An Audiophile Workout: Philips Golden Ears Training

Last year I had the pleasure of reporting on the work done at Philips Research Labs in Belgium. One of the cool things we got to see was a demonstration of Philips software used to train listeners in the organization. At the time, the program was available only to people involved with audio research within the company. I was told there would be a consumer version available sometime in the future. Well, the future is now!


Philips Golden Ears Training is a comprehensive on-line tool that allows you to work your way through 4 levels of listening acuity training: Basic; Bronze; Silver; and Gold. I sat down this morning to see how far I could get through it in a half day before writing a little introduction.

Each major level has 5 stages: Timbre; Details; Spaciousness; Bass; and Loudness. Prior to each test there is a learning stage where you can control and listen to various artifacts. For example, in the Basic stage timbre test, one of the challenges is to hear "coloration" on the test track. Here's a screen grab from that page.


Here, a number of changes to EQ are made to the sample track and as the test proceeds you select "A", "B", and "C" buttons to listen to various samples, one of which is the "colored" sample. Similar tests are executed for: spectral balance; noise detection; distortion; stereo width; bass level; and loudness matching. I had little trouble and it took me about a half hour to get through the Basic level.

I entered the Bronze level next, and it was only slightly harder. In the Timbre section of the Bronze level rather than testing for overall EQ tilt (Basic level), it tested for your ability to hear "Boominess"—a hump in the bass centered around 200Hz.


The Bronze level had tests in: boominess; noise band identification; high-frequency extension; left/right balance; bass extension; and loudness change. This level took me about 45 minutes.

Rowan Williams told me I would probably breeze through the first two levels, but things would start to get tough in the Silver level. THe first two tests in the Timbre category were pretty simple: spectral balance and treble amplitude. The third test was Details and asked me to identify MP3 artifacts. I've often commented that 320kbs MP3s are very good and it's hard to tell the difference between it and 44.1/16 files. Well, let's just say I got totally stuck at this point in the test. I had been using the Ayre Acoustic QB-9 DAC, AURALiC TAURUS MKII amp with my Focal Spirit Professional, so I got out my HD 800, Audeze LCD-X, and NAD VISO HP50.


Maybe I'd just been at it too long this morning, but I gave up. I'll have to give it another go tomorrow...but I'm gonna do it. You know why? Because it's never too late to improve ones listening skills!

I encourage you to do the same. Go to the Philips Golden Ear Training site to brush up your listening skills. It really is easy and fun...until it gets tough and frustrating. And that's the point where you really start to improve your listening skills. Enjoy!!!

Argyris's picture

I suspect large quantities of crow and humble pie will be consumed by confident audiophiles who just know they can hear minute differences in source material and equipment, only to discover after working through this tool that they can't (at least without a lot of training). cheeky

I'll be sure to see how far I can get myself. I'm pretty good at spotting coloration, since I've had plenty of practice listening to pink noise and using a parametric EQ to notch out problem areas.

Jazz Casual's picture

Then it's a good thing that I'm not an audiophile. Doing this test and listening to pink noise is the antithesis of what I enjoy about this hobby. I'd rather listen to music thanks.

Willakan's picture

Well, that's a nice little false dichotomy. One might conceivably listen to non-musical signals to help with equalising a system to sound nicer playing music, no? Indeed, listening to said signals would probably help you equalise it faster so you can spend more time listening to said music. Of course, you could forgo equalisation altogether, but extrapolating from the implications of that position why bother with hi-fi at all :D?

Jazz Casual's picture

Well that's presumptuous and you certainly have extrapolated wildly. Would you understand my post better if I were to add that I have no inclination to undertake such a test, listen to pink noise or perform sine sweeps because I prefer to listen to music and evaluate hi-fi gear in my own way, which I find much more enjoyable? I don't think that is incompatible with having an appreciation of hi-fi. You accuse me of setting up a false dichotomy and then go on to create your own. It would seem that I take a less rigid approach to this hobby than you, or am I being presumptuous? ;) 

Willakan's picture

No, not at all. I am suggesting your approach is inefficient at best. Your first post implied some misplaced sense of superiority due to this. You are welcome to feel that all approaches to hi-fi are equally great, but insofar as hi-fi can be defined I tend to disagree.

