Evidence of Headphone Break-In?

The nice folks over at AKG decided to send me a bunch of Quincy Jones headphones, including three Q701s! That's good news for both you and me. For me, because it will allow me to do some experiments I've been wanting to do for a long time. The AKG K701, now reborn as the Quincy Jones Q701, is notorious for needing hundreds of hours of break-in before they sound right. I want to see if that can be measured, and in this article I'll show you the first sets of measurements.

It's good news for you, because when all is said and done, I'll give away two pairs of the Q701 headphones to some lucky reader.

So, let's talk about break-in, shall we...

What is Break-In All About?
Separating voodoo from science in the audio world isn't easy. The human hearing system is exquisite, but it also has the annoying habit of learning and changing, so what we hear one day may not be what we hear in the same circumstances the next day. The placebo effect is strong in listening experiences. Was there really a change when you swapped cables? Or were you fooling yourself? It's not easy to tell.

One area where voodoo and science struggle mightily is over the idea that audio products in general, and headphones in this particular case, "break-in" and change over time. This idea seems logical enough for headphones: the new driver in the headphone needs to wiggle around some and exercise its flexures to settle into its norm. Problem is, if you're trying to hear it you only get one shot; once something has hours on it, you can't take them back off to compare. Headphone enthusiasts get one try to listen for changes during break-in of brand new cans, and over-attentive listening is one very likely path to fooling yourself.

I rarely hear break-in. I reckon the diaphragm does change a bit, but by the time it's got a couple of hours on it, it's done settling in. There is one exception in my mind, however, the AKG K701 is notorious for needing long break-in, and I've heard it a number of times. It's got me convinced that there is such a thing as a break-in period for headphones, and some need it more than the others.

I decided to see if it could be measured.

The Experiment
I took a brand spanking new pair of AKG Quincy Jones Q701 headphones; put them on the dummy head; and measured the frequency response over time, playing pink noise in between measurements. I measured the cans fresh out of the box; immediately after the first test (so it had about 5 minutes on them); then after 25; then 1 hour; 2 hours; 5 hours; 10 hours; 20 hours; 40 hours; 65 hours; and finally after 90 hours of break-in.

Once all the data was gathered, I really couldn't tell what the differences where by eye, so I plotted the data as differences. I used the 90 hour data as the reference, and plotted how the data in each set was different than the 90 hour data. My assumption was that the first measurement out of the box would be most problematic, and that the data should settle in the direction of the longest burn-in time.

Any good mathematician will tell you that this method is a recipe for making the data look like it's settling toward the reference set. Just because the line in these graphs is getting less wiggly over time, doesn't mean that the frequency response is getting smoother. It just means that the frequency response is changing over time and moving in the direction of the 90 hour data. That's fine, because all I'm looking for here is a clear trend where the data changes smoothly from the start to the end reference, which might indicate a change in the sound over time possibly due to break-in. Whether it is break-in or not is another story. I just want to see if I could see trend in the data.

Let's take a look shall we?


Gatepc's picture

Great article I love that you really got into the details. Can you confirm if these are in fact the same drivers inside the K701/702s?

Oh what would be great is if you managed to test some high end cables in as much detail as this. You know the ones that are sparkled with fairy dust and dipped in the liquified remains of extinct animals? I think that would be very interesting.

Thanks for the post.

dalethorn's picture

Those spikes in the chart are most likely Quincy's telepathic reaction ("ouch") to all of that pink noise. For break-in I generally take the Pasadena Symphony's binaural recording of ASZ (first segment 1m 56s), set it to loop, turn up the volume, and let it play for a few hours. If nothing else, it'll warm up the voice coils pretty well.

xnor's picture

Three literally (well, one so far) smokin' hot Q701s to experiment on? The guys at AKG don't know who they're dealing with. *dr-evil-maniacal-laugh*

Nah, jokes aside. I really like the article, especially the summary surprised me. Only a fool would jump to conclusions but you pose a couple of very interesting questions instead and animating your readers to think for themselves.
While the changes you measured definitely exist and are supported by the trends we have to keep in mind how small they actually are. Half a dB here, quarter of an ohm there - audibility is questionable or at least very far from those unrealistic claims made in other forums. Still, those measurements are quite an accomplishment and exactly what I had hoped for a year ago in the measurements lab thread.

Hmm, actually all I wanted to post was a simple "first" but looks like I'm getting old and slow so I had to write a few more lines. If you read my bo.. eh amazing post, that only can win, to this point you should know that I'm giving my prize to the second best poster, hah! :-P

Tyll Hertsens's picture
... but that's okay.

The data is stimulating, don't you think?


maverickronin's picture

Did you run the full suite of tests at each point in time, only the ones that you had graphs of, or did you only show graphs of the data that changed much over time?

I ask because it always seemed to make sense to me that the greatest changes would be at the limits of the diaphragm's excursion and that (I think) would be best measured by a low frequency square wave. Am I on the right track or do the THD+N numbers encapsulate that as well?

maverickronin's picture

Now that the full reports are up and I've had a chance to look at them there seem to be some slight fluctuations in the amount of overshoot on the 30hz wave forms but it mostly goes back and forth so it doesn't look like any firm conclusions about break in existing can be made from that data either.

