HIFIMAN Susvara Headphone Review

[amirm]

The model is not by me but shown in the paper I attached. It shows your frequency response analysis is an oversimplification.

The argument that we are trying to model a headphone with frequency response measurements is yours. This is what I said is imagined. You imagine that research attempted to create a model for a headphone where there is not an ounce of it there. Once more, the research started with listening tests. Frequency responses of the headphones were then correlated with listening tests results and hypothesis was created for the most preferred frequency response curve. That hypothesis was put to test and was shown to be correctly predicting listening preference based on adherence to that target (within some error rate). Nothing about this is about modelling a headphone, transducer, etc. which you keep saying we are doing.

 

[amirm]

Sound coming from headphones is way more complex than in a simple LTI system, where you can just compare the frequency response to a benchmark curve.

But I'm sure you know that, and this might just be a strawman argument.

You are following textbook definition of a strawman by taking frequency response measurements and claiming we are trying to model the entire system using that. It is absurd as I keep telling you we are not doing that. A frequency response is just that: frequency response. It determines tonality of a headphone, speaker or any transducer for that matter. It happens to be the thing we are most interested in when we talk about these devices. Other attributes exist which I explore through distortion, impedance, and group delay but they are by far the low order bits.

It doesn't matter if the system is LTI or not. Yes, some distortion can cause brightness but again, compared to huge variations in frequency response we see, it is not what we are focused on as top level assessment of headphones and speakers.

All of your confusion is caused by not understanding the research and really fundamentals of why we measure what we measure.

 

[Coverpage]

Okay, I’ll just agree to disagree as there’s no new arguments. Thanks Amir for your efforts in responding.

 

[plumpudding2]

While the Harman research shows frequency response to be the major indicator of listener preference, it says nothing about whether various headphones are identically preferred when equalized to the same target.
Amir himself has stated the HD800 has special spatial and separation qualities and I think a harman style study would show the majority of listeners would prefer an HD800 equalized to target over the truthear zero red equalized to target.

If the hypothesis is true that some other quality is a strong predictor of listener preference after eq, and given the fact that equalization is easy to do, it would make more sense to purchase headphones based on these "other" qualities rather than tonality.

Hopefully some day someone will conduct such a study.

 

[amirm]

While the Harman research shows frequency response to be the major indicator of listener preference, it says nothing about whether various headphones are identically preferred when equalized to the same target.

There is zero emphasis on anything being "identical." As I post earlier, Harman equalized a surrogate headphone/IEM and showed that it garnered similar preference to the original headphone.

That aside, yes, there are differences outside of tonality which I appreciate and was not explored in the research. For this reason, we need to get past frequency response variation being the "feature" and let these other components (spatial effects, dynamic range, etc.) be the compelling differentiation.

 

[Sengin]

The underlying mathematics and sources of non-linearities and distortions differ significantly from say, electrostatic headphones. Each headphone type, and even individual designs within a type, has its own set of characteristics that contribute to Total Harmonic Distortion (THD). Thus, citing THD alone is insufficient to accurately describe a headphone's sound. It's incredibly opaque.

No one ever said that THD alone is sufficient to accurately describe a headphone's sound, so I'm not sure why you bring this up.

Are you trying to say that given an identical frequency response and an identical phase response*, an electrostatic headphone will sound different from another type of headphone? If so, I want to challenge this. If it did, there would be something that shows up in the frequency response or phase difference. If one headphone has a resonance at say 8KHz that adds (for this example) a perfect +2dB bell at 12 db/oct slope and Q factor of 1.0, this will show up in the frequency response with a bump at 8KHz. It also means you can add EQ to match that perfect +2dB bell, 12 db/oct, Q of 1.0 on the other headphone with perfect FR and phase response and they will sound identical. In other words, there is no 'magic' that isn't on the FR/phase graphs you could use to say "headphone A is electrostatic." Otherwise it will show up in some way - that is, those non-linearities are both measured and part of the graphs (e.g. in distortion). I think some people get used to how certain headphone types are tuned, or how they distort, or what phase characteristics they have, and assume that is a property of the headphone type.

