Soranik MEMS 3S 2025, the way to smooth treble in IEMs ?

[RogerSmith]

Although not everybody may agree on that, it seems very likely to me that the sound of IEMs is to a huge amount, if not 100%, determined by it's frequency response. The last years have brought many cheap and amazingly sounding IEMs with good frequency response and extremely low distortion. And since most of them have up to about 6-10 kHz a very smooth frequency response, one can EQ them to any targe one desires, be it HarmanIE2019 or the "new meta" or whatever.
As I see it, the only issue that remains to be addressed is a smooth frequency response in the treble, as many otherwise very fine IEMs have huge swings or peaks or resonances there, at least a behavior that is not easy to EQ, especially as these peaks might appear for each individual at different frequency, based on the shape of ears and ear canals. Now there is the new driver technique of MEMS, which contrary to DD, BA, Planar or EST seem to be able to deliver a very contained and extended treble behavior. The vietnamese boutique brand Soranik has a quite pricey ($2500!, https://www.soranik.com/product/mems-3s-2025/) model using these, which are fabriquated by USOUND, the company oratory1990 works for, as far as I know. Unfortunatey one currently needs an additional energizer for these. So it remains to be seen if these drivers really deliver and will become much cheaper of the years.

 

[staticV3]

Now there is the new driver technique of MEMS, which contrary to DD, BA, Planar or EST seem to be able to deliver a very contained and extended treble behavior.

Personally, I'd reserve judgement of the Soranik until we have 5128 measurements.

In the meantime, we have some traditional IEMs with smooth treble response:


And some xMems IEMs with peaky treble response:

 

[RogerSmith]

Personally, I'd reserve judgement of the Soranik until we have 5128 measurements.

In the meantime, we have some traditional IEMs with smooth treble response:
View attachment 420391View attachment 420392View attachment 420393

And some xMems IEMs with peaky treble response:
View attachment 420395View attachment 420396

True, you could also add the Truthear Nova (especially with foam tips) to the smooth treble example. But still, that would amount to four altogether from thousands of IEMs, so really a tiny fraction. I am not saying that it cannot be done currently, but that it requires a very careful implemention with current driver types and might be better or easier to achieve with MEMS. Also, the two negative MEMS example you have given may just signalize that it needs a bit more experience. Furthermore they both use xMEMS (the company), whereas the Soranik uses both, xMEMS and the USOUND one. So there are many open questions, but maybe an interesting new direction in further improving IEMs.

 

[mc.god]

Another example of smooth treble response is the Aful Explorer:

Just to say that here we have examples of all different treble driver types able to achieve the result:

Variations: 2 EST
Meteor: 4 Micro Planar
Galileo: 1 BA
Explorer: 2 BA

 

[RogerSmith]

Another example of smooth treble response is the Aful Explorer:

View attachment 420731

Just to say that here we have examples of all different treble driver types able to achieve the result:

Variations: 2 EST
Meteor: 4 Micro Planar
Galileo: 1 BA
Explorer: 2 BA

That is correct, but the great majority of IEMs using these drivers don't have good treble implementations, and the ones mentioned here, seem to fall off much steeper and are well less extended (this might have to do with the desired target curve, but still):

 

[oratory1990]

Although not everybody may agree on that, it seems very likely to me that the sound of IEMs is to a huge amount, if not 100%, determined by it's frequency response.

Frequency response and impedance (acoustic impedance), yes. That's my understanding of it as well.

So it remains to be seen if these drivers really deliver and will become much cheaper of the years.

The speakers themselves are not expensive - they're actually cheaper than comparable BAs. It depends on the exact model, the projected quantities and also through which path you place the order (local sales rep, local or global distributor, direct order, ...), but even the most expensive ones are still in the single digits.

Unfortunatey one currently needs an additional energizer for these.

USound also makes MEMS speakers that do not require DC to work - but they *do* require much higher voltages (about 10x higher voltages / 20 dB higher levels) as their voltage sensitivity is lower (because the piezoelectric force is lower than the electrodynamic force)

Now there is the new driver technique of MEMS, which contrary to DD, BA, Planar or EST seem to be able to deliver a very contained and extended treble behavior.

As much as I would love to take credit for that, it's not fully true. The treble response of an in-ear headphone is determined by resonances created through the shape and dimensions of the front volume (tube length, tube diameter, cavity resonances) as well as any damping qualities of the involved materials (metal tubes will not damp the length modes, silicone tubes will a bit), and of course through damping materials (mesh or foams). The type of transducer has very little to do with that, other than the obvious (BA speakers physically contain a volume of air and a tube, which will inevitably create a Helmholtz resonance somewhere).

USOUND, the company oratory1990 works for, as far as I know

Not anymore! My last day was in November 2024, but I did work there for over 9 years.

And some xMems IEMs with peaky treble response:
View attachment 420396

Well, the Creative Aurvana Ace doesn't use the MEMS speaker a lot. It only really comes into play above 10 kHz (and the resonances you see above 10 kHz are length modes of the measurement setup, not resonances of the actual speaker).
But since the MEMS speaker they use is so small, they can use it to extend the frequency response of which the main speaker (a dynamic moving coil speaker) might not be capable on its own, without having to significantly increase the size of the earphone.

it requires a very careful implemention with current driver types and might be better or easier to achieve with MEMS.

I would actually say it's a bit harder with MEMS, because their acoustic impedance is higher than most other speakers (except for some BA tweeters). They don't react to added damping the same way that more traditional speakers do.
Which isn't really an issue if you design the earphone using simulation models, but if you do it the old fashioned way (trial and error, adding tubes/damping based on previous experience) then it makes your job a bit harder than you might like.

