HIFIMAN Susvara Headphone Review

[Coverpage]

The system's excess phase is close to 0 per frequency. You can find excess phase measurements of few headphones, it's usually flat or close to flat. (Oratory have some) The headphones as an isolated system is linear. As long as both headphones have the same leakage tolerance, if you EQ them to the same target next to the eardrum of one individual, they'll sound same.(ignoring distortion here).

leakage tolerance is how FR of headphones change when their air seal is broken. In this case EQ setup I described above, only leakage tolerance between headphones can make a difference.

Minimum phase means that the zeros and poles of the system TF is on right side of the s-plane. It is basically saying the system is stable (for example the output doesn’t increase to infinity) and causal (output don’t depend on future input).

The use of the idea of minimum phase and frequency response to describe the entire system assume the system is linear and time invariant. This allows superposition to be true. That is, if I input signal 1 and get signal A as output and input signal 2 and get signal B as output then if I input signal 3 which is just the sum of signal 1 and 2, the output will be the sum of signal A and B.

This allows the frequency response curve to be useful. Because you can break down the input to its frequency components, and the output will be the addition of those frequencies multiply by the frequency response curve.

This superposition property is possible when the system is linear and time invariant.

Due to the complex physics of the headphones, there are non linearities. With the driver having a voice coil or linear motor, and uneven force constant due to the magnetic flux densities in the air gap, the geometry of the diaphragm which is of a complex shape as it vibrates, resonance of the headphones physical construction, the friction, the acoustic of how sound wave travel from the driver to the ears, AB amplifier has cross over frequencies etc.

If it is LTI, like you have said, you can transform the output of a headphones to match another. So a simple experiment would be to choose one planar headphones, one dynamic and one electrostatic just to make the difference very obvious although I don’t think you need to go even to this length to get the result. And like what you suggested EQ then to make them sound the same. I’ve tried, it’s impossible.

Part of it is detail retrieval, some headphones has higher resolving capability especially apparent in electrostatics. In an LTI system you should be able to EQ this detail, for example increase the treble and should hear the details. But you can try, it’s very obviously impossible.

So if a frequency response curve does not describe the headphones because it’s not LTI, can we base our buying decision solely on it, and how we adhere to the Harman curve? We can’t.

And that’s why we get conclusions like the IEM sounds better than the Susvara. If you have listened to Susvara you’ll know how very unlikely this is.

Just to be transparent, do I think Susvara is overpriced. Yes, I personally think there’s so much subjectivity in sound that we should not have these ultra expensive headphones even from DCA for that matter. But that doesn’t detract from the idea that the way we evaluate headphones by relying only on these measurements is just not good enough.

The measurements are just not sufficient to predict how good a headphone sounds. It’s more than frequency response curve.

 

[amirm]

Part of it is detail retrieval, some headphones has higher resolving capability especially apparent in electrostatics.

There is no such thing as "detail retrieval." This is a made up term by audiophiles for headphones which has no grounding in engineering or science of sound reproduction. It classic lay intuition terminology used by people to claim technical knowledge and value when there is none.

 

[amirm]

So if a frequency response curve does not describe the headphones because it’s not LTI, can we base our buying decision solely on it, and how we adhere to the Harman curve? We can’t.

Yes we can. Research shows that we can. And my own experience of some 120 headphones and IEMs, together with 300 speakers says equalization absolutely, positively works to create a neutral sound or frankly, any tonality you desire. I trust you don't have a fraction of this experience, nor are you familiar with the research.

By your logic, there is no way for a designer to create any frequency response. It would all have to be accidental.

 

[Soria Moria]

There is no such thing as "detail retrieval." This is a made up term by audiophiles for headphones which has no grounding in engineering or science of sound reproduction. It classic lay intuition terminology used by people to claim technical knowledge and value when there is none.

Would you ever consider making a video explaining this? Could be useful. I don't have a good grasp on it.

 

[Coverpage]

Yes we can. Research shows that we can. And my own experience of some 120 headphones and IEMs, together with 300 speakers says equalization absolutely, positively works to create a neutral sound or frankly, any tonality you desire. I trust you don't have a fraction of this experience, nor are you familiar with the research.

By your logic, there is no way for a designer to create any frequency response. It would all have to be accidental.

Sorry we can do what? I really don’t catch the first part.