*I am not including distortion as a function of volume for simplicity - of course a headphone that distorts at say 114 dB and one that doesn't won't sound the same, but it will also affect the frequency response of that headphone at 114 dB so wouldn't match the 'perfect example' I posit.

However, based on the predictions made, like the Zero 1 and 2 IEMs potentially sounding better than the Susvara and others, we can form judgements. As someone who has experienced these headphones firsthand, I find these results highly suspect. And I'm sure I'm not alone.

People can enjoy the spatial qualities that different phase on the left channel and right channel at different frequencies produce. Given that a headphone is supposed to accurately reproduce the sound that's recorded, I would argue that adding these spatial qualities means it is not a good headphone. It certainly doesn't mean that it sounds bad though. You could add these spatial qualities, if you wanted, to any other headphone if it takes EQ well though. Or you could decide you wanted it on for this song, or that artist, but not for these others. Or you could listen as the artist intended. The susvara does not look like a headphone that produces audio in a manner the artist intended - but if you enjoy that there is nothing wrong with it.

 

[Sokel]

The main issue with this HP is that it's price was suppose to reflect a sole value:engineering excellency as it does not have the pedigree or the country of origin to be considered luxury or status item to justify the price.
And it's in that exact value that it falls short as the review shows.

If the deviation is deliberate it should be clearly stated by the manufacturer.
(I can't imagine resonances,distortion and inability of high dynamic range to be deliberate thought)

 

[krumpol]

The spider diagram had me laughing. Plankton is so figurative I had this instant mental image of Schiit amps sufficiently wet to have plankton swimming inside them.

Found the batch! There must have been some plankton and wetness without thickness in this shipment:

 

[solderdude]

The Susvara has an unusual thin and lightweight membrane.
I reckon to develop and produce it is costly and part of the price is to recuperate development costs (and keep the cigars smoking).
It was (I reckon) all in an attempt to reach a similar lightweight membrane as can be achieved with electrostatics.
The problem with electrostats is the capacitance and the difficulty of driving said capacitances at >20kHz which requires current and voltage.

For electrostatics making the membrane conductive is an easy task. Not so easy to 'attach' the traces on such a membrane without weighting it down too much.
I suppose that's where most of the money went into. So yes, this driver is probably 'faster' than any stat (that is driven by electronics).
This also is the reason for the unusual low efficiency of this driver.
The problem is not many measurebators measure above 20kHz for headphones (for obvious reasons) while for speakers and electronics this is easy to do.

The HE1000 and Susvara all measured well above 30kHz at 90dB SPL on my fixture without even lowering response where a lot of headphones start to drop above 18kHz or so, some even lower. This is something the cheaper models don't do. Also the response from the Susvara driver is much 'smoother' (as in lesser peaks and dips) in the treble region.
I think this is the 'high-end' part including materials used (on looks).

Does this have to cost € 6k when produced in China ? I don't know, I have no insight in small number production of such fragile membranes that have to be tensioned correctly, fixed and has to have lightweight traces on it too.

Does all of this come at a cost (efficiency, resonances) ... probably yes. Would this be SQ compromising at 'easy listening levels' ? I don't think so and most will be using headphones this way.
Is it wise to drive it to deafening levels ? I would use more sensitive headphones for that.

That said, Dan C also makes his membranes and uses some really smart trickery to get better 'swing' and seems to be able to tune (pads etc) to conform to a target better than Fang B seems to be able to do and only now trickles down to € 2k where hifiman already had very similar performing cheaper models (HE 1k, Ananda) etc that aren't that different from the Susvara in almost all aspects but don't achieve the smoothness of the treble of the Susvara.