Shike's picture

Unlike the headphones, I much enjoyed seeing you run pink noise through them for the sake of at least trying to get to the core of the burn-in argument.

The problem, unfortunately, is this is only one down. The curse this headphone has left is one you and your inbox will feel, as people start questioning whether their own headphones burn-in. "Do the HD800, HE6, LCD-2" they'll cry. The 701's will be long gone, but you will be stuck with requests locking you up in the measuring chamber for eternity.

Oh, and Gatepc forgot the main ingredient for boutique cables: virgin tears gathered on the top of the Himalayas . . . that's how they achieve "quantum tunneling" you know?

Currawong's picture

It's almost amazing that it has taken so long to do some measurements like this. What would be interesting now is if you found two different pairs with one supposedly having a lot more bass than the other and measured their frequency response.

Fred76's picture

Would be interesting to do the same experiment with amps and/or DACs, by ear my DIY gear has usually taken 5 to 10 hours before the low frequencies started to kick in, many don't believe in capacitor break-in though.

LFF's picture


Was having a conversation about this on Saturday! I was commenting to another head-fi'er that I wished you would do an article on burn-in and see what the measurements said.

I check this the next day and BAM! Wish granted. Awesome job by the way and the depth of analysis is fantastic. Loved reading the article.

I'm off to tell another head-fi'er how I wish you will give me a pair of the Q701's to fill the hole left by my HF-1's. :-p

goldstarsteve's picture


I don't normally rush to comment but if the prospect is a free Q701 I will not only comment but even run the spell checker for you,

Anyway I wanted to suggest you try some controls. For example try measuring several times in a one hour period and plotting it to see what happens. Try taping the headphones in place with sellotape (seriously), try same experiment on different headphones that you don't think need to break in and most importantly - run same set of measurements on headphone you have "broken in" and see if you continue to get these weird changes. Why weird - because there is no scientific reason for break in to occur with headphones so my guess is we are seeing some kind of noise here but I am not sure where from (your equipment, movement, whatever...).

All power to your new site and I appreciate the combination of measurements with your subjective listening. I would like to see an article on how to match headphone curves to speaker curves so that one can know which kind of headphones sound similar sonically to speakers one likes. I don't think this has ever been done before - it is an interesting idea to develop.



Tyll Hertsens's picture
Sorry I forgot this link, the full measurement spreadsheets for the entire test is here. Article updated.

Thanks for all the comments folks!

Tyll Hertsens's picture
The scale is wrong on the THD measurements.

I will fix the page, but will place a copy of the previously published erroneous THD data here.



Total Harmonic Distortion + noise Differences
Let's look at differences in Total Harmonic distortion. These are differences over time in %THD+noise at the 100dB setting.

Fig. 11 shows the difference over time for the left channel THD+noise at 100dB vs. frequency.

Fig. 12 shows the difference over time for the right channel THD+noise at 100dB vs. frequency.

Some expanded views:

Fig. 13 shows the difference over time for the left channel THD+noise at 100dB between 10Hz and 220Hz.

Fig. 14 shows the difference over time for the right channel THD+noise at 100dB between 10Hz and 220Hz.

Notice the features between 25Hz and 35Hz. Total guess here, but at this low frequency we might be seeing the diaphragm material breaking-in and changing its resistance to bending at the excursion extremes.

Now note the feature at around 200Hz on both charts. I have no idea what it is. I can tell you the weird data points at 220Hz is the meter in the AP changing ranges and creating a measurement artifact.

I also measured the difference in impedance

Fig. 15 shows the difference over time for the left channel impedance.

Here you can see the generally moving trend with time, and the exception of the data set at the 20 hour mark (light blue).

Summary Did I show break-in exists? No. There are too many variables still. Was it simply movement? I don't know. If I did it again to another brand new pair would I get the same results? I don't know. If I did it to an already broken in pair would I get the same results? I don't know.

What I do know is that during the course of these measurements some things changed. While the data showed only very small differences, the data was clearly above the noise, and a general trend observable. The data also showed a discontinuity around the 20 hour mark in both the FR and THD data. While, it seems to me, much of the change observed could easily be due to movement, especially in the frequencies above 5kHz, some changes seem more likely due to break-in. In particular, the changes in frequency response around the fundamental resonance of the driver at 80Hz, and in %THD+noise at the same frequency and at around 25Hz.

There were also things I have no idea about: the changes in FR at around 1100Hz; and the THD change at around 200Hz.

The full measurement spreadsheets for the entire test is here.

Alrighty then, time for some fun, let's give away a pair of broken-in Quincy Jones Q701s. Within the next month or two, I will be willing to set the chamber and test system aside again for another four day experiment. Between now and the time I publish the next article, I will be reading the comments here. The comment I find most enjoyable (could be technical ... could just be funny) will win a Q701.

Have fun!

Stirrio's picture

There once was a man from Waukegon
Whose headphones took too long to break in.
So he listened and moaned
Till the sound of his 'phones
Got him up off his chair and butt-shakin'.

dalethorn's picture

So today I got a second set of the $50 (or is it $59?) Beyer DTX-300p 'phones, this time in white. I was very curious about break-in since my red set was fully broken in. I figured that at least in this price range I should hear something, a difference in response, driver matching, i.e. Q/C variations et al. Nothing. I use mostly 320k CBR MP3's with a lot of detail across the freq. and dynamics ranges. Nothing. I couldn't detect a sliver of difference, and that's interesting for the Q/C aspect if nothing else.

