How to equalize your headphones: A Tutorial

[leeperry]

Quote:

Originally Posted by Dr. Strangelove /img/forum/go_quote.gif I would wager most of the quality transducers out there (specifically headphones) cannot reproduce sounds outside the audible range with accuracy. What is the point if the data cannot even be reproduced?

all I can say is that all my 24/96 recordings sound more detailed/natural than CDDA, whatever the Bjork DVD's or my DVD-A bootlegs.

and same goes for the "In The Court Of the King Crimson" HDCD, which sounds miles away from the CDDA track or any older copy I had...or these HDCD Beatles bootlegs, damn the reverb trails sound SO natural and not "quantized"(don't shoot!) like on the CDDA tracks.

and same goes for my Cubase mixouts...whether this is a projection of my brain is something I can live w/

anyway, thanks to Piccolo's tutorial I've managed to lower the treble resonances of my 770 to very acceptable levels, this is a far cry from no EQ I can tell you that

I'd love to kill all these resonances, but as long as they don't come in the way in an outrageous way(as they did before)...I'll learn to live w/ them

all headphones have resonances anyway, as I understand it...you can't expect to put drivers in a closed plastic can and not resonate

 

[gregorio]

Quote:

Originally Posted by Dr. Strangelove /img/forum/go_quote.gif I would wager most of the quality transducers out there (specifically headphones) cannot reproduce sounds outside the audible range with accuracy. What is the point if the data cannot even be reproduced?

Sorry that this thread has been hijacked but the mis-information being given is quite stupendous and I couldn't let it pass. These people seem to have gained their knowledge of how digital audio works through pure supposition, with absolutely no reference to the facts. More must be better therefore 24bit is better than 16bit. Higher resolution obviously means better, seems reasonable and logical on the face of it but without reference to how digital audio actually works, it's complete nonsense. Of course 24bit is higher resolution, this isn't in question, the question is how does this resolution manifest itself. Unfortunately for the suppositionists the higher resolution does not manifest itself in terms of audio quality!!

I realise you already know this but it really winds me up when people assert for others, and then argue and insult from a position of pure ignorance.

G

 

[Dr. Strangelove]

You cannot say for sure whether or not you are simply hearing a better mix, mastering, etc. As I said you would have to down-sample the 24/96 and do a blind test... that you rule out everything save the bit depth and sampling rate. Also would have to make sure volume was the same as well.

Btw why not dampen the cups?

 

[leeperry]

Quote:

Originally Posted by Dr. Strangelove /img/forum/go_quote.gif You cannot say for sure whether or not you are simply hearing a better mix, mastering, etc. As I said you would have to down-sample the 24/96 and do a blind test... that you rule out everything save the bit depth and sampling rate. Also would have to make sure volume was the same as well. Btw why not dampen the cups?

well I already encountered the problem many times, when downsampling my 32float/44.1 masters to CDDA....so many tiny bits of details simply disappear(even w/ UV22HR blabla)

convert a CDDA track to 8bit, that's the same thing...but even worse of course

good point, I might try to dampen them again...it did wonders on my 770Pro :


but most of these resonances are happening in the outer ear anyway

after some major nitpicking, this sound quite flat...even on that Nick Cave song, talk about major accidents

 

[geremy]

What is the typical rail-rail voltage of an audio source?

 

[gregorio]

Quote:

Originally Posted by leeperry /img/forum/go_quote.gif well I already encountered the problem many times, when downsampling my 32float/44.1 masters to CDDA....so many tiny bits of details simply disappear(even w/ UV22HR blabla) convert a CDDA track to 8bit, that's the same thing...but even worse of course

32bit float is not really the best way of going about the task, fixed integer is more accurate. However, the noise floor of 32bit is theoretically at -192dB. What sounds do you have at that level which are noticeable? Answer, none. If there were sounds in these least significant bits and you could hear them, when you played back material in the most significant bits you would literally die pretty much instantaneously, although in practice of course no DAC or sound system is going to be able to resolve anywhere near this level of dynamic range anyway. So, there is not and cannot be anything useful in the least significant bits of 32bit. In practice, for the vast majority of commercial recordings there is nothing but noise in at least the last 4 or 5bits in 16bit, let alone 24 or 32bit. So if anyone asks you the difference between 16bit and 24bit audio, the correct answer is 8bits of noise.

