Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Graphic EQ calibration using loudness matched bandlimited pink noise (Read 10271 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Graphic EQ calibration using loudness matched bandlimited pink noise

The idea being that the test tones should be loudness matched according to the equal loudness curve and that the tones should not be pure tones but rather bandlimited noise covering the whole band affected by the EQ (to account for wild variations in loudness within a band, especially treble bands)

http://www.head-fi.org/t/656208/calibratin...qs-the-easy-way

Graphic EQ calibration using loudness matched bandlimited pink noise

Reply #1 –
Quote
...until they all sound equally as loud as the 1000 test sound.
IMO -  Totally the wrong approach.  By ear, no two people are going to make the same adjustments, or if you try again the next day, you'll probably make different adjustments.

I'd say...

1. Adjust your system for flat response.  Use a calibrated instrumentation-measurement microphone, not your ears. 

2. If you want to use loudnesss compensation , apply the standard loudness compensation curve as you reduce the volume from standard-normal listening levels.


P.S.
It's perfectly OK to adjust EQ by ear according to personal taste/preference, or to correct/improve a particular recording.  But, that's different from calibrating your system/room.  And that's done with music, not with pink noise or test-tones.

Graphic EQ calibration using loudness matched bandlimited pink noise

Reply #2 –
... I'm talking about headphones and EQ on stuff like poweramp for android though, hardly situations where you'd expect people to make detailed measurements. As for tuning by ear using music, well, let's just say that I've gained more insight into the workings of poweramp's eq in ten minutes of playing with my test sounds than I have in months of playing with it by ear with music. And I've been able to EQ some very difficult phones to within the ballpark of the sound I get with my desktop's parametric EQ using the poweramp eq. I needed some very counterintuitive settings on poweramp to balance the sound, settings I would never have made without the help of these test sounds even though I already knew the frequency response I wanted--because of the quirks of the poweramp eq.

OTOH do you have pointers on how to do measurement-based EQing using free software and cheap hardware? I was planning to try out drc-fir but was intimidated by the amount of hardware apparently required and now their wiki site has disappeared.

Graphic EQ calibration using loudness matched bandlimited pink noise

Reply #3 –
Quote from: Joe Bloggs on
The idea being that the test tones should be loudness matched according to the equal loudness curve and that the tones should not be pure tones but rather bandlimited noise covering the whole band affected by the EQ (to account for wild variations in loudness within a band, especially treble bands)

http://www.head-fi.org/t/656208/calibratin...qs-the-easy-way


Since "equal loudness curve" is not just one thing but a family of curves, which curve should be used?

I suspect the answer to that would be the curve corresponding to the preferred SPL of the listener which should be the SPL used for setting up the eq.

Band limited noise may not be the best stuff in the world for setting levels.  A worthwhile alternative may be warble tones operating over the same bands.

Another alternative would be groups of multitones.  The multitone frequencies would probably be the easiest to live with if centered on musical notes or fractions of them.

BTW multitones are generally the easiest things to use for measuring frequency response.

Graphic EQ calibration using loudness matched bandlimited pink noise

Reply #4 –
So which member of this family of curves does replaygain use?

Graphic EQ calibration using loudness matched bandlimited pink noise

Reply #6 –
Quote from: Arnold B. Krueger on
Quote from: Joe Bloggs on
The idea being that the test tones should be loudness matched according to the equal loudness curve and that the tones should not be pure tones but rather bandlimited noise covering the whole band affected by the EQ (to account for wild variations in loudness within a band, especially treble bands)

http://www.head-fi.org/t/656208/calibratin...qs-the-easy-way


Since "equal loudness curve" is not just one thing but a family of curves, which curve should be used?

I suspect the answer to that would be the curve corresponding to the preferred SPL of the listener which should be the SPL used for setting up the eq.

Band limited noise may not be the best stuff in the world for setting levels.  A worthwhile alternative may be warble tones operating over the same bands.

Another alternative would be groups of multitones.  The multitone frequencies would probably be the easiest to live with if centered on musical notes or fractions of them.

BTW multitones are generally the easiest things to use for measuring frequency response.


I started by crafting the curve after the equal loudness curve at 60dB but the bass and treble ended up being way too loud so I toned them down.  I admit it's not very scientific but I've got independent verification from a few testers that the standard 60dB curve would be too loud in the bass and treble (leading to overcompensation and too weak bass and treble in the resulting EQ)

How is band limited noise not good for setting levels?  The only problem I've noted so far is that the level may not be steady if the noise is filtered over a narrow band in the low bass range, say 1/3 octave below 60Hz.

Wouldn't multiple pure tones still be vulnerable to the problem of being unrepresentative of all frequencies in the band?  You'd understand my worry if you look at the unsmoothed frequency response of some full size headphones:

graph of Beyer T1, for example

And what about beating resulting from the pure tones being close to each other in frequency?

Graphic EQ calibration using loudness matched bandlimited pink noise

Reply #7 –
The main problem is that you'd have to make sure the eardrum SPL matches that of the specific equal loudness curve your using.
 

Graphic EQ calibration using loudness matched bandlimited pink noise

Reply #8 –
Quote from: Arnold B. Krueger on
Since "equal loudness curve" is not just one thing but a family of curves, which curve should be used?

I suspect the answer to that would be the curve corresponding to the preferred SPL of the listener which should be the SPL used for setting up the eq.

The correct loudness correction curve would be the one that corresponds to the change change in hearing sensitivity at the specific playback SPL as referenced against the original mix level.  The correction curve assumes that the frequency response of the playback system and recording monitor system are at least similar, so what you're correcting for is the change in spectral balance resulting from playback and a level that is offset from the original monitor level.

It's only a family of curves so they can graphically represent what is actually a change in hearing sensitivity that varies over both level and frequency.

The actual correction would be the curve that corresponds to the instantaneous SPL at any moment in time.  In other words, it's not a fixed curve but a dynamic process that changes along with the change in playback level.  To get this right you have to know what that level is at any given moment.  Knowing the system gain won't tell you enough, and knowing the volume control knob position is pretty much useless.  You have to have the entire system calibrated, then calculate the SPL based on that calibration data plus a known system gain offset plus a measurement of the input signal.  When it's all calibrated, as you approach reference level at which the original material was mixed, the dynamic curve changes vanish since you are then hearing the material at the original mix level and no level-dependant spectral imbalance occurs. 

There's only one system I know of that actually nails this dead on: Audyssey Dynamic EQ, found in AVRs with Audyssey processing.  If you want it for headphones, and have an IOS device, grab their "amp" app, which not only has Dynamic EQ but precision model specific headphone EQ, BTW, no small deal.

There's a small problem in calibration with audio-only music as music studio mix levels are not standardized, and thus not always exactly known, but even with some assumptions the biggest areas of correction are at volume levels way below reference or mix levels, so they end up close anyway.  For film audio, mix reference levels are tightly standardized, and are therefore known, so no issue and the correction is dead on at all levels.