Audeze LCD-X Over Ear Open Back Headphone Review

[amirm]

Just measuring and reporting FR on one particular test rig isn't all the info that is needed.

I didn't just measure and report the frequency response. I confirmed its correctness using both subjective testing (EQ and listening) and verification with the company's own EQ suggestions. The latter was done with measurements of the correction. For this headphone at least, the full story of what it is about completely fell out of the data presented. Nothing more is needed as we know all we need to know about it now.

Use of a standard measurement rig for which we have listener preference data was critical to know what was missing. This was a problem with B&K 5128 Hats I used earlier where we did not have this so interpreting the results was very difficult and vague. I have tested a number of headphones with this GRAS fixture and it is remarkable how beneficial EQ is when one follows what the measurement show.

 

[Cahudson42]

+1 on all the thoughts 'let see how the testing develops' - as opposed to the 'lets not bother, it's already done, nothing will be added' school..

Though I have nothing specifically in mind, I'll bet all Amir's past experience with electronics and speakers brings something to the party that others might not have had..

I thought this when I saw the new Dashboard - and the whopping SINAD discrepancy. We can debate if it means anything, but here clearly was a difference in measurement results worth discussing..

Looking back at past early DAC and speaker tests, I always miss the Reconstruction filter graph - and speaker distortion ones. As they got added later - after experience.

The same will happen here. Just wait..

 

[Jimbob54]

I think their sales pitch is for pro use, where you can plug them into any point in the chain and get acceptable volume level. The closed-back version has much better FR and you’d think would be way more appropriate in a pro setting. It would be interesting to see an analysis of the compromises they had to make to hit the impedance/sensitivity target.

Not sure it holds true for all revisions but at one point I think the x and xc were the same cans except for the cups and padding. Which raises the question which one did the manuf prefer.

 

[Cahudson42]

How about a consolidated list of 'things that might be measured' (in addition to FR) that might make a difference?

'Pulse response'? (the 'fast bass' discussion..)
Rate of change of SPL vs. frequency? (FR derivative)...
Rate of change of distortion vs. SPL? vs. F?
And??? (Including any other 'transitional' rather than 'steady state' measurements?)

 

[DualTriode]

seriously? I have it, it's decent but I never fell in love with it. it seems to be more for relaxed listening or something.

The key word is equalize.

Harmon developed a list of equalization curves to apply to the chameleon (my description) Harmon K712 headphone.

We are talking about research done by a PhD.

They wanted to control for the visual and physical difference among all the different headphones tested.

They used a GRAS test fixture to measure the frequency response curve for each of the tested headphones. Then they developed an equalization curve for each of the tested Headphones to apply to the chameleon K712 headphone.

Now they have a single headphone flexible enough to equalize to the signature Frequency Response of each of the list of tested headphones.

If you do not have a file full of equalization curves you experience with the K712 will be off the shelf stock.

Thanks DT

 

[Degru]

The point is fair, but this is one of the few headphones that uses passive electronic filtering.


Rather than think of the transfer function of some distant sound source, as HRTF is normally considered, this is more about reflections off the enclosure, the uniformity of the driver radiation in the near-field, and the like. Most of the transfer gain hump is tied to direction-independent resonances of the ear canal.

The passive filter only gets rid of a narrow peak at 3khz. The rest of the FR is still tuned the traditional way. Besides, I don't think there's anything wrong with using passive filters for tuning if applied appropriately as was done here.

 

[solderdude]

I didn't just measure and report the frequency response. I confirmed its correctness using both subjective testing (EQ and listening) and verification with the company's own EQ suggestions. The latter was done with measurements of the correction. For this headphone at least, the full story of what it is about completely fell out of the data presented. Nothing more is needed as we know all we need to know about it now.

Use of a standard measurement rig for which we have listener preference data was critical to know what was missing. This was a problem with B&K 5128 Hats I used earlier where we did not have this so interpreting the results was very difficult and vague. I have tested a number of headphones with this GRAS fixture and it is remarkable how beneficial EQ is when one follows what the measurement show.

Yes I have no doubt this is the case. I merely commented on the raw measurements + target response which are difficult to interpret.
It is easier for people to see compensated plots where a straight line is 'audibly neutral' at say 80dB SPL, everybody else uses these and are intuitive.
When the target curve is known this should be really easy to produce.

I have spent literally days refining measurements to arrive at the set you see here. The Audio Precision is far easier and more powerful to use in this regard than other solutions.

I believe this without a shadow of a doubt. Only showing 40Hz and 100dB SPL, however, is just a tiny, tiny bit of the entire distortion puzzle is what I was trying to explain. I am not against FFT as a plot.
I too have used multitones with headphones and it is a visible mess. Decided not to publish hem. You can hardly believe some headphones that show such a mess can still sound unbelievable good.

