New Audeze LCD3
Jan 4, 2012 at 9:57 AM Post #2,971 of 11,521
Just an observation of the SR009 measurements. Here are two measurements from the exact same headphone (before and after burn-in). I am not a big proponent of headphones changing much after burn-in; but there appears to be up to a 5dB change at 6kHz (for the worse after burn-in), at 10dB change at 10kHz (also for the worse), and approximately 4dB-ish just after 1kHz.
 
http://www.innerfidelity.com/images/StaxSR009SZ91278.pdf
 
http://www.innerfidelity.com/images/StaxSR009SZ91278afterburnin.pdf
 
Now is that due to burn-in or issues with the measurement system (people+gage)? Hard to say. Some don't believe in burn-in; I personally find that changes are minimal, but in this instance, I would venture to say that a good part of the variations seen are from the measurements themselves. But without concrete statistical data, it is hard to say. Again, sorry to others for this boring 6 Sigma babble, but when it comes to critical measurements, it can be well....critical.
smile.gif

 
Jan 4, 2012 at 10:02 AM Post #2,972 of 11,521


Quote:
Tyll, 
 
you heard the man: now go on and measure the same headphone 200 times and do not come back until we know that you can get down to .1dB. After all cm, decibel, ruler, neuman head, headphone, airplane lever, it's all the same business, isn't it?! And don't forget to fill in your QC report please :wink:
 


It's not 200 times. But yes, it can be a lot of work. Please read my previous post.
 
Honestly, I've seen many an R&D effort to understand a phenomena that was not really there, but something that the measurement system wasn't up to par on and showed something that was not there. I'm not saying there isn't anything going on with Audeze headphones (or any other manufacturer's from a variability stand point), just that from a statistically significant point of view, it's much harder to judge.
 
I think we all get hung up on measurements so much that we forget that we listen with our ears too.
 
 
Jan 4, 2012 at 10:26 AM Post #2,973 of 11,521
Catching up on this thread, it seems to me to be pretty clear that there is some sort of manufacturing or other variability issue with the LCD-3. When you couple together the fact that listeners seem to disagree significantly over some aspects of the sound in a way that seems to go beyond listening preferences, that some people who have listened to two pairs one after the other have reported they sounded different, and some different looking measurements, I think we almost have to conclude that this is being caused by some sort of sample variation, although as has also been pointed out this could be anything - it could be pads just as easily as drivers- we don't and as Peter (MH) points out CANNOT know the cause, since we lack the ability to determine it.
 
I personally am a bit relieved to understand why my own experience was different from that of some other experienced listeners. I also think that this needs to be addressed by Audeze.  They may decide that some variability is just the way it is, and then buyers have to decide how to factor that in.  Or they may already be trying to address it for all we know.
 
As MH pointed out, there is a thread that was running a long time about "two versions" of the Beyer T1.  And supposedly there was a running change to the HD800 that alleviated the 6kHz peak at least somewhat.  So it is not like high end headphones have never had such an issue.  Heck, Sony changed the R10 from "bass light " to "bass heavy", let's also remember.
 
Head Fi is serving a great purpose here.  A potential issue has been raised.  Now we shall see how it unfolds from here. 
 
Jan 4, 2012 at 10:56 AM Post #2,974 of 11,521
It's disingenuous at best, misrepresentation at worst on Audeze's part to claim QC has improved over LCD-2 when cursory evidence suggests production issues. The fact that the mids issue has consistently been reproduced by many members (myself included, on one of the 2 pairs I had) points to more than your average defect ratio.
 
They will probably address it in the form of LCD-3 rev. 2... circa 4 months before announcing the even better LCD-4 
bigsmile_face.gif

 
Quote:
 
Head Fi is serving a great purpose here.  A potential issue has been raised.  Now we shall see how it unfolds from here. 



 
 
Jan 4, 2012 at 11:09 AM Post #2,975 of 11,521
Quote:
Now is that due to burn-in or issues with the measurement system (people+gage)? Hard to say. Some don't believe in burn-in; I personally find that changes are minimal, but in this instance, I would venture to say that a good part of the variations seen are from the measurements themselves. But without concrete statistical data, it is hard to say. Again, sorry to others for this boring 6 Sigma babble, but when it comes to critical measurements, it can be well....critical.
smile.gif


That actually looks like pretty good evidence (and so far AFIK the only "good" evidence) of burn-in.  The overshoot on the square waves increased noticeably after the burn in which would seem to indicate some sort of stretching or loosening of the diaphragm.  That's not a huge difference by itself.  The difference in the THD+N on the left channel would probably be clearly audible in an ABX using pure tones and is much more important IMO.
 