Jazz Casual's picture

I thought not. "Ineffiicient at best" - who's sounding superior here? Sorry, but I find your use of "inefficient" laughable in this context. Perhaps I should engage your services to help me make some efficiency gains in this area. Would you start with a time and motion study? ;) Presumably you believe there is a better approach (most likely your own) that is somehow more valid than mine, even though you don't know what mine involves. So I'll dismiss your assertion as arrogant and baseless "at best".

Let's not lose sight of the fact that we are discussing listening, evaluating and hopefully enjoying hi fi here - not the design and manufacture of it, which is very different.

I think it's safe to assume that most of us who visit this site know what high fidelity means. And it is a worthy ideal that many audio manufacturers aspire to attain in a variety of ways, which consequently leads to different results. Their approaches are not uniform, so it seems entirely appropriate to me that ours (the listeners and consumers) wouldn't be either. I'm trying desperately not to use the subjective word here. Oops! I've gone and done it.

It would appear that you and I come to this hobby from very different perspectives, which is fine. You're probably an objectivist who takes a pseudo-scientific approach and I'm not. ;) Happy to agree to disagree or continue arguing with you interminably.  

Case's picture

"Doing this test and listening to pink noise"

Except the test uses clips with actual music.

Jazz Casual's picture

And you've missed the point. :) 

Dano91's picture

Silver done, going to try gold!

ODAC+O2+HD 600, 23y old

EDIT: all done from gold except that frequency bands in timbre part, that one.. no way.. :/

EDIT2: gold finally completed, huray!

KyleHuff's picture

I did everything except the Golden Ears Timbre challenge. This is not something I'm used to doing, so to be able to reliably pick the right frequency bands, I'd have to train for far too long time on that short insipid clip. It doesn't seem likely to pay off.

The MP3 codec one was tough. I ended up focusing on the high-hat. Adding some EQ helped. The differences are there if you know what to concentrate on.

I had many failures to load sound clips. The errors are silently eaten and the UI continues as if nothing is wrong. I had to reload the page to get it working again.

thisiscc's picture

I was unwilling to put in the time to actually learn the frequency differences by heart so I cheated by having two browsers open at the same time to reference the modified file...still took me ages. This test took me possibly longer than all sections combined. I was also doing it all in one sitting latenight.

Really important to let your ears rest for a few seconds if you're having difficulty too.

MP3 codec one was a breeze in comparison. That being said I'm impressed with the quality of 128kbps, although I don't think it was 'that' hard to distinguish it from reference. Glad they didn't put 320kbps vs. FLAC because I would have failed that with flying colors.

Agree with people who say equipment doesn't matter much. Using Xonar STX --> HD800.

thelostMIDrange's picture

I had to focus on the vocals to get through it. low bitrates make them more one dimensional and congested. Otherwise a non eventful spending of an hour of my life I'd like back,

Hjelmevold's picture

That was a fun excercise, and overall it was a very well-designed test.

I had some slight trouble with the bitrate tests (listening for smearing of transients and irregularities in the reverberation sound field), but noticing changes to anything above 12kHz was my biggest challenge. And I haven't even turned 40 yet! frown

When I saw the intro to the final quiz, I thought: "Yes! Finally I get to listen to a before/after sample, and then pick what attribute that had changed." But no, it was the other way around, which I frankly found a bit disappointing, and the changes were much more obvious there than on the other gold challenges. So it was an anticlimactic end.

I felt like the whole test was a little too robotic, being told what attributes to listen for, and then pick your answer from a set of just a few (sometimes just two) multiple choice answers. But luckily there were a few positive exceptions, such as the ones where you had to identify a boost or cut somewhere within the frequency spectrum.

Here's how I would have designed the questions:

  1. On Easy levels: Listen to three samples, where one is different. Follow-up question: Which sample was different?
  2. On Difficult levels: Listen to two samples, with a follow-up question: Did you hear a difference? (with a chance for no change at all)
  3. What attribute was different?
  4. How was this attribute different, or where was it different? (Relax, I'm not going to ask "how much")
  5. Higher difficulties have less amount of change
  6. On the highest difficulties there's a probability that you can't toggle what sample to listen to - you just hear before/after (this was done on some of the loudness tests, and I loved it)

Finally, as an audio professional, I wish there were questions for the following attributes:

  • Amount of dynamic compression (both for properly and improperly applied compression). I find this omission symptomatic of the current loudness wars angry
  • Shape of dynamic compression (compressor attack and release)
  • Phase irregularities (flipped phase, comb filters...)
  • Congestion in frequency (elements occupying similar frequencies)
  • Congestion in space (elements occupying similar positions in the sound field)
  • Loudness changes on single tracks in a mix
  • What about pitch? Wow/flutter, singing/playing in tune...
bogdanb's picture

Silver ears, in 2-2and 1/2 hours.

