HIFIMAN Susvara Headphone Review

[Blockader]

I would say that not everyone likes full Harman bass, or it can greatly depend on the recording, and I can say that my in-ear quasi-anechoic and near-field measurements exhibit flat bass sans room tilt.

Anyways, my focus was on my having not seen many Susvara frequency response curves that overshoot the ear gain target like in the measurement that you focus on.

In regard to ear gain levels, compare my HE1000se and Sennheiser HE-1 measurements to my outdoor 30-degree neutral Genelec 8341A measurement:

View attachment 341535
Figure 1: Sennheiser HE-1 in-ear measurement. Invalid phase response measurement.

View attachment 341536
Figure 2: HE1000se in-ear measurement.

View attachment 341537
Figure 3: 30-degree HRTF from calibrated Genelec 8341A after subtracting the measurement of the lone in-ear mic. The phase respond measurements through the Genelecs are invalid. Note the differing locations of the treble nulls; the headphone ones are consistent with my 90-degree speaker HRTF measurements.

View attachment 341538
Figure 4: Indoor in-ear response for both 30-degree channels playing in phase.

View attachment 341539
Figure 5: Not far from my centered speaker HRTF (after subtracting the lone mic response).

tl;dr: Neutral speakers are brighter than many "bright" headphones. Strings in live concerts can also be surprisingly bright compared to Harman playback.

If you noticed, I did not mention harman full bass. I do not agree with harman bass either and Sean Olive also noted that some people prefer less bass.

my point is basically, with any true neutral target curve Susvara bass will appear recessed and mids/ear gain boosted. That's how they sound too. I owned Susvara for couple of days.

Also, outdoor measurements are pretty much useless.(ground plane outdoor measurements can be viable to give an idea about anechoic bass response of speakers) What makes monopole speakers bass boosted is the reflections. Outdoor = no reflections. Music is not produced with speakers on outdoors, speaker outdoor targets are not suitable to form an idea on how headphones should sound.

 

[Robbo99999]

If you noticed, I did not mention harman full bass. I do not agree with harman bass either and Sean Olive also noted that some people prefer less bass.

my point is basically, with any true neutral target curve Susvara bass will appear recessed and mids/ear gain boosted. That's how they sound too. I owned Susvara for couple of days.

Also, outdoor measurements are pretty much useless.(ground plane outdoor measurements can be viable to give an idea about anechoic bass response of speakers) What makes monopole speakers bass boosted is the reflections. Outdoor = no reflections. Music is not produced with speakers on outdoors, speaker outdoor targets are not suitable to form an idea on how headphones should sound.

Albeit Figure 4 is his in ear response of his Genelecs in a room at a 30 degree angle. He'd probably be best off to apply a least some smoothing to the measurements and to overlay them on the same graph so it can be more easily seen the differences between his in-ear headphone measurements vs his in-ear speaker measurements. I'm assuming he could create an EQ for his headphones that would equal his in-ear response of his Genelec speakers, that would be interesting....and all he might need to do after that to finetune the EQ further would be to boost the bass a little in the headphones due to less tactile bass felt with headphones.

 

[Mr. Haelscheir]

I'm assuming he could create an EQ for his headphones that would equal his in-ear response of his Genelec speakers, that would be interesting....and all he might need to do after that to finetune the EQ further would be to boost the bass a little in the headphones due to less tactile bass felt with headphones.

That is exactly what I have done. XD And if could involve sitting still while literally tethered to my PC with these very uncomfortable in-ear mics in my head among other discomforts, but the results are absolutely worth it. The smoothed graph is this guy here:

That is for the left ear, my right ear possibly indeed having some physiological differences or at least something that causes the in-ear microphone to be consistently seated differently from the left. I did some small per-channel peaking filter adjustments to correct any major perceived channel imbalances within a sine sweep. I do like to apply a 5 dB bass shelf at 100 Hz, whereby some recordings with binaural head-tracking can still present quite the bass impact. I haven't found much advantage to applying a general tilt, whereby this should suffice for near-field levels. Classical recordings differ enough in how they were mixed or whatnot anyways.