The big difference is in the THD+N plots.  That's clearly not the sort of thing which is very dependent on chance.  It such a large difference in THD+N could be blamed on measurement error you probably wouldn't see much consistency between the many of the models with multiple samples or even between TOTL and bargain basement 'phones.
 
With the LCD-3s you seem to be mostly concentrating on the FR while that's the least important difference.  The higher the frequency the more difficult it is to plot an accurate amplitude because of exact placement changes the interactions between the headphone and the ear causing peaks and nulls but that doesn't change the validity of what the tests indicate goes on at lower frequencies.  The distortion plots and 300Hz square wave are the real issue with that second pair of LCD-3s, not just a slightly rougher FR.
 
Jan 4, 2012 at 11:11 AM Post #2,976 of 11,521


Quote:
Head Fi is serving a great purpose here.  A potential issue has been raised.  Now we shall see how it unfolds from here.


 
I think the "great purpose" being served here is not so much the particulars of the LCD-3 investigation as it is an overall change in the way we go about evaluating cans. I think Purrin, Arnaud, and myself have brought an additional view into the headphones we listen to: objective measurement.  I'm very happy that there's a skeptical reception of these measurements, and I'm very glad there's dialog questioning the relationship of our listening to the measurements. This tension between what we hear and what we see in the measurements is very healthy. 
 
Previously there was no tether to the rampant "he said, she said" subjective opinion. Measurements bring in another view that allows us, at least in part, to compare listening experiences with objective measurement. Neither approach will lead us to a clear center ... because there isn't one. Subjective experience and personal preference will always deliver a somewhat fuzzy take on what's "good." But the measurements and experienced listeners can nicely reinforce what's bad, and poor performers will get incisively culled from the heard.
 
While we'll always rumble about cans at the top of the heap, the crap at the bottom will be quickly identified, and manufacturers will be handily slapped upside the head when they deliver inferior product. 
 
Manufacturers of good product will forever be tweaking there offerings, they wouldn't be good manufacturers if they didn't keep trying to improve. It doesn't bother me that things might be changing with particular skus at Audez'e or Sennheiser or Beyer --- they're just doing there job.
 
What bothers me is crap being pawned off on the public as good stuff. And that is being addressed with the new feature of objectivity we're dialoging about.
 
Great job folks!
 
Jan 4, 2012 at 11:34 AM Post #2,977 of 11,521
Hey Tyll, I remember you saying you periodically check your measurement rig for consistency by retesting your pair of HD800s.  Since I think the THD+N plots are the most significant difference between the two pairs of LCD-3s I was wondering how much variation you see in that measurement on multiple runs of a 'phone know to be working properly.
 
Is it as consistent as I'm inferring or am I reading too much into your graphs?
 
Jan 4, 2012 at 11:52 AM Post #2,978 of 11,521


Quote:
That actually looks like pretty good evidence (and so far AFIK the only "good" evidence) of burn-in.  The overshoot on the square waves increased noticeably after the burn in which would seem to indicate some sort of stretching or loosening of the diaphragm.  That's not a huge difference by itself.  The difference in the THD+N on the left channel would probably be clearly audible in an ABX using pure tones and is much more important IMO.
 
The big difference is in the THD+N plots.  That's clearly not the sort of thing which is very dependent on chance.  It such a large difference in THD+N could be blamed on measurement error you probably wouldn't see much consistency between the many of the models with multiple samples or even between TOTL and bargain basement 'phones.
 
With the LCD-3s you seem to be mostly concentrating on the FR while that's the least important difference.  The higher the frequency the more difficult it is to plot an accurate amplitude because of exact placement changes the interactions between the headphone and the ear causing peaks and nulls but that doesn't change the validity of what the tests indicate goes on at lower frequencies.  The distortion plots and 300Hz square wave are the real issue with that second pair of LCD-3s, not just a slightly rougher FR.

I'm not just fixated with the FR, but there are differences. The SR009 plots I showed were off the same headphone and revealed differences on most of the graphs. The THD+N was well. Now as mentioned, I've never been a big believer in burn-in....please feel free to venture over to the Science Forum for further discussions as I don't want to get into them here. As well, I'm not sure how the burn-in phenomena affects stat drivers over orthos or dynamics.
 
I think the deeper we dig, we might find some really interesting things.
smile.gif

 
 
 
Jan 4, 2012 at 12:03 PM Post #2,979 of 11,521


Quote:
I'm not just fixated with the FR, but there are differences. The SR009 plots I showed were off the same headphone and revealed differences on most of the graphs. The THD+N was well. Now as mentioned, I've never been a big believer in burn-in....please feel free to venture over to the Science Forum for further discussions as I don't want to get into them here. As well, I'm not sure how the burn-in phenomena affects stat drivers over orthos or dynamics.
 