Converting to 8bit is (with proper dithering) going to produce a perfect recreation of the analogue input as it would at any bit depth. With 8 bit though, there are so many quantization errors that the noise is quite intrusive. The theoretical noise floor for 8bit is only 48dB.

BTW, UV22 is a very old method of dithering, popular professionally about 15 years ago. There are far better dithering algorithms available these days.

G

 

[Quaddy]

Quote:

Originally Posted by gregorio /img/forum/go_quote.gif Wow, someone who can quote Google, a real expert, for a 5 year old possibly!! 24bit for recording isn't becoming the standard, it has been the standard for more than a decade and I've been using greater than 16bit for professional recordings since 1992. For recording and mixing there are benefits to using 24bit. But for playback there are none!! You can't play it back anyway, you show me a single DAC on the market which can reproduce 24bits of dynamic range or for that matter, show me a single human being who could survive it! G

patrick82?

 

[leeperry]

Quote:

Originally Posted by gregorio /img/forum/go_quote.gif Converting to 8bit is (with proper dithering) going to produce a perfect recreation of the analogue input as it would at any bit depth.

hah

well don't throw rocks at me, but I like to decode my lossy audio in 24bit instead of 16.......to avoid dodgy roundings.
I've been doing it for a long while because I think it sounds better, and I've recently found out that this link also advised the same thing : How accurate are the 24-bit mp3 decoders?

Quote:

If you round it to 16-bits, you add a small amount of extra distortion to this reconstructed signal, getting even further away from what was on the original CD. If you round it to 24-bits, you're still adding distortion, but it's 256 times quieter than that added by rounding to 16-bits.

Quote:

It's not that you could hear the last bit of a 24-bit signal on its own - it's too small a difference, or too quiet a signal for the human ear to detect. However, measuring larger signals with that fine degree of accuracy makes them sound more realistic.

or maybe we're all just daydreaming?

I can check whether all the bits are being used through Ozone4 or Wavelab, and they actually appear to be so....but yeah, prolly noise

 

[gregorio]

Quote:

Originally Posted by leeperry /img/forum/go_quote.gif hah well don't throw rocks at me, but I like to decode my lossy audio in 24bit instead of 16.......to avoid dodgy roundings. "If you round it to 16-bits, you add a small amount of extra distortion to this reconstructed signal, getting even further away from what was on the original CD. If you round it to 24-bits, you're still adding distortion, but it's 256 times quieter than that added by rounding to 16-bits."

"Rounding", you're joking? Dealing with re-quatisation by rounding died out professionally at the beginning of the nineties. Well before we entered the C21st this problem was eliminated with the use of dither. Dither creates no distortion at any bit depth but does introduce un-correlated noise at roughly the least significant bit level. However, the up to date method is to use noise-shaped dither which re-distributes this noise so that it cannot be heard even at the least significant bit level. In practice this increases the dynamic range of 16bit to as much as 120dB. Now you tell me, how many recordings do you know of that have 120dB dynamic range?

"However, measuring larger signals with that fine degree of accuracy makes them sound more realistic." - NO IT DOES NOT!!!! If you increase bit depth by one bit, you double the number of potential values available for encoding the signal. This extra resolution results in half the number of quantisation errors. Quantisation errors are represented as un-correlated noise, so halve the errors and you halve the digital noise floor (ie. The noise floor is extended by 6dB per bit). CD is 16bit therefore 16 x 6 = 96 (CD has a 96dB dynamic range) and 24bit x 6 = 144dB dynamic range. Resolution in bit depth ONLY relates to the dynamic range there is nothing else, no magical extra qualities or sublties, just the dynamic range. The ONLY difference between 16bit and 24bit is 48dB of dynamic range.

G

 

[PiccoloNamek]

What the hell guys? Show some decorum. Couldn't you have taken this to PM?

Is there a mod out there who can remove the posts on digital audio (etc) by gregorio and vvanrji?

 

[Ashirgo]

Well, but why? Gregorio's posts are quite informative, as well as vvanrij's, but the point of their disagreement is rather philosophical.

Leeperry, you cannot hear the difference between SACD and CD, sorry for that. And if you're willing to say that you do hear it, just prove it, some folks here are not satisfied with only verbal/written statements

PiccoloNamek, a great thread indeed! I also use some equalisation with a bit of solid theory behind; at first I was going to disagree with your claims etc., but after some testing I find them very close to the truth, just with a grain of salt.