More details in this wall of text and its links: https://www.reddit.com/r/headphones/comments/189dpt3/_/kbs5o4q
I have yet to compose the Head-fi post presenting my findings after also EQing my HE1000se with large aperture NMD (NTRAX Mod Design) pads (custom "Type A6" parameters) to the same target (as limited by upper treble variations) where the phase responses and hence impulse and step responses were very similar, though there were still differences in group delay (the HE1000se still had a cleaner bass group delay than the Meze Elite), CSD, and of course distortion, but they did sound very similar regardless other than subjective effects of pad feel, and single-sample transients still sounding more "incisive" on the HE1000se (and Arya Stealth) at least at more extreme volume levels (probably just excitation of resonances and hence technically an overshoot distortion). But this mind my binaural head-tracking impressions go into an off-topic rabbit hole covered by those linked threads.

What I can say is that at Roy Thomson Hall, experiencing Mahler 5 for the second time, but from the mezzanine (moderately loud compared to the ridiculousness of hearing it from Row O of the orchestra level), and this recent Saturday hearing Brahms Piano Concerto No. 1 from Row C, the similarities to the best recordings I have heard were uncanny. My remaining discontents other than hoping to get a better HRTF measurement regards the limitations of the HRTF rendering implementation's (I use SPARTA AmbiBIN in Reaper) ability to convincingly image the stereo field further away, and the accuracy of the phase response processing and hence the combined magnitude response from playing from both virtual speakers in phase, this differing between different HRTF renderers, some making pink noise sound brighter than the other though the channels in isolation may measure the same.

 

[Robbo99999]

That is exactly what I have done. XD And if could involve sitting still while literally tethered to my PC with these very uncomfortable in-ear mics in my head among other discomforts, but the results are absolutely worth it. The smoothed graph is this guy here:

That is for the left ear, my right ear possibly indeed having some physiological differences or at least something that causes the in-ear microphone to be consistently seated differently from the left. I did some small per-channel peaking filter adjustments to correct any major perceived channel imbalances within a sine sweep. I do like to apply a 5 dB bass shelf at 100 Hz, whereby some recordings with binaural head-tracking can still present quite the bass impact. I haven't found much advantage to applying a general tilt, whereby this should suffice for near-field levels. Classical recordings differ enough in how they were mixed or whatnot anyways.

More details in this wall of text and its links: https://www.reddit.com/r/headphones/comments/189dpt3/_/kbs5o4q
I have yet to compose the Head-fi post presenting my findings after also EQing my HE1000se with large aperture NMD (NTRAX Mod Design) pads (custom "Type A6" parameters) to the same target (as limited by upper treble variations) where the phase responses and hence impulse and step responses were very similar, though there were still differences in group delay (the HE1000se still had a cleaner bass group delay than the Meze Elite), CSD, and of course distortion, but they did sound very similar regardless other than subjective effects of pad feel, and single-sample transients still sounding more "incisive" on the HE1000se (and Arya Stealth) at least at more extreme volume levels (probably just excitation of resonances and hence technically an overshoot distortion). But this mind my binaural head-tracking impressions go into an off-topic rabbit hole covered by those linked threads.

I see, but you've not overlaid your in-ear response of the Susvara with the in-ear speaker response on the same graph so we can see how different or close they are. I thought that's how this whole conversation started, did you want to make the point that the Susvara was already close to what your in-ear response is for your speakers in a room?

EDIT: sorry, you don't have the Susvara now that I read back your other posts. So I've lost the point that was originally being made? Either way, it would be interesting to see some of your in-ear headphone responses overlaid on your in-ear speaker response.

 

[Mr. Haelscheir]

I see, but you've not overlaid your in-ear response of the Susvara with the in-ear speaker response on the same graph so we can see how different or close they are. I thought that's how this whole conversation started, did you want to make the point that the Susvara was already close to what your in-ear response is for your speakers in a room?

Oh. I never said I had a Susvara. This was just an overblown debate on my impression of typical online Susvara FR graphs giving me the impression of their having a bit less ear gain than the HE1000se or Arya for example, and my having read that some find the Susvara more relatively "laid-back" (maybe it was Joshua Valour, whose opinions I can also find iffy) compared to the HE1000se, for example. My free-field in-ear measurements at least demonstrated what real speaker ear gain looks like.

Anyways, here is the overlay you asked for:

These magnitude plots are normalized around 1 kHz except for the cyan plot which is match with the purple plot from 120 Hz to 800 Hz.