I think the deeper we dig, we might find some really interesting things.
smile.gif

 
Part of my discussion with Sankar involved burning in the LCD-3. He indicated that every headphone they shipped had been sufficiently burned in at the factory, and that the sound quality should not substantially change from how it sounds at unboxing.
 
Now, whether this is actually the case or not is a matter of individual interpretation -- but I am personally inclined to believe the manufacturer's claim on this one.
 
Jan 4, 2012 at 12:30 PM Post #2,980 of 11,521
Most likely explains why I've experienced minimal effects with burn-in. I would think for $5k, Stax also does some factory burn-in (also to weed out any early life failures...typically where the majority issues are), so that would lead us back to why the SR009 measurements of the same pair of headphones measure so differently?
 
 
 
Jan 4, 2012 at 2:47 PM Post #2,981 of 11,521

Edit: sp. & repair link
Quote:
Just an observation of the SR009 measurements. Here are two measurements from the exact same headphone (before and after burn-in). I am not a big proponent of headphones changing much after burn-in; but there appears to be up to a 5dB change at 6kHz (for the worse after burn-in), at 10dB change at 10kHz (also for the worse), and approximately 4dB-ish just after 1kHz.
 
http://www.innerfidelity.com/images/StaxSR009SZ91278.pdf
 
http://www.innerfidelity.com/images/StaxSR009SZ91278afterburnin.pdf
 
Now is that due to burn-in or issues with the measurement system (people+gage)? Hard to say. Some don't believe in burn-in; I personally find that changes are minimal, but in this instance, I would venture to say that a good part of the variations seen are from the measurements themselves. But without concrete statistical data, it is hard to say. Again, sorry to others for this boring 6 Sigma babble, but when it comes to critical measurements, it can be well....critical.
smile.gif



 
Sorry to be not agreeing again but those 2 measurements look very reasonably similar and I don't see large variations like you claim except if picking very narrow peculiar bands.
 
One issue with non smoothed frequency response functions is that you're bound to see some dips in the response at some frequencies above 5k or so because you happen to hit a zone of cancellation at the mic location for an acoustic resonance at that frequency.
 
Space averaging the results should help but it's not sufficient in the 10kHz example you highlighted. Maybe marv will comment but these narrow deeps are not easily perceived by hearing and, attributable to the measurement error or not, you can simply ignore them.
 
As I posted originally, the 2 LCD3 plots don't look anywhere near. For example see the 5dB difference between 4-5kHz range and 6-7kHz range for sn2260 ( http://www.innerfidelity.com/images/AudezeLCD3sn2312260.pdf ) while sn2454 ( http://www.innerfidelity.com/images/AudezeLCD3sn2312454.pdf ) is quite smooth between 5 and 9kHz. If you look at raw data (grey) you see a sharp resonance at 5-6kHz in sn2260 (such 5dB sharp peak is clearly audible) but not in sn2454. Another key difference in the raw data is sharp peak just above 3k for sn2260 vs a much more damped resonance centered around 3.5kHz for sn2454.
 
Yet another difference is relative levels between upper mids (4-8kHz) and treble (13-20kHz). Sn2260 shows similar level while sn2454 show 3-4dB increase in the treble. 
 
Basically two different LCD3 headphones...
 
 
Jan 4, 2012 at 3:16 PM Post #2,982 of 11,521


Quote:
Edit: sp.


 
Sorry to be not agreeing again but those 2 measurements look very reasonably similar and I don't see large variations like you claim except if picking very narrow peculiar bands.
 
One issue with non smoothed frequency response functions is that you're bound to see some dips in the response at some frequencies above 5k or so because you happen to hit a zone of cancellation at the mic location for an acoustic resonance at that frequency.
 
Space averaging the results should help but it's not sufficient in the 10kHz example you highlighted. Maybe marv will comment but these narrow deeps are not easily perceived by hearing and, attributable to the measurement error or not, you can simply ignore them.
 
As I posted originally, the 2 LCD3 plots don't look anywhere near. For example see the 5dB difference between 4-5kHz range and 6-7kHz range for sn2260 ( http://www.innerfidelity.com/images/AudezeLCD3sn2312260.pdf ) while sn2454 ( http://www.innerfidelity.com/images/AudezeLCD3sn2312454.pdf ) is quite smooth between 5 and 9kHz. If you look at raw data (grey) you see a sharp resonance at 5-6kHz in sn2260 (such 5dB sharp peak is clearly audible) but not in sn2454. Another key difference in the raw data is sharp peak just above 3k for sn2260 vs a much more damped resonance centered around 3.5kHz for sn2454.
 
Yet another difference is relative levels between upper mids (4-8kHz) and treble (13-20kHz). Sn2260 shows similar level while sn2454 show 3-4dB increase in the treble. 
 
