HIFIMAN Susvara Headphone Review

[ergre]

I agree with you. And anyone who have tried knows it’s not possible.

The point is it’s not an LTI system. Frequency response through a sine sweep cannot completely describe the system.

Why not? It is a single driver no crossover and it is very close to the ear. for A loudspeaker I can understand but for headphones?

 

[solderdude]

One thing that could be deduced from the "static" sound amirm was hearing is that the poor Susvara's diaphragm was being smashed onto the magnets due to the SPL of 114db, bearable to neither human ears nor headphones.

highly unlikely as the largest excursions would be present for lowest frequencies only not at 1kHz to 5kHz where the excursions are much smaller.
Besides it seems resonance related.
And the reason it is so insensitive is because the magnets are further apart (from the membrane) so another reason that this cannot be the case.

 

[solderdude]

Why not? It is a single driver no crossover and it is very close to the ear. for A loudspeaker I can understand but for headphones?

It can't. An artificial pinna and earcanal will differ from those of actual humans which usually do not comply to standards.
Then there is the angle of the driver, break-up, resonances, distortion, dynamic compressions, seal to name but a few aspects.

 

[ergre]

It can't. An artificial pinna and earcanal will differ from those of actual humans. Then there is the angle of the driver, break-up, resonances, distortion, dynamic compressions, seal to name but a few aspects.

They would influence the the respons. Those are more about the precision of the measurements Not the completeness. It is not the same situation as with a loudspeaker where indirect sound has an influence on what you hear and a frequency response would be incomplete.

 

[solderdude]

Unless you measure the response at various levels you can not detect dynamic compression for instance as the response would differ so a single FR won't tell you that.
Nor will you see differences in distortion.
Then there is smoothing of measurements and different measurement fixtures and standards, positioning on the fixture, averaging or not.
Then there are conditions like impedance over frequency which can have a tonal influence and requires more than just an FR plot.
Resonances often can be seen in the FR but not always clearly. For instance it could show as a dip or peak so FR won't say everything there is to know here either.

A single FR plot simply does NOT say everything. If it would Amir would only show a single FR plot only.

 

[Blockader]

Why not? It is a single driver no crossover and it is very close to the ear. for A loudspeaker I can understand but for headphones?

A basic sine sweep cannot reveal how the frequency response of headphones will alter when they interact with the acoustic impedance of an individuals ear canal. That has been said, planar magnetic headphones(especially Hifiman's) have exceptional frequency response consistency across various individuals' ear canals due to their very low resonance frequency. (also as solderdude said dynamic compression is a thing and I'm sure these headphones change their FR around 114db because the driver is about to bottom up at that point)

Let's admit it. These headphones suck. I bought them at some point because of how much hype was surrounding them. I decided to sell them after 15 minutes of listening.

 

[ergre]

Unless you measure the response at various levels you can not detect dynamic compression for instance as the response would differ so a single FR won't tell you that.
Nor will you see differences in distortion.
Then there is smoothing of measurements and different measurement fixtures and standards, positioning on the fixture, averaging or not.
Then there are conditions like impedance over frequency which can have a tonal influence and requires more than just an FR plot.
Resonances often can be seen in the FR but not always clearly. For instance it could show as a dip or peak so FR won't say everything there is to know here either.

A single FR plot simply does NOT say everything. If it would Amir would only show a single FR plot only.

The dynamic compression is a form of nonlinear distortion and I see that Amir measured that at different levels indeed. I am also pretty sure that someone can estimate that and that it wil be very low. Impedance has more to do with what the amplifier sees and only is audible when the amplifier is not good enough to drive the headphone. Also it what You would hear is a change in frequency response.
There May be a lot of details and potential problems but the frequency response is essentially all you need for the linear part. I don’t see a mechanism like indirect sound for loudspeakers that would make this statement problematic.