• Red trace: HE1000se with stock pads.
• Green trace: Meze Elite with hybrid pads EQed to my previous "V3 PEQ" based on taking my by-ear-adjusted headphones.com-based Harman EQ and refining it with in-ear mics.
• Teal trace: The original smoothed EQed free-field left ear and left speaker 30-degree response ("R 30 L" means turning my head right 30 degrees relative to the single Genelec speaker and measuring my left ear) through the Meze Elite with hybrid pads as based on my outdoor Genelec measurements; SPARTA AmbiBIN was set to simulate a virtual speaker in an anechoic room, whereby changes in direction are EQed around this base response.
• Purple trace: The latest measurement of that free-field response, whereby sometime in November, something changed with my measurements or I had I lost track of a bass shelf filter causing deviation from the original free-field bass response below 120 Hz by around 3 dB at 20 Hz; I haven't confirmed whether this was measurement drift or my having forgotten to reenable a filter.
• Orange trace: Same as the previous, but with "Filter: ON LS Fc 100 Hz Gain 5 dB" applied on Equalizer APO. Had the purple trace's bass measured just like on the teal trace, then this bass shelf would have brought the sub-bass up to that of the HE1000se in this normalization, the remaining lower levels from 100 Hz to 500 Hz probably being comparable to my "Harman clarity EQ".
• Cyan trace: The right virtual speaker as measured at the left ear.
• Salmon trace: The actual "L 30 L" measurement from the Genelec 8341A with 1/12 octave smoothing, its having clear differences from the result of EQing the "R 30 L" measurement on top of the EQ applied by AmbiBIN for that virtual speaker and letting it apply the EQ for sound coming from the other virtual speaker. I believe https://www.earfish.eu/ normalized the measurements to some SOFA standard rather than providing a file that allows the original measurements per direction to be EQed onto headphones that have been EQed perfectly flat with the same measurement gear.
• Magenta trace: The combined response of both the left and right 30-degree virtual speakers playing simultaneously as measured through my left ear. My indoor Genelec measurements show that the ear gain should not be so reduced, so this is a phase error with the HRTF rendering software, but maybe helps the music sound "smoother" compared to Binauraliser NF whose measured combined response is brighter.

Considering the graph below (I sourced this copy from https://forum.hifiguides.com/t/grap...n-ear-over-ear-and-harman-linear-in-room/9270), my free-field response is expectedly brighter than Harman over-ear levels, whether or not the lack of crossfeed has a bearing.

 

[Sebby]

Oh. I never said I had a Susvara. This was just an overblown debate on my impression of typical online Susvara FR graphs giving me the impression of their having a bit less ear gain than the HE1000se or Arya for example, and my having read that some find the Susvara more relatively "laid-back" (maybe it was Joshua Valour, whose opinions I can also find iffy) compared to the HE1000se, for example. My free-field in-ear measurements at least demonstrated what real speaker ear gain looks like.

Anyways, here is the overlay you asked for:

View attachment 342138

These magnitude plots are normalized around 1 kHz except for the cyan plot which is match with the purple plot from 120 Hz to 800 Hz.

• Red trace: HE1000se with stock pads.
• Green trace: Meze Elite with hybrid pads EQed to my previous "V3 PEQ" based on taking my by-ear-adjusted headphones.com-based Harman EQ and refining it with in-ear mics.
• Teal trace: The original smoothed EQed free-field left ear and left speaker 30-degree response ("R 30 L" means turning my head right 30 degrees relative to the single Genelec speaker and measuring my left ear) through the Meze Elite with hybrid pads as based on my outdoor Genelec measurements; SPARTA AmbiBIN was set to simulate a virtual speaker in an anechoic room, whereby changes in direction are EQed around this base response.
• Purple trace: The latest measurement of that free-field response, whereby sometime in November, something changed with my measurements or I had I lost track of a bass shelf filter causing deviation from the original free-field bass response below 120 Hz by around 3 dB at 20 Hz; I haven't confirmed whether this was measurement drift or my having forgotten to reenable a filter.
• Orange trace: Same as the previous, but with "Filter: ON LS Fc 100 Hz Gain 5 dB" applied on Equalizer APO. Had the purple trace's bass measured just like on the teal trace, then this bass shelf would have brought the sub-bass up to that of the HE1000se in this normalization, the remaining lower levels from 100 Hz to 500 Hz probably being comparable to my "Harman clarity EQ".
• Cyan trace: The right virtual speaker as measured at the left ear.
• Salmon trace: The actual "L 30 L" measurement from the Genelec 8341A with 1/12 octave smoothing, its having clear differences from the result of EQing the "R 30 L" measurement on top of the EQ applied by AmbiBIN for that virtual speaker and letting it apply the EQ for sound coming from the other virtual speaker. I believe https://www.earfish.eu/ normalized the measurements to some SOFA standard rather than providing a file that allows the original measurements per direction to be EQed onto headphones that have been EQed perfectly flat with the same measurement gear.
• Magenta trace: The combined response of both the left and right 30-degree virtual speakers playing simultaneously as measured through my left ear. My indoor Genelec measurements show that the ear gain should not be so reduced, so this is a phase error with the HRTF rendering software, but maybe helps the music sound "smoother" compared to Binauraliser NF whose measured combined response is brighter.

