measurements

home
back to headphones

post separation

The measurements are made on a home made and not calibrated measurement ‘rig’ and not with an official (and very expensive) HATS. They are made to satisfy my own curiosity.

DSCN0013
As this test-rig is a flat-bed type and doesn’t have a fake Pinna nor simulated ear canal and not much ‘compensation’ is needed aside from the microphone itself which has a small boost around 16kHz. This is compensated for in all the plots.
Because a flatbed measurement rig creates a perfect seal bass response measurements may not match what some people are hearing. This could be caused by a small leakage which may be present on a human head and not being present on this test-rig.

Note that some of the measured headphones are NOT in stock form and thus will also measure differently from those in stock form.

Aside from the small correction in the treble also some compensation is used for the lower frequencies.
Speakers that measure flat in anechoic conditions will reproduce the most realistic sound. These speakers, however, are never used in anechoic circumstances but always in living or other listening rooms. Those rooms are all echoic.
What happens is that all sounds emitted from the speaker arrive at our ears directly AND bounced off from walls, ceiling, floor and stuff in the room.
The higher the frequency the more directional the sound becomes. Also our hearing is quite capable to separate the direct sound from echo’s. This works well from a few 100Hz and up to about a few kHz.
Below a few hundred Hz our brains cannot separate direct from reflected sound. The long wavelength combined with the speed of sound and distances involved is the root cause.
The lower frequencies fold around the speaker and becomes a point source (omni-directional). Therefore the lowest frequencies bounce of walls, floor and ceiling and all ADD where mids and higher frequencies do not add because the brain can separate direct from reflections. Because of this a speaker that MEASURES flat in an anechoic room will have a raised bottom end in a ‘normal’ room.
This too will be the case in a studio when doing the final mix.
When one considered that ideal speaker in a normal room as ‘neutral’ that means that the bass will be raised.

Headphones do NOT exhibit the same behavior. All sounds are coming directly to the ear so bass is not added. A headphone that measures flat in the bass thus sounds bass shy compared to a flat speaker in a room.
To compensate for the absence of the room reflections for the lower frequencies this part of the frequency response has to be elevated in order for the headphone to sound as good as a good speaker in a normal room.

This phenomenon has also been studied by Sean Olive and Todd Welti. They came up with a +4dB boost below 60Hz (for headphones) which is quite similar to my findings and those of Golden Ears. Personally I find the Harman curve a bit too much so the plots on this website have slightly less compensation and slightly different in curve.

Rtings (no spelling error) has a nice video up explaining issues they found using their (Kemar) HATS regarding bass and treble response that is certainly worth a look.
They are doing a better (and faster) job researching how to measure with HATS than Tyll does. Tyll (InnerFidelity) is working on this too though.
The only other site I know that has a pretty good correction curve for their HATS is ‘Sonarworks‘ who has a digital version of the Kameleon basically.

Below the plot of the applied compensation for measurements on this website.
A headphone that would measure ‘flat’ as in equal sound-pressure for each frequency will look exactly as shown below. A little bit too bass-shy as it were.

corr plot

This means that when a headphone measures as a ‘flat’ line in the plots on this website the actual measured response had about 4dB more bass. A ‘flat‘ line in the frequency response plots on this website sounds ‘flat’.

LINKS to measurements of the following headphones can be found below.

AKG: K181
AKG: K400
AKG: K500
AKG: K518
AKG: K550
AKG: K551
AKG: K612

audio-technica: ATH-M40X
audio-technica: ATH-M50 (old version)
audio-technica: ATH-MSR7 (fake)

AudioQuest: NightHawk

Ausdom: ANC7
Ausdom: M05

Beats: Studio Wireless

Beyerdynamic: DT150 (250Ω)
Beyerdynamic: DT250 (250Ω)
Beyerdynamic: DT770 Pro (250Ω)
Beyerdynamic: DT990 Pro (250Ω)
Beyerdynamic: DT990 (600Ω)
Beyerdynamic: DT1350
Beyerdynamic: T50P

Bose: QC25

Fostex: T50RP(mkII)
Fostex: T50RP (mk3)
Fostex: TH-X00 (Mahogany) Massdrop

Grado: SR125i

HIFIMAN: HE350 (Massdrop)
HIFIMAN: HE400i

Jays: V-Jays

Koss: KSC35
Koss: KSC75
Koss: PortaPro

Philips: A5 Pro
Philips: Fidelio X1
Philips: SBC-HP400
Philips: SHB-9850NC
Philips: SHL-9505 (uptown)
Philips: SHP-9000
Philips: SHP-9500

