Perception of "detail" in headphones
Sep 18, 2013 at 6:38 PM Thread Starter Post #1 of 34

Solshock

New Head-Fier
Joined
Aug 12, 2013
Posts
27
Likes
10
Hello all,
I'm a guy who originally had some dynamic headphones, and have since bought 2 pairs of electrostatics.  Throughout the whole journey, there have been electrostatic fans that have claimed that stats retrieve more "detail" due to their virtually massless drivers.  However, impulse responses of the Stax Omega 2's for examples measure "worse" than much MUCH cheaper dynamics.  Is there any truth to the claim that stats are more detailed or "effortless" that can be explained by quantifiable measures, or is it mostly confirmation bias or differences in FR preferences?
 
Sep 19, 2013 at 3:45 PM Post #3 of 34

Solshock

New Head-Fier
Joined
Aug 12, 2013
Posts
27
Likes
10
As far as I'm concerned, "revealing detail" is 100 percent a function of having a flat frequency response and low distortion. Whichever device measures better for those two parameters wins.

--Ethan

Hi Ethan,
Thanks for taking the time to chat.  I enjoyed watching your audio myths seminar online. 
 
In this case, the HD600 would be the most "detailed" pair of headphones in existence if the FR curves are to be believed, and should be the end game set.  I suppose some people call that pair "veiled" because of the relatively muted treble response (i.e. neutral?).  As for distortion, is the difference between distortion between most pairs of headphones even audible? 
 
What aspect of neutrality makes a pair of transducers "detailed" exactly?  Cant boosted treble also make certain details more audible and thus more"detailed" sounding?
 
Thanks.
 
Sep 19, 2013 at 10:31 PM Post #4 of 34

briskly

100+ Head-Fier
Joined
Dec 12, 2011
Posts
214
Likes
50
Hi Ethan, Thanks for taking the time to chat.  I enjoyed watching your audio myths seminar online. 
 
In this case, the HD600 would be the most "detailed" pair of headphones in existence if the FR curves are to be believed, and should be the end game set.  I suppose some people call that pair "veiled" because of the relatively muted treble response (i.e. neutral?).  As for distortion, is the difference between distortion between most pairs of headphones even audible? 
 
Like most open dynamic headphones, the HD600 has a mid-bass hump, in a short, well-damped curve. Calling it veiled is a real stretch of the word, I think. Concerning distortion, most mid level dynamic headphones have fairly low distortion in the midrange and up, of around 0.3% or so at 90db. The 0.3% (or -50db below the fundamental) is generally with the relatively benign 2nd harmonic, with higher orders typically even lower, so it seems unlikely to be particularly audible. The bass may be a different story, as distortion is significantly greater down low.
 
What aspect of neutrality makes a pair of transducers "detailed" exactly?  Cant boosted treble also make certain details more audible and thus more"detailed" sounding?
 
Neutral and detailed go together simply due to the masking effect in the frequency response. Sounds in different frequency ranges won't overshadow each other unnecessarily if response is even. Boosting the treble will bring forward harmonics in that zone, things like shimmer and air and whatnot, but you lose out on the midrange.
 
Thanks.

 
Sep 20, 2013 at 1:52 AM Post #5 of 34

proton007

Headphoneus Supremus
Joined
Feb 9, 2012
Posts
3,518
Likes
178
I also take a look at the square wave response. It can tell a great deal about the treble to mid and mid to bass transition, or simply, how accurate the headphones are.
 
Sep 20, 2013 at 5:38 AM Post #6 of 34

stv014

Headphoneus Supremus
Joined
Jul 17, 2011
Posts
3,493
Likes
269
Like most open dynamic headphones, the HD600 has a mid-bass hump, in a short, well-damped curve. Calling it veiled is a real stretch of the word, I think. Concerning distortion, most mid level dynamic headphones have fairly low distortion in the midrange and up, of around 0.3% or so at 90db.

 
It can in fact be lower than that, depending on which measurements you believe. This one shows 0.3% at 90 dB SPL, but not much more than 0.1% at 100 dB. One would expect the THD to be less at lower SPL, especially with a "smooth", mostly second order distortion (distortion that increases at lower levels - e.g. crossover distortion or quantization - usually also produces high order harmonics). Therefore the 0.3% THD+N figure at 90 dB is likely to be dominated by noise.

 
Sep 20, 2013 at 9:04 AM Post #7 of 34

wakibaki

1000+ Head-Fier
Joined
May 26, 2011
Posts
1,088
Likes
63
I think it's very difficult to discuss these issues, because although you can go to Stereophile and read their glossary of audio terms, not everybody has done so, and the meanings of words like 'detailed', 'veiled', 'shimmer', 'air' &c. are not well defined in this context, there could be other glossaries with conflicting meanings, it's not as if there are multiple established dictionaries like the 'Oxford English' 'Collins' or 'Websters' carrying these words with some correspondence between their definitions, so it's entirely possible to carry on a conversation where people believe themselves to be agreeing, but are in fact disagreeing, and vice versa.
 
I don't really understand why anybody finds it necessary to use terms like 'detailed' when measurable quantities such as distortion and frequency response predate this new vocabulary by a considerable time, and it's always been my view that its invention dates from a time when audio reviewers foresaw their own redundancy because electronics, with the exception of transducers such as microphones and speakers (vinyl cartridges thankfully being on the way out), had started to reach a point where it could be built to be largely beyond reproach.
 
Or to put it another way, when people resort to using these words I tend to tune out, because I think it's all a load of guff, which I think you'll find is a word which is quite adequately defined and understood.
 
w
 
Does this qualify as a rant, I hope?
 