Valor is right in that case: HE1000 is more "tiring".

 

[Robbo99999]

Oh. I never said I had a Susvara. This was just an overblown debate on my impression of typical online Susvara FR graphs giving me the impression of their having a bit less ear gain than the HE1000se or Arya for example, and my having read that some find the Susvara more relatively "laid-back" (maybe it was Joshua Valour, whose opinions I can also find iffy) compared to the HE1000se, for example. My free-field in-ear measurements at least demonstrated what real speaker ear gain looks like.

Anyways, here is the overlay you asked for:

View attachment 342138

These magnitude plots are normalized around 1 kHz except for the cyan plot which is match with the purple plot from 120 Hz to 800 Hz.

• Red trace: HE1000se with stock pads.
• Green trace: Meze Elite with hybrid pads EQed to my previous "V3 PEQ" based on taking my by-ear-adjusted headphones.com-based Harman EQ and refining it with in-ear mics.
• Teal trace: The original smoothed EQed free-field left ear and left speaker 30-degree response ("R 30 L" means turning my head right 30 degrees relative to the single Genelec speaker and measuring my left ear) through the Meze Elite with hybrid pads as based on my outdoor Genelec measurements; SPARTA AmbiBIN was set to simulate a virtual speaker in an anechoic room, whereby changes in direction are EQed around this base response.
• Purple trace: The latest measurement of that free-field response, whereby sometime in November, something changed with my measurements or I had I lost track of a bass shelf filter causing deviation from the original free-field bass response below 120 Hz by around 3 dB at 20 Hz; I haven't confirmed whether this was measurement drift or my having forgotten to reenable a filter.
• Orange trace: Same as the previous, but with "Filter: ON LS Fc 100 Hz Gain 5 dB" applied on Equalizer APO. Had the purple trace's bass measured just like on the teal trace, then this bass shelf would have brought the sub-bass up to that of the HE1000se in this normalization, the remaining lower levels from 100 Hz to 500 Hz probably being comparable to my "Harman clarity EQ".
• Cyan trace: The right virtual speaker as measured at the left ear.
• Salmon trace: The actual "L 30 L" measurement from the Genelec 8341A with 1/12 octave smoothing, its having clear differences from the result of EQing the "R 30 L" measurement on top of the EQ applied by AmbiBIN for that virtual speaker and letting it apply the EQ for sound coming from the other virtual speaker. I believe https://www.earfish.eu/ normalized the measurements to some SOFA standard rather than providing a file that allows the original measurements per direction to be EQed onto headphones that have been EQed perfectly flat with the same measurement gear.
• Magenta trace: The combined response of both the left and right 30-degree virtual speakers playing simultaneously as measured through my left ear. My indoor Genelec measurements show that the ear gain should not be so reduced, so this is a phase error with the HRTF rendering software, but maybe helps the music sound "smoother" compared to Binauraliser NF whose measured combined response is brighter.

Considering the graph below (I sourced this copy from https://forum.hifiguides.com/t/grap...n-ear-over-ear-and-harman-linear-in-room/9270), my free-field response is expectedly brighter than Harman over-ear levels, whether or not the lack of crossfeed has a bearing.