Yoga Flame's picture

Thanks for doing this. Great job again, as usual.

When all the measurements are done, might I suggest playing the same song through both the burned-in and pristine headphones, and recording the results. Then use Audio DiffMaker to create an audio file carrying only the differences between the two.

The numbers so far show some differences as burn-in (and possibly other factors) takes effect. As you noted, the numbers are small. But are they too small to be heard? For those of us who aren't great with numbers, a wav file or YouTube audio clip of the actual difference might resonate (heh) better.

Simeon718's picture

I have a pair of headphones for dubstep, now I just need a pair for my classical music. [Paradox? You decide!]

Buyer no 1's picture

Great test!

Guess after 20 hours you needed some fresh air, opened a window, and then changed the air pressure in the room, thereby changing the sealing of the HP..... ;-)

svyr's picture


Nice job with the Q701s... Like the rest of the lurkers from head-fi I was also curious about someone providing detailed measurements of the hp FR, THD+N vs length of the break-in...

Tyll, it would really help if you didn't try to map 10 data sets on one chart that ends up looking like a fruit salad. (higher res graphics might also be useful... I know headphone.com still makes really tiny graphs, but the web has sort of moved on and on 1920x1200 screens it looks tiny and blurry )

Perhaps supplementing the existing plots with plots of (say) the start, 5h, 20h, 65h would be more illustrative. Or maybe start 20h, 65h separately. It'd also be interesting to see a plot of the last 3-4 sets, to figure out whether the headphones settle on an end frequency or not. (although your interpretation seems to suggest it's just moving about, not settling?)

It would be nice to see measurements of the same pair after a few days/weeks(?) of 'cool-off' after the final measurement, since some people say break-in goes away ;) .

I'd also be curious to know whether the pair permanently sits on the measuring set-up (and how tightly it's fixed to it :D? ) or it might move off a bit if you put it on differently each time (or slide off if it's not quite attached firmly), and what the resolution and accuracy of the measuring gear/(perhaps)number of runs per each measurement to average are? (since on some plots we're talking 0.1db differences).

This is a bad example, but while debugging my old SRM-1/mk2 I at some point measured L-/L+ voltages on 1khz tones (w a fluke RMS 187 multimeter. Not perfect but ok) and there was random but persistent(?) drift of about 1-2v for around 280v measured L-L+ signal voltages. (although in retrospect with the quoted accuracy for AC of ± (0.4%+40) that sounds about right :D )

svyr's picture

oh, the other thing is, how much of the difference measured is attributable to variation in the output of whatever is driving the Q701? (amp warm up, thermal drift, mains voltage fluctuation, etc) (perhaps the same question for the measuring equipment)

>What would be interesting now is if you found two different pairs with one supposedly having a lot more bass than the other and measured their frequency response.

Why? Wouldn't he attribute that more to unit variation of the Q701 drivers, etc Or do you mean the owner thinking his pair is burned in and as some head-fiers say 'omg-the-bass-liek-rly-bloomed' (cough-placebo) :D ?

Finally, I don't like the title of the page for the article: "Evidence of hp burn-in"...It's more of a set of measurements and a discussion of... As the summary suggests: 'Did I show break-in exists? No. '

Tyll Hertsens's picture
"Finally, I don't like the title of the page for the article: "Evidence of hp burn-in"...It's more of a set of measurements and a discussion of"

So I added a question mark.

svyr's picture

:) lol... Was it there before and I haven't noticed, or do you mean you just did?

PS edited the first two posts a bit, clarifying some things. Please re-read.

PPS questioning in this case seems to be a lot easier than measuring ;) ...

Although, the magnitude measured makes me somewhat pleased re: what people a lot of time claim with burn in for their HP (the 'omg the bass is so sick now' argument)... Still, maybe it could be different for other HP ;) :D

Tyll Hertsens's picture
On your comment:

"I'd also be curious to know whether the pair permanently sits on the measuring set-up (and how tightly it's fixed to it :D? ) or it might move off a bit if you put it on differently each time (or slide off if it's not quite attached firmly), and what the resolution and accuracy of the measuring gear/(perhaps)number of runs per each measurement to average are? (since on some plots we're talking 0.1db differences)."

The headphones just sit on the head like normal and are not moved during the whole test. The door to the chamber is closed during the entire test over four days. The APs measurement resolution and accuracy is high enough to show the data trends, but if you look closely on some graphs you can see the resolution limit. When each data point is acquired, it takes a bunch of measurements and waits for the measurement to settle. There are a variety of settling algorithms available, the one I use looks for measurements settling in an asymptotic pattern and once it reaches a certain settled value, it takes three readings and averages them.

svyr's picture

does it account for pads wear-in over the longer time in the test :P?

svyr's picture

PPS to that on the subject of measured magnitudes ;) . On the AUDIBLE burn-in claim 'and blooming post burn-in bass'. I find it very hard (impossible) to even tell less than 0.5dB volume difference side by side on the ABX test site when I have the opportunity to compare to samples of both.