 

[Robbo99999]

Yes I have no doubt this is the case. I merely commented on the raw measurements + target response which are difficult to interpret.
It is easier for people to see compensated plots where a straight line is 'audibly neutral' at say 80dB SPL, everybody else uses these and are intuitive.
When the target curve is known this should be really easy to produce.

I massively prefer to see the raw frequency response overlaid on the Harman Target Curve, which is the way Amir chose to show the measurement. I find that a lot more intuitive. It's easy to see if the measured response is above or below the target curve at any given point, it's not rocket science. It's also good to know from a knowledge and intuitive point of view that a flat sounding headphone response is not a straight line. It also gives more relateable information when comparing against other target curves. I think it's greatly more useful for people to become familiar with reading the actual frequency responses rather than some kind of compensated frequency response which only shows deviation from the target.

EDIT: I wouldn't mind if a compensated graph was given in addition to the raw frequency response, but I would prefer all discussion and "annotations" to be based around the raw frequency response and I'd keep the compensated graph on the level of a footnote add on so to speak.

 

[Vini darko]

I massively prefer to see the raw frequency response overlaid on the Harman Target Curve, which is the way Amir chose to show the measurement. I find that a lot more intuitive. It's easy to see if the measured response is above or below the target curve at any given point, it's not rocket science. It's also good to know from a knowledge and intuitive point of view that a flat sounding headphone response is not a straight line. It also gives more relateable information when comparing against other target curves. I think it's greatly more useful for people to become familiar with reading the actual frequency responses rather than some kind of compensated frequency response which only shows deviation from the target.

EDIT: I wouldn't mind if a compensated graph was given in addition to the raw frequency response, but I would prefer all discussion and "annotations" to be based around the raw frequency response and I'd keep the compensated graph on the level of a footnote add on so to speak.

I agree

 

[Theriverlethe]

Here is one application where the LCD-X would be great due to clean output and low distortion with heavy EQ.

They would be the worst choice in the Audeze lineup, given the massive amount of correction needed.

 

[imagidominc]

I'm so excited for more headphone reviews! I have the Elex, HD800, HD6XX, HD600, HE400i, and DT770. I can't wait to see how they all perform, for better or for worse. I don't like to EQ (I can listen to the HD800 stock and not be bothered by the treble spike), so it's good to know that me skipping the LCD X was the right choice for me.

 

[Feljor]

amirm could you please check your sennheiser HD 650, Heather CX and Hifiman HE-400i

 

[amirm]

amirm could you please check your sennheiser HD 650, Heather CX and Hifiman HE-400i

Yes. I have already tested them once but not using the latest version of the tests.

 

[BYRTT]

Thanks review Amir and the nice pro graphs as always, sounds a bit as you happy about the loaned Gras 45CA unit so is it a keeper ..

.....I was surprised how well this worked. I thought surely it would be too much but it was not.....

Have graphed your EQ filter string below for left channel, it looks a good eyed dial in provided Roon probably cant help read any target curves..

Thanks for the benchmarks on AP where LCD-X had so low distortion verse HD650, for that low distortion reason if you interested hear LCD-X convoluted to Harman curve as seen below i can offer 24/32/64bit IR Wav-files at whatever rates for Roon, though if interested i would like you share left & right channels txt-files plus the Harman target because quality of tracing your curves seen above and below is not superprecise with all that combing in high end and also because one curve is dotted, plan is then support either a stereo convolution wav-file where frequency above 1kHz is a avarage of left + right channels, or each a dedicated mono convolution wav-file per channel but for example in Jriver it gets complicated using mono files in one then have to create a formatted config file that points to those mono files and if procedure is the same for Roon suggest to save on your time use the avaraged stereo version..
S

 

[amirm]

Thanks for the benchmarks on AP where LCD-X had so low distortion verse HD650, for that low distortion reason if you interested hear LCD-X convoluted to Harman curve as seen below i can offer 24/32/64bit IR Wav-files at whatever rates for Roon, though if interested i would like you share left & right channels txt-files plus the Harman target

Man you are so good at creating these graphs! Yes, love to have the exact convolution for it. I have enclosed the target and measured response.

 

[amirm]

And yes, a single file would be fine. Some of the differences in the measurements is due to fit which would exist as a variable regardless.

 

[DualTriode]

@amirm and All,

I am glad to see this review of the LCD headphone.

It seems to me that it is as much a review and discussion of test procedures, the AP hardware & APx500 software, as well as the GRAS 45CA.

I am thinking that a set of tests would include raw Frequency Response data plotted on top of the Harmon Curve, today this is pretty much a standard that most understand. Thanks for posting the GRAS, Harmon Curve data points.

What also should be included is an equalized curve.

I will be looking at a Log Sine Sweep, chirp, and H2 and H3 plotted vs. frequency for a distortion test and plot. Maybe try a Multi-Tone as well.

For disclosure there is an APx555 and APx1701 on my bench. About 4 weeks ago I placed an order for a GRAS 45CA-9 that will arrive about the first of November. I plan to participate and share some measurements as well.

