Archsam
Senior Member
So now who is going to send Amir the Clear? I am not letting my Clear Pro go away from me for one second.
It's a long way to ship mine from the UK....
So now who is going to send Amir the Clear? I am not letting my Clear Pro go away from me for one second.
Audeze LCD headphones non-EQed preference scores:
81 - LCD-2 Closed (crinacle)76 - LCD-2 Closed (oratory1990)74 - LCD-1 (oratory1990)71 - LCD-XC (crinacle)70 - LCD-2 (oratory1990)70 - LCD-2 Fazor (oratory1990)65 - LCD-1 (crinacle)52 - LCD-4 (oratory1990)51 - LCD-2 Classic (oratory1990)50 - LCD-X (oratory1990)48 - LCD-2 Classic (crinacle)40 - LCD-X (crinacle)31 - LCD-3 (crinacle) I have these... and rather like them.
Scores: https://github.com/jaakkopasanen/AutoEq/blob/master/results/RANKING.md
Martin
Whoa I didn't realize my Onkyo A800 scored that high.Audeze LCD headphones non-EQed preference scores:
81 - LCD-2 Closed (crinacle)76 - LCD-2 Closed (oratory1990)74 - LCD-1 (oratory1990)71 - LCD-XC (crinacle)70 - LCD-2 (oratory1990)70 - LCD-2 Fazor (oratory1990)65 - LCD-1 (crinacle)52 - LCD-4 (oratory1990)51 - LCD-2 Classic (oratory1990)50 - LCD-X (oratory1990)48 - LCD-2 Classic (crinacle)40 - LCD-X (crinacle)31 - LCD-3 (crinacle) I have these... and rather like them.
Scores: https://github.com/jaakkopasanen/AutoEq/blob/master/results/RANKING.md
Martin
Not sure what the stated accuracy is, but I would assume the preference ratings for headphones is not as accurate as with speakers. And, unless Amir does a whole suite of measurements (phase matching, group delay, etc.), we currently can’t do a deep dive to see how well the headphones image, how wide their soundstage is, how well the emulate the sound as being in front of us and not in our heads, etc.
And of course you have to actually wear the things so a good chunk of people will choose worse sounding headphones if they are noticeably more comfortable.
Unless I’m mistaken, here is the formula for AE/OE:
114.49-(12.62*SD)-(15.52*AS)
Where:
SD is standard deviation from 20Hz-10kHz
AD is absolute value of slope 20Hz-10kHz
There isn’t even a bass frequency portion like there is with the speaker preference rating.
Here is the accuracy of their model:
View attachment 88138
Compared to the preference rating for speakers:
View attachment 88139
For headphones, there were only 3 models that got scored >6/10 by humans, compared to ~15 models for speakers
- The IE target curve was developed using a G.R.A.S. RA0045 IEC60318-4 compliant coupler. Other common fixtures for testing IE headphones such as B&K 4128, B&K 5218, G.R.A.S. KEMAR, etc. have not been validated against the target and may not produce results that correlate 100% to the predictive model.
- The AE/OE target curve was developed using a G.R.A.S. 45CA headphone test fixture with a custom pinna. Other common fixtures for testing AE/OE headphones such as B&K 4128, B&K 5218, G.R.A.S. KEMAR, etc. have not been validated against the target and may not produce results that correlate 100% to the predictive model.
Huh, I find it odd that the in-ear has higher correlation, as that fully bypasses our pinna so there should be more “uniqueness” to our personal target curves. As you have IEMs from the likes of Etymotic that severely lack in bass compared to the Harman IEM curve, and yet they are still loved.The correlation between predicted and actual preference rating for the over-ear headphone formula is 0.86, exactly the same as the speaker formula (and even higher at 0.91 for the in-ear formula). Imaging and soundstage are primarily functions of a track's recording/mixing, headphone frequency response (including channel balance, sure), the listener's PRTF, and to a lesser extent phase matching and group delay as you mention (which I do think would be valuable to include in reviews). Speaker imaging and soundstage have similar contributing factors (room acoustics and directivity interactions with this being a loose analogue to PRTF), but speakers are tested in mono due to frequency response being the major relevant factor. I don't see why this should be drastically different for headphones.