YMMV with golden ears?

trickztastic's picture

If I don't win the AKG q701's, I'm gonna break-in somebody's face. I'm looking at you Bret Michaels.

3DPRO's picture

great article Tyll!

The detail of your experiment is really great. My brain just exploded after reading this.


hehe. :P

and I downloaded the spreadsheet to read on the train to my class tomorrow. it's gonna be fun!

dvltn's picture

I read this entire article and found no facts to support that your headphone nor living quarters were broken into. There is no mention-of or evidence of robbery, forced entry, nor burglary. This is extremely misleading.

dalethorn's picture

Good thing there wasn't a burn-in. That would be playing with fire.

DrL's picture

... I'm wondering if the phenomenon of break-in, or what people perceive as break-in, may have more to due with one getting used to the sound of a particular headphone than an actual physical change... I mean, I've had similar experiences with many music albums - they sound ok at first, but they keep getting better and better every time I play them...

An interesting experiment would be to give two groups of people a K701 (or Q701 or another headphone) and ask them to judge how much the headphone's sound change over time. One group would be given brand-new headphones and the other group would be given headphones with several hundreds of hours on them. Obviously, the test persons would be unaware of which group they were in. That should make for an interesting comparison. Maybe AKG would be willing to sponsor even more headphones for the noble cause :-)

svyr's picture

Placebo, getting used to the sound of your HP, and expectations are well explored in the multitude of related threads on head-fi and I've mocked them above ;) ... (search for the coathanger wire cable folklore :D, it's fairly amusing)

I'd be curious to see Tyll do a similar test on a pair of Orthos, BA IEMs, Beyer's Tesla line HP and electrostatics (some people think Tesla Drivers take a very long time to burn in, others say they don't at all. Many people say BAs and ES don't burn in. No idea about Orthos)

dalethorn's picture

There are so many things missing in the simple 'placebo' explanation that I could write a book, and I know less than anyone here. One, there were changes with the AKG. Two, we assume the AKG's were not broken-in by the mfr. Three, we assume that the AKG is representative of all headphones. Four, we assume the freq. response measurement is the end-all of sonic difference. Five, we assume that what we can't hear consciously we also don't hear subliminally.

And on it goes. This is a great start, and as I noted above I broke in a new headphone and heard no difference at all (more evidence against break-in), so as Bruce Willis would say, there ya go.

SAS's picture

You didn't test two headphones, but you did test two drivers. When you compare them, the results only really look similar in the low frequencies. That might well be evidence for break-in at around 80 Hz as you suggest. When looking at the higher frequencies, if there is break-in, it is unpredictable.

That both drivers show the odd behavior at 20 hours and 40 hours is just as likely to mean that they are accurately reporting a change in your measurement system as that the measurement system is reporting a change in the drivers.

Rather than the tangled mess of lines, it would be informative to take specific frequencies where you think that things are occurring, and plot amplitude on the Y-axis against break-in time on the X-axis . That will make it easy to see whether the diaphragm is settling into its broken-in state or changing randomly.

If you calculate a total amplitude difference across all frequencies between the measurement at any given time and the 90 hour measurement, does that difference progressively decrease?

Jamey Warren's picture

Who listens to Pink Noise?

Just kidding but I'm curious if different types of music cause different amounts of break in.

Would a hip-hop lovers headphones break in faster?

Would a ukulele listener's headphones ever break in?

Please ship Q701's to Livingston or meet me downtown Bozeman for a drop off. Do you still eat taco's?

Newb89's picture

What does Star Trek and toilet paper have in common?

....Both go to Uranus to wipe out the Klingons!

Hope you enjoyed the joke as much as I enjoyed the article.

PMM's picture

Hey, Tyll; I'm wondering about certain things, if you wouldn't mind sharing:

Is there a reason why you didn't conduct cumulative spectral decay measurements, or impulse response measurements of different frequencies and decibel levels? Do you have equipment to create waterfall charts, and don't you think it's worth doing in this case?

Are you aware of any 'standard test sounds', similar to the 'standard test images' (e.g. http://en.wikipedia.org/wiki/Lenna ) used in visual editing fields? Would you ever consider using them?

The variance in impedance that you found in figure 15 kinda fascinates me. Do you think those fluctuations would be enough to affect transient responses?

In the end, it seems to me that 'break in', if it exists in some material capacity, is going to have more of an effect on transients perceived as timbre and decay than it will on a stable tone at a stable volume.

No matter what, thank you sincerely for conducting this long-form experiment, and putting in the time to share the results with us.

Also, completely unrelated, but did you ever hear the 100% compostable Sun Chips bags that were released for a short while in America last year? 92 decibels at close range! http://www.youtube.com/watch?v=vWzQJfjAyP0 (I am saving an unopened bag for posterity.)

Tyll Hertsens's picture

... I don't have the gear to do cumulative spectral decay plots. I'm afraid they might not tell us much as cabinet and residual driver noise is likely less of a problem in headphones as they are so small physically ... but I'd love to give it a try. Someday.