Roland: RH-300

Sennheiser: HD25-1
Sennheiser: HD215
Sennheiser: HD238
Sennheiser: HD320
Sennheiser: HD438
Sennheiser: HD600
Sennheiser: HD650
Sennheiser: HD800

Sony: MDR-1A
Sony: MDR-100AAP
Sony: MDR-XB500

Stax: SR-5 with SRD-6 energizer

Superlux: HD562
Superlux: HD572-SP
Superlux: HD581 (proto)
Superlux: HD662-EVO
Superlux: HD681
Superlux: HD681-EVO
Superlux: HD685
Superlux: HD687 (proto)
Superlux: HD688 (proto)

t.bone: HD 990 D

Tascam: TH02-B

Teufel: Aureol® Real

Ultrasone: HFI-780

‘Stratocaster’ headphones
SFI-973F
Skyerdynamic

Veclan: Monk Plus
Silvercrest: KH-2347
Intempo: Pulse
Nameless

 post separation

The reasons why I use my ‘own’ measurement method/rig.

I have been asked in the past what ‘compensations’ I use and what my measurement method is. Often this question originates from my plots differing from those of others.

My thoughts on this are as simple as can be but need a lot of words to explain how I got there.

One can use expensive HATS for measuring speakers as the ears (Pinna) and ear canal are ’emulated’ for these kind of measurements anyway.  I don’t want to spend a load of money on these things though nor do I think it is essential. With headphones the sound is coming from the sides (directly firing into the ear canal) in a very small and sometimes hermetically closed (depends on pads and headphone type)  ‘small room’. Slightly angled drivers are still firing directly into the ear canal by the way.
This calls for different compensation than is needed when sounds are coming from the front. These types of measurement rigs (HATS) are very expensive and when used with the wrong compensation (Tyll’s plots alas) can give very ‘confusing’ plots that don’t seem to have a real relation to ‘reality’ above 1kHz.

To me a graph should show a horizontal line when the sound is flat (i.e. receive the same amplitude of SPL at all frequencies). So headphone with too much bass should show more lower frequencies compared to the mids. With the mids I mean a ‘band’ between 400Hz and 2kHz roughly. When the treble is shown to be ABOVE these average mids it will sound ‘brighter’ and when it is below that ‘mids band’ they sound darker.
This is what the plots on THIS website represent.

I experimented with small mics placed in the ears but the results weren’t nearly as accurate as real life and also weren’t easy to repeat so abandoned this project pretty fast.

DSCN2405I kept my rig ‘simple’ and used 2 wooden boards covered with closed-cell foam rubber trying to emulate skin.
The planks are spaced 17 cm apart, which is about the same as that of my head.
This way the pads are compressed about the same amount as on my head (not entirely so but IMO close enough). Disadvantage is that because ‘wooden plank couplers’ are flat and not curved like a real head or a HATS the seal I may be getting may be better as one could get  in real life. It does make measurements more consistent though.
There are markings on the sides of the rig so re-positioning the headphones in the same repeatedly place is relatively easy to do. Lows are compensated acc. to various research about perception and the used capsule (WM61A) is known to have a peak of about +3dB centered around 16kHz. Both are ‘removed’ (compensated for) using hardware in the form of analog filtering BEFORE entering the ADC at line level. Therefore the frequency response plot I obtain is ‘RAW’ and ‘audible flat’ at the same time. No further compensations or HRTF are made/needed afterwards.

DSCN0013

The absorption of sound from the skin and refraction of sound as well as the amount of air inside the measurement chamber of my rig will DIFFER from the flat foam rubber surface, without a Pinna, from my rig.

This is something that simply is a disadvantage of a very simple and flat rig and needs to be taken for granted as an absolute error possibly being present depending on the headphone.

I do not calibrate 1kHz to be exactly 90dB SPL (as many others do) but do measure with average levels AROUND 90dB SPL in order to create a decent ‘distance’ to ambient noise level and to ensure both the microphone and headphone stay within proper operational conditions.