Sorry, yes, I got half a conversation. Thanks for putting that altogether on a graph after I asked about it. It's quite a lot to take in re what you have done, but it's not the topic of the thread, so I won't ask about it or comment on it here. I'll just say that if I was doing it I'd be basing it around the speaker in-room response rather than anything Anechoic / Freefield or Diffuse. I'd measure the speakers in a room using in-ear mic (which would become the Target) and then equalise my headphones to that in-ear response, and would experiment with putting in a small bass shelf on top of that. I've never attempted to do this, but that's where I'd start. (And you probably already know about The Impulcifier Project, and also the Smyth Realiser as other avenues, neither of which I've tried either.)

 

[Mr. Haelscheir]

Sorry, yes, I got half a conversation. Thanks for putting that altogether on a graph after I asked about it. It's quite a lot to take in re what you have done, but it's not the topic of the thread, so I won't ask about it or comment on it here. I'll just say that if I was doing it I'd be basing it around the speaker in-room response rather than anything Anechoic / Freefield or Diffuse. I'd measure the speakers in a room using in-ear mic (which would become the Target) and then equalise my headphones to that in-ear response, and would experiment with putting in a small bass shelf on top of that. I've never attempted to do this, but that's where I'd start. (And you probably already know about The Impulcifier Project, and also the Smyth Realiser as other avenues, neither of which I've tried either.)

As a final chart, here is a comparison to my indoor near-field measurements with the speakers around 1 m from each ear at my desktop. The in-ear speaker measurements have 1/12 octave smoothing preserving the shape of the ear gain region.

• Teal trace: Left GLM-calibrated Genelec 8341A 30 degrees left of center as measured from the left ear.
• Purple trace: Left ear free-field EQ for simulating a virtual neutral speaker 30 degrees left of center in an anechoic chamber. This is the base EQ profile that SPARTA AmbiBIN modifies. It isn't far off, and note that I had EQed down 4 kHz and 7 kHz for channel-balancing sine sweeps.
• Salmon trace: Right GLM-calibrated Genelec 8341A 30 degrees right of center as measured from the left ear.
• Cyan trace: Left ear free-field EQ for simulating a virtual neutral speaker 30 degrees right of center in an anechoic chamber ("L 30 L"). Quite the difference.
• Red trace: Both GLM-calibrated Genelec 8341As playing in phase as measured from the left ear. This isn't too different from the measurement of the left channel alone.
• Magenta trace: Left ear measurement of both virtual neutral speakers playing in phase. SPARTA AmbiBIN's crossfeed phase implementation and the error in the "L 30 L" EQ leads to quite a difference or relative relaxation.
• Gold trace: Gear turned 90 degrees right relative to the left GLM-calibrated Genelec 8341A. The two treble nulls are consistent with my typical headphone measurements. My ideal is that headphones should have shallow treble nulls so that they are easy to EQ to flat prior to applying HRTF EQ.

Part of my doing the measurements outdoors was to start with a good reference with minimal comb filtering that would be easy to derive a smooth headphone response for. Yes, real rooms including mixing rooms would have comb filtering, but I wouldn't consider their presence to be an "ideal" or "most 'resolving'" listening experience, and I wouldn't expect mixers to EQ tiny perturbations, and if ever there is a wider dip or boost, I would consider that an error in their monitoring system. I have yet to acquire room treatment to see if these Genelecs can sound clearer than my binaural head-tracking setup.

You can EQ each ear to match the respective speaker response, but you would also need to EQ the crossfeed from measurements to project the sound like true speakers. Simple crossfeed like bs2b for example fails to project certain treble content forward.

That should be it unless you manage to suggest something else I have already done. ;D

 

[muslhead]

Valor is right in that case: HE1000 is more "tiring".

The se is, the v2, is not

 

[the_brunx]

It’s amazing how some people can read and tell themselves stories over and over into almost permanently believing that this average performing susvara is the best headphone ever made. Sad! Just goes to show how the human perception can be easily mislead by marketing and is unreliable.

 

[Mr. Haelscheir]

At worst, it is high-end audiophiles doused in wonky tunings hearing something close to a natural HRTF for the first time and being convinced that anything else with a similar tuning still sounds "worse" or cannot be achieved with EQ. Now, the "transient quality" for extreme transients is still something I would like to explore objectively.

 

[solderdude]

One simply can't compare in-ear mic measurements with a standard fixture for several reasons.
It will not show the actual ear gain because the biggest contributor (the earcanal gain) is not taken into consideration.
Possibly it will tell you something about your pinna and directivity of your personal hearing assuming the mic is accurate, calibrated and designed for in ear usage.