 

[Chagall]

A basic sine sweep cannot reveal how the frequency response of headphones will alter when they interact with the acoustic impedance of an individuals ear canal. That has been said, planar magnetic headphones(especially Hifiman's) have exceptional frequency response consistency across various individuals' ear canals due to their very low resonance frequency. (also as solderdude said dynamic compression is a thing and I'm sure these headphones change their FR around 114db because the driver is about to bottom up at that point)

Let's admit it. These headphones suck. I bought them at some point because of how much hype was surrounding them. I decided to sell them after 15 minutes of listening.

I think while @solderdude and @amirm have different conclusions in their reviews the most awkward thing is that both of them are right.

But considering the price and the competition it's safe to say that the headphones are outdated.

 

[solderdude]

The dynamic compression is a form of nonlinear distortion and I see that Amir measured that at different levels indeed. I am also pretty sure that someone can estimate that and that it wil be very low. Impedance has more to do with what the amplifier sees and only is audible when the amplifier is not good enough to drive the headphone. Also it what You would hear is a change in frequency response.
There May be a lot of details and potential problems but the frequency response is essentially all you need for the linear part. I don’t see a mechanism like indirect sound for loudspeakers that would make this statement problematic.

I'll simply repeat: a single FR plot on some standard fixture does NOT say anything else than how that particular one measured on that particular fixture under specific circumstances.
It does say something about the tonal balance of that headphone on that particular fixture.

It would be great if you were able to tell how much distortion there is from an FR plot, how severe cone break-up or partial resonances on a membrane are, how well it couples to one's head, how position sensitive it is, how long resonances may last etc.
Good luck with that. In the 10 years I have been measuring and listening to headphones I have never found the correlation you say is there.

 

[Drengur]

Ah... yes. The old "since this industrially manufactured consumer products does not do well in objective tests, we need to change the tests." My suggestion is that we ditch the Harman curve for the new Susvara curve, that way, those pesky 20 dollar in-ears will deviate as far as they deserve. While we are at it, I've always found the exhaust-fumes of the '76 Bronco to be quite revitalizing and if it does not fit the current emission criteria, well... have you inhaled it yourself?

I understand that owners of these expensive headphones need to find some reason for their preference, and that it can be difficult to cope with bad measurement results. There are a few classical retorts that have not been used yet in this thread so I came up with a couple of discussion points, a cope-hope if you will:

1. The Topping is a toy, you should be using the Le Arnaque Mk2 with the Falsk Kötbulle cable
2. But what did the artist intend? Do you think Beethoven cared about distortion? He was deaf you know.

 

[ergre]

I'll simply repeat: a single FR plot on some standard fixture does NOT say anything else than how that particular one measured on that particular fixture under specific circumstances.
It does say something about the tonal balance of that headphone on that particular fixture.

It would be great if you were able to tell how much distortion there is from an FR plot, how severe cone break-up or partial resonances on a membrane are, how well it couples to one's head, how position sensitive it is, how long resonances may last etc.
Good luck with that. In the 10 years I have been measuring and listening to headphones I have never found the correlation you say is there.

I think you need to think a bit harder about the model of the system and how that relates to measurement. I am not saying that I know but you seem very unclear about that for someone with your experience.
maybe that is something that can help you make more sense.

 

[Art of sound]

This is a review, listening tests, EQ and detailed measurements of the HIFIMAN Susvara Planar Magnetic Headphone. It was kindly drop shipped by a member and costs US $6,000.
View attachment 336679
Owner was advised that the Benchmark AHB2 power amplifier would be the right device to power it so I tested it with that as well as my normal Topping A90 headphone amplifier.

Hifiman aces the design of the Susvara coming in a modern and elegant looking package. It is rather lightweight for its size as well, making it comfortable to wear. The only minor issue there was feeling the strap on my head but otherwise, it was excellent and I almost forgot I was wearing it. You can't see it but the metal strap is rectangular which I am not fond of in general but it kind works here. Whether you are getting $6,000 worth of luxury here I don't know but a good effort has been made to make you feel you are getting a flagship headphone.

If you are new to my headphone measurements, I highly recommend that you watch my video tutorial on headphone measurements which ironically uses Hifiman HE-6 as an example.