So take the plots in this section for what they are:
An amateur’s attempt to ‘measure’ the frequency response and obtain plots that have a fairly reasonable resemblance to ‘reality’ in the sense that a flat FR on these plots corresponds to an accurate reproduction of ‘well made recordings’. That may NOT coincide with how someone else perceives similar headphones.
Think product variances, sealing issues, head/ear shape or differences caused by the absence of a Pinna in the measurements.
The plots on this website differ considerably from those from Tyll (+ headroom) but differ less from those from Golden Ears and rtings.
To make things clear I really LIKE Tyll’s website and efforts in the headphone world as do I like Golden Ears and rtings each for their own reasons.
In no way do I claim my measurements are more accurate or much less accurate but they do DIFFER. IF you find they do not jive with your impressions of a headphone or have other views than don’t bother to peruse the measurements on this website other than just for ‘fun’.

 

post separation

These measurements have a different (larger) vertical dB scale compared to those of other websites so they might appear to have ‘worse’ FR behavior than other measurements out there. Always look at the vertical dB scales when comparing measurements between different websites. An example of this is given below. The top and bottom plot are the exact same sets of data (T50RP without FR compensation circuit). The top one is like the plots shown in this section the bottom one is as shown on a lot of websites or in a way a vendor would present it’s plot in order to boost sales (look… our headphone is very flat and accurate).scalesThere are no measurements of ear buds or IEM’s as the measurement rig does not have an ear canal simulator. Also NO measurements of expensive headphones (aside from the HD800) as these do not fall in the scope of this website and don’t have access to them.

The CSD plots shown on this website lack resolution and thus are not revealing enough. Only severe ringing can be seen and that’s what I was after, not pursuing accurate CSD plots.

Also note the scales may differ (vertical and horizontal) as I am still just playing with it and primarily use the rig to determine what the changes are (so relative to earlier measurements made) and not to compare with other measurements.

For accurate measurements I can recommend Golden Ears . Some plots appear to have measuring artifacts though. Golden Ears seems to be abandoned, no new measurements have been added. This really is a shame as it was a valuable resource.
Another interesting site is rtings with plots that have a high relation with how I perceive (and measure) headphones.
Tyll’s pdf’s (Innerfidelity) are also a very valuable source of information. Just remember his plots are downward sloping above 1kHz.
The plots do NOT have a high relation to how they sound above 1kHz. The plots slope downwards from there making them hard to evaluate.
Headroom’s plots are also quite valuable but you have to set the ‘graph maker properties’ to ‘compensated’).
(Changstar & SBAF) newer plots (the ones with harmonic distortion plotted as well).
They have a number of people submitting ‘measurements’ and some of them are quite ‘off’ from reality.
Also these Russian measurements can be used to extract info on FR. A
Rin Choi‘s blog has some interesting plots on his website (this site seems to be abandoned though).

Here is an article with some info on making measurements with home made ‘rigs’ and the differences between some alterations in the measurement set-ups. It also shows something of the home made rig I used to obtain the results below.

post separation

home
back to headphones

 

Comments
  1. coffeegal says:

    The picture you show by your test results of the Sennheiser HD 230 most definitely isn’t that model. The description also sounds completely off the mark. The HD 230 is a very transparent sounding set of headphones with a lot of treble extension and quite good bass. There was never any orange foam but black leather effect earpieces.

    I would expect the measured response to be very much better than that shown.

    A good picture can be found at http://i1153.photobucket.com/albums/p501/OldskoolRS/AV%20stuff/AV%20Room/HD230Phones_zps8620350e.jpg

    • coffeegal says:

      I’ve just fired up my HD 230 headphones and running them from a MacBook Pro headphone output (which has pretty poor bass).

      The test file I have found at http://www.youtube.com/watch?v=jSkH45nGuM8 gives a detectable (by my ears) effect at 14Hz and is uncomfortable to listen to at 16Hz. I think that means the bass output is pretty good 🙂 They do have separate concentric bass and treble drive units. Spare earpieces (containing drive units) are still available at £84 each. I’m not sure what the headphones you have masquerading as HD 230s are.

      Sarah (audio geek)

    • Solderdude says:

      You are absolutely right, I made a typo it should have said HD320 instead of HD230 (1993).
      Thanks for the pointer !

      I don’t own this headphone, but know the owner and he wanted it repaired and I measured it.