Sep 20, 2013 at 10:01 AM Post #8 of 34

briskly

100+ Head-Fier
Joined
Dec 12, 2011
Posts
214
Likes
50
  It can in fact be lower than that, depending on which measurements you believe. This one shows 0.3% at 90 dB SPL, but not much more than 0.1% at 100 dB. One would expect the THD to be less at lower SPL, especially with a "smooth", mostly second order distortion (distortion that increases at lower levels - e.g. crossover distortion or quantization - usually also produces high order harmonics). Therefore the 0.3% THD+N figure at 90 dB is likely to be dominated by noise.
 

The decrease in harmonic distortion with increasing SPL seemed odd; the distortion comparison between the Headroom measurements and Innerfidelity isn't quite consistent.  For certain headphones (e.g. the LCD-3) the decrease in THD+N with increasing SPL corresponds almost exactly to the increase in volume. We might conclude Tyll's system has a fairly high noise floor from that, but his measurements for Headroom show much lower harmonic products for each headphone and a much lower floor. With the LCD-3 at headroom, using a sine at 90db, the harmonics and noise are all below -80db. No headphone gets anywhere near that level of distortion, including other LCDs, at innerfidelity.
 
This really makes me wonder what Tyll does differently for each site.
 
Sep 20, 2013 at 10:30 AM Post #9 of 34

stv014

Headphoneus Supremus
Joined
Jul 17, 2011
Posts
3,493
Likes
269
  We might conclude Tyll's system has a fairly high noise floor from that, but his measurements for Headroom show much lower harmonic products for each headphone and a much lower floor.

 
On an FFT plot (HeadRoom), the noise floor looks lower than the total - especially unweighted - RMS noise level (InnerFidelity), particularly if the window size is large. For example, TPDF dithered 16 bit PCM audio has a noise level of -93.3 dBFS without weighting, but an FFT with a 65536 sample Blackman-Harris window shows the noise floor at about -130 dBFS on average (doubling the window size reduces it by 3 dB). That is obviously because the total noise power is distributed over many FFT bins.
 
Sep 20, 2013 at 11:25 AM Post #10 of 34

MatsGyver

Head-Fier
Joined
Jul 20, 2013
Posts
53
Likes
13
As far as I'm concerned, "revealing detail" is 100 percent a function of having a flat frequency response and low distortion. Whichever device measures better for those two parameters wins.

--Ethan


I actually agree with you Ethan, but detail is not all.

Music is about feeling something. Either it is trough the lyrics, the instruments voices or a combination of all. Music is therefore about more than just transparancy and detail. And when lots of recordings of today are less pleasing on the ear, when played trough neutral headphones, the notion that the best headphones is the ones that measures flat and has low distortion falls flat.

Seeing distortion and frequency response in relation to nothing makes no sense. You have to take into account the music genre, and dynamic range of the music you are listening too, to see the full picture. Without this factor, the best measuring headphone with the lowest distortion will be the best. It is just not the case.

A headphone with neutral sound and low distortion is usually a headphone with a light diaphragm and hard damping. This makes it very detailed, but not universally better sounding. You have to adapt to the recordings thats out there, not make the recordings adapt to your equipment. The former is a battle you are destined to loose. Stax-esqe damping and detail level is not in my oppinion universally best.

I think this is more relevant in headphone listening, because of the general detail level in listening experiance in the technology tend to be higher. With speakers it is harder to get it to sound too detailed because of room acoustics. This is not as big of a problem with headphones, and the barrier of "detailed enough" is reached and passed with headphones like Stax SR009 and HD800. With most recordings these headphones are too detailed. Even though they are great headphones. For great recordings with 25-30dB + dynamic range, and SPL 90-110dB, I dont think they are too detailed. In fact I think they are sublime for that kind of recordings.
 
Sep 20, 2013 at 11:34 AM Post #11 of 34

Solshock

New Head-Fier
Joined
Aug 12, 2013
Posts
27
Likes
10
So based on what I'm reading, could I hypothetically equalize a pair of koss portapros to match the FR of an HD600 and get equal performance?  Nothing else plays into the "detail" or transparency of the transducer?
 
Sep 20, 2013 at 11:36 AM Post #12 of 34

stv014

Headphoneus Supremus
Joined
Jul 17, 2011
Posts
3,493
Likes
269
Music is about feeling something. Either it is trough the lyrics, the instruments voices or a combination of all. Music is therefore about more than just transparancy and detail. And when lots of recordings of today are less pleasing on the ear, when played trough neutral headphones, the notion that the best headphones is the ones that measures flat and has low distortion falls flat.

 
Already having an accurate/flat/detailed headphone, it would make more sense to use signal processing (equalization, crossfeed, etc., even reverb to simulate room acoustics) to alter the sound for bad recordings, rather than to buy several headphones. Of course, if one only has bad recordings, then the high quality headphone might be wasted.
 
Sep 20, 2013 at 11:37 AM Post #13 of 34

MatsGyver

Head-Fier
Joined
Jul 20, 2013
Posts
53
Likes
13
THD, (distortion) would limit the quality of the sound. But in theory, as long as the physical construction makes it possible, I guess you could do that. But the THD would be so high that it would sound bad.
 
Sep 20, 2013 at 11:40 AM Post #14 of 34

stv014

Headphoneus Supremus
Joined
Jul 17, 2011
Posts
3,493
Likes
269
  So based on what I'm reading, could I hypothetically equalize a pair of koss portapros to match the FR of an HD600 and get equal performance?

 
You would need very accurate equalization (I mean something like an inverse filter created from measurements and/or many bands of parametric EQ), and also correct the distortion if it is significant - this would be a difficult mathematical problem with complex frequency dependent distortion, and I do not think it is something that is actually done in practice. So, it is not really practical to make a low quality headphone identical to a high quality one. But you can still get surprisingly good results. There is already software like Accudio for this purpose.
 

Users who are viewing this thread

Top