Of course you can design speakers with a specific frequency response. But this doesn’t describe the entire sound of the speakers. And not the sole reason for people buying the speakers when auditioning.

 

[amirm]

Of course you can design speakers with a specific frequency response. But this doesn’t describe the entire sound of the speakers. And not the sole reason for people buying the speakers when auditioning.

If we can design a speaker with specific frequency response, then your logic that such a response cannot be created with EQ as part of it is false. Here is the response of a speaker that uses such EQ:

Same as this:

By your logic, even a crossover doesn't work since the underlying drivers are not linear.

Really, this is an absurd notion. You have taken a theoretical concept used in engineering and confused it with what we are doing here. The entire world of high-fidelity sound reproduction relies on frequency response measurements like above. Please don't keep saying none of this is possible because the system is not "LTI."

 

[Rhamnetin]

Part of it is detail retrieval, some headphones has higher resolving capability especially apparent in electrostatics. In an LTI system you should be able to EQ this detail, for example increase the treble and should hear the details. But you can try, it’s very obviously impossible.

When I had a Stax SR-009, I tried looking for this but there were no instances in which the SR-009 was able to resolve a detail that my HD 6XX couldn't.

 

[majingotan]

When I had a Stax SR-009, I tried looking for this but there were no instances in which the SR-009 was able to resolve a detail that my HD 6XX couldn't.

Even the 7Hz Zero will resolve every information in the track and heck it’s even easier to perceive it than headphones due to isolation from outside noises

 

[Coverpage]

There is no such thing as "detail retrieval." This is a made up term by audiophiles for headphones which has no grounding in engineering or science of sound reproduction. It classic lay intuition terminology used by people to claim technical knowledge and value when there is none.

Even without this intact my argument still stand.

But I’d argue it’s possible to measure this. I don’t have the time to delve into this atm. But I’ll look into it. I remember Dan Clark was talking about detail in his presentation in the video you posted. I’ll revisit that too.

If we can design a speaker with specific frequency response, then your logic that such a response cannot be created with EQ as part of it is false. Here is the response of a speaker that uses such EQ:

Same as this:

By your logic, even a crossover doesn't work since the underlying drivers are not linear.

Really, this is an absurd notion. You have taken a theoretical concept used in engineering and confused it with what we are doing here. The entire world of high-fidelity sound reproduction relies on frequency response measurements like above. Please don't keep saying none of this is possible because the system is not "LTI."

Nobody is saying FR curve means nothing.

What you are saying is FR curve is everything. That is, it’s a complete description of the entire system and so if fits the harman curve with low THD (never mind where this THD comes from) they are good sounding headphones.

I keep repeating myself, but all you need to do to proof that FR is enough is to make two different headphones sound the same through a filter.

And if you can’t, there are other aspects that you haven’t modelled and perhaps those are the important part, as your methodology predicted the Susvara will sound worst than the iem to most people.

 

[Blockader]

Minimum phase means that the zeros and poles of the system TF is on right side of the s-plane. It is basically saying the system is stable (for example the output doesn’t increase to infinity) and causal (output don’t depend on future input).

The use of the idea of minimum phase and frequency response to describe the entire system assume the system is linear and time invariant. This allows superposition to be true. That is, if I input signal 1 and get signal A as output and input signal 2 and get signal B as output then if I input signal 3 which is just the sum of signal 1 and 2, the output will be the sum of signal A and B.

This allows the frequency response curve to be useful. Because you can break down the input to its frequency components, and the output will be the addition of those frequencies multiply by the frequency response curve.

This superposition property is possible when the system is linear and time invariant.

Due to the complex physics of the headphones, there are non linearities. With the driver having a voice coil or linear motor, and uneven force constant due to the magnetic flux densities in the air gap, the geometry of the diaphragm which is of a complex shape as it vibrates, resonance of the headphones physical construction, the friction, the acoustic of how sound wave travel from the driver to the ears, AB amplifier has cross over frequencies etc.

If it is LTI, like you have said, you can transform the output of a headphones to match another. So a simple experiment would be to choose one planar headphones, one dynamic and one electrostatic just to make the difference very obvious although I don’t think you need to go even to this length to get the result. And like what you suggested EQ then to make them sound the same. I’ve tried, it’s impossible.