I can understand the love for the Susvara from owners that probably also wear expensive watches etc.
I can understand the love for $ 25 IEMs for what they do well.
They are not over-ears though and bypass the pinna which is part of the auditory system and much more complicated to measure than sticking an IEM in a tube.

 

[DuncanDirkDick]

The main issue with this HP is that it's price was suppose to reflect a sole value:engineering excellency as it does not have the pedigree or the country of origin to be considered luxury or status item to justify the price.

I think the larger issue is that most people don't understand (consumer) product pricing and targeted spending groups. It it priced as a luxury good with decent enough performance that the targeted audience (not your everyday joe, not an accoustics engineer either) wouldn't outright call it a scam. The country of origin argument seems a bit naive, oversimplified and outdated to me.

 

[Coverpage]

No one ever said that THD alone is sufficient to accurately describe a headphone's sound, so I'm not sure why you bring this up.

Are you trying to say that given an identical frequency response and an identical phase response*, an electrostatic headphone will sound different from another type of headphone? If so, I want to challenge this. If it did, there would be something that shows up in the frequency response or phase difference. If one headphone has a resonance at say 8KHz that adds (for this example) a perfect +2dB bell at 12 db/oct slope and Q factor of 1.0, this will show up in the frequency response with a bump at 8KHz. It also means you can add EQ to match that perfect +2dB bell, 12 db/oct, Q of 1.0 on the other headphone with perfect FR and phase response and they will sound identical. In other words, there is no 'magic' that isn't on the FR/phase graphs you could use to say "headphone A is electrostatic." Otherwise it will show up in some way - that is, those non-linearities are both measured and part of the graphs (e.g. in distortion). I think some people get used to how certain headphone types are tuned, or how they distort, or what phase characteristics they have, and assume that is a property of the headphone type.

*I am not including distortion as a function of volume for simplicity - of course a headphone that distorts at say 114 dB and one that doesn't won't sound the same, but it will also affect the frequency response of that headphone at 114 dB so wouldn't match the 'perfect example' I posit.

I am on mobile so sorry for my brevity and typos. One way to understand the non linearity is to look at one non linear part of the system. Let’s say the diaphragm modes.

When you do a frequency sweep it’ll give you the magnitude and phase plot. But due to the behavior of the diaphragm when there is a multi frequency input, based on the spectral content of the input it’ll go into a specific mode. Each mode can have different property and it can struggle to output certain frequencies for the multi frequency input based on the characteristics of a mode.

So this is a non linearity, the behavior of the system changes based on the input. So there can even be even lost spectral information, and it won’t show up on the FR graph.

You can hear this. It’s not something subtle. Two different headphones sound very different no matter the EQ you apply to make them have the same FR (based on predefined input a chirp or sine sweep).

People can enjoy the spatial qualities that different phase on the left channel and right channel at different frequencies produce. Given that a headphone is supposed to accurately reproduce the sound that's recorded, I would argue that adding these spatial qualities means it is not a good headphone. It certainly doesn't mean that it sounds bad though. You could add these spatial qualities, if you wanted, to any other headphone if it takes EQ well though. Or you could decide you wanted it on for this song, or that artist, but not for these others. Or you could listen as the artist intended. The susvara does not look like a headphone that produces audio in a manner the artist intended - but if you enjoy that there is nothing wrong with it.

We don’t even know what’s missing when solely using FR curve. It should be much more than spatial information, but what we know is that Harman target give rise to very weird results when used as the main measurement of the quality of headphones.

So do we turn a blind eye to these strange IEM-is-better-than-susvara results and say FR is enough because Harman research (which has its obvious limitations) says so. that’s what’s in contention.

 

[olieb]

One way to understand the non linearity is to look at one non linear part of the system. Let’s say the diaphragm modes.

Resonances and modes are linear. (They can bring out nonlinearities more pronounced but that typically is because of high level/excursion.)
Therefore resonances are imprinted in the FR but can be difficult (up to impossible) to EQ when narrow band and high Q.

