[donkeywalker]

One thing that continuously frustrates me is the version/batching theories of headphone production lines.
 
Two HD800s or T1s may sound different out of box [1] , or maybe they sound similar initially but different after some time of use[2].
 
From my own experience, the new HD800 I got from Amazon and the used HD800 I got from headfi did sound very different in the bass compartment (seller had less then 100 hours use) with the later serial number more bass heavy, and now I just got a little overly bright T1 here which again made me wonder if there is a version issue because there are so many contradictory listening experience in headfi (of course, there is even a thread about dampening T1 [3]). 
 
It's hard to find official announcements , reports/statistics , or any explanation of these phenomenon from headphone producers. I'm wondering whether the headphone manufactures by default want us to assume that headphones of the same model should sound nearly the same. It seems like a serious issue if the producers can't control the quality to the degree that there is so observable difference in major frequency reproduction, or they want to adopt to new market needs without announcing an official version change.
 
Just want to know what you guys think about it and your experience 

 
 
[1] http://www.head-fi.org/t/433059/sennheiser-hd800-certificate-for-frequency-response-arrived/255
[2] http://www.head-fi.org/t/508836/the-official-beyerdynamic-t1-owners-club-appreciation-discussion-thread/2910
[3] http://www.head-fi.org/f/4/headphones-full-size

 

[ssrock64]

It's been an issue that hassn't been discussed much in recent Head-Fi history (apart from the LCD-3 wood fiasco of late), probably because nobody really knows exactly the issue.

 

[donkeywalker]

Yeah, I think there is lots of perspectives in this issues. Things like production process/difficulty, the different process used by the brands. Although not a big fan of IEMs but I've heard that UE TF10 has versions that sound drastically different.
 
I thought the frequency chart that comes with the HD800s is a showoff of the production quality, which possibly means that maintaining the same response curve is insanely hard to do and that level of individual differences is supposed to be good enough? 
 
Shipping with frequency response curve for each individual phone will solve all the mysteries. If it takes $20 extra to do so, the brands might be able to find some demands of those charts, though that might be the last thing they want to do 

.  

 

[ssrock64]

The thing about FR charts is that they tend to be completely unreliable. It all depends on seal, the testing equipment, amping, and so much else. That's why an InnerFidelity chart and a Golden Ears chart rarely match up.

 

[vid]

Quote:

The thing about FR charts is that they tend to be completely unreliable. It all depends on seal, the testing equipment, amping, and so much else. That's why an InnerFidelity chart and a Golden Ears chart rarely match up.

 
Frequency response graphs are no more unreliable than any human impressions you read on the internet, in print, or hear from your friend sitting next to you. And your own impressions are unreliable to yourself in that they're likely to change depending on context.

 

[TickleMeElmo]

Consider a company like Intel that produces millions of chips and then bins those chips by clock speed/voltage capability; you will still get chips that run at much lower stock voltages than others. This is a company that literally spends billions on R&D and strives to get better yields on their products. CPUs on a given process can range from 300-1000 dollars and still have different capabilities. Intel owns its own silicon fabs, and does all of its engineering in house.
 
Now consider a company like Beyerdynamic, that has about 300 employees. There is no way that a company with limited resources like that, which probably has to source components from various manufacturers, has as rigorous engineering processes as Intel. Furthermore, if a T1 driver is slightly below average performance compared to the product line there is a good chance that beyer will use it since they can't put the driver into a cheaper headphone and sell it. If Intel finds a lower performing chip it can laser cut some of the silicon and sell the 8 core enterprise chip as a 6 core consumer chip. The point I'm trying to drive home is that it's unrealistic to expect all headphones of a certain model even of a certain batch to sound the same.
 
I know that driver manufacturing and silicon fabrication are different but it highlights issues in tolerances and yields.

 

[analogsurviver]

As posters before me indicated, headphone testing is  demanding and hard to do reliably, particularly so on industrial scales with measuring of each sample. I have experience with QC of phono cartridges which, as imperfect as they still are, are as group of transducers far more accomplished than headphones. Recent work is making some real progress in making this gap less broad, but any decent reasonably priced cartridge will be about at the  level of technical acomplishment best  headphones have achieved to date.
 
