• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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 Sundara Review (headphone)

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
Balanced only gives 3dB more due to current limiting in the output stage (one can expect +6dB for > 100ohm cans)
When we are working close to max. then 3dB is 3dB !
My Chord Mojo has 400mW at same impedances, and it can power the Sundaras, but I have to put the volume (at times) towards the max, Hugo2 @ 740mW will not struggle, but still, volume control is way up..
By that notion 40mW (to me) is on the low side.
I , unlike you, am not lenient on power near clipping :D.
 

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
It's a gain issue rather than power issue.
OK explain it to me.
we have two amps, one has 112W power into 8 ohms.
the other 450W.
the input to both is 2Volts.
at rated power amp A produces 30V , while amp B produces 60V.
if you were the designer, would you give them same gain?
with it be 15 or 30?
and how would the volume control work?
Would amp A go into clipping at half volume?
or would amp B never reach rated output?
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
if you were the designer, would you give them same gain?

No.

with it be 15 or 30?

it would be 15 or 30 when the goal is to have an amp that starts to clip at 2V.

But here's the thing. I actually am a designer and would not give the amps 15x or 30x respectively but probably closer to 30x and 60x gain and I can assure you they would sell better and in subjective AB tests would win.

It's a bit like the L30 story... People saw huge output power numbers and expected the amp to reach this when using a standard 2V out DAC from the same brand. Almost everyone was universally disappointed the 'promised' output power was not available.
To make things worse most people wanted the extra power because they wanted to use it to boost lows using digital EQ and when they did they had to use negative pre-amp gain.
And there you had the volume control issue (volume control maxed out, highest gain, inefficient headphones) while they bought the L30 because of its variable gain, high output power, lowest noise and distortion.

So.. there you have it. When the design criteria would be 2V input voltage I would design design those amps 15x and 30x gain.
When I did not have this restriction I would either add a gain switch or an input voltage range control and would have a more usable amp.

Now... for another design... say DAC+amp combo that does not have line inputs. It would make sense to just reach max. output power at 0dBFS.
Almost the same issue arrises as the L30 story except there is no gain control and max. power can be reached. Great for the specs.
When using EQ the average max SPL will now not be so high. Yep, it won't distort or just touch clipping levels at max volume which is a good thing but means when playing at max volume the volume control would be at max. position.
No one likes that.


and how would the volume control work?

Just like any other it would have (semi)log volume control. One can even choose or manufacturer your own taper.

Would amp A go into clipping at half volume?
or would amp B never reach rated output?

See explanation above.
It's why my amps have high gain.
Yes they will have higher distortion and worse S/N ratio than designs build to have best specs but no one bitches about not having enough power and you can even drive it from a phone/dongle with 1V or lower output and use digital EQ and get the promised output power without having to max out the volume control.
Downside is the bottom section of the volume control is problematic in channel balance. Thankfully there are various gain options.


But this is about the Sundara and am happy to report that the practical design of the amp in question can reach almost 2W into the Sundara.
Yes... there will be tube distortion so it will measure as utter crap and end up in the 'poor' SINAD section so I hope it is never reviewed here :) (really)
With its max 20x gain (tube dependent) you can reach max. output power with 0.75V so you can drive it with just about any source you want and get full power out of it. Not really needed for Sundara but very handy for Susvara or HE6 (original), K1000 etc.
 
Last edited:

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
No.



it would be 15 or 30 when the goal is to have an amp that starts to clip at 2V.

But here's the thing. I actually am a designer and would not give the amps 15x or 30x respectively but probably closer to 30x and 60x gain and I can assure you they would sell better and in subjective AB tests would win.

It's a bit like the L30 story... People saw huge output power numbers and expected the amp to reach this when using a standard 2V out DAC from the same brand. Almost everyone was universally disappointed the 'promised' output power was not available.
To make things worse most people wanted the extra power because they wanted to use it to boost lows using digital EQ and when they did they had to use negative pre-amp gain.
And there you had the volume control issue (volume control maxed out, highest gain, inefficient headphones) while they bought the L30 because of its variable gain, high output power, lowest noise and distortion.