@ standard tests - There are various standard test sounds in audio testing. Maximum Length Sequence is a special noise-like sound from which impulse/phase/frequency response can be derived; there numerous intermodulation distortion standardized test signals (SMPTE RP120-1983, DIN 45403); my AP comes with a variety of multi-tone tests to do rapid frequency response measurements for QA testing in manufacturing environments; and, of course, pink noise, white noise, and the lesser known brown noise (note related to the brown note). None are nearly as sexy as Lenna though. :(

@ transient testing - I'm convinced measurable effect of break-in, if any, is small. I agree that I'll need to make sure there's a significant battery of tests done the next time around. I will be sure to include tests that excersize the headphone under transient conditions (square wave, impulse response).

@ your thanks -- Glad to put in the time doing it ... it's my job. :)

@ sunchips -- cool :) --- friggen marketing people --- always making noise.


PMM's picture

Exciting! I really look forward to the next round.

Re: spectral decay plots: I agree that the headphone cup/cabinets wouldn't contribute anything significant to this topic, but I'd still be curious to see if the decay changes at all as the impedance changes; and I don't think we should immediately rule out residual driver noise, since that's still within the purview of 'breaking in a driver'.

Shouldn't matter for now, though. Depending on what results we see from the transient response tests, we'll get a hint as to whether a waterfall chart is worth the trouble.

Re: standard test sounds: Ahh, of course the QA gear would come with prepackaged noise and tones that it's calibrated to detect. I was thinking of test clips of actual things, like bells or voices, which could be compared by spectrogram or something -- but yes, obviously, recording stuff like that in a standardized fashion would be an ordeal all by itself... http://thesoundmyheadmakes.blogspot.com/2010/12/handheld-recorder-shooto...

Then again, handheld recorders are a completely different animal compared to, say, an Earthworks M Series. http://www.earthworksaudio.com/our-microphones/m-series/ Heh. Just another thing to put on the wish list.

(Bonus! While researching the topic of spectrogram comparisons, I came across this cute but amazingly useless video of Lenna turned into a spectrogram and played back: http://videosonar.com/2/video/spectrograph/x8b900.html )

gummydogs's picture

did you ever try listening to either of your three Q701s while doing measurements?

I've spent some time listening to the broken-in pair, but I'm leaving the other two in the box until I get through the tests. I won't be listening to the second one at all before its testing. I'm currently doing some experiments with the burned-in pair to get some control data ... it's been quite interesting and I'll probably post about it next week.

Once I've got a good set of control data, I'll test the second pair. Once the data is massaged and I see what's in it (if anything), then I'll perform some listening tests on the last pair. Hopefully the measurements will give me an idea of what to listen for.

This thing will take a while to work through, but at the end I hope to have the answer to whether there is measurable evidence of break-in.

Subjectively, I think there is break-in effect, but I'm getting more and more convinced that it will be barely measurable, if at all.

edit: Well I guess I'm Tyll Hertsens now :P
Tyll, I think you might have accidentally edited over my comment when you tried to quote it.

Double edit by Tyll: Oh geez! I pushed a wrong button. Sorry about that.
(I'm such an idiot sometimes.)

PMM's picture

Edit number 3. Man, what an adventure this comment became. I realized that, duh, Tyll already recorded impulse responses in the PDF file even though he didn't show them in the writeup here. So hey, let's see what the difference is between minute 1 and hour 90...


Oh my. That's significant! At this point I'm thinking eureka, we've done it. But I'm a perfectionist, so I wonder what it looked like at half the break-in time. There isn't a 45 hour reading, so I go with the 40 hour reading.


"Holy crap," I think. "This is it," I think. I get excited. This looks like clear progressive evidence of a break-in effect, and even correlates with claims that improvements are modest for tens of hours and then suddenly get noticeably stronger around the triple digit mark.

But again, I'm a perfectionist, and I want to see how smooth that progression is. I go back to the start of the PDF, and move from 1min to 5min, then...



Hypothesis is now fully blown to smithereens, confounded beyond comprehension, and I can barely begin to guess why 40 hours of break in looks more like 1 minute than 5 minutes does, and 5 minutes looks more like 90 hours than 40 hours does.

At least I gave it the old college try.

Tyll Hertsens's picture
Thanks for that, it was fun ... pointless in the end, but fun. Thanks.

What's happening (I think) is that the signal strength is changing a bit as each test is run. I'm not entirely sure where it's coming from, but in the initial part of the test the voltage get's regulated. I noticed some of the level changes and rewrote the test so that the voltage gets set once, and then never changes subsequently.

I have been working on getting more data, and slowly tweeking the test so that the next run through should be better controlled.

PMM's picture

So the differences may be coming from the signal generator, and not represent the driver becoming more sensitive to voltage.

Even after a stable voltage from the generator is established, is there a way that I can interpret these readings as decibel equivalents? I'm not sure what "0.02 volts" means in terms of a wearer's actual listening experience, and for that matter that exact voltage probably results in different sound pressures on different headphones anyway.

I'm thinking out loud here and don't want to presumptuously heap a bunch of new work on your plate, but wouldn't it be more informative to pick a reasonable listening volume like 80dB and set the signal generator to whatever achieves that? I don't listen to headphones based on absolute voltage levels, I listen based on "loud enough" -- which probably goes for you and everyone else, too.