Part of it is detail retrieval, some headphones has higher resolving capability especially apparent in electrostatics. In an LTI system you should be able to EQ this detail, for example increase the treble and should hear the details. But you can try, it’s very obviously impossible.

So if a frequency response curve does not describe the headphones because it’s not LTI, can we base our buying decision solely on it, and how we adhere to the Harman curve? We can’t.

And that’s why we get conclusions like the IEM sounds better than the Susvara. If you have listened to Susvara you’ll know how very unlikely this is.

Just to be transparent, do I think Susvara is overpriced. Yes, I personally think there’s so much subjectivity in sound that we should not have these ultra expensive headphones even from DCA for that matter. But that doesn’t detract from the idea that the way we evaluate headphones by relying only on these measurements is just not good enough.

The measurements are just not sufficient to predict how good a headphone sounds. It’s more than frequency response curve.

Nice copy pastes first of all.

''Due to the complex physics of the headphones, there are non linearities.''

excess phase measurements would show that. Excess phase shows how much a system deviates from the theoretical min phase behavior. The mysterious complex physics of the headphones are not so mysterious as you would like to think.

even HD600 has incredibly linear behavior throughout the whole frequency range.

''If it is LTI, like you have said, you can transform the output of a headphones to match another. So a simple experiment would be to choose one planar headphones, one dynamic and one electrostatic just to make the difference very obvious although I don’t think you need to go even to this length to get the result. And like what you suggested EQ then to make them sound the same. I’ve tried, it’s impossible.''

You tried but you didn't do it with right tools. You did without knowing what you are doing. I told you how it should be done. Did you use a binaural mic? Did you take your HRTF measurements? Did you calculate how the acoustic impedance of your ear canal would couple with the acoustic impedance of the headphones?

''Part of it is detail retrieval, some headphones has higher resolving capability especially apparent in electrostatics. In an LTI system you should be able to EQ this detail, for example increase the treble and should hear the details. But you can try, it’s very obviously impossible.''

The detail retrieval is a result of the relation between FR of headphones next to eardrum, auditory masking function of the human hearing(loud sounds mask the quieter sounds around their frequency range) and the spectra of the music.

Excellent video about auditory masking, he was also my teacher.

''The measurements are just not sufficient to predict how good a headphone sounds. It’s more than frequency response curve.''

I don't know how many times I heard this nonsense. You are just coping, and you know that. I used to have Stax 009S too and compared them to all the cheap headphones I had.

FR consistency of headphones between individuals + leakage tolerance + FR gives everything you need about how headphones sound.

 

[Chagall]

I keep repeating myself, but all you need to do to proof that FR is enough is to make two different headphones sound the same through a filter.

And if you can’t, there are other aspects that you haven’t modelled and perhaps those are the important part, as your methodology predicted the Susvara will sound worst than the iem to most people.

@amirm Please make a video about this. I guess a lot of people are confused about how can FR be the only differentiation. I know I am.

 

[usern]

Part of it is detail retrieval, some headphones has higher resolving capability especially apparent in electrostatics.

What is detail retrieval? What is resolving?

 

[Coverpage]

Nice copy pastes first of all.

''Due to the complex physics of the headphones, there are non linearities.''

excess phase measurements would show that. Excess phase shows how much a system deviates from the theoretical min phase behavior. The mysterious complex physics of the headphones are not so mysterious as you would like to think.

View attachment 337514
even HD600 has incredibly linear behavior throughout the whole frequency range.

''If it is LTI, like you have said, you can transform the output of a headphones to match another. So a simple experiment would be to choose one planar headphones, one dynamic and one electrostatic just to make the difference very obvious although I don’t think you need to go even to this length to get the result. And like what you suggested EQ then to make them sound the same. I’ve tried, it’s impossible.''

You tried but you didn't do it with right tools. You did without knowing what you are doing. I told you how it should be done. Did you use a binaural mic? Did you take your HRTF measurements? Did you calculate how the acoustic impedance of your ear canal would couple with the acoustic impedance of the headphones?

''Part of it is detail retrieval, some headphones has higher resolving capability especially apparent in electrostatics. In an LTI system you should be able to EQ this detail, for example increase the treble and should hear the details. But you can try, it’s very obviously impossible.''