And given today's costs, you can believe me that  FR record chart of your sample will run you far in excess of $ 20. Not necessary because doing this measurement costs that much, but because it means for the manufactures to execute constant control over anything from incoming raw materials through every step in the production process to the final QC. And it means more rejects, ultimately driving the cost of the product to the end customer up. Good manufacturers would not allow B or C samples ever to reach the market - after all, nobody is buying headphones, particularly expensive ones, for looks
only. After hearing any fully functional A grade sample you would find hard to accept anything less.
 
On the other hand, it is quite possible to find good set of phones in a model of relative moderate cost but built in greater numbers, allowing for the teething troubles in production to be ironed out long before you purchase your set.
 
If sonic description of the same model reviewed by different users display too much divergence in general terms, this is indication there might be problems in QC, allowing sub par samples to reach the market.
 
Best regards,
 
analogSurviver

 

[donkeywalker]

Thanks for the long answer analog survivor. It seems like there has been a weak push of QC from the consumer end in terms of SQ. the complexity of personal taste, sources and songs used in testing builds up a larger bias to the headsets. Would be happy to see reviews of the same headphone produced in different run or factories. I'm also wondering if anyone ever attempted to create some cheap and consistent measurement solutions just for the physical nature of the headsets.

 

[ev13wt]

I bought a custom MMX 300 600Ohm from Beyer. In the comments field I wrote: "Have a great day at work and if you find the time, pick me some beautiful drivers please. "

It sounded cleaner and better than my 770 250 by a nice margin, and cleaner than my 880 600Ohm. I've sold both and various other gear in preparation for something new.


Make of it what you will, and I bet its purely subjective, but that MMX did everything better.

 

[analogsurviver]

Quote:

Thanks for the long answer analog survivor. It seems like there has been a weak push of QC from the consumer end in terms of SQ. the complexity of personal taste, sources and songs used in testing builds up a larger bias to the headsets. Would be happy to see reviews of the same headphone produced in different run or factories. I'm also wondering if anyone ever attempted to create some cheap and consistent measurement solutions just for the physical nature of the headsets.

No such thing as either cheap or consistent method at the same time will probably be devised ever. First, measuring equipment is expensive, second methods do vary a bit, third headphones are notoriously difficult to measur due t quality and consistency of coupling to the ear or ear canal - slight leak in IEMs will totally kill the bass, etc.
 
Any really decent manufacturer will probably run far more complex/comprehensive measurements that he/she will ever be comfortable publishing. The most acid test used in transducer production I know of is used by Quad for their ESL speaker, from and including model 63. Equidistant from a reference microphone are positioned reference speaker and production sample and both are fed with a square wave. If the speakers are exactly alike, the outputs from the speakers at the microphone should cancel, producing no output at all. That tests for sensitivity, frequency response, distortion, you name it ... And if production sample does pass this test, it means it will form a matched pair with another next production sample that passes test - that means practically each and every speaker should be the same. Anyone who did any amount of measuring will know what a feat this accomplishment is.
 
Yet I never saw headphone manufacturer doing anything similar...

 

[ev13wt]

On Beyers booth I've seen them build custom (Manufaktur) headphones right there. Each headphone does go on some sort of a box and they press a button. I am guessing some sort of sinus tone or sweep. Then they plug it into another port and have a short listen. I guess if you listen to 100+ headphones a day - the person has the experience to sort out bad drivers.

 

[analogsurviver]

Quote:

On Beyers booth I've seen them build custom (Manufaktur) headphones right there. Each headphone does go on some sort of a box and they press a button. I am guessing some sort of sinus tone or sweep. Then they plug it into another port and have a short listen. I guess if you listen to 100+ headphones a day - the person has the experience to sort out bad drivers.

Yes - humans, especially female, are very discerning listeners and can be very good at doing this. Too much in some cases - rejecting far less imperfect drivers than any given spec/QC requirements demand.