I had extreme trouble understanding Reverberation. I don't kow why?! And actually I have understanded it better in testing than in training. For me the sound was mor rich and actually I preferred the reverb.

I was using optical audio from a mac, yamaha reciever rx v567 and Nad Viso HP 50. I tried ca couple of times with apple ie, seemed in a way more clear the reverb, ut I choosed to try all with the viso HP50.  Is it possible than the NAD hp50 where responsible for that?

I'm going to try golden ears now. 

AsSiMiLaTeD's picture

I flew through the tests until I hit that damn test in the Golden Ears section with all 18 EQ boosts and cuts, got stuck there for about half an hour trying to make my way through it.  Then after that, flew through the rest of the tests.  My ears must be sensitive to compression, because I had no issues with the mp3 tests.

I agree that the final quiz was a bit dissapointing, I too was hoping for "Listen to these two samples and figure out which of the attributes has changed".  By the time I got to that final test I was so good at picking out the diferences that I found it to be one of the easier parts of the whole ordeal.

I wish it would print out a total time or something on that last page, should have timed myself but think I spent around 1.5 hours on it.  I took the test with my B&W P7 headphones plugged directly into the headphone jack on the iMac, didn't even need the DAC for this.

Jim Tavegia's picture

Especially the level tests.  I knew it would be.  Often the connection from my ears to my brain is not DSL. 

Azteca X's picture

Well good news, my friend: you can still enjoy music all you want.  As an audio professional I will take a crack at this stuff to help me pick out little things better, and I want people with very good hearing to help voice headphones, speakers, etc.  But ultimately, you can just listen to what you like in the way you like.  Listen on!

devhen's picture

I breezed right through everything until I got to the frequency bands in the golden ears section. That part is tough and I quickly got frustrated and bored with it but I'm eager to try again a little later and see if I can improve.

The tests up to that point were a lot easier than I expected them to be and I'm starting to think maybe I haven't ruined my ears yet after all. Of course, my performance could have something to do with the fact that I'm currently demo'ing the LCD-X. laugh


HRT Music Streamer II+ DAC > Lake People G109 amp > Audeze LCD-X

The Headphone Viking's picture

Well I'm curious after seeing all the crazy rigs some people have used for this, how far i can get with just a MacBook and a myST Nail1 in ear... Gonna switch to HD800 if the going gets tough, but so far no problem in silver, lets see about gold :P

As a reviewer I thought this would be a good test to see how capable i am :)

c61746961's picture

No need for TOTL stuff, I too breezed through almost everything with a $30 TASCAM TH-02 (with its problematic midrange), the test feels a bit watered down as to not destroy audiophile's egos, you just have to focus a bit on some things.

The Headphone Viking's picture

I've blasted through everything no problem, except golden timbre part two, the freq band thing... Damn its frustrating...

arnaud's picture

Hi Tyll,

Thanks for the entertaining read and link, I had (am having actually) fun taking the test. I made only 1 error until the MP3 test in silver section. Just like you, got stuck. For me, 112kbps was so easy to pick, 128 much much harder. Surprising as I typically can pick up 128 from lossless with my own cds.

Eventually got through that MP3 hurdle (took me a good 20 min though) and the rest was easy until the 8 bands cut/boost test where I need more training to memorize which band corresponds to a particular change. That indeed feels like a training at this moment...

Very nice of Philips to open the test to all. And, it might be a good way to populate a database and see how much discerning the most "demanding" part of your market (us audiophiles lunatics ;) is ;).



miceblue's picture

...or were these tests ridiculously easy? The only one I really struggled with was with the timbre frequency response one on the way to gold level.

Setup for all tests but the last timbre frequency response: MacBook Pro Retina Display + V-MODA Crossfade M-100

Nothing fancy at all.


For timbre, I used MacBook Pro Retina Display + ODAC + O2 + Alpha Dog.

c61746961's picture

I don't believe these are as challenging as the tests used to train their personnel. This seems a lot more consumer oriented, and... you would not want to discourage your consumers, would you?

clohmann's picture

I can omly use my laptop's soundcard for the audio online and it's resolving power sucks for the MP3 section. Everything else I can tell the difference! If I could engage my outboard Bryston DAC, no problem. Boo!

clohmann's picture

I can hear the differences as hardware burns in, including, esp. headphone cables and innerconnects are concerned.

MrOrtler's picture

All done! Wasn't too bad, except for the mp3 artifacts and reverberation sections in silver and the frequency bands section in gold. You shoulden't need "god tier" headphones, heck not even "mid-fi" headphones as I was able to complete the test using only a pair of UE 600's powered directly by my macbook pro.

Bennyboy's picture

I did it while dropping industrial quality acid and discovered I can hear God's stereo. 

The idiot listens to Phil Collins.  I knew it.

igormsk-yandex's picture

The mp3 test is highly dependent on headphones quality - I had a really hard time hearing a difference with 128kBit with my Shure SE215 (I passed it but after lots of trial and error - it was more luck than actually hearing the difference) but with Sony XBA-3 plugged in the same laptop I passed the test the next day easily on the first attempt.

One year back I did a bling A|B test between 320kBit mp3 and flac – from 20 attempts I was right in 16. I used HD600 plugged in EMU 0404usb and a the first 20-30 sec of 2 or 3 part of the 9.Symphonie Beethoven’s  for testing.


With the bass test in gold I clearly heard a change with 60Hz rolloff with SE215 but 50Hz was impossible to pass for me with that headphone. With XBA-3 it wasn’t easy either but I did it with only 1 mistake. Now I know exactly how to test headphones on bass extension J. It’s a pity there is no 40 or even 30Hz test.


Now it’s time to finally try the frequency test in gold section.

Bennyboy's picture

"One year back I did a bling A|B test between 320kBit mp3 and flac....."

I guess you really are a Goldenears.

timmyw's picture

I had no trouble with most of the tests, with one exception. The noise tests at -36db and lower. Even at -32db I had a hard time. In fact I just guessed the last one and got it right through sheer fortune. 

I didn't try Gold yet I just mucked around with it last night. there were some tests that were so easy I just clicked the thing almost right away. The distortion tests in particular I found quite easy.

Which is why I am a bit confused as to why the noise ones stopped me in my tracks. Maybe it's time to see the audiologist.

AstralStorm's picture

No, plain noise is just that hard - and you need very quiet surroundings and amplifiers for such a test. Typical office settings have 40 dB noise, quiet ones 30 dB.

That and music will mask some of it too.

Tonmeister's picture

It is nice to see other audio companies promoting listener training to consumers. We did something similar a few years ago releasing a version of How to listen here

ckilner's picture

I'm new to critical listening and enjoyed this link. I didn't expect to do well with $50 UE350 in-ears through an iPad, but cruised through 3 levels - haven't tried Gold yet. I found the timber and MP3 tests the most challenging, but would love to see how much easier these would be with quality headphones. I'm over 50 and know that my hearing beyond 12KHz is shot, so I better get around to those better headphones soon! 

starstarman's picture

I finished gold with KEF M500 on V800 + V200 combo. Some of the challenges are definitely very gear dependent, likely the MP3 artifacts test. So I revisit that section to redo the challenges with different cans on the same DAC/amp combo. And here is my feeling in difficulties on them :

Beyerdynamics DT1350 - Incredibly hard
KEF M500 - Hard
PSB M4U 2 - Very hard
Sennheiser HD650 - Hard
Shure SRH-1840 - Very easy
Beyerdynamics T1 - Too easy to the point I flew through it
Sennheiser HD800 - Very easy
Fostex TH900 - Easy
Denon AH-D7100 - Easy
Hifiman HE-6 - A little bit hard (probably due to inefficient amping)
Sony XBA-H3 - A little bit hard
Denon AH-C400 - Very hard
KEF M200 - Hard
Etymotic Research ER4PT - Very easy
Ocharakuk Flat-4 Sui - Hard
Ultimate Ears UE350 - Very hard
Denon AH-D400 - Incredibly hard to the point almost impossible
Shure SE215 - Very hard
Denon AH-C100 - Very hard
Klipsch X10 - Incredibly hard to the point almost impossible
Klipsch reference S4 - Incredibly hard to the point almost impossible
Denon AH-C710 - Incredibly hard to the point almost impossible
Denon AH-D5000 - Easy
Denon AH-D2000 - Easy

I don't know what caused that but flat FR response doesn't seem to play the main role there.

