• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Hifiman Edition XS insane mod + measurements

Markelian

Member
Joined
Apr 24, 2022
Messages
19
Likes
42
If you just want the actual mod and arent interested in the journey just skip to post 2.
Here are some experiments you might find interesting anyway.

Individual measurements below, where you can see the harmonic distortion and group delay.
Note that the cursor line is the same in all the screenshots so it might give you an idea relative between the photos.
I'm also showing group delay, which is a measure of phase distortion.
When a signal bounces around and reflects from objects (such as the inside of the cup, etc), its arrives later than the direct signal and with a phase shift. In theory good design advocates we should minimise this bouncing as it can worsen imaging.

I'll add the individual measurement images below in spoilers, just to keep the clutter to a minimum.
Stock - starting point
1740520036176.png

Group delay
1740520050015.png
Rear grille removed
1740516874179.png

Group delay
1740516891476.png
Cotton between the driver and rear grille (that whole space filled with regular cotton)
1740516500036.png

1740516122230.png
Cotton on back, cloth over the cotton, no grille
1740516285018.png
1740517161762.png

Results:

1740526676126.png


Compared to stock, just putting a lot of cotton for driver back damping reduces output, however it reduces it much more on the lower frequencies than the high, resulting in an overall "brighter" image. notice also the high-Q dip is no longer so sharp, but has reduced in size. Adding a cloth over the cotton raises the high frequencies, it is rather "breathable" fabric (and you can easily breathe through it) so maybe it impedes lower frequencies less, but the high frequencies suddenly see high acoustic impedance and reflect back.

1740527091923.png


We can see in the group delays how well just the cotton does at taking this energy that would have bounced around, and just dissipating it as friction between the fibers. Adding the cloth adds back some GD, corroborating the theory that it is a barrier to higher frequencies and causes them to reflect, but if you have your theory i'm always happy to learn.

1740527373211.png

In terms of THD %, roughly the same performance, with cotton+ cloth having a possibly notably lower overall THD.

Regarding the grill vs no grille question:

1740527570135.png


Without grills the FR is raised, repeating both measurements many times always shows the same result.
The thing is removing the grills isnt just about them being resonant, there's also a structural question. Normally the driver seals on the grills back side, then the grills front side seals on the outer shell plastic. With them removed, you can assemble the headphones just fine but now the headphone seals against the plastic directly, and sits 1mm deeper inside the shell. Insignificant? Well maybe not according to measurements, but someone else should try to replicate this result. Notice also the bass rolloff curve is much steeper, probably to do with this sealing. Probably you can add some gasketing material, if you still wish to remove them.
The resonance of the grille has a fundamental starting at 1.2k and moving up, there is no indication on any of the graphs any ringing is taking place, so removing the grill for this reason as some have theorised is unecessary.
Despite that it seems that removing the grilles does seem to reduce some reflections in the midrange.
1740527637566.png


Here is the kind of cotton and cloth used:
1740528071019.jpeg
 

Attachments

  • 1740515967505.png
    1740515967505.png
    670.2 KB · Views: 41
  • 1740517223964.png
    1740517223964.png
    670.1 KB · Views: 595
  • 1740517241955.png
    1740517241955.png
    349.7 KB · Views: 588
  • 1740520475700.png
    1740520475700.png
    214.4 KB · Views: 31
  • 1740520536963.png
    1740520536963.png
    216.3 KB · Views: 29
  • 1740522586765.png
    1740522586765.png
    210.9 KB · Views: 34
  • 1740522730051.png
    1740522730051.png
    210.8 KB · Views: 33
  • 1740523483586.png
    1740523483586.png
    217.3 KB · Views: 32
  • 1740526503822.png
    1740526503822.png
    171 KB · Views: 33
  • 1740527062475.png
    1740527062475.png
    220.5 KB · Views: 35
  • 1740527235719.png
    1740527235719.png
    213.3 KB · Views: 30
  • 1740527344126.png
    1740527344126.png
    255.5 KB · Views: 43
Last edited:
So onto the real finding.
You might have noticed on the above graphs a terrible resonance at around 5.8 kHz, showing as a high-Q dip in the frequency response, and very high distortion. This basically cannot be EQ'd out as you're just feeding energy into the resonance.
I wanted to see if its due to some interaction with the cup, or just inherent to the driver.

Here is the actual raw driver unit measurement, just in free air, with no reflections.
1740530419407.png


You can see the distortion profile is pretty much inherent to the driver construction (membrane tension, thickness, dimensions etc). It is not uncommon for planars to have a nasty resonance mode, and its relatively hard to get rid of. Increasing the tension of the membrane pushes it higher, and vice versa, but it will always be somewhere, and for a full range driver it will always be problematic. Hifiman probably settled on there because it gives the overall best compromise, so what can we do about it now?

I was experimenting a bit with the reflections coming off the back of the pads. Then i realised λ = v/f , v in air is 343 m/s, so a standing wave at 5.8khz would have a corresponding physical dimension of about 5.9cm. Then i remembered, the pads themselves are about 6cm, from inside wall to inside wall.