My point? A softly shaped peak in the range of 4600-6400 Hz, another around 9000 Hz to 10 100 Hz and the last 13500 - 15 400 are a great way to improve highs of any headphone with shortages of it. What is more, recessed range of 1800 Hz to 3300 Hz is also beneficial (natural sound as a result etc.). Finally, very, very sharp, pointed peaks in the range 50 Hz to 150 Hz (two or three) are great if you want to get a punchy bass.

Just check that attachment for a graphical representation

And, needless to say, some headphones may already have such peaks and valleys in their frequency response, determine that before you make any changes

As to these sharp-pointed peaks in the lowest register, probably only Stax Omega exhibits that unique phenomenon (as I remember).

 

[PiccoloNamek]

They are informative, yes, but also very off topic. Perhaps they could be split into another thread somewhere in the sound science forum?

Quote:

PiccoloNamek, a great thread indeed! I also use some equalisation with a bit of solid theory behind; at first I was going to disagree with your claims etc., but after some testing I find them very close to the truth, just with a grain of salt.

My claims are accurate, scientifically sound, and can be repeatably shown to be true in tests.

Quote:

My point? A softly shaped peak in the range of 4600-6400 Hz, another around 9000 Hz to 10 100 Hz and the last 13500 - 15 400 are a great way to improve highs of any headphone with shortages of it. What is more, recessed range of 1800 Hz to 3300 Hz is also beneficial (natural sound as a result etc.). Finally, very, very sharp, pointed peaks in the range 50 Hz to 150 Hz (two or three) are great if you want to get a punchy bass.

If your headphone is equalized to flat as my tutorial guides you to do, you will not have any shortages or excesses of bass, mids, or treble. Of course, not everyone likes a flat response, that's fine. But even then, equalizing to flat first and then applying your normal EQ will help you to get better results and more accuracy out of from effort.

 

[Ashirgo]

Well, for now I just do not trust any kind of "flat" response. To be honest, I use that equalisation (that you can see in the picture) with my HD555 so that they could be as close to HD 800 in frequency response as possible (save for that valley around 6800 Hz - to eliminate resonance).

But something inside of me tells me that I am likely to test them with a flat equalisation. Just with that aforementioned valley; and I do believe that recessed 1800 to 3300 Hz is an essential thing, common with HD 800, Stax Omega and Sennh. Orpheus (could they be mistaken to place that there?).

 

[PiccoloNamek]

Quote:

Well, for now I just do not trust any kind of "flat" response.

Not even your own personal flat response?

Quote:

To be honest, I use that equalisation (that you can see in the picture) with my HD555 so that they could be as close to HD 800 in frequency response as possible (save for that valley around 6800 Hz - to eliminate resonance).

But have you heard the HD800? If not, how can you know how to equalize? Wouldn't it be easier to achieve your goal if you were starting out with a response that was already perceived to be flat?

Quote:

But something inside of me tells me that I am likely to test them with a flat equalisation. Just with that aforementioned valley and I do believe that recessed 1800 to 3300 Hz is an essential thing, common with HD 800, Stax Omega and Senn. Orpheus (could they be mistaken to place that there?).

You know you want to.

Also, the only headphone I've had to make reductions to in the midrange is the ER4S. My STAX do not have any noticeable peaks or hills in that area, so I left it as it is.

 

[Ashirgo]

Quote:

Originally Posted by PiccoloNamek /img/forum/go_quote.gif Not even your own personal flat response?

Truly, I do not. I am afraid it may be not optimal with the most of recordings, which are not mixed so as to sound best in "flat freq." environment, but to be listened to with our "uneven freq. response" equipment.

Quote:

Originally Posted by PiccoloNamek /img/forum/go_quote.gif But have you heard the HD800? If not, how can you know how to equalize? Wouldn't it be easier to achieve your goal if you were starting out with a response that was already perceived to be flat?

I refer to HeadRoom frequency response graphs in my equalization and I am sure for now that my headphones sound lovely with that particular equalization (a fruit of many hours

) and I do not need to acquire any new ones (the only goal would be to pick some with less distortion and higher comfort, and maybe freq. response curve easier to shape with my equalizer).