So.. there you have it. When the design criteria would be 2V input voltage I would design design those amps 15x and 30x gain.
When I did not have this restriction I would either add a gain switch or an input voltage range control and would have a more usable amp.

Now... for another design... say DAC+amp combo that does not have line inputs. It would make sense to just reach max. output power at 0dBFS.
Almost the same issue arrises as the L30 story except there is no gain control and max. power can be reached. Great for the specs.
When using EQ the average max SPL will now not be so high. Yep, it won't distort or just touch clipping levels at max volume which is a good thing but means when playing at max volume the volume control would be at max. position.
No one likes that.




Just like any other it would have (semi)log volume control. One can even choose or manufacturer your own taper.



See explanation above.
It's why my amps have high gain.
Yes they will have higher distortion and worse S/N ratio than designs build to have best specs but no one bitches about not having enough power and you can even drive it from a phone/dongle with 1V or lower output and use digital EQ and get the promised output power without having to max out the volume control.
Downside is the bottom section of the volume control is problematic in channel balance. Thankfully there are various gain options.


But this is about the Sundara and am happy to report that the practical design of the amp in question can reach almost 2W into the Sundara.
Yes... there will be tube distortion so it will measure as utter crap and end up in the 'poor' SINAD section so I hope it is never reviewed here :) (really)
With its max 20x gain (tube dependent) you can reach max. output power with 0.75V so you can drive it with just about any source you want and get full power out of it. Not really needed for Sundara but very handy for Susvara or HE6 (original), K1000 etc.
Yes agree on all, but missed my point.
gain is related to rated power, volume control is a power throttle .
going back, of course an amp with twice the power had twice the gain, that is to be expected.
but a pair of headphones requiring most of that power, will need to be towards the end of the volume setting.
it is not gain related, it is power related.(as to Sundara needing almost max. Vol on Hugo2)
I do know you are a designer, we have had interactions before, on DIY Audio. Respect.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
Yes agree on all, but missed my point.

My point was that when the Mojo would have had more gain (and thus clip at full volume) the volume control would not have maxed out but be set much lower. The same is true for Hugo.

Of course, I do get your point about the maximum power output but the difference between the Mojo and Hugo2 is only 2.6dB so when somewhat similar tapers are used and clipping points (0dBFS) is reached at full power the difference in volume control settings would not differ much.
Here too depending on 'taper' as they will not be the same.
 

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
Of course, I do get your point about the maximum power output but the difference between the Mojo and Hugo2 is only 2.6dB so when somewhat similar tapers are used and clipping points (0dBFS) is reached at full power the difference in volume control settings would not differ much.
I understand.
Hugo2 may only have 3dB roughly more power, but when it comes to transients it has got more in reserve.
It is such a good DAC/Amp - expensive? yes for sure, but very good indeed.
Shame Amir does not see eye to eye with Mr. Watts - regardless of his odd rantings, his designs surely work.
 

Rayman30

Addicted to Fun and Learning
Forum Donor
Joined
Dec 5, 2019
Messages
789
Likes
627
I would be curious to see how the new Edition XS compares, both objectively and subjectively.
 

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
I would be curious to see how the new Edition XS compares, both objectively and subjectively.
I have modded mine already, so I have to go by memory.
Are they better, in a word - yes.
Read my review here on another forum.
Modding, improves on Midrange and stereo imaging so much, I can not see myself ever going back, but I understand it is not for everyone.
I wrote the review before any modding attempt.
My Mod for Edition XS.
 

asdfqwer2354

New Member
Joined
Dec 31, 2021
Messages
2
Likes
0
I just picked up a pair of Sundaras and I'm trying to figure out if I can make these sound better. To me something seems really off but I can't quite put my finger on it. I loaded up some of the EQ's and I'm playing with them now but they don't seem to do exactly what I want. ( Jaakk, O1990 Hifi, O1990 Harman, O1990 Diffuse, AMIR )

One thing that I might be honing in on is the 's' type sounds. Overall vocals sound really bad but when I come across singers that lisp their s's it's almost physically painful - Chevelle does this pretty hard. I think I can mitigate this a bit by nuking the 5300~ region.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
The issue is caused by a peak near frequencies where Oratory's and Amir's test fixture dips so energy there is not shown even when present.
sundara-fr.png

Not everyone's ear has the same response at 10kHz as the pinna and ear canal of the test rig. There are people that do not hear the treble peak (when they happen to have a dip in that region similar to that of the test fixture. Some have this higher or lower and are bothered by it but cannot find it on plots made on specific fixtures.