Hifiman Susvara Headphone Measurements
Let's start with frequency response of the Susvara and compare it against our desired target:
View attachment 336680
I was very surprised by the highly variable nature of the measurements. It is beyond noisy. Other measurements that use high level of smoothing don't show this as much and with it, hide potential issues there. That aside, we lack bass which is typical of a number of other headphones. Response is also insufficient in the 1 to 3 kHz which likely is there to keep the headphone form sounding too bright (given the lack of bass). Developing eq to correct the high frequency response is going to be a bit challenging but we will try using the deviation for our target response:
View attachment 336681

I was disappointed to see high levels of distortion where our hearing is most sensitive:
View attachment 336682
Narrow spikes usually indicate resonances so they are audible beyond their distortion characteristics and point to design issues. We can also see how fast they escalate. 94 to 104 dBSPL causes a 3 3X or so increase but go to 114 dBSPL and they shoot through the roof. What's more, if you go just 1 dB beyond, the drivers bottom out and just buzz! Wanting to rule out Topping A90's own distortion I switched to Benchmark AHB2. Alas, I was not able to get above 113 dBSPL as AHB2 ran out of voltage drive. Taking this into account, we see that the problem is the headphone and not the amps:
View attachment 336683

To make a more level matched comparison, I tested again but at 104 dBSPL:
View attachment 336684
Ignoring minor deviations, we see that the response is identical so there is really no need for the Benchmark AHB2 in this application. The Topping A90 is actually more capable of driving the Susvara. As noted, this is due to AHB2 not having enough voltage drive due to rather "high" impedance so SUSVARA (as compared to a speaker):
View attachment 336685
The Susvara is very insensitive so you definitely need a very capable amplifier to drive it:
View attachment 336686

Back to our distortion, here it is in absolute levels:
View attachment 336687
Even at 94 dBSPL we are exceeding our target between 4 and 5 kHz. I searched and there is at least one other measurement on the web that shows a similar issue so please don't say this sample must be broken. There is clearly a design issue here. The super messy Group Delay points to multiple sound sources (i.e. resonances):
View attachment 336688
I don't think I have ever seen such broad messiness across almost the entire audible band above bass.

Let's agree that objectively we are not doing well here. But "how does it sound?"

Hifiman Susvara Listening Tests and Equalization
I was listening to music on my everyday headphone and decided to just switch to Susvara without going back to my reference test clips. First thing I noticed -- which was odd -- was the high frequencies being distorted. When I went back to my standard tracks, this did not stand out. There, the tonality was "OK" but nothing exciting for me. So out came the EQ tools to correct the frequency response errors:
View attachment 336689
Ignore Band 5 for now. Addition of the two filters at 1900 and 4500 Hz filled in the treble area but now the headphone was bright. So I added a pair of filters to boost the bass region with similar slop to the inverse of Susvara frequency response. Once there, overall response was far better with good bit of bass and much improved spatial qualities. Alas, I was still sensing brightness. I pulled down the response at 4500 Hz and that helped but it was still bright. So I broke my usual rule of not messing with > 8 kHz and dialed in a negative notch filter at 12.7 kHz. That tailored the tonality to what I liked.

I sat back and started to enjoy the fidelity which was at times quite excellent. That was at low and medium levels. With my dynamic tracks I crank up the volume and I was quite surprised when I started to hear static when I did not even have it super loud. Suspecting EQ being the problem, I defeatured the whole thing and problem remained (although required a bit more volume to get there). This is a showstopper in my book as the problem is not even the elevated bass with EQ where the problem usually occurs. I think the issue is the high frequency issues we have found.

I know some will claim to not hear this or complain that the above level is too loud. I have my standards of how much power a headphone needs to be able to handle and the Susvara falls way short of that. I want perfection at this price point yet even if this was a $400 headphone, I would complain about this problem.

Conclusions
Hifiman nails the look and overall industrial design of the Susvara delivering a headphone that puts a smile on my face every time I look at it. It is also very comfortable to wear. Alas, objective measurements show issues across the board. Response is quite variable beyond typical noise. Has deficiencies against our target. And importantly, appears to have serious resonance issues that manifest themselves in distortion graphs. Listening tests confirm those distortion finding with the headphone creating static at levels that should not remotely be possible with a flagship headphone.

