• 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 Susvara Headphone Review

Rate this headphone:

  • 1. Poor (headless panther)

    Votes: 229 57.5%
  • 2. Not terrible (postman panther)

    Votes: 75 18.8%
  • 3. Fine (happy panther)

    Votes: 33 8.3%
  • 4. Great (golfing panther)

    Votes: 61 15.3%

  • Total voters
    398
What does it say to take a legendary piece of equipment (who has been reviewed by thousands of critics over the past 7 years) and take a dump on it. take a vague measurement, and suddenly consider it measuring(=sounding) like a 20$ iem.
It didn't measure the same or sound as that $20 IEM. It did much worse on both fronts. As to thousands of critics, they are all over the place. Here is the first review I found in my google search: https://soundnews.net/headphones/full-size/the-ultimate-hifiman-susvara-review/

"9. Benchmark AHB2 Power Amplifier – My Daily Driver V2.0

Holy mother of Thor, this is dynamics, speed, transient response, resolution and transparency incarnate into a single metal box. There isn’t a clearer, a more resolving or a more transparent sounding amplifier out there, this is a wire with gain amplifier that doesn’t add anything into the mix. It might appear as slightly thin and not as organic sounding, hence scoring much less when it comes to tonality and it never sounded as big as S300 or Volot. You can’t get a more impactful, a faster or cleaner sounding amplifier. You should pair it with a warmer source for a perfect match.


  • Resolution/Details: 10
  • Transient Response: 9.5
  • Dynamics: 9.5
  • Tonality: 7.5
  • Power/Control: 9
  • Stage Size: 8
  • Overall Score: 8.92"

I have highlighted tonality for you. He gives it 7.5 out of 10. This reads to you as the best there is in his book? Seems like without using an amplifier that cooks its response (supposedly), it sounds just like I measured: thin which is due to lack of bass. And not organic like.... OK I give up on this one. :) But maybe that is due to lack of lower treble response.
 
In the 80s, the band "The Art Company" had a song called Susanna. Just for fun's sake, y'all can listen to it again and replace Susanna with Susvara, lol. During the song, there was a phone call and you can imagine it as measurement calling hahaha....


 
I think the big reason why people have trouble with this review is because they equate "I like $headphone" with "$headphone must be good" and then want measurements to confirm their taste.

Of course, measurements aren't always exactly the same due to positioning variation, smoothing, rigs, methodology and may not accurately represent the FR at the ear drum (especially in the treble area). However, the general tonality of the Susvara in this measurement is consistent with measurements done by other reviewers. Having also listened to it, the flat bass area and the generally recessed 1-3kHz cannot be argued with. The distortion measurements are probably of less practical relevance at safe listening volumes. It is easy to see that this kind of tonality is not the most pleasing (in the sense of conforming to average listener preference) and may not reproduce studio recordings faithfully. It's far from horrible, but could certainly be described as flawed.

Still, a lot of people love this headphone. Of course, this could all be dismissed as "bias" or "audiophoolery", but I think there is something different at play too:

The Susvara with its tonality, its reversed polarity and its very "open" design makes music sound different, and for people interested in the Susvara (who are most likely headphone audiophiles who have listened to a lot of headphones), this can be very interesting and desirable. If I had to own only one headphone, I'd choose something roughly Harman-compliant, but I own quite a few headphones, some of which don't measure all that well. I can recognize them as flawed, but I still enjoy listening to them a lot, because they make me appreciate music differently. The same can easily be true for the Susvara and it is possible to see why some people enjoy it as much (besides being impressed by price and reviews). The problem is with being offended when someone doesn't share that positive opinion or insisting that the measurements are somehow wrong because they don't look as good as you think they are. There's nothing wrong with enjoying something that doesn't measure well.
 
How do you know this sales data?

My interpretation of Amir's measurements is that Hifiman tried to do something with phasing effect. This effect could be just enough for the review and hifi people to recommend it over other products. But as we can see, the effect brings downsides in distortion - distortion peaks, phase and FR steps align with each other at multiples of 300 Hz. Reviewers don't measure distortion or phase so these things are opaque and now when they are thoroughly measured this tradeoff is revealed. I think it's failure of Hifiman to explain this effect and its tradeoffs and feature of hifi reviewers only focusing on frequency response and not testing limits when doing measurements. Seems Amir is only one who tests limits - how loud you can push a headphone with distortions below audibility threshold.
Shhh, you might be unraveling Hifiman secret sauce, lol
 
It didn't measure the same or sound as that $20 IEM. It did much worse on both fronts. As to thousands of critics, they are all over the place. Here is the first review I found in my google search: https://soundnews.net/headphones/full-size/the-ultimate-hifiman-susvara-review/

"9. Benchmark AHB2 Power Amplifier – My Daily Driver V2.0

Holy mother of Thor, this is dynamics, speed, transient response, resolution and transparency incarnate into a single metal box. There isn’t a clearer, a more resolving or a more transparent sounding amplifier out there, this is a wire with gain amplifier that doesn’t add anything into the mix. It might appear as slightly thin and not as organic sounding, hence scoring much less when it comes to tonality and it never sounded as big as S300 or Volot. You can’t get a more impactful, a faster or cleaner sounding amplifier. You should pair it with a warmer source for a perfect match.