The detail retrieval is a result of the relation between FR of headphones next to eardrum, auditory masking function of the human hearing(loud sounds mask the quieter sounds around their frequency range) and the spectra of the music.

Excellent video about auditory masking, he was also my teacher.

''The measurements are just not sufficient to predict how good a headphone sounds. It’s more than frequency response curve.''

I don't know how many times I heard this nonsense. You are just coping, and you know that. I used to have Stax 009S too and compared them to all the cheap headphones I had.

FR consistency of headphones between individuals + leakage tolerance + FR gives everything you need about how headphones sound.

I used to teach control theory, so it’s not copy paste, let’s keep it civil and don’t resort to ad hominem attacks.

Excess phase is measured in the same way you measure frequency/phase response? How does it measure if a system is non linear?

 

[Coverpage]

What is detail retrieval? What is resolving?

In the interest on not wanting to argue on ambiguities, you can omit that and everything else will stand.

 

[usern]

In the interest on not wanting to argue on ambiguities, you can omit that and everything else will stand.

Well, you seem to know more than youtube reviewers who use same lingo so perhaps you could describe it better than "this headphone is resolving, this not"
Comparative definition is gatekeeping, because most people don't have access to such headphones.

 

[MacClintock]

but I also disagree a bit here, the susvara can do it better it has magical mids but only with a good matched amp like a flux FA10 for example.

The amp just needs to be powerful enough and measure good, otherwise doesn't matter which one.

If you combine a sterile and cold sounding topping A90

Cold? Or even freezing? To me the A90 sounds red, almost like a rainbow. But with extra salt. Also liquid.

for example with a Susvara you won't have the same magic in the mid range, neither the synergy in the bass range. My appologies for sounding like a lunatic, for most here

We are kind of used to this, especially in reviews of "sacred" hifi gear. Just mention Susvara or Chord Dave and then automatically people like you sprout like mushrooms after a rainy day.

 

[TurtlePaul]

@amirm Please make a video about this. I guess a lot of people are confused about how can FR be the only differentiation. I know I am.

I think Amir explains in his review: you don’t only hear the frequency response, you can also hear the resonances.

All headphones and speakers generate sound by vibrating air. The mics measure and record these vibrations and translate the linear response to a frequency response graph and the non-linear response to a distortion graph. Frequency response seems to be pretty dominant in what we hear. Because of auditory masking we can endure quite a bit of low order distortion.

There are other data that can be generated. For example, the impulse response shows which frequencies are delayed relative to others. However, you can’t hear impulse response unless you have a speaker with the tweeter several feet closer to the listening position than the woofer.

There is also a waterfall plot that shows how quickly each note decays. This can be useful to show resonances. However, resonances also appear in the distortion plots and in the frequency response plots (as peaks). Driver resonances can also be seen in the driver impedance plots. Resonances are widely accepted as audible.

The one contention with headphone measurements is that a simulation head and ears are used to simulate human head and ears. For most headphones this works pretty well.

All aspects of sound can be measured with a microphone. Some which aren’t shown in Amir’s data have not been shown to be predictive of speaker/headphone preference.

 

[Tachyon88]

A company that sells headphones for 6k, that don’t even measure well, is laughing all the way to the bank. Even if they did measure well a 50% margin is a 3k gross profit. There’s no way they have 3k of cost in that product. Makes no sense.

I read on the other form that Fang went around a canjam or some other show and asked some vendors what the maximum they think they could charge for a headphone. Not sure if its true or not, but the susvara does cost 6k. You can get it for almost half that price from certain outlets, but still...

 

[JCollins]

Wow. I was not expecting that outcome. Glad I went with the DCA Stealth.

 

[Chagall]

''The measurements are just not sufficient to predict how good a headphone sounds. It’s more than frequency response curve.''

I don't know how many times I heard this nonsense. You are just coping, and you know that. I used to have Stax 009S too and compared them to all the cheap headphones I had.

FR consistency of headphones between individuals + leakage tolerance + FR gives everything you need about how headphones sound.

FR consistency of headphones between individuals - this we don't know until putting the headphones on and measuring the response in situ
+leakage tolerance - same thing I would think as with FR consistency
+ FR gives everything you need about how headphones sound - based on measurements that aren't precise and are usually an average - video

So what do we know?
I'm not trying to be difficult, I just don't get it.