Here is roughly the construction
1740532477965.png

I know it was totally unecessary to draw but maybe it helps someone.
The idea was that cotton would dissipate instead of reflect some of the energy going directly to them from below (because they overhang a bit above the driver), and secondly that eventually the mode bouncing between the pads would also get trapped and dissipated there.

The result is easily the best performance yet:
1740532810195.png


Notice the overall FR hasnt changed very much, but the 5.8k resonance is gone, as well as the 10k to some extent. Notice overall how the FR is "smoother" and less undulating, this is due to killing the various complex standing mode effects. In this case the 5.8k one must have been a cancellation mode. The mod also helps fill our 2k hole (just slighly).

1740532977710.png


THD % improved by a large margin, especially inthe critical midrange where the vocals are. Even the 5.8khz driver resonance is improved.
1740533238987.png

Showing how effective it was at killing reflections
1740533377556.png


Lower distortion, group delay, improved FR, for the cost of a bit of cotton, and completely reversible. I rest my case.
Listening to it the sound is clearly improved, much more defined imaging, just a cleaner sound. For the cost i'd say a no brainer.
 
Thanks for your work on this! I have no desire to mod mine for sound, but I appreciate knowing things in case I decide to later on.

My mod for comfort was to put a spandex? (thin stretchy) headband around the thing, just under the wide part of the band where the adjustment slide is. That serves me well as a suspension headband. Tightens up the clamping just slightly. Since it can cover the entire forehead as a headband, I can even widen the fabric after it is on and distribute the force more, if things start bothering me over time.
 
Thanks for your work on this! I have no desire to mod mine for sound, but I appreciate knowing things in case I decide to later on.

My mod for comfort was to put a spandex? (thin stretchy) headband around the thing, just under the wide part of the band where the adjustment slide is. That serves me well as a suspension headband. Tightens up the clamping just slightly. Since it can cover the entire forehead as a headband, I can even widen the fabric after it is on and distribute the force more, if things start bothering me over time.
Nice solution as well, as your username implies very pragmatic :)

Bonus

You might have noticed that when you get your hand closer to the back, even covering the grille, the sound becomes more and more "closed in" but otherwise still plays kind of fine.
I was curious what the physical mechanism for it was, basically its causing a massive low-Q dip at around 2kHz, and those are the frequencies important for spatialisation/"presence". Often you will see people experimenting boosting there to try to increase "soundstage" (to varying success), but looks like this does the opposite, and the result checks out.

Hand covering over the driver
index.php

index.php




Also notice the great amount of group delay, caused by the signal bouncing off the palm, back through the headphone, and arriving to the mic later than the direct signal did.

Next thing to look into should look into are pads. The seal relies on the slight bit of leather on the bottom , sealing against the plastic.
But bottom of earpad has a thicker portion where the seam is, which results in a slight gap being opened (shine a light through it and you'll see).
 
Interesting!


I used to do a similar mod with dental purposed cotton with my old HD 6XXs, but with the opposite intention: instead of nulling destructive resonances, I wanted to distance the driver and increase the ear gain and treble region to compensate for old pads.

It helped a ton but in the end buying new ones always offered the best result.

One point of contention I had with the XS were the high frequency peaks, specially around 11-14khz. From the graph it seems that those regions are amplified. What's the in person effect?
 
You call this a successful mod, I say that you found and fixed a design flaw. Respect! I hope Hifiman is taking notes.
 
Good work.
That's why im saving pennies to buy some clone iec711 and kb501 to measure and mod headphones by myself.
 
I used to do a similar mod with dental purposed cotton with my old HD 6XXs, but with the opposite intention: instead of nulling destructive resonances, I wanted to distance the driver and increase the ear gain and treble region to compensate for old pads.
Haha, i used to do a similar thing for my old pads as well, but i used folded toilet paper instead. I experimented putting it only on the portion behind the ear, so it raises only that side and gives an angled pads effect. Thinking you know, maybe it changes spatialisation, changes pinna interaction or whatever. Not sure if i would definitively say what the result of that was but it did work to compensate for pad squish at least.
It helped a ton but in the end buying new ones always offered the best result.
One point of contention I had with the XS were the high frequency peaks, specially around 11-14khz. From the graph it seems that those regions are amplified. What's the in person effect?
You should take anything above 8-9kHz in most non-professional measurements, mine included, with a big grain of salt. You will see a similar trace in some other's measurements as well and up there its too easy to just get artifacts from how the headphone behaves with your coupler. In general this aspect of absolute accuracy should be left to professionals, you should only glean relative differences between my measurements, that is really all i hoped to portray here.
 
This is very interesting. Thank you for doing this and taking measurements. A few days ago I cut up some old clothes and put a bit of them under the pads of my he400se exactly like you did with the cotton; but with the intention to try and reduce the amount of creaking the pads do. I never even considered the impact it could have on the sound. In hindsight I obviously should have.
 
Back
Top Bottom