Where MEMS speakers really shine is with fully automated assembly methods (they can withstand high temperatures, so they can be used with reflow soldering techniques and don't require hand soldering)

 

[markanini]

Smooth graphing treble in IEMs has a questionable significance IMO. Anecdotally I was impressed with the treble of Moondrop Lan in 2023, so I ordered Truthear Zero after it expecting a further improved treble since it was flattened on the graph. But the treble sounded clearly worse to me.

An example of a very smooth graphing treble is Truthear Nova. Yet it's suspiciously absent from most discussions about smooth graphing treble. Maybe because few users found the subjective treble performance to excel in a way like the graph would suggest. Count me as one of those people.

 

[mc.god]

Smooth graphing treble in IEMs has a questionable significance IMO. Anecdotally I was impressed with the treble of Moondrop Lan in 2023, so I ordered Truthear Zero after it expecting a further improved treble since it was flattened on the graph. But the treble sounded clearly worse to me.

View attachment 422279

An example of a very smooth graphing treble is Truthear Nova. Yet it's suspiciously absent from most discussions about smooth graphing treble. Maybe because few users found the subjective treble performance to excel in a way like the graph would suggest. Count me as one of those people.

View attachment 422280

We already agreed that treble response on graph can be far from what it is really happening into our ear canals due to different shape, length and width, so it is not surprising. Then it sums up to our preferences, one can surely like some peaking more than a smooth descending curve, especially if the smoothness is associated with limited extension. But imho here is exactly where smoothness can be an added value, giving you the ability to better tune with PEQ, being it with a simple high shelf or with specific peaking not disturbed by other peaks present in the iem response.

 

[RogerSmith]

Smooth graphing treble in IEMs has a questionable significance IMO. Anecdotally I was impressed with the treble of Moondrop Lan in 2023, so I ordered Truthear Zero after it expecting a further improved treble since it was flattened on the graph. But the treble sounded clearly worse to me.

View attachment 422279

An example of a very smooth graphing treble is Truthear Nova. Yet it's suspiciously absent from most discussions about smooth graphing treble. Maybe because few users found the subjective treble performance to excel in a way like the graph would suggest. Count me as one of those people.

View attachment 422280

Well, maybe it is a good advice to not only rely and the measurements of crinacle. There are several measurements out there pointing to the ruggedness of the Truthear Zero at higher frequencies, including the one of Amir with a huge channel imbalance above 15kHz.
The Truthear Nova, on the other hand, does have extremely smooth treble, even more so when using foam tips, and I also don't think that it is true, that not many people mention it or praise it. I personally find the treble excellent, at par with the Variations.

 

[markanini]

Well, maybe it is a good advice to not only rely and the measurements of crinacle. There are several measurements out there pointing to the ruggedness of the Truthear Zero at higher frequencies, including the one of Amir with a huge channel imbalance above 15kHz.
The Truthear Nova, on the other hand, does have extremely smooth treble, even more so when using foam tips, and I also don't think that it is true, that not many people mention it or praise it. I personally find the treble excellent, at par with the Variations.

Treble features depend highly on depth of insertion. The most accurate graph from the most trusted operator would solve what exactly? More likely a graph past 3kHz will bias you into thinking that you are hearing something which can't verified or guaranteed due to biometric diversity.

 

[RogerSmith]

Treble features depend highly on depth of insertion. The most accurate graph from the most trusted operator would solve what exactly? More likely a graph past 3kHz will bias you into thinking that you are hearing something which can't verified or guaranteed due to biometric diversity.
View attachment 422388

I don't know who did these measurments on which rig. But assuming they are valid, it can be seen that with a medium insert, which most likely corresponds to normal usage, the graph has the least swings. Also most likely no foam tips were used, which would dimish that still. To really judge these measurments, comparison of other IEMs would be needed, I would speculate that the swings are for most of them much wider still.

 

[Heinrich]

I don't know who did these measurments on which rig. But assuming they are valid, it can be seen that with a medium insert, which most likely corresponds to normal usage, the graph has the least swings. Also most likely no foam tips were used, which would dimish that still. To really judge these measurments, comparison of other IEMs would be needed, I would speculate that the swings are for most of them much wider still.

I recommend to look for un-smoothed head related transfer functions of real people, but especially for the standard HATS measurement rig(s)—raw, not, as they say, compensated. You may wonder how smooth these are, not.

There‘s a nice amount of irritation with interpreting the measurements on rigs for headphones. When it comes to me the industry failed miserably in this regard. It started, to my best knowledge, with German academic Günther Theile introducing „diffuse field compensation“. And then they longed for perfection, in the average, instead should have followed an incremental approach.

Tl;dr? IEMs shall have peaks and dips.If you think otherwise, your fault ;-)) just kidding

Second to that, how should a special driver tech avoid peaks and dips?

Add.: As stated in an other thread, thankfully started by somebody else, we could discriminate peaks originating in the insertion of the IEM to the (outer) ear versus peaks to be expected from the HRTF—which are smoothed out in the „compensation“ for the measurement rig. Example: the „Harman“. The smoothing is where the confusion begins, it‘s a pitty. So the highly involved hifi-guy more often than not does not acknowledge the peaks as necessary. The IEM circumvents, or messes with, parts of the ear that make the natural peaks. So it has to generate the peaks all by itself, but with appropriate intensity at the right frequency, as to fit the response to the individual listener.
—Please tell me if I‘m wrong …