If you don't exceed medium loud levels distortion would not impact you. And maybe you are OK with this type of tuning. Neither is my cup of tea, leaving me with the sadness that such a great looking headphone is saddled with serious technical problems.

I can't recommend the Hifiman Susvara headphone. I also don't recommend pairing it with Benchmark AHB2. Even though it can almost drive it, it costs a lot more than a capable headphone amplifier.

P.S. I know some are going to be upset with the findings in this review. I highly recommend you watch my video tutorial on headphone testing before commenting.

------------
As always, questions, comments, recommendations, etc. are welcome.

Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/

is that a clamping issue on susvara below 200hz. i thought this is an endgame headphone for most people. very interesting to see that in amir's take.

 

[IAtaman]

Why not? It is a single driver no crossover and it is very close to the ear. for A loudspeaker I can understand but for headphones?

A headphone on an ear is a variable and a high impedance source driving a variable and a high impedance load. What will be the final FR of the coupled systems depends on the headphone and the ear, and can be very different than what the measurements show.

 

[flowjm]

The Emperor has no clothes.

 

[Robbo99999]

I have yet to hear a good sounding planar magnetic headphone really, each one I've heard sounded wrong - hollow in the bass and tinny in the treble, very fatiguing to listen to.

Which ones you listened with, and did you use EQ?

 

[Robbo99999]

Can you EQ one headphone to sound the same as another headphones?


That is the TF (represented as the FR Curve) of headphone 1, vs the TF (represented as the FR Curve) of headphone 2, going through a filter should sound the same for an LTI system.

Unit to unit variation will get in the way of that pretty significantly unless you're measuring the actual units you're using on a GRAS that happens to be sitting in front of you. But even then if you have 2 different models of headphone I don't think you can get them to sound exactly the same due to what's going on above 8khz which is where these rigs lose their accuracy, and additionally I think there's specific interactions of your own anatomy which will further make 2 different headphone models sound different. So in conclusion, I think you can get two different models of headphone to sound broadly similar, but I doubt you can get them to sound exactly the same.

 

[ergre]

A headphone on an ear is a variable and a high impedance source driving a variable and a high impedance load. What will be the final FR of the coupled systems depends on the headphone and the ear, and can be very different than what the measurements show.

That is not what I was asking about at all. For example when you measure a box you need 3 measurements because of the idea of a box we have. I can think of all kinds of problems when I measure that box. Maybe it is not that rigid or something. When you think about it more in theory what measurements do you need and why in excess of what Amir is measuring.

 

[nyxnyxnyx]

Don't think so. For $6k I expect excellence in every respect. That simple. And this headphone just does not cut it.

Fair point. For this price and/or this level of prestige, I hoped that it was truly a technical marvel and a wonderful listening experience.

 

[jody2k]

I don't think anyone ever claimed the Susvara was the king of anything, because of it's high pricepoint this headphone has always been tested in a very critical way. But I did my own research I found most trustable review sources (and with this I don't mean youtube) claiming this headphone until today is still considered in the top of the list. Like hasen't the Susvara been on the market forever? Extensively tested by many experts in the last 7 Years? So why does this "measurement problem" come up all of a sudden only in this forum in this topic? Why only here by 1 person?

 

[solderdude]

That is not what I was asking about at all. For example when you measure a box you need 3 measurements because of the idea of a box we have. I can think of all kinds of problems when I measure that box. Maybe it is not that rigid or something. When you think about it more in theory what measurements do you need and why in excess of what Amir is measuring.

A headphone is a technically highly flawed device. Even much more so than a speaker.
Just measuring amplitude per frequency in a single position doesn't tell it all.
For instance dynamic behaviour is not measured and distortion isn't as telling as people think it is.

The sad part is headphone measurements are indicative at best and FR and distortion do not tell the whole story.
There is interaction with one's head and there is the highly variable brain with all its quirks when it comes to relating measurements to how it sounds.
It isn't about tonality only but also about compensation for HRTF/HRIR as well with headphones. For speakers this is already 'embedded', not so with headphones.