  • Resolution/Details: 10
  • Transient Response: 9.5
  • Dynamics: 9.5
  • Tonality: 7.5
  • Power/Control: 9
  • Stage Size: 8
  • Overall Score: 8.92"

I have highlighted tonality for you. He gives it 7.5 out of 10. This reads to you as the best there is in his book? Seems like without using an amplifier that cooks its response (supposedly), it sounds just like I measured: thin which is due to lack of bass. And not organic like.... OK I give up on this one. :) But maybe that is due to lack of lower treble response.
Hi Amir,

thanks for your work.

How do you explain Crinacle making incredible 20$ iems and then rating the susvara at the top of his list?
 
It didn't measure the same or sound as that $20 IEM. It did much worse on both fronts. As to thousands of critics, they are all over the place. Here is the first review I found in my google search: https://soundnews.net/headphones/full-size/the-ultimate-hifiman-susvara-review/

"9. Benchmark AHB2 Power Amplifier – My Daily Driver V2.0

Holy mother of Thor, this is dynamics, speed, transient response, resolution and transparency incarnate into a single metal box. There isn’t a clearer, a more resolving or a more transparent sounding amplifier out there, this is a wire with gain amplifier that doesn’t add anything into the mix. It might appear as slightly thin and not as organic sounding, hence scoring much less when it comes to tonality and it never sounded as big as S300 or Volot. You can’t get a more impactful, a faster or cleaner sounding amplifier. You should pair it with a warmer source for a perfect match.


  • Resolution/Details: 10
  • Transient Response: 9.5
  • Dynamics: 9.5
  • Tonality: 7.5
  • Power/Control: 9
  • Stage Size: 8
  • Overall Score: 8.92"

I have highlighted tonality for you. He gives it 7.5 out of 10. This reads to you as the best there is in his book? Seems like without using an amplifier that cooks its response (supposedly), it sounds just like I measured: thin which is due to lack of bass. And not organic like.... OK I give up on this one. :) But maybe that is due to lack of lower treble response.

Amir, it sounded worse than IEM even after EQ at normal listening levels?

The tonality of HD600 is better than HD800s, but after EQ - HD800s is a better headphone to me.
 
Nope. We use the so called "Farina Chirp." This is a continuous sweep using an exponentially increasing frequency tone. Hence the reason it can sound like a bird "chirp." An unusual mathematical aspect of this sweep is that both frequency response and harmonic distortions get measured at the same time as they are literally separated in time. You get one impulse in positive time and a number in "negative time." The negative time impulses represent various order harmonics. Perform an FFT on them and you get the frequency response representation of those Harmonics as well:

index.php


There were NO independent measures for distortion. All came from from the same sweep that produced the frequency response.

There is a good tutorial on this online: https://thinksrs.com/downloads/pdfs/applicationnotes/SR1_SweptSine.pdf

"The log-sine chirp also has the unique advantage of being able to separate distortion response from linear response...."

Professor Farina also had published some videos on it, teaching the concept in one of his classes. Suggest seeking that out if you want to understand more.
I think you're not catching my point.

You agreed that the system is likely to not be LTI. Because you can't apply a filter (perhaps by deriving the inverse transfer function) to make all headphones sound the same. We agree here.

Your argument is lower THD at lower volume, which probably means the system is close to linear. Therefore, the FR curve is still valid. This is where we don't' agree.

A lower THD when you measure the system doesn't mean the system is close to linear. The value of the THD measurement is for that particular input signal.
Non-linear or time-variant (or both) systems mean that superposition does not work. So when you play music (a new set of input), you can't decompose the spectral contents and apply them to the FR curve to derive the output.
 
Hi Amir,

thanks for your work.

How do you explain Crinacle making incredible 20$ iems and then rating the susvara at the top of his list?
Thank you for kind words. On that list, they would not stand to scrutiny for a moment let alone permanently. There is a reason you don't see me making such lists. In a controlled test, he would surely rank headphones differently than in his list.

There is also audiophile nonsense in there such as for HE-1000 that he says Susvara is an improvement on: "A planar that comes unnervingly close estat speed, but not quite. Solid technicalities with excellent tone."

As I just explained there is no such thing as "speed." It is just nonsense terms audiophiles have spun, confusing themselves and others.