Tyll Hertsens's picture
... dB is a power measurement, and Impulse response is a waveform measurement in time, so you really can't show the impulse in dB. It's not how loud the impulse is, but what it looks like as a signal.

The level that I test at is usually 90dB. which is significantly louder than I would recommend listening, but it is a listenable level. I think the problem of the varying waveform height is that when I was setting the 90dB level for the impulse response, the signal generator was using the MLS signal, which basically sounds like pink noise. So I was using noise to set the level, which results in some error of the actual level set. I think I've fixed this in the break-in tests already, and I will fix it in the data sheet program soon.

Bottom line: the only thing you're really worried about in the impulse response test is the shape of the impulse. You really can't compare the amplitude headphone to headphone because, theoretically, they all contain the same amount of energy, but the headphones distribute that energy differently into a different waveshape. So it's not the initial amplitude peak that's important, but the way the headphone distributes the impulse energy into a waveform.

MIG29's picture


Very interesting thread. I have myself experimented with burn-in of headphones.

Inspired by the Synergistic Research cables burn in procedure i have experimented with utilizing af tesla coil for speeding up burn in. And with quite som sucess i might add. My theory is that the coils in the transducers are loaded in by the tesla coils in such a way that the micro structure is re-structure in the process. Anyway they measures amazingly flat after this procedure. Practically no variation in amplitude. Below i have included a picture of the process. I realize that it is not the same cans, but i think it might also work great with the flat coils of the Q701´s


If it is of interest i would be happy to post the measurments.


Tyll Hertsens's picture
Have you played around with EMPs from nuclear fission reactions yet. Please post pix.
PMM's picture

I've wondered for some time how signal quality would be affected by the resonance-tuned single wire transmission lines that Tesla thought up while he was still more sane than not. http://en.wikipedia.org/wiki/Single-wire_transmission_line

It would be funny if something like a cheap unshielded tin wire outperformed ~Single Crystal Silver 9 Layer Shielded & Impregnated With Harpies' Tears~ cables as soon as it was resonance tuned.

PMM's picture

Alright. So in the case of the 1m vs 90h shot, http://i.imgur.com/Z1qFS.gif , we can tell that the basic impulse shape looks the same aside from their vertical scales being off, and the scales were only off due to that 90dB of power being distributed differently from measurement to measurement. Maybe the voltage went higher but the amperage went lower, or something like that. So from this we can pretty much predict that once impulse voltages are uniform, we're not likely to see any evidence of break-in from this test.

One more question for clarification: Since you said it was set at 90dB, that much loudness should be what a user perceives regardless of vertical scale on the voltage axis, yes?

MIG29's picture

Well here goes. What is break in ? Well it is most likely a change in the mechanical behavior of the transducer. For big transducer as used in traditional speakers, I postulate that a large part of the breaking in is due to the material in the surround of the bas/midrange loosening up, and to some degree for the cone to find the breaking up pattern (which to some degree is determined by the degree to which the material of the cone can be considered to be uniform both in mechanical behavior and density. There are of course other effects, but I don’t believe they are significant in relation to what we are looking fore here.

Well how does this translate to the AKG Q701´s. Well they use the concept of Varimotion diaphragm, which is a way of using a single diaphragm in a dual concentric fashion. AKG achieves it by varying the diaphragm's thickness. Its "80µm center zone is virtually pistonic for well-balanced HF response while the 40µm perimeter is highly elastic to accommodate the increased excursion required for accurate midrange and LF reproduction. The workings of this is explained in http://www.ejjordan.co.uk/PDFs/Jordan%20Manual%202011%20Chapter%205.pdf
Please note that this concept is largely dependent on the diaphragm material acting elastically (within its elastic boundary that is), especially in the perimeter of the diaphragm.

Further, the diaphragm is a two component diaphragm. I will bet you that the behavior of the cones concentric modes are controlled either with mass (difference in weight) between the two materials, or by thickness variations.(A bit like the spider wed that is seen on some diaphragms.) That means that the diaphragm is designed to break up in an controlled manner, which is especially important for the outer parts (the most flexible parts of the diaphragm), which controls the bass response. Exactly where the odd behavior occurs at around 20hours of use. So what changes. My theory is that the cone is so cleverly designed that a charnier like mechanism occcurs, this alters the overall breakup pattern from then on forward. his again alters the local cancelation (Lets call it higher order flex) pattern of the outer parts of the diaphragm, which causes the drastic change in amplitude seen at 20-40 hours. Or put in a nother way You can also say that the eigenmodes of the diaphragm is forever altered. From then on it has settled in and the overall break pattern is maintained onwards.

Well at least i think that is what happens. So i think break in is real.

One last comment. The behaviour of the transducer might be somewhat dependent on the temperature, so an important parameter like temperature of voicecoil should be fixed. An indication of that being the case could be the voltage needed to gain the chosen db level, or the impedance. Due to the size of the transducer, and the force needed to move a relatively light diaphragm like this, even small variations may cause detectable changes here. well enough of the babbeling.