And thanks for linking. Had fun doing them.

Tyll Hertsens's picture

...and an interesting list.  Is there something you would point to as the determining factor that made it easy or hard? Or are there too many things to generalize?

starstarman's picture

Thanks for the response. I have just re tried some of the "controversial" results and I'll summarize them. Please take my personal experience with a grain of salt.

What I listen when I'm trying to find the lower conversion artifacts:
1. Low impact (stated as lack of precision on impulsive sounds in the training description)
2. Out focused vocal
3. Trimmed high freq

So with the gears:
1. Cans with overwhelming but fluid low freq res tends to make picking up the "impulsive trait" very hard.
2. Closed/narrow sounding (not closed cup actually) will defeat the "out focus trait".
3. Weak or unnatural (peak/dip) treble response cans will make picking up the "high freq trait" very hard.
4. Cans with noisy background will completely be a disaster even for trying to differentiate the 96k mp3.

a. Denon D7100 : It has very unnatural FR and bad measurement and while I thought it will be difficult to test with this can, it actually turns out quite easy. The low impact is significantly severed on it when bit rate degrades. Also, the elevated treble around 10k helps a lot to identify the "high freq trait". The stage focus is also easily picked up.

b. Denon C710 : This is the disastrous one due to the noisy background I couldn't even 100% pick out the 96k mp3. The original file sounds low res enough. It is terribly hard to do the challenge on them.

c. Beyerdynamics DT1350 : I though it would be easy on them but no. At 128k the difference is very hard to be picked out on them. I guess it's caused by a combination of unnatural treble response and closed staging.

d. Beyerdynamics T1 : It is really easy on them. The easiest of all. Everything is crystal-clearly different when conversion rate changes : impact on low is gone, focus is out and the high freq percussions timbre changes.

e. PSB M4U 2 : As I redid the challenge with them, I had a really hard time trying to get past the consecutive 128k test. They were considered one of the best sounding mid-fi for me but what they sounded on the test file is like some fluid bass, hard flat stage and recessed treble. Strange enough.

AstralStorm's picture

Guess what, T1 is not the best there either. Try something like Hifiman HE-6, esp. equalized. (Stax would be good too, if not for the uncorrectible bass cut making for problems with bass traits.)

T1 and HD800 will make the test artificially easier because they have major highs boosts. This is especially true in MP3 stage, as that will throw psychoacoustic models for a loop. You will find that stage becoming harder once the headphones are equalized.

Many in-ears have trouble with high frequency traits as they have major roll-offs there.
Generally any headphone when equalized to a flat line, specifically perceptually flat (not to some "roomy" target like Harman now claims is most preferred) will tend to be more revealing of such issues.
Perhaps also Hifiman's RE272, RE0 and RE-ZERO as well as VSonic GR07 and UE Reference Monitor can fight.

320 kbps MP3 test would be quite unfair. That is audible almost solely on specific samples and often you need intact high frequency hearing to boot. Granpappys of headphone hobby excluded, sorry Tyll. Most rock concert patrons are also excluded. Similarly many musicians. That's why musician earplugs have been invented.

Veydar's picture

It's easy for me to hear the difference between 112kb and reference but I completely fail at 128kbs using a Xonar Essence STX and HD 600's. I guess the positive aspect here is that I will never have to spend an awful amount of money on audio equipment since it would be wasted on my ears. I never trained my ears for this, so I guess it's not much of a surprise. Then again, I usually don't fail on such a basic level seeing that it's just Silver Ears so it's sort of a hard pill to swallow ; ) 


Edit: Made it through that part. Took me quite a while to figure out during which part I can actually hear a difference and subsequently whether it's vocals or instruments but once I had that down I could reliably tell the difference. 

egosumlux's picture

I used my trusty old mc mini 2006 with three different headphones. A Superlux HD 300, my Vmoda M80 and my Samson RH600 so no fancy stuff such as external DACs and TOTL 400 dollar headphones where neccesary