Basically two different LCD3 headphones...
 


Sorry, I don't follow...please review the 10kHz dips on both graphs (before and after burn-in). Aren't they averaged out as well? So why do they appear to be quite different (10dB at 10kHz)? Why should we ignore things that are measured? And this is on the same pair of headphones. Or how about the 4dB bump at 1kHz ish? Makes little sense to me...either its there or it isn't and why did the measurement system pick it up if it wasn't there? And this is on the same pair of headphones. I'm worried that we are analyzing things that are not there (or as big as we perceive them to be) and just limitations of the measurement system. Trust me, this wouldn't be my first time. Last year I had to cancel an R&D project that looked like there was some trending going on that turned out to be noise due to a limitation of our measurement system. I'm not saying that there is not variability pair to pair. Just remember to hot headed debates of the HD800s or T1s have treble spikes vs. no they don't.
 
I think we're only scratching the surface of this one.
 
EDIT: To not bore others, please feel free to PM me if you would like to discuss further.
smile.gif

 
Jan 4, 2012 at 9:04 PM Post #2,983 of 11,521


Quote:
It's disingenuous at best, misrepresentation at worst on Audeze's part to claim QC has improved over LCD-2 when cursory evidence suggests production issues
 


 



Holy cow, this thread has gone off the rails.  Since no one posting in this thread is privy to the internal Audeze design specifications, quality procedures or key process indicators for the LCD-3 or the company at large, I don't think that any such claims like the above can be made.  Who says that Audeze has not met internal guidelines for improved quality?  Maybe it was worse before?  Do you have access to every pair of LCD-2 and -3 internal quality test results?  I believe it is more likely that due to the nature of the product, Audeze may have quite a wide band of what is considered acceptable and that the shown performance of a majority of these units indeed fell within that acceptable band, thus the quality is significantly improved compared to the past.
 
Look, like it or not, these are bleeding edge products, made in very small batches and assembled with significant amount of human processing touches at multiple manufacturing steps.  If there wasn't variation in these first lots, I would be beyond surprised, personally.  I find it likely that Audeze and their suppliers were under tremendous pressure to get these products into the market as soon as the majority of processes and materials were close to, but not quite, perfected yet.  This includes sourcing new materials/vendors that may not be consistent and hiring new personnel to build to build them, who may not yet have the expertise in all phases in order to have the robotic R&R needed for these plot to line up the way some of you seem to demand.  As soon as the LCD-3 was announced, they were damned if the released too soon (for perceived "quality" issues) or damned if they didn't (for raising expectations and then delaying to market).
 
For these reasons and more, discussions of expected "quality" and six sigma for such a product at this point are ludicrous in my opinion.  You may as well apply such principles to batches of single malt whiskey, which are appreciated by the cognoscenti for their variations.
 
This is the price that you pay for being an early adopter: you get your hands on the product first but you also are more likely to see some/all the warts up close, before smoothing occurs in all manufacturing phases.  If you want to complain about being an early adopter, don't be one.  Really - if you're surprised or complaining about any of this, what did you expect?
 
Also, as noted, there is likely personal taste variations in the expectations of what the Audeze house sound exactly consists of for each individual. 
 
Finally, I do appreciate the measurements that have been done, but since there is no common recognized calibration source, method or standard for the various groups presenting data, this too must be greeted with some skepticism, as has been ably pointed out by Tyll and MH.
 
My personal feeling is that this particular avenue of discussion has veered into post market surveillance for Audeze, as has been noted by others.  Introducing objectivity into an inherently subjective experience will cause this unless all participants are equally well informed, equipped and experienced.
 
edit: In other words, my comment is: treat everything presented here with a huge "FOR REFERENCE ONLY" instead of treating it like a gospel for this product and a praise/condemnation of the company.  At this price point, you should try before you buy.
 
Jan 4, 2012 at 9:32 PM Post #2,984 of 11,521
@smeckles: there's no smoke without fire. I think we've seen quite a bit of the proverbial smoke, myself first-hand with two different-sounding pairs, and measurement notwithstanding, there are issues with the phones.
 
I can, and will, complain about any issue on a $2k product, regardless of whether it's leading edge, high-tech, whatever you want to call it. I happen to work hard for my money, and won't be some sort of guinea pig for a half-baked product.
 
I don't know how closely you've followed the discussions on their products, but this isn't the first time they had production issues.
 
Edit: Just like with any product or service, the minority experiencing issues is more vocal than the happy majority, so feel free to factor that in as well.
 
Jan 4, 2012 at 9:38 PM Post #2,985 of 11,521
All we know for certain mwilson, is that you received a defective LCD-3.
 

Users who are viewing this thread

Back
Top