or Connect
New Posts  All Forums:Forum Nav:

I was wondering why there is a positive and negative on the dB axis in many frequency response curves. How could something have a "negative decibel" amplitude?

The dB scale is dependent on a reference value. Negative dB is below the reference.

Oh, which is what? The source?

Exactly 0db across the whole graph would be a completely flat response. The item being tested would play the frequencies at the correct levels. When you see + or - db. It means the item being tested is either playing that frequency louder or softer than the source.

This all makes sense, but on question I'd still have is how does one normalize the output of the driver being tested so that 0dB is actually what is expected?

Since the decibel scale is relative, there's nothing you expect 0 dB to be until you define your reference value. Do check the Wikipedia article on decibel.

0 db is the reference level. So say they play a tone sweep. They set the volume so that it should play all frequencies to a reference level (say 70db just as an example) as the sweep plays if the spl starts to go above or below 70db that is where you see + - db.

The reference point is not that important, frequency response is about balance! If the graph is aligned to 0dB or -20dB doesn't matter too much. What a FR graph is useful is to compare how it compares at X frequency range against Y frequency range in that same frequency response.

Well OK, what I'd personally define a reference point if I was hosting a site such as goldenears site would be to that the line would correspond to a "average frequency amplitude", to me that is the true reference point of a frequency graph. Like for example if the frequency response was a diagonal straight line from say +10 to -10dB then 0dB would be the reference point as it's the average frequency amplitude. Too bad frequency response graphs are hardly perfectly straight though so determining accurately the average frequency amplitude is usually more difficult, then you have to weight the bumps vs the valleys and it's more of an integral calculation, that is the most accurate way to compare headphones vs each other if the graphs are aligned this way.

EDIT: Actually goldenears like I used as example seems to do this so thumbs up to them for realizing this. :)

Example:

"Golden ears target" represents "average frequency amplitude" in this case as I explained above. It's against this line I compare how much of a bass bump there is or recession or boosts at given frequency response range. When it comes to presenting the data, I personally fully agrees with Golden Ears way of doing it, I find it even more useful than Tyll's personally.

Edited by RPGWiZaRD - 12/18/12 at 12:06pm
Just to add to what's been said RPG's post (and the graph) are elaborating on what CDewey is saying.

Generally speaking with FR measurements or DNR measurements, you'll see negative values over positive values, where 0 is "reference" or "maximum" as opposed to "minimum" as most would likely imagine. FR isn't quite related to output intensity, which is less relative and rarely measured (as most any headphones made can easily get loud enough to cause hearing damage).
New Posts  All Forums:Forum Nav:
Return Home