Try to use the Oratory EQ and add a -6dB filter at 9kHz and you will probably be free of sharpness (sibilance is in the 5-7kHz region, sharpness in the 8-12kHz region). Maybe this could help. I know a lot of people that were bothered by these peaks which most hifiman seem to have.
 

TinFoilCap

Member
Joined
Jan 3, 2022
Messages
5
Likes
8
Grabbed a brand new (but opened) pair of these off FB Marketplace yesterday. Asking $300, I offered $250. Seller countered with $200! Needless to say, I was there in 15 minutes cash in hand.

They sound fantastic, definitely different sound than my Beyer DT-770's. I can't hear any sharpness in the vocals using oratory EQ, Mobo DAC, and DIY Whammy amp. But I am still developing my headfi ears.
 

asdfqwer2354

New Member
Joined
Dec 31, 2021
Messages
2
Likes
0
@solderdude Thanks! Amazing information. I would never have figured that out.

I'm going to try mixing:
  • O1990-Harman
  • Suggested -6db@9k~(Q7?)
  • Remove low-shelf at 105 so the bass boost doesn't muddy the mids
  • Add the bass boost to the lower 2 bands
  • Boost some of the vocal fequencies to make them less flat-subdued
Open to any other suggestions that you guys have tried.
 

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
for what it is worth, below is my setting, arrived by ear! :
75Hz q=1 +2dB low shelf
1500Hz q=0.9 +2dB peak
5900Hz q=4 -2dB peak
By default I look at EQ as necessary evil, so I always opt for the minimum amount I can get away with, hence the above.
 

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
The issue is caused by a peak near frequencies where Oratory's and Amir's test fixture dips so energy there is not shown even when present.
sundara-fr.png

Not everyone's ear has the same response at 10kHz as the pinna and ear canal of the test rig. There are people that do not hear the treble peak (when they happen to have a dip in that region similar to that of the test fixture. Some have this higher or lower and are bothered by it but cannot find it on plots made on specific fixtures.

Try to use the Oratory EQ and add a -6dB filter at 9kHz and you will probably be free of sharpness (sibilance is in the 5-7kHz region, sharpness in the 8-12kHz region). Maybe this could help. I know a lot of people that were bothered by these peaks which most hifiman seem to have.
Proof!
For those who believe graphs are "Bible"!
Here is a few graphs of different headphones measured by Oratory, all showing dips at 10kHz (& gentle bass slope) - common denominator here is Oratory test gear.
If interested, look at a few of Amir's headphone graphs, you see similar pattern.

1641306158786.png
 

Attachments

  • 1641305866921.png
    1641305866921.png
    132.5 KB · Views: 64
Last edited:

RHO

Major Contributor
Joined
Nov 20, 2020
Messages
1,182
Likes
1,087
Location
Belgium
Proof!
Here is a few graphs of different headphones measured by Oratory, all showing dips at 10kHz - common denominator here is Oratory test gear.
If interested, look at a few of Amir's headphone graphs, you see similar pattern.

View attachment 176647
According to @Resolve , that dip is supposed to be there. Something to do with the anatomy of the ear which is replicated by a measurement rig with concha simulator.
As @solderdude explained, not everybody has the same resonance there. But on the same rig the resonance should stay fairly constant over multiple types of OE headphones.
Solderdude's rig does not have a concha simulator. So it doesn't show the dip.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
Indeed this is proof the HATS used clearly doesn't measure correctly at 10kHz which is a known fact.
Look at ALL plots made by Amir and Oratory you will see that this is in the measurement rig.