On Susvara he says: "Subtle refinements to the already-great HE1000 makes this arguably the best planar in the world." "Arguably? Not sure then? And what I am supposed to do with that declaration? It is not like he can prove that.

He also seems to prefer flat bass as that is his target:

Susvara-S2.jpg


And if that is his target, what about the shortfall in 1 to 2.5 kHz? How could a headphone that misses that region be at the top of his list? Either his target matters, or it doesn't. It can't be both.

Note that in my case, the target fully characterizes my preference.
 
I think you're not catching my point.

You agreed that the system is likely to not be LTI. Because you can't apply a filter (perhaps by deriving the inverse transfer function) to make all headphones sound the same. We agree here.

Your argument is lower THD at lower volume, which probably means the system is close to linear. Therefore, the FR curve is still valid. This is where we don't' agree.

A lower THD when you measure the system doesn't mean the system is close to linear. The value of the THD measurement is for that particular input signal.
Non-linear or time-variant (or both) systems mean that superposition does not work. So when you play music (a new set of input), you can't decompose the spectral contents and apply them to the FR curve to derive the output.
You somehow think that being non-LTI is a good thing for headphones?
 
You agreed that the system is likely to not be LTI. Because you can't apply a filter (perhaps by deriving the inverse transfer function) to make all headphones sound the same. We agree here.
No we don't. "Sound" the same has no meaning. You have to identify what you mean by "sound." If you mean tonality, I stated that you can arrive at statistical tie in controlled listening tests if you EQ. If you mean spatial qualities, while EQ can help a lot there, ultimately there are other factors it can't change (such as driver size, distance to your ear, angle, etc.).
 
You somehow think that being non-LTI is a good thing for headphones?
It's not a good thing. If it's LTI all of us can easily get perfect reproduction of our music, and we don't have variations of headphones.

Because you can transform one headphones into another through a simple filter.

But sadly, this is reality. The physics of headphones and microphones mean that modeling software (there's a lot for microphones) simply doesn't work. So people actually tried for microphones; you simply can't get the same details between microphones. And there might be modelling software for headphones too.
 
Non-linear or time-variant (or both) systems mean that superposition does not work. So when you play music (a new set of input), you can't decompose the spectral contents and apply them to the FR curve to derive the output.
For what purpose? If you mean a human listener, I absolutely can do that. Professional monitors with DSP do this day and and day out (linearizing frequency response) and are used in production of music. By your notion their measured frequency response must be an illusion which of course is not.

Take finite element analysis. We can take a complex, 3-D shape and divide it into small enough segment that we can treat them as flat (i.e. "linear"). Once there, we can perform a ton of analysis using simulation. By your notion that wouldn't work either.
 
No we don't. "Sound" the same has no meaning. You have to identify what you mean by "sound." If you mean tonality, I stated that you can arrive at statistical tie in controlled listening tests if you EQ. If you mean spatial qualities, while EQ can help a lot there, ultimately there are other factors it can't change (such as driver size, distance to your ear, angle, etc.).
Exact same output between two headphones through the designed filter for the same given input.
 
Then they can decide from themselves whether a poorly performing /distortion effect box system is preferable to their subjective taste or not
I think it's also important to note that there is another option: if the cheaper option that adheres to the target FR (e.g. Harman) is also tunable without distortion, then you could add EQ to match your preferred target* and save money.

I'm guessing based on reading through these replies that the susvara has some other differences (e.g. phase) that you can't get with an EQ but you can get elsewhere. In general, the phase of a signal means nothing to our ears. However, phase differences between left and right are very noticeable. One trick to add spatial width in the audio production world is to use a low pass filter to roll off the highs but with different slopes on the left channel and the right channel, but at the same frequency (e.g. 12 dB/oct on L and 24 db/oct on R). The different slopes affect the phase at the target frequency differently, causing not only different phases but this difference also changes. This adds to the sense of space, and could increase what people describe as 'soundstage' or 'etherealness' (you can do this to the lows on each track as well but it's not always desirable - e.g. kick or bass guitar you generally want mostly mono and in the center or it can more easily sound muddy and messy).

Of course, doing this will affect amplitude. However, there is a way to affect phase without affecting amplitude: all pass filters - these reverse phase at the target frequency (so 180 degrees) and roll off the phase change nearby without affecting amplitude. There's probably other math-y filter stuff you can do, as e.g. MFreeformPhase (https://www.meldaproduction.com/MFreeformPhase) lets you change phases as-desired per-frequency. So if you were persistent enough, you could probably set up an EQ plugin and a phase response plugin into Peace/Equalizer APO and get a response much closer to susvara than just EQ.