With regard to the fission reactor, my DIY kit unfortunately did not pan out. So now i am using af mix of Sweedish neuclear power, coal power and the occasional wind power, to drive my EMP generators (Cans). I do recomend windy days for listening as about 20% windpower in the mix greatly enhances sonic character, and creates a cohearence in the frequency range that is quite unik

PMM's picture

Some things come to mind: I don't hear a lot of people saying that headphones have warm-ups and cool-downs, or that a headphone "sounds better in the desert" or anything, and I presume that each test that Tyll conducted lasted about the same period of time from start to finish, so voice coil temperature seems like a wild goose to me.

But you did bring up the voltage situation -- which Tyll mentioned that he will be addressing in the next round (http://www.innerfidelity.com/content/evidence-headphone-break#comment-48...), and I mentioned that aside from the scale the actual shape of the impulse doesn't seem to change at all (http://www.innerfidelity.com/content/evidence-headphone-break#comment-48...).

I haven't made an animation, but the shapes of the 30Hz and 300Hz square waves don't change much from 1m to 90h either. That's kind-of a bugger.

All the usual "absence of evidence is not evidence of absence" stuff applies, and I'd be glad to see every possible test conducted, but I'm no longer as hopefully curious about transient response as I was a week ago.

(Also, on the topic of EMP generators: Might I recommend attempting a DIY headphone rendition of these fine helium plasma ear tanning brain roasters? http://members.cox.net/jayphil/Plasma%20history.html)

MIG29's picture

Hehe . . one of my friends actually played around with the Magnat corona discharge loudspeaker. Lots of fun, but i never really liked the sound. But it has a cool mad scientist thing about it.

With regards to the temperature, you might be right. I am new to micro speakers, so what applies to large transducers might not be true for small ones :-)

And i think you might have a point with regards to the missng change in shape of the impulse response. I guess you would expect to see some change here if my theory was in play.

Well never mind I think it is fun thinking about these things, and hearing others thoughts (especially when sung and acompanied by music).

Look foreward to see the results form the new test.

Tyll Hertsens's picture
PMM wrote:
Some things come to mind: I don't hear a lot of people saying that headphones have warm-ups and cool-downs, or that a headphone "sounds better in the desert" or anything, and I presume that each test that Tyll conducted lasted about the same period of time from start to finish, so voice coil temperature seems like a wild goose to me.

Looks like we've all got our thinking caps on.

Since this test is going to need to be very sensitive, and since the damned test takes so long I want to rule out as many things as possible before running the next test. In the previous tests, the headphone is burning in with pink noise continuously and the test itself starts immediately after the noise shuts down. I'm thinking that the driver may be heating up after a while and may have some feature due to the long term heating.

A voice coil is very small, so I would think it would achieve it's highest temperature rather quickly and then stabilize. None the less, I want to rule out the problem, so I'm going to let the driver sit for 15 minutes of silence prior to each test for the next go round.

PMM's picture

MIG29: What aspect of the Magnats didn't sound proper to you? I'm genuinely curious to know how people react to such exotic speakers.

On the page I linked, there were two versions of the Hill Plasmatronics speaker measured, a stock version which produced a range that looked little better than a regular speaker, and then a modified version which was almost uncannily flat. It's fully possible that a Magnat could have an unimpressive frequency range when unmodified too. And even if it wasn't the frequency range, the plasma-fast impulse responses could end up sounding too analytical or what-have-you.

Tyll: Fabulous. I know I've said that I'm excited and thankful that you're doing this, but it almost can't be stressed enough. Roughly a decade ago, I was lamenting that nobody really measured headphones at all except for this one dude who worked for HeadRoom (hah!), and he only seemed to be interested in frequency response back then. All this time later, now that "one dude" is on his own, and ready to take these measurements as far as he possibly can. It's got this wonderful full-circle feeling to it.

PMM's picture

Coming back to something that I wondered about in my first comment, and the reason why I brought up the reference image of Lenna: Is there a way for you to record an actual audio sample clip with your test hardware? Not squares and sines and impulses, but some reference-grade snippet of actual recorded audio that you think represents realistic listening.

I'd like to see if any differences come up in a spectrogram or wavelet or something... http://stevehanov.ca/wavelet/ Since this wavelet software is free, we'd "just" need decent recordings of something realistic to compare.

Tyll Hertsens's picture
... of square waves is what you're looking for ... I think.

A square wave essentially contains the fundamental tone, and an infinite series of odd harmonics. A 30Hz square wave is the sum of a 30 Hz sin wave plus sin waves (in proper phase) at, 90Hz, 150Hz, 210Hz, 270Hz, 330Hz, 390Hz, 450Hz ... etc. All the odd harmonics. So the cool thing about square waves is they have components all through the audio spectrum, and it's easy to know what they should look like, and how they're different from the ideal.

So I do think they are "audio that you think represents realistic listening."

PMM's picture

I'm not wholly convinced one way or the other, and just hope that we don't run into a mistaking-the-forest-for-the-individual-trees moment -- but I will gladly trust your experience in the meantime, and hush the heck up now so you can do your thing.

MIG29's picture

PMM: Well the thing is that exotic is the keyword. I heard the darn thing in a workshop placed upon and accompanied by a very old bas/midrange transmission line, that quite frankly left a lot to desire. So i dont think i am in a position to say something relevant on the sound.