I don't really understand your point about there being no 'bass frequency portion' (LFX?) in the headphone formula. The need for this in the speaker formula was due to the other parameters ranging only from 100 Hz upwards - the headphone formula doesn't require this as its parameters already cover the bass region. I don't get the comfort objection either - it's purely a sound preference formula, just like the speaker formula, so of course it doesn't include things like comfort, just as the speaker formula doesn't include non-sound related buying criteria like size, shape, aesthetics etc.
And I don't think it's much of a surprise that fewer headphones than speakers received very high ratings. (The graph you posted is for in-ears not over-ears by the way - the latter had 6 headphones rated >6/10 by listeners, not 3, so not as big of a difference to speakers.) On average headphone manufacturers are even more behind the (Harman!) curve than speaker manufacturers, due to headphones' more recent consumer ubiquity and the (I'd argue erroneous) refusal from many companies to adopt something similar to the Harman target. Then there's the PRTF issue which means you'd have to be incredibly lucky to find any headphone that perfectly matches your specific pinna response, as well as the fact that headphones produce no (or at the most minimal) tactile bass, both problems speakers don't have. And then you have to consider most music is mixed and mastered on (and to sound best using) speakers, not headphones.
EDIT: I've noticed on that Listen Inc. page under Directions > Notes on Running the Sequence, they say this:
So it looks like in order to have the most accurate preference ratings, the RA0045 coupler should be used (as Oratory does), not the newer 'hi-res' coupler.
Remember that the Harmon Curve is measured for over the ear headphones.
The in ear phones are another story.
DT
Huh, I find it odd that the in-ear has higher correlation, as that fully bypasses our pinna so there should be more “uniqueness” to our personal target curves. As you have IEMs from the likes of Etymotic that severely lack in bass compared to the Harman IEM curve, and yet they are still loved.
I find myself close to both targets, it’s just hard to wrap my head around how the in-ear can be more accurate than the over-ear.
As for the bass portion aspect, that was me just saying that doing a simple log scale of the frequency band and treating every band as similar also seems more simplistic than I would imagine.
Of course comfort can’t be part of the formula, I was just saying that comfort plays a large role (as well as clamping force, pad size, etc.), so it’s more personal than speakers and thus more info is very welcomed.
I just don’t want Amir to get too bogged down with headphone measurements where sufficient measurements exist elsewhere, unlike with hardware and speakers where his AP and the Klippel allow for more detail than pretty much anywhere else. That once/if Amir buys that test unit and makes sure it’s all calibrated and ready to go, that it’d be more useful to do headphones/IEMs where measurements don’t exist, or maybe not to a sufficient degree. They are much easier/cheaper in regards to shipping to and fro though, so users are more likely to send in one for measurements, rather than say a tower speaker.
I had a pair of Audeze LCD-2's, which I sold on. Their upper mids / low treble was just missing in action- - I am guessing that the LCD-2's showed a similar suckout in this range to the LCD-X in your test.
But the BASS of those LCD-2's! Such clean, tight lows! Just the best low end I ever heard from 'phones. Nonetheless, I could not live without upper mids and low treble so I sold them.
I did use some DSP EQ but I still wasn't satisfied- the frequency response was corrected but they lacked detail in that same range. I suspect that in addition to the frequency dependent magnitude issues there are also some serious time domain flaws.You just needed to apply a little EQ to them.
No, I can only generate an impulse response.@amirm it would be great to see the step response of headphones measured to see the time domain issues. The first couple of milliseconds typically show how good (in this case, bad) the time domain behaviour is and how far it deviates from the ideal minimum phase response. Not sure if your gear can do that?
No, I can only generate an impulse response.
