The Effects of Small Changes over a Frequency Range and Big Changes over a Frequency - Page 3

Quote:

Originally Posted by hulawafu77 

even the slightest shift in change in placement can have a dramatic change in measurements.  How do you know for example, that the headphone was placed for measurement on his tests EXACTLY where it was before for his break in tests?  You don't.  And be very surprised if he does. 

For these particular tests, I believe he did not move the headphones off the test head for the duration. 

This is a very deep subject, and my upcoming Audio Book has an entire section on this. The short version is:

* Small changes in level (EQ) are more audible at midrange frequencies than at very low and very high frequencies.

* Low-Q boost and cuts (wide bandwidth) are more audible than high-Q boost and cuts.

* For any EQ change to be audible, those frequencies must be present in the source. This may seem obvious, but very narrow changes might miss some frequencies depending on the key of the music.

* I know this is a headphones forum, but when listening in a room, having absorption at all of the early reflection points makes it much easier to hear small changes in EQ. Not only EQ, but also small changes in reverb and panning, which applies more for mixing engineers than for hi-fi listeners. In a well-treated room anyone can reliably hear level changes as small as 0.5 dB.

--Ethan
 

Quote:

Originally Posted by tinyman392 

OK, up until now, it has been widely believed that a 1 dB change could not be heard, and the test to this pointing users to test it themselves by increasing the entire amplitude of a sound by 1 dB…  That data then goes on to FR graphs where we are widely made to believe that these same changes will make no audible change…

The current version of EQu still gets levels wrong, at least on an iPhone 4S or an iPad. If it says it's boosting by 3 dB it's really boosting by 6 dB. Measure the output if if you want to verify this. That 1 dB reading on your iPod may very well be 2 dB. Probably the developer is using 10log for level calculations, instead of 20log, which is what it should be when it comes to voltage gain. EQu's output also isn't very clean--put a 1 kHz tone through a spectrum analyzer and you'll see two sidebands +/- 85 Hz from the main tone, ~50 dB down. The Equalizer app doesn't have these artifacts, and it doesn't get the dB levels wrong either. You may have to redo the experiments.

Quote:

Originally Posted by tinyman392 

I think you have the apps switched...  I did a review of both of these apps (I'm an app and headphone reviewer for another forum) a while back.  When you are describing EQu, it sounds like my description for Equalizer, and when you describe Equalizer it sounds like my description of EQu.  Equalizer was the equalizer that boosted much higher than EQu did.  Again, this is based on the review of both I did.  Also to add, Equalizer added much more distortion at a lower EQ compared to EQu.  I think you have these two apps switched; I've used both, and extensively compared both of them (further than just audio).

Nope, no mistake. I tried it again. EQu gets the levels wrong and it has the artifacts. Equalizer (the one with 7 parametric EQs) gets it right and has clean output as long as you avoid clipping. So please verify with measurements, or the results of your experiment cannot be trusted. Try it with a pure tone and measure the level (with a voltmeter if you want). Then try boosting that frequency by 3 dB. When I try it, the voltage doubled--a 6 dB increase. If you have a spectrum analyzer (software and a sound card will do), you might see the artifacts surrounding the main tone. This, however, won't affect your experiment because the sidebands are still there even if you leave EQu flat.

I got both apps for the iPad in December 2010. It's been almost a year since I last did the test, and today, EQu still behaves the same way on an iPhone 4/4S, and an iPad.

Quote:

Originally Posted by tinyman392 

Well, last time I said this, I got chewed out by about 4-5 members on this site...  Each of which said any of the following: [...]

Quote:

Originally Posted by tinyman392 

 

The only reason why I included the use of Tyll's graphs was because it was stated that the 2-3 dB spikes in the treble along with the 1 dB increase in the mids weren't audible.  I just wanted to see if there was a difference between the two.  I know for a fact that I would not be able to hear these changes over the course of 300 hours though.

 

I see. The key here is that you have to separate a couple of things. First off, sub-1db differences can be noticed in an ABX test settings, the usual threshold quoted for making sure of volume matching is .1db (not that every change is audible down to a .1db difference, but that at that point the difference becomes negligible). With headphones, on the other hand, the situation is more complicated as even minor positioning changes of the fit can easily alter the response by 3db or even more! Even the cushions wearing in (and wearing out) can effect difference of a couple of db because of the difference in the seal. This makes measuring headphones an annoyingly rigorous procedure (as I'm sure purrin or Tyll can attest to). I've seen two graphs of measurements from the same headphone which showed a huge suckout, like a 15db dip in response which disappeared after nudging the phones over a bit.

In connection to the burn-in differences of a couple of db that you mentioned I think it's important to maintain perspective. These differences could be measurement artifacts to begin with, not to mention the differences in fit and the batch differences, which are often greater than those that came up in the burn-in graphs. So it's simply that the measured differences may also be negligible, even though there's a decent change that someone could ABX the difference (which isn't possible, due to the minute differences in fit and the difficulty of convincing someone to wear a headphone for 300 hours straight without touching it for fear of invalidating the results). As such it is both premature and foolish to assume that those results substantiated any claims of burn-in, even though they were the product of a great effort.

As far as EQ, I think what you're doing is awesome. I know from personal experience that it takes a delicate touch to fix response deficiencies.

a old JAES paper gives the following just detectable difference thresholds - as you can see it is a function of how big and how wide both, as a function of frequency too

http://home.provide.net/~djcarlst/abx_crit.htm

Quote:

Originally Posted by jcx 

a old JAES paper gives the following just detectable difference thresholds - as you can see it is a function of how big and how wide both, as a function of frequency too

 

http://home.provide.net/~djcarlst/abx_crit.htm

 

That really depends on the testing conditions though, for example ABXing a pure sinewave of 440 Hz through EQ with a 1/24 th octave band centered at 440 Hz should be quite easy even if the amplitude of the said EQ is just 1 dB.

That become impossible (or very near impossible) with music or pink noise.

We just need a quadriplegic volunteer...