Look here... different measurement rig... surprise no 10kHz dip.
Then look at consistency and notice how some people have dips and others show peaks above 10kHz.
Its the reason why most plots are greyed out above 8kHz to 10kHz.
This is the reason why some people are bothered by it and you (lucky card) cannot hear it.
 

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
According to @Resolve , that dip is supposed to be there. Something to do with the anatomy of the ear which is replicated by a measurement rig with concha simulator.
As @solderdude explained, not everybody has the same resonance there. But on the same rig the resonance should stay fairly constant over multiple types of OE headphones.
Solderdude's rig does not have a concha simulator. So it doesn't show the dip.
Are you telling me that different manufacturers and models all have a short fall at 10kHz?
Or
Are you telling me it is a natural phenomena because of the way the ear is constructed?
If the latter, then the brain has learnt (over few thousend years) to compensate for that, and it should no longer be shown or at least included in the Harman curve!
The "concha simulator" has ignored the brain/ear combined effect.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
RHO (and me) are telling you that the (substantial) dip at 10kHz, caused by pinna+ear canal simulation are always at the same frequency in a HATS.
The concha and earcanal in the HATS are chosen to be somewhat close to reality. Not as close as the BK5128 but reasonable average.
Look at measurements made by Tyll... they ALL have a peak at 10kHz (which also is due to the concha simulator) which also is not correct.

The reality (my ears and those of others) due to all pinna being different (as different as fingerprints I am told) and not having the same dimensions (pinna keeps growing all your life) the dip in someone else's ears may be higher or lower than that of the HATS.

Also your brain is used to the dip you have near that frequency as well.

It is not included in the Harman target. I have never gotten a clear answer why not other than the curve is averaged (very, very heavily smoothed)

Do yourself a favor and ignore everything above 8kHz in Amir's and Ora's measurements and consider that above 6kHz accuracy is already out the door. You should not EQ this either (which Amir repeatedly remarks in his reviews.
measurement 'error' or rather 'artefact' or 'inaccuracy' and downside when using a HATS for headphone measurements.

For Sundara... it does have a peak but in a spot where the HATS has a dip so you can't see it in the plots.
Depending on how different your ear is from the fake concha and ear canal from the HATS you may hear the peak or not.
Your ear seems to be pretty close to that from the HATS so you don't hear it as it cancels out.
 
Last edited:

Ken Tajalli

Major Contributor
Forum Donor
Joined
Sep 8, 2021
Messages
1,995
Likes
1,772
Location
London UK
RHO (and me) are telling you that the (substantial) dip at 10kHz, caused by pinna+ear canal simulation are always at the same frequency in a HATS.
The concha and earcanal in the HATS are chosen to be somewhat close to reality. Not as close as the BK5128 but reasonable average.

The reality (my ears and those of others) due to all pinna being different (as different as fingerprints I am told) and not having the same dimensions (pinna keeps growing all your life) the dip in someone else's ears may be higher or lower than that of the HATS.

Also your brain is used to the dip you have near that frequency as well.

It is not included in the Harman target. I have never gotten a clear answer why not other than the curve is averaged (very, very heavily smoothed)
I got that already.
But since the way the Ear/brain combo's inner working is not an open book, showing such a dip is misleading, along with other stuff it should be ironed out.
Basically if a perfect transducer should exist, with a ruler flat FR, then showing a "graph of the transducer" should be a flat line, and any irregularity of the ear canal, microphone response curver etc. should not be included.
If as you say, 'some' are not sensitive to it and 'some' are, then it is an individual thing, and should not be in the response curve.
After all the question is ' Is the headphone having a dip at 10kHz?'
It doesn't end there, there are other trends in Oratory curves , at different frequencies - these are indicators that the Oratory curves are not fully compensated.
Her is another snapshot;

1641308924199.png


Look closely;
- all have rolling down slopes at bass
- Most have peaks at about 5kHz
- look at 8kHz
- all have rising curves towards 20kHz
- gentle hump at 2-3 kHz
 
Last edited:
Top Bottom