I sent my LCD Xs to Riga in 2015 and use the custom for Ref 3, and yes, they really sound great after the correction.First reason suggest a precision filter was because of the unique low AP benchmarked distortion for LCD-X verse HD650, next reason is in support files method can save Amir some research time well other for case than he had to send me the real files all he need to do now is a listening test and even if it fails or succeed its probably also interesting stuff for him in process can probably also hint a bit about his loaned lab gear and used methods, for example about dedicated correction myself had very positive experience order head phones with a dedicated correction from Sonarworks even those set of cans was low cost compared to for example HD650, that said i try hold bit back comment on your personal Q rules and to other user comments never boost only do cuts because there is so much into EQ in modern days where real many system chains has excess phase verse in the past using pure analog circuits plus its not always sure a digital equalizer claimed or suspected to be of minumum phase is performing 100% minimum phase, and yes acoustic EQ is hard to execute right because most often measurements are executed with less precision than in a electric domain but if we up that and magnitude deviations are linear then EQ in acoustic domain is as precise as in electric or digital domain. It was not my meaning get started on EQ subject but i did a bit long story there sorry and now cant stop hang on below dedicated corrections for those low cost cans below and notice there is didicated corrections L verse R channel in all audioband including above 10kHz area its the upper one that is the dedicated to my serial number and the lower one is their avarage profile, these cans for me are unlisteable raw and think most will agree, they okay listenable using avarage profile and they shine using the dedicated..
View attachment 88407
View attachment 88408
Most normal advantage is we can do whatever FIR filter there but for this case advantage is it takes less than five minuttes create precise inverse Harman target verse the raw LCD-X curve and doing same precision curve using PEQs would probably take hours and maybe use numbers of PEQs up in three digit area.
I can only export the graph data, not the impulse.That will work, can you export it as a full resolution .wav?
Thumb up for that based my experience of dedicated correction for DT 770 Pro 250Ω, my HD650 run on the avarage profile and i have a set of Status Audio OB-1 that are unsupported so plan is some day send them both HD650/OB-1 and get dedicated corrections, that said when i synthetic modeled Sonarworks avarage correction to Amir's HD650 measured on the Bruel & Kjær 5128 HAT it was very close of a avarage profile so for HD650 because of selective transducers there is maybe not so much to come for as there was for DT 770 Pro 250Ω but lets see.I sent my LCD Xs to Riga in 2015 and use the custom for Ref 3, and yes, they really sound great after the correction.
That all sounds great, and I agree the NAD HP50 are good headphones once EQ'd, I like them better than my HD600 for instance - better bass extension, a less "in your head" experience (strangely enough considering they're closed back phones), and almost as good in terms of detail in vocals, etc, also a smoother headphone. But the bit re binaural recordings, I'm not sure that is valid. If it's trying to approximate measuring the frequency response of your headphones "close" to your eardrum......if that's the aim then of course the frequency response will not be a straight flat line but will be shaped similarly to the Headphone Harman Curve in terms of it reacting to the fact that your ear & ear canal amplify or attenuate certain frequencies - so by extension if you've made a recording that reflects this and listen to it back on speakers then the frequency response will be totally messed up in terms of it not sounding right because when you listen to a speaker in a room you want a "flat" frequency response of course, not a Headphone Harman Curve type response. So I don't think binaural recordings of headphones are valid unless I've missed the aim or missed out some kind of functional processing step of the recording.I listened to and measured the Audeze LCD-4z. I also made some binaural recordings and compared them to both the original recording and recordings made using the NAD VISO HP50. One can download and listen as it is a fun comparison.
The LCD-4z also have the strange peaks and dips and sound strange to these ears. Not my cup of tea. Especially if one compares to the HP50's in the binaural recordings linked above.
Re: convolution versus PEQ's. It depends on which DSP software is used to create the convolution filters. Top end DSP FIR filter Designers give you full control over every audio parameter, even taking out reflections/time domain anomalies. But you need to know what you are doing. Also, it does not have to correct for every possible frequency response bump. You can vary the Q, not only overall, but frequency dependant as well. Meaning one can be more aggressive in the bass, but less so as the frequency rises or even apply a partial correction. Same applies for the time domain correction.
@amirm it would be great to see the step response of headphones measured to see the time domain issues. The first couple of milliseconds typically show how good (in this case, bad) the time domain behaviour is and how far it deviates from the ideal minimum phase response. Not sure if your gear can do that?