However, I can describe the experience. It is like seeing you friend flying a kite in a thunder storm. And when the lightning finally (it took some time to get up and running) crawls down the line, he capturs it inside a small cage. After a while the lightning starts speaking to you in a very high pitched voice! What i am trying to say it is hard not to focus on the spectacle of it all.

Would it be something for me? No. I would not like to carry around the fobia of having the GhostBuster suddely braking in to my late night listening sessions sucking the life out of my high-frequency drives with their proton packs ;-)

And again good luck with the next experiments

kevinmogee's picture

Having just bought a pair of new headphones, I was searching for articles on burn-in/break-in and came across this page. I read the entire article and came away with a LOT of information. Some of it I understood and some of it has me scratching my head in confusion, but the one question I have is.....Were there any winners picked? I read through the comments and didn't see where you announced any winners. Just curious.

Tyll Hertsens's picture
The winner announcement is at the bottom of this page.
kakez's picture

Tyll. You're a great dude man. My lil story; I make hip-hop music, but emphasize in making GOOD music, I study many things, been on a real 80's kick, lots of good music then, especially anything Quincy Jones produced. A month ago I bought the AKG K240's, and compared them to some audio technicas, shure over ears and some sennheisers, cant tell you what models they were.. well because I bought the AKG's! It was the first time I'd really ever talked about AKG, I had some in-ear shures for a long time (6 or 7 years). After having these AKG's my life has changed. These were only $100, and feel more comfortable and sound better than stuff up in the $250.00 range..

I then kept researching (as goes with these kinds of things, music, audio, headphones, etc) I found the Quincy Jones AKG HEADPHONES! I was in awe, and freaked out, also a lil skeptical (check out the beats by dre.. ahem) but.. I started looking up reviews.. You-tubed.. found YOURS and felt that you did such a great unbiased review I at first couldn't tell if you liked them.

After more research I saw your other videos, and also you had mentioned 'burn-in' a new phrase for me.. More research.. and well here I am.

You have schooled-me in many things. I feel in one month my life is changed (FOR THE BETTER) obviously.. haha. My financial aid is coming this next week, and I'm more than happy to throw $400.00 at such a great product, and I cannot wait to have them on my desk, with pink noise through them, burning them in, as much as possible, then sitting back in awe of what greatness will envelope my being, transforming my world into one that I'd imagine Renaissance painters were in whilst painting your favorite piece, still viewed today, and studied by many.. 500 years later..

Okay, rambled a bit. Sorry bout that. Had to inform you, and THANK you for informing me! I'm 23, so still a young'n at least, but glad I haven't lived much longer without these Quincy's. I haven't even heard em, but just know, they will be world changing. I don't care about TONS of bass, or whatever the complaint is, I want the best listening experience, comfort, style, and obviously.. sound. I am curious what amp you'd recommend? Or any other tips and tricks you may bestow upon your faithful (new) padewon of the 'headphone-force' lol. (ps. those green ones, so so dope! the ones I'm going to get)
Have a great night, or day, or whatever the time it is there, or when you read this.

asymcon's picture

I've noticed you've missed to measure the changes in, as 'expert' says, "inaudible" range above 20kHz. As you may know the Q701s frequency range extends to 39.8kHz. I'm sure that all the "break-in" sound material should include those frequencies as well, because they're no-less important as the "audible" ones.

I've just purchased K702 and I'm ready to break-in. As source material I've chosen my pristine recordings of acoustic guitar with range far beyond 20kHz (in fact it's even higher that 48kHz, but it should be fine)

mikeru22's picture

OK, so this is a very interesting article, but there is only one more thing I need for convincing.


This was after 90 hours of continuous playback. I would imagine the test headphones are pretty warm by then, and nice and low mechanical impedance by then. This explains the rise in bass response over time. If you let the headphones rest for a few days and run the test signal through them once more, do they stay where they were after this "break-in", or do they return to their original response?

Ant60's picture

Hello guys, here my experience with K702 Burn-in.
Before I read this interesting article I was searching news about Burn-in for K702 because after purchasing my headphones I have noticed that they was harsh with treble, too strong and tiny bass, no body. I read that was necessary more of 60 hours burn-in period. So I let they plays for two nights drived by a Beresford DAC-headphones amplifier called Caiman (gatorized) and some other hours during the day. After the second night the headphones played with a very natural sound with bass articulate and present and incredible, the harshness of treble was disappeared. All things was at right place. So I have calculated in approximative mode that twenty hours was a break-line time for K702. Now I read this article that matches perfectly with my experience.
Also note that on my desktop station I also have two Concertino (Sonus Faber tiny loudspeakers) with that I could do benchmark the headphones in contemporary because the DAC has two output, fixed for a class D little amplifier driving the Concertino and variable output with volume knob for the output of headphones. When I was listening to Concertino and then immediately I fit the headphones (with zero volume for the Concertino) the scene was very similar, just more bass extension for the K702, as natural it should be, while out of box the headphones was very different. This in my opinion indicates that the Headphones was arriving at their best performances. Now the K702 play very good and I think, with a tube amplifier the things should be even better.
My listening experience in Hi-End rigs counts 30 years of honorable career, and ever I happened to see so much difference in a few hours of burn-in.
At the end, also for me 20 hours is a very important goal for burn-in the k702.

Let it plays two couples of days and enjoy the K702.