Thanks DT

 

[Robbo99999]

Thanks review Amir and the nice pro graphs as always, sounds a bit as you happy about the loaned Gras 45CA unit so is it a keeper ..


Have graphed your EQ filter string below for left channel, it looks a good eyed dial in provided Roon probably cant help read any target curves..
View attachment 88332
Thanks for the benchmarks on AP where LCD-X had so low distortion verse HD650, for that low distortion reason if you interested hear LCD-X convoluted to Harman curve as seen below i can offer 24/32/64bit IR Wav-files at whatever rates for Roon, though if interested i would like you share left & right channels txt-files plus the Harman target because quality of tracing your curves seen above and below is not superprecise with all that combing in high end and also because one curve is dotted, plan is then support either a stereo convolution wav-file where frequency above 1kHz is a avarage of left + right channels, or each a dedicated mono convolution wav-file per channel but for example in Jriver it gets complicated using mono files in one then have to create a formatted config file that points to those mono files and if procedure is the same for Roon suggest to save on your time use the avaraged stereo version..
SView attachment 88333

The one problem I see with these "automated" convolution files is that you are correcting for every little deviation in the measurement, and given the measurement above 10kHz is unreliable in both measurement terms (headphone position impacts over 10kHz greatly) and also in translation to how it would fit an individual (as individuals have wildly different responses above 10kHz) - then I see large & high Q corrections above 10kHz as likely to be doing more harm than good in terms of improving the sound. I think you'd want to limit above 10kHz changes to something similar to High Shelf Filters or broad peak filters eg Q1-2. I also feel a bit uncomfortable about (completely) filling in massive sharp dips as seen at 4kHz, I feel that this could be detrimental to the sound, I don't think high Q filters should be used if possible and even more so in the treble region.

 

[BYRTT]

Man you are so good at creating these graphs! Yes, love to have the exact convolution for it. I have enclosed the target and measured response.

Thanks files and warm note..

And yes, a single file would be fine. Some of the differences in the measurements is due to fit which would exist as a variable regardless.

In a read up on Roon website could see DSP engineroom is 64bit as for Jriver and 64bit IR corection was supported for wav-extension so dialed them as stereo channel 64bit IEEE wav-files and seperate 44,1/48/88,2/96/176,4/192kHz rates into zip-folder, should it work interesting could see its also possible for Roon make automatic shift per rate instead of manual point to whatever of the 6 unzipped files, it looked fair easy enough ala Jriver that one create a config file per rate but strange enough then one should zip all the files and point to the zipped folder. Probably best startup Roon without cans connected and check up how gain structure looks in that 4kHz boost is more than normal hot, and a practical note is correction is of pure minimum phase although i set center to minus 10mS causing a small buffer delay, have fun and look forward whatever analyze brings for that dedicated trial and if for example you like something edited its okay, for example as Robbo99999 suggest we could omit correction above 10kHz or whatever. For other LCD-X owners feel free use the attached convolution files but note they dedicated to Amir's analyzed sample so no garantee and maybe wait for Amir's feedback because if he state they useless for the dedicated sample then its most probably worthless corection for any other samples.

First graph is the raw L/R channels and their avarage curve to base the reverse convolution filter when divided with the Harman targets curve, well there is also a realistic low end high pass filter to hinder we ask too much of transducer think it was of 3rd order @22Hz, second graph using nice ASR colour scheme is the raw avarage curve that times reverse EQ form a 22Hz high passed ideal looking Harman curve..

The one problem I see with these "automated" convolution files is that you are correcting for every little deviation in the measurement, and given the measurement above 10kHz is unreliable in both measurement terms (headphone position impacts over 10kHz greatly) and also in translation to how it would fit an individual (as individuals have wildly different responses above 10kHz) - then I see large & high Q corrections above 10kHz as likely to be doing more harm than good in terms of improving the sound. I think you'd want to limit above 10kHz changes to something similar to High Shelf Filters or broad peak filters eg Q1-2. I also feel a bit uncomfortable about (completely) filling in massive sharp dips as seen at 4kHz, I feel that this could be detrimental to the sound, I don't think high Q filters should be used if possible and even more so in the treble region.

Can follow most points no problem but also think its the perfect place to do this test for many reasons as Amir is trained listener and also have lab gear to quality check should he feel for it, plus members can debate their view on whatever result Amir will feedback.

 

[Robbo99999]

Can follow most points no problem but also think its the perfect place to do this test for many reasons as Amir is trained listener and also have lab gear to quality check should he feel for it, plus members can debate their view on whatever result Amir will feedback.

Sure, there's no problem in trying them, but the research is there to show that there's no point (& detrimental by extension I suppose) making fine corrections above 10kHz, and also filling in all those sharp peaks in the treble I also don't agree with, but no harm in trying it of course.

What are the advantages of convolution vs regular parametric EQ filters?