Dispersion can really change the way a speaker images. Domes sound different from cones. Seriously, you need to hear those Bliesma dome mids. Or the flat cone Al-Sandwich drivers. And the off axis response can show issues that aren't visible on axis. Like how soft domes go into breakup at audible frequencies and as a result have sharp notches in the top octave off axis.

Or how my OB speakers sound totally different from the cardioid monitors which have a totally different sound projection from the LS50 Meta.

The LS50 Meta is as beamy at 1kHz as Hathor is at 100Hz! Yet in the treble the LS50 Meta is actually the beamier speaker. In this sense you could say that Hathor is more of a constant directivity speaker. I think the way Hathor images is very, very special (good).

 

More like a 4.5" driver, even. Larger drivers will be more beamy. It's the baffle width that's the limiting factor, I think.

I meant at a show or something. Marten Coltrane Quintet uses those drivers. I thought it was very good.

However 240Hz projector sounds awesome!

 

I've blind tested myself on just my dx240 (ESS 9038pro) dig filters and I always opt for #5 which is APODIZING FAST LINEAR PHASE or something like than. Doesn't matter the earphone or amp board as long as it's one I like.

No idea why.

 

Idle wondering.

Operating under the presumption that it IS all FR at eardrum level then wouldn't the differences between DACs and DAC filters be distinguishable at the transducer level using an acceptable quality HATS like the 5128, as it's apparently far more representative of a realistic load on transducers?

Taking Golden's running theory that it's all ultrasonics up for a bit, aren't most amp+transducer combinations going to be more severely bandwidth limited compared to DACs? I vaguely remember that they addressed that in the video (don't feel like rewatching it) by saying you needed a certain tier of gear for DAC differences to be perceptible. Still curious how gross the differences at transducer level are going to be.

 

Miniscule FR differences like passband ripple and the limited bandwidth of DACs should be detectable at the transducer level. But they're swamped by the rest.

As far as distortion goes, as long as it's sufficiently low you shouldn't be able to see anything.

I really think it's something else, I mean why would an aluminum driver always have a different presentation from a beryllium driver? Sure, they do measure differently, but sometimes more similar than not. I'm of the opinion that materials have their own colorations and peculiarities that can't really be seen in the measurements.

There are other things like why would different feet change the sound of electronics. I doubt you can measure anything there, however there are theories as to why it would make a difference.

 

To be fair I just remembered that I had these. First image is HD600 out of the Unity (YELLOW trace) vs out of the Piety (BLUE trace). If you think the voicing looks really weird, please do keep in mind this is a MiniDSP EARS rig and these pads are nearly 10 years old, no exaggeration. If there are minor differences between traces here, please keep in mind that I had to manually adjust potentiometers to align things.



Second image here is the HD600 out of the same amp (*probably* the Unity if I was tryharding, but the Piety elsewise) exhibiting the differences between OS mode (MAGENTA trace) vs NOS mode (BLUE trace) on the Modi Multibit 2. I just left the headphones untouched on the rig and toggled the filter. The upper treble differences here ARE pretty evident, but I replicated this with EQ as best I could and didn't feel like it approximated the sound, really.



So it's possible to argue both sides given this BUT given the fact that my hearing tops out at 17kHz nowadays and the fact that the Piety and Unity couldn't sound more different from one another within the same tier of gear, I think it's definitely not the case that FR explains the lot of things, at least not within the audible bandwidth here.

Last one here, HD600 out of FiiO BTR7 SE out (RED trace), MMB2 into Piety high gain (GREEN trace) and out of my Arturia Minifuse 2 interface's headphone out (BLUE trace). Some differences here but IMO nothing that tracks with subjective impressions of each. The differences HERE I can attribute to the Minifuse and BTR7 having piddly power delivery and the Minifuse having pretty high OI from what I remember, but again I don't see how that'd track in FR here alone.




Supplementary, here's the Andromeda compared out of the FiiO BTR7 (in wired and AAC mode) vs the MMB2 > Piety stack. The difference in FR between the BTR7 and the desktop rig is attributable to differences in OI.

As with the HD600 I just left the Andros on the coupler and minimised jostling. I even swathed this one in lots of jackets and shirts to minimise outside noise. I'm pretty impressed that FIR differences were this visible at the transducer level.:









P.S.
I just realised what I SHOULD have done: Unity internal ES9018 DAC vs MMB2, both feeding the Unity amp. AGH. Getting on that later this week after I let the Unity warm up again. I was *just* using it after a long while last week but I switched back to Piety for fun sound. Might be better to use the Andromeda for those measurements cuz better isolation.

 

Okay I was a bit impatient but here's results comparing the MMB2 in OS and NOS mode against the integrated ES9018 DAC chip in the Magni Unity. I decided to stich with the HD600 because the Andromedas suck at ultimate air extension which was the whole point of this. I've been letting the Magni bake as an AIO for about 4+ hours and the MMB2 stays on 24/7 unless I run into a power interruption.

I set the potentiometer to my usual target (95dBZ full-scale pink noise) and simply swapped the same USB cable between both DAC modules while keeping both potentiometer and headphone static on an isolated desk. There may be differences that are non-representative of my normal use case because I measured these using 48kHz sampling rate to try and see how it'd handle ultrasonics high up when my usual diet is Redbook audio.

There ARE differences in the very high treble between all three configurations as evident below, but as mentioned above I did try approximating these with EQ only to find that the results were dissimilar in terms of perception. Also the differences in FR between NOS and OS settings were more pronounced than the difference between either MMB2 mode and the ES9018 chip; this does not correlate with subjective differences between the three converters where the ES9018 module sounds much more diffuse and smeary than either MMB2 setting (if we're conflating upper treble response with resolution then the MMB2's NOS mode ought be the least detailed, but that's not the case).

 

For folks using loudspeakers, it doesn't nearly matter as much as: 1) room acoustics 2) ground-plane noise reduction 3) "signal ground-plane" noise reduction and 4) quality power distribution.

As the brilliant Nelson Pass once put it: "The ear is not a microphone and the brain is not a tape recorder".