 

[Galliardist]

So do we turn a blind eye to these strange IEM-is-better-than-susvara results and say FR is enough because Harman research (which has its obvious limitations) says so. that’s what’s in contention.

But does Harman research actually say that FR is enough? Or is it simply that the preferred target, outside of the bass area, is a key criterion?

Consider that the FR curve does rule out accepting a lot of other issues. A bad resonance, for example, is an obvious deviation from the target.

 

[splitting_ears]

That aside, yes, there are differences outside of tonality which I appreciate and was not explored in the research. For this reason, we need to get past frequency response variation being the "feature" and let these other components (spatial effects, dynamic range, etc.) be the compelling differentiation.

This! Imagine a world where you could choose a pair of headphones based on looks, comfort, isolation, portability, repairability and anything other than frequency response/deviation. How much easier it would be.

Now we have to ask ourselves silly questions like "would I trade less sub-bass for better high-mid accuracy" when almost all listeners could be 100% happy with decent Harman compliance + a simple bass/treble tone control. And parametric EQ for us nerds, of course

 

[BDWoody]

I get confused and need things explained simply to me.

You're done trolling this thread.

 

[Sokel]

I think the larger issue is that most people don't understand (consumer) product pricing and targeted spending groups. It it priced as a luxury good with decent enough performance that the targeted audience (not your everyday joe, not an accoustics engineer either) wouldn't outright call it a scam. The country of origin argument seems a bit naive, oversimplified and outdated to me.

Luxury good the Hifiman?
The people that buy luxury goods wouldn't touch it with a stick.
Maybe so in it's domestic market but even there I know for a fact that luxury goods comes mainly from Europe or USA (even if they have to make a round trip).

I have seen them in Louis Vuitton getting the 20k euro travel case and fill it with another 20k cost of bags and not one of them,but each of a group!
So nitche?Maybe.Some kind of rarity?
But not luxury even by a long shot and not by price,there are cheaper ones that can be bought as luxury,not these ones.

 

[francerex]

You're done trolling this thread.

Can you ban me too? I was also trolling.

Getting to be point where people like Dan Clark, Crinacle, Fang Bian, etc - that are actually out there making amazing headphones liked by thousands - are swiftly dismissed like crazy is honestly insane. It shows beliefs that are more akin with faith and scientism than proper science.

When I first started getting intersted in headphones I stumbled across this forum. As an academic myself I was happy to have found such a rigourous environment. Oh boy, how wrong I was.

At this point, I honestly believe that this forum might be more harmful than anything else for someone approaching the hobby.

No offense intended

 

[usern]

Getting to be point where people like Dan Clark, Crinacle, Fang Bian, etc - that are actually out there making amazing headphones liked by thousands - are swiftly dismissed like crazy

Disagreeing with something is not "dismissing like crazy". Their work gets praise here too.

No offense intended

Lies, you intend to offend and provoke if you are trolling.

 

[BDWoody]

Can you ban me too? I was also trolling.

Sure can.

 

[plumpudding2]

Can you ban me too? I was also trolling.

Getting to be point where people like Dan Clark, Crinacle, Fang Bian, etc - that are actually out there making amazing headphones liked by thousands - are swiftly dismissed like crazy is honestly insane. It shows beliefs that are more akin with faith and scientism than proper science.

When I first started getting intersted in headphones I stumbled across this forum. As an academic myself I was happy to have found such a rigourous environment. Oh boy, how wrong I was.

At this point, I honestly believe that this forum might be more harmful than anything else for someone approaching the hobby.

No offense intended

I've been following this thread from the start and if you can filter out the gleeful exclamations of the superiority of 20 dollar iems you're left with some quite interesting discourse on headphone design and measurements and their limitations.

Head-fi suffers from the same poor signal-to-noise ratio (pun intended) but for different reasons, as there are zealots on both sides.
I hang around on both forums as it is my personal belief that they both help in guiding newcomers to a sound that makes them happy.