I still find it amusing that people think they can actually say something about sound quality by looking at some plots.
Sure.... it can say something about tonality (which is correctable) on a standard fixture but that's about it.

What headphone measurements are good for is showing problems in performance. Believe it or not but some problems are not as audible as some people think and there are factors that aren't shown in (acoustic) measurements yet can impact sound.

The only thing Susvara is really guilty of is price/performance ratio. And even that is in the eyes of the beholder.

 

[Mr. Haelscheir]

One simply can't compare in-ear mic measurements with a standard fixture for several reasons.
It will not show the actual ear gain because the biggest contributor (the earcanal gain) is not taken into consideration.
Possibly it will tell you something about your pinna and directivity of your personal hearing assuming the mic is accurate, calibrated and designed for in ear usage.

I still find it amusing that people think they can actually say something about sound quality by looking at some plots.
Sure.... it can say something about tonality (which is correctable) on a standard fixture but that's about it.

What headphone measurements are good for is showing problems in performance. Believe it or not but some problems are not as audible as some people think and there are factors that aren't shown in (acoustic) measurements yet can impact sound.

The only thing Susvara is really guilty of is price/performance ratio. And even that is in the eyes of the beholder.

I am aware of the limitations of blocked canal entrance measurements, and in other places where I have shared these have noted how they would exclude the canal gain and in this case show a dip where said canal gain belongs (3 kHz for my ears for 30-degree incidence); I am not directly comparing them against standard fixtures. These measurements still show me well what my actual HRTF looks like relative to measurements of certain headphones via the same setup. I still consider it feasible to conclude from these relative measurements that free-field and near-field speaker response possesses a matters brighter ear gain than your typical headphone, and true diffuse-field and free-field do exceed Harman over-ear ear gain.

After that, my measurements sate the curiosity of whether more dollars really does afford better distortion or a smoother frequency response (I care much about the "EQability" of a headphone) and a better control of driver resonances/damping or internal reflections, even if those improvements are barely audible, and the curiosity regarding the causes of subjective impressions of headphone transient response.

 

[solderdude]

Yes, relative to your own measurements it is fine.
It was the mentioning of the Harman curves that made no sense as one would have to know the differences between those and your measurements (especially above 1kHz)
One cannot make any assumptions about the sound of the Susvara though.
When measured in the same way one could deduct a little info compared to other headphones that are measured that way.

 

[the_brunx]

Performance wise, Why would anyone want to buy a Susvara over a He400se? They are both similar with the 400se being a notch better, Its craziness.

 

[the_brunx]

This headphone wont even be able to render binaural recordings accurately due to cables being soldered the wrong way in. What a Joke. This is the best engineered headphone? Please. best marketed headphone?. Probably, along side Abyss the money pit.
And all you hear is, you need a better cable, a better amp, a longer burn-in etc to make it sound good.

 

[solderdude]

Performance wise, Why would anyone want to buy a Susvara over a He400se? They are both similar with the 400se being a notch better, Its craziness.

You must not have heard them side by side is all I can say.
These do have somewhat similar tonality (all open planar hifiman do) but there is a definite difference in out of the box sound quality.
This is how most 'purists' are likely to use them by the way.
Is the difference worth almost 6k$ that's another question. It is not to me but I do not have such an income that 6k$ does not even make a dent in my wallet.
It is another thing to think the HE400SE (owned one) is the better sounding one. It isn't. It is a much more economical choice for sure.

With EQ they are more similar and when EQ'ed the difference is smaller.
And yes, price performance ratio is quite different (excellent to poor) but there is a difference in the way the music is rendered.

 

[solderdude]

This is the best engineered headphone?

No of course not.
Funny is that the HE400SE has the exact same issue (inverted polarity) but hear no one whine about it.

To me the Susvara is overpriced and not the best headphone around for sure.
It does have some very desirable qualities when you actually listen to them with and without EQ.
This is not obvious from just FR and distortion measurements only.

 

[the_brunx]

If the HE400se came in a susvara shell and they named it the susvara V2, im sure many of the same people would be selling their susvaras to get the upgrade.

 

[solderdude]

Not if they actually listened to them. There is quite an audible difference in the treble quality.