      The HD230 is 600 Ohm and seems to have been the top model at that time (1985).
      It comes from around the time the legendary HD414X was launched (1986)
      Had no idea about the existence of the HD230 and it seems to be a rare one as it is indeed a 2 way system. I am not aware of many 2 way systems, aside from many IEM’s with 2 drivers, these even exist with up to 10 drivers tucked away in a very small earpiece.
      The AKG K340 also is a 2-way system but uses an electret driver for the highs.

  2. justus says:

    hi, i coulnt find a description of your measurement rig. An 1/4″ omni flush w a baffle like the standard rig in the linked article? great stuff guys!

    • Solderdude says:

      There isn’t a detailed description of it. The last link in the post above (just above the comments) shows a picture of it. 2 WM61A mics are used but generally use one 1 of them. The mics are compensated for the boost in lows they get being baffle mounted as well as the 16kHz peak all these mics have so the output signal of the pre-amp has a ‘linear’ FR signal that does not need software compensation. The pre-amp is a bit noisy so I cannot perform (meaningful) distortion tests with it. It is a very basic setup but works for me and produces results that closely resemble what I hear (in most cases). Because there are no pinnae nor tubes used I cannot measure earbuds nor IEM’s (they are difficult to measure correctly anyway)

  3. Justus says:

    Cool, thanks! I like the simple setup (esp. since coupler are no good, and artificial ears too expensive). How did you decide to compensate for this LF boost?I dont see this boost in artificial ear measurements. How do you decide its headphone or mic? In free or diffuse field you’d get a boundary effect 6dB boost, but this is not applicable here afaik. Justus

    • Solderdude says:

      Hi Justus,

      I had noticed my measurements were ‘off’ in the lows compared to those made with calibrated HATS.

      The lows correction wasn’t that difficult to do as I have several headphones that have been well documented and measured with known calibrated set-ups.
      All I needed to do was measure a reference headphone (in this case the HD650) which measures repeatable, and compare everything below 3kHz with those made with a real HATS.
      This gave me the needed correction curve which was simply applied in the form of an RC filter.
      This of course is a very cheap way of correcting but accurate within 1dB.

      The missing 6dB in the lows is a psycho-acoustic phenomenon that doesn’t bother experienced listeners in most cases.
      For those who want to now what the missing 6dB effect is:
      When listening to ‘flat’ speakers in a room and then putting on a headphone which plays at the same SPL (so equally loud) measured at the listening spot will give the ‘impression’ some bass is missing in the headphone when they are equally ‘flat’ in reproduction.
      For this reason most people prefer headphones with some boost in the lows.
      This effect becomes less at higher listening levels so seems partly psychoacoustic/acoustic.

  4. outkayn says:

    Hi!
    I need to buy studio headphones (to control my mixes, I have studio monitors and room treatment too). Im considering AKG K271 MKII or Audiotechnica ATH-M40x.
    I didn’t found test of the akg k271 on your site. Maybe u can tell me which pair of headphones will be more flat, and better for the mix in studio??

    Cheers.

    • Solderdude says:

      These headphones are good for monitoring. A ‘wonky’ frequency response can even be a blessing for monitor headphones.
      For mixing you need ‘flat’ and recommend the Beyerdynamic DT250-250 or if it can be an open headphone the Sennheiser HD600 or even Focal Spirit Professional.
      The HD600 is also excellent for hifi, followed by the FSP and DT250-250.
      The latter is not as refined in the highs.

  5. Pierre says:

    Just an idea. It is possible using those silicone ear model to simulate ear canal? Here’s some of the product link that I think could work so that it will mimic our ears more accurately.

    https://www.amazon.com/Life-Like-Silicon-Ears-pair-Life-like/dp/B0058VOL4C

    • Solderdude says:

      I had thought about using fake Pinnae and even thought of making a mold from mine.
      At the same time I google pictured ears and found a wide range of ear shapes and sizes with different Concha’s.

      With some headphones adding a fake Pinna has more influence then others.

      The Pinna does the ‘most’ for sounds coming from in front of you. It has a different Concha gain then when sounds come from the side.
      It also ‘shields’ sounds coming from the rear of the driver.

      I decided that all I want to know is what the driver measures like at the ear canal entrance with about the same ‘reflection and absorption’ of sound as human skin.
      Sounds we hear are ‘processed’ and ‘compensated’ by the brain anyway so the ear canal does not need mimicking and subsequent (and far from perfect) ‘removal’ again.
      I do plan to do some more research on Pinna effects as around 2-3kHz there is some boost by the Concha. But not nearly as much as sounds coming from the front.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s