I wonder what it would sound like to capture an impulse response of the susvara using a dirac spike, then plug that into a reverb plugin that takes an IR, and play back music on a headphone that aligns really closely to harman (e.g. the stealth) - how close to the susvara would it sound? The IR should capture the phase per-frequency as well right? Would be interesting but I don't know if it's possible.

This all means you could find a (possibly cheap) headphone and create your own tailored soundstage by messing with phase differences at different frequencies on the left and right channels (the more 'objectively better' headphone, the more accurately you can tailor your soundstage to your subjective liking). Maybe you find something you like better than the susvara? But I definitely get the sense that some people/reviewers find these things in headphones and like them, and because of that they attribute it to "must be what the artist intended" or "this was always here in the music but I couldn't hear it until I put these headphones on" when that isn't the case. Non-linearities, distortion, and phase differences can sound pleasing - it's why tape saturation is extremely popular for audio producers (people pay big bucks for sought-after analog tape machines to mix their songs on and why there are so many saturation plug ins) and why people like tube amps. There's nothing wrong with liking it, but there's also nothing wrong with buying a good quality cheaper headphone and adding saturation, distortion, phase changes, EQ, etc yourself.

*This implies you are able to, e.g. you are on a computer, or your DAC supports EQ with enough resolution, etc. You may also just happen to like the response of the susvara and don't want to fiddle with stuff and just plug them in and listen - nothing wrong with that either.
 
Last edited:
The physics of headphones and microphones mean that modeling software (there's a lot for headphones) simply doesn't work.
No one is aiming to model a headphone. We are trying to estimate its tonality and we can do so with high level of success. Indeed a lot of Harman research was performed by using a surrogate headphone to emulate the frequency response of other headphones. Before running with this, they verified the efficacy of the method:

1703641109409.png


As you see, the surrogate headphone equalized to the real headphone's response produced statistically even preference from listeners in controlled tests.

So both in theory and matter of actual testing, your assumption is wrong here.

Beyond tonality, the problem gets hard but tonality is so important and fundamental to headphone performance.
 
Exact same output between two headphones through the designed filter for the same given input.
You can't get that out of the same headphone. Move it one millimeter and the response changes. By your logic, you don't know how that headphone sounds even when wearing it!!!
 
For what purpose? If you mean a human listener, I absolutely can do that. Professional monitors with DSP do this day and and day out (linearizing frequency response) and are used in production of music. By your notion their measured frequency response must be an illusion which of course is not.

What you are trying to make a case for is that the non-LTI system, can be modelled as a LTI system through your method of measuring FR.

I agree. But how well the model holds up depends on how non-linear the system is. So, case in point, putting any filter on the headphones doesn't make 1 headphone sound like another headphone. Which mean that this linearized model is not accurate.

Take finite element analysis. We can take a complex, 3-D shape and divide it into small enough segment that we can treat them as flat (i.e. "linear"). Once there, we can perform a ton of analysis using simulation. By your notion that wouldn't work either.
This is the perfect analogy.

Finite element tries to use equations for physics that are linearized and apply them to a 3D or 2d geometry to simulate what will happen in real life.

So you have to do 3D or 2D "drawing of the driver, diaphragm, electrical circuit etc) and it tries to simulate the magnetic flux, electricity, mechanical vibrations, and even acoustic. So this is available for multi-physics FEA software, which is obscenely expensive.

So you're right, you can use FEA to model a headphone, but the result of the simulation when comparing input and output won't be linear.
 
You can't get that out of the same headphone. Move it one millimeter and the response changes.
Firstly, this shows the weakness of the FR measurements you make to evaluate the headphones, if they change by one milimeter.

Secondly, you can get this out of the headphones if they are LTI. Just have them on two separate rigs, measure the FR for both and apply the filter on either.

We both know they'll sound different even after the filter. (have different outputs)

By your logic, you don't know how that headphone sounds even when wearing it!!!
I honestly don't I understand what you mean.

We do know, they sound different even if the filter is applied.
 
Of course, doing this will affect amplitude. However, there is a way to affect phase without affecting amplitude: all pass filters - these reverse phase at the target frequency (so 180 degrees) and roll off the phase change nearby without affecting amplitude. There's probably other math-y filter stuff you can do, as e.g. MFreeformPhase (https://www.meldaproduction.com/MFreeformPhase) lets you change phases as-desired per-frequency. So if you were persistent enough, you could probably set up an EQ plugin and a phase response plugin into Peace/Equalizer APO and get a response much closer to susvara than just EQ.
Nice, I will look into using some VST with APO to mess with phase. Do you know more free (or cheap) VST plugins that can do that besides MFreeformPhase? APO actually has all pass filters also and they do change sound if you split the channels before applying it!
OHUYdql.png


@amirm did you measure group delay for only one channel?
 
Last edited:
Back
Top Bottom