|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
173.57.125.189
In Reply to: RE: Averaging of multiple locations posted by Dave_K on December 09, 2014 at 15:56:26
Room measurements can be averaged within a small area, and that will obviously yield a smoother curve for that area, which may or may not be adequate, depending upon the measurement goals. But to really understand a speaker's performance in a particular room, and the room itself, averaging - any amount of averaging - blurs essential details which may be needed in order to improve the room or the speaker positioning.
In a home playback environment, averaging the response over several feet, and producing a single "average" curve, may be fine. In a tracking/mixing/mastering control room, not so much. There, details matter.
:)
Follow Ups:
On the contrary, in a home environment it might be "good enough" to take a single measurement, but in a control room I think it's more important to make measurements over the full listening area. First, because performance is more critical. Second, because a working engineer is not stationary; they need to move around a bit especially in front of a big console.
Even if you lock your head in a vice at the sweet spot, your ears aren't in the same location. They are ~7" apart and if you place the microphone at the location of your two ear openings, there will be differences between the two responses which are visible on plots of amplitude and phase vs. frequency. These differences are meaningless. And in every system I've listened to, with the possible exception of extreme nearfield setups, I can move my head left and right a foot and the only significant changes in the sound are a shift in the center of the soundstage because you're not equidistant between the speakers, and subtle shifts in treble balance as the effective toe-in angle changes. The measured responses at the left-of-center and right-of-center positions will show changes in the pattern of peaks and nulls, but these differences aren't audible.
My point is that a single unsmoothed, unaveraged in-room measurement is dominated by features which are inaudible. These aren't details, they are 'noise' obscuring the things that are actually important. You want to suppress this 'noise' so that the 'signal' is more apparent. That requires smoothing the response with a filter, or averaging multiple measurements from different locations around the listening position, or a combination of both. Spatial averaging helps separate the response deviations that are really audible from those that aren't, and it also allows you to use less smoothing. A spatially averaged result with 1/12 octave or even 1/24 octave smoothing will reveal more meaningful detail than a single point measurement with 1/3 or 1/6 octave smoothing.
"On the contrary, in a home environment it might be "good enough" to take a single measurement, but in a control room I think it's more important to make measurements over the full listening area. First, because performance is more critical. Second, because a working engineer is not stationary; they need to move around a bit especially in front of a big console."
WTF? I already said that detailed measurements in a control room are essential, and that averaging multiple locations doesn't produce better data. Did you even read or understand what I wrote?
nt
Good points on measuring -But two things were and (still might be) limitations:
-Swept tones and pink noise generate "reflections" and (were) corrupting the data. Does this still happen ?
-Testers were using 1/3 octave smoothing - which was low-resolution. Are they doing higher-rez now ?
Edits: 12/11/14
-Swept tones and pink noise generate "reflections" and (were) corrupting the data. Does this still happen?
If you're trying to measure and optimize in-room response, the reflections ARE the data. If you're trying to measure the speaker response alone, e.g. if you're designing speakers, then you don't want the reflections.
Anyway, using pink noise and a real-time analyzer to make frequency response measurements is not a good way to go. What you want to do is measure the impulse response from a full range sine sweep or maximum length sequence (MLS). From the impulse response, you can derive lots of things including magnitude and phase response vs. frequency, RT60, waterfall plots, spectrograms, etc. This is how most room and speaker measurement software works (e.g. REW, R+D, XTZ, OmniMic).
If you have the impulse response, you can time gate (truncate) it to remove some or all reflections. However, gating limits the lowest frequency, the resolution, and signal-to-noise ratio of the response. If you want to remove ALL reflections, including floor bounce, then you need a very short gate, typically <= 1msec for measurements at the listening position, which limits the response to 1 KHz and above. Or you could choose a gate length that includes the first arrival + early reflections, but excludes reverberation & modal ringing. That would likely yield a response from ~150 Hz and above.
But if you want to look at the bass region, then you can't use time gated measurements. There's no way to separate the direct & reflected sound in the bass because the wavelengths are long. Nor would you want to, because you don't hear them separately.
So the usefulness of time gating depends on what you are making the measurements for. If you're doing it to optimize the location of speakers and/or the listening position, you usually shouldn't use any time gating. An exception might be if you're only adjusting toe-in, but I generally do that by ear. On the other hand, if you're making measurements to generate EQ settings for the midrange or treble, or to optimize crossovers, then you absolutely should time gate.
-Testers were using 1/3 octave smoothing - which was low-resolution. Are they doing higher-rez now ?
Any halfway decent software will have options for smoothing, ranging from no smoothing at all up to 1/1. I usually prefer no smoothing when looking at data below 100 Hz. When looking at full range frequency response plots, I use 1/6 octave for single measurements and 1/12 octave for spatially averaged measurements.
Thanks for this -I guess Stereophile's published data - if high-resolution, is not so bad after all.
My gripe (now) would be that no magazine indicates *internal* problems of loudspeakers - like voice-coil inductance (motor distortion) and then, system noise below signal (in db, like 30-40 which many are).
Probably too time consuming - but the noise (I would think) they can do...
Edits: 12/12/14 12/12/14 12/12/14
And yes, it's common to see THD+N only 30-40 dB down, sometimes as little as 20 dB down in the bass.
The measurement I would like to see is dynamic compression.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: