CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Headphone Measurements => Topic started by: Marvey on December 10, 2011, 08:43:10 PM
-
While I don't have enough data to support this assertion, it seems that headphones with badly matched drivers tend to sound bad. They end up having the worst of either channel. Two good examples I can think of are the T1 and LCD2. I've heard some good sounding ones bad sounding ones (this is all anecdotal of course.) Take for instance the following T1 and LCD2r1 samples (BTW, some of the other admins/mods have heard these.)
Beyer T1 with badly matched drivers:
(http://www.changstar.com/index.php?action=dlattach;topic=30.0;attach=27;image)
Beyer "Good" T1
(http://www.changstar.com/index.php?action=dlattach;topic=66.0;attach=284;image)
The T1 with the badly matched driver sounded bright, thin, and lacking in bass. Although there's nothing in the FR that would necessarily indicate such, especially when compared to the "good" T1 below. Maybe the human brain/ear mechanism hears the worst of both channels (the relative treble plateau on the left channel and the high midrange plateau of the right channel.)
Audeze LCD2 r1 with badly matched drivers
(http://www.changstar.com/index.php?action=dlattach;topic=48.0;attach=144;image)
Audeze "Good" LCD2r1
(http://www.changstar.com/index.php?action=dlattach;topic=17.0;attach=170;image)
LCD2s, especially the revision 1s, are known to have a very shelved down high-midrange and treble - and be sound very pleasant for forgiving. The ringing on the r1 above wasn't necessarily excessive, but the badly matched unit above sounded hollow and thin. There was quite a lot of harshness and the bad kind of sibilance with vocals, even on male voice. I suspect the ringing at 6.5kHz and 8.5kHz (not shown explicitly in the FR graphs here, but on the CSD plots) contributes to a double whammy where the ear/brain mechanism end up hearing both - and at their full glory too. The differences in the bad r1 measured above seem to far apart to be accounted for by pads and placement on the measurement rig.
Of all the headphones I've heard and measured, these two stood out as particularly bad, that is not typical and not representative of other samples, during subjective listening tests. I think it's interesting what we see in the measurements of these.
BTW, I think it's rather unacceptable for ~$1000+ headphones to have such poor QA.
-
Good stuff. Beyond driver matching, how might other factors influence the frequency response? Let's say you are turning wooden enclosures for T50rp for example, how would variations in enclosure volume affect the frequency response of each side?
-
Enclosure volumes are going to affect the bass response. Since there's so little volume to work with to begin with, it's important to be precise. At least this is what I've found with the half-dozen or so T50RP mods that I've measured.
The T50RPs tend to have some variation in the midrange to treble. Can't really do anything about that as this is more related to the driver itself.
-
Aww man, and I was looking forward to use the wooden cups I have on hand too... Unfortunately, the cups have a slight mismatch in size, and the cup volume is different too.
-
Aww man, and I was looking forward to use the wooden cups I have on hand too... Unfortunately, the cups have a slight mismatch in size, and the cup volume is different too.
Not a problem. You can easily reduce volume on one side to match the other by displacing it (with slivers of wood, etc.) Just fill the cups with water and see how close you can get them.
For speakers, I once used stacks of cans of dog food to fine-tune the final size of the enclosure.
-
For speakers, I once used stacks of cans of dog food to fine-tune the final size of the enclosure.
For the woofer? :D
-
Could this explain why I get a slight ringing after listening to the LCD's? I had the LCD2 rev2.
-
For speakers, I once used stacks of cans of dog food to fine-tune the final size of the enclosure.
For the woofer? :D
LOL!
-
Could this explain why I get a slight ringing after listening to the LCD's? I had the LCD2 rev2.
That would be mild tinnitus(ringing sound perceived in the absence of an actual external sound impulse) that occurs after loud and or long exposure to sound/music...usually subsides after sometime unless there is some serious noise related damage to the auditory apparatus
-
In the world of CIEMs... driver matching is hell..
I think my Heir8.As are matched well but I also think fitment and the overall shape of a person's ear has alot to say about matching. Even when I've made sure both sides are equally well isolating by doing the 'rubbing fingers beside the ear' ritual, the right channel always bug me.
They seem to be slightly less detailed than the left channel. And sometimes the centre imaging can skew a little to the left. Or maybe its just Justin Bieber not standing in the middle of the recoding apparatus.
-
Good post. I can add some anecdotal (read: worthless) data to it. I never measured my LCD2s but I am fairly certain they had poorly matched drivers. The soundstage always sounded like it was rotated in left of center. No software crossfeed would fix it.
-
Good post. I can add some anecdotal (read: worthless) data to it. I never measured my LCD2s but I am fairly certain they had poorly matched drivers. The soundstage always sounded like it was rotated in left of center. No software crossfeed would fix it.
Software (or hardware) balance controls might fix it. However, IMO a ~$1k headphone with severe driver mismatching issues is a defective product.
-
That'd work if the two drivers had slightly different sensitivities. If the problem is driver mismatch in certain frequency bands, it'd be hit or miss depending on the track. :(
-
Indeed, then you have a dedicated equalizer per channel nightmare. If the drivers on a single headphone are off, one can only wonder about the driver differences between it and another same model headphone...Life is like a box of chocolates (http://www.changstar.com/index.php/topic,167.0.html)
-
I've measured AKG studio cans that had as much as 9 dB different sensitivity than specified and 6 dB from one headphone to the next of the same model. I guess the +/- 3 dB some manufacturers provide is no exaggeration.
What this means for channel matching... I can only guess.
-
I've tested a bit on driver mismatches, here's what I found:
- if it's just volume, then obviously standard balance control/channel gain works very well - I had only one such case though
- if there are major differences in some parts of frequency response, all-pass alone (bandlimited time delay) works better than IIR equalization.
- minimum phase FIR is equivalent to standard IIR in this matter, linear phase... depends, sometimes it's better, sometimes it's worse. Might be dependent on the CSD, but I don't have reliable enough setup to measure this.
- apparently ringing is more important than FR for perception of balance
- a combination of balance plus the all-passes works extremely well of course
- the choice of channel balancing technique is unimportant if the difference is less than 3 dB. At least to my ears. At larger differences differing balance technique becomes important.
- applying crossfeed after EQ does not change the choice of the method - the spatial difference is actually less audible.
- IEMs are more resolving when it comes to channel imbalance, fortunately most IEMs are well balanced.
- narrow peaks of imbalance are much less audible, but they affect subjective "locatedness" of the sound - make the soundstage seem fuzzy and less defined - especially if they are numerous
- highs imbalance affects positioning much less, but messes with the depth of the soundstage
- bass imbalance affects positioning the most, but...
- mids imbalance is the most audible most of the time
- equalization is more accurate when done separately on each channel than centering a mono source - slightly better results, mostly in "locatedness".
- correcting resonances based on plain CSD results in wrong, weird sound. Manually correcting based on *tweaking* the phase response using the CSD works very well, but is very time consuming.
Caveats:
- I can hear misbalances down to 1.0 dB, straight dips and peaks down to 0.75 dB directly, 0.5 dB with many tries.
- The imbalances in IEMs were confirmed using a 5mm x 65mm silicone tube coupler with a silicone "horn mouth" fed into a measurement mic. (volume matches my ear canal volume - ~1.2cc; also checked with 1cc and 0.75cc.)
- No such confirmation for headphones yet. My current ghetto coupler (with silicone foam) is too resonant on its own, giving invalid CSDs and not quite valid THD. Needs more mass and back surface less reflective than plywood.
- Correction was applied in three ways: rebalancing only using mono source; mono equalization plus rebalancing; stereo equalization
- All filtering done using a Python reimplementation of algorithms ccirls (FIR) and mpiir_l2 (IIR) in this thesis: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.6.9336
Verified against CAPS equalizer and Electri-Q - these algorithms provide audibly better results. FIR one is not great for minimum phase though.
Minimum phase FIR was achieved with better results by applying mixed phase (33%) CCIRLS with windowed cepstrum method to convert to true minimum phase.
- CSD correction was attempted only on one pair of AKG K271. Very time consuming, took 2h to get a good filter.
- Program used for these corrections will be released soon.
These results were derived using:
- 3x AKG K271 headphones - notable ringing, some dB imbalance in bass and subbass; ringing differs between channels and pairs
- 5x Superlux HD681, unmodded - also ringing, general very audible imbalance in 2 out of 5; many narrow imbalance blips.
- modded Hifiman HE-500; HE-6 - no real ringing - well balanced - some narrow blips
- Paradox - some imbalance around 2 kHz, otherwise very even.
- Spiral Ear SE-5 - very imbalanced area in highs, rest is ok.
- Hifiman RE-ZERO, 2x RE-600, 2x RE-400 - quite well balanced out of the box, all have long "flat" reverb ringing; wide shallow imbalances
- Hifiman RE-300a - some mids imbalance, no ringing
- modded Sony MH1 - highs ringing, noticeable midbass imbalance and difference in amount of air for each channel
- One "dark chocolate" Beyerdynamic DT-1350 - "wavy" balance everywhere, but good general balance
Couldn't test with these due to no imbalance being detectable at all. Great QC is great! p:3
- VSonic GR07, GR07 mkII, 2x GR06
- Brainwavz M2 (this sounds like VSonic GR06, same OEM?)
- Brainwavz B2 (well matched TWFKs)
- Sennheiser HD600, 2 pairs, old driver
- Sennheiser HD650, new driver
- Hifiman RE-272 - one narrowish imbalance at 1 dB/2.5kHz is all I think this has, I couldn't ABX the difference.
- Beyerdynamic DT-770 - new, wide blip in highs, 1.5 dB, but couldn't ABX the difference.
Couldn't fully test these due to only tiny loan:
- Audeze LCD-2 - balance was very good
- Beyerdynamic T-70 - excellent balance, but very uneven FR
Couldn't fully test these, because I couldn't stand the sound:
- everything Ultrasone (680, 840, Platinum)