Originally Posted by JohnPM
I would strongly advise people not
to attempt measurements 60dB above the background noise level in their environments. Measuring at high levels risks damage to equipment, hearing and relations with your neighbours and pushes equipment out of its linear range, increasing distortion and degrading measurement quality.
Measurement software can see well below the acoustic noise floor of your room. The important signal-to-noise (S/N) ratio in measurements is the S/N of the recovered impulse response. To see that, open the Impulse plot with the vertical axis set to dB FS - by default this plot is normalised, which means the peak value is at 0 dB FS, and the noise floor of the impulse response will be apparent when zoomed out so the plot spans a second or so - if the noise floor is at -50 dB FS, for example, the measurement has 50 dB S/N. If the impulse response S/N is below 40 dB or so you may start to see noisier frequency response plots and the floors of waterfalls and spectrograms will start to show more noise, but the measurement still remains useful.
If the S/N of the impulse looks poor and you are using a microphone and preamp or a line input adjust the input gains to keep the signal at a good level, with the peak during a measurement getting to around -10 dB or so (this is shown on REW as the "Headroom" figure when you measure, it should not be too high). USB mics will usually show lower levels as they tend to accommodate wider input ranges, but should still give good impulse response S/N values.
The best way to improve S/N in the measurement is to use a longer sweep, each doubling of sweep length improves S/N about 3 dB. If you are already at the maximum sweep length, try multiple sweeps, though the improvement from each doubling of the number of sweeps is usually not quite as much and some soundcards do not maintain synchronisation between sweeps, causing invalid measurements (evident in multiple peaks at or near 0 dB FS in the impulse response, one per sweep).
I 100% agree. I think the suggestion to measure at 90db or so is rather dangerous. People don't realise how loud a constant 90db is.
And it is unnecessary (depending on what you are measuring).
I currently live on a rather busy road so the noise floor is much higher than I would like. I also have a new baby.
Yet the FR response from my speakers as seen by REW doesn't vary whether I use 70dB or 90db. Occasionally I remeasure because of an obvious bus induced false reading. When I get the warning from REW about S/N I always save the measurement to compare. Not much difference.
Even waterfalls and ETC curves don't show drastic differences when measured in peak hour or at 3AM.
Admittedly I have active dipoles which don't like really deep bass and don't have any protection for individual drivers, but I try not to go above 75db. The ARC in my Anthem prepro picks a similar level.
As an aside, I haven't checked this thread for a while, but it is disappointing to see it become polluted with arcane hobby horses. They may or may not be founded in fact, but I thought this thread was to help beginners take the leap into measuring their room and speakers.
95% of 'audiophiles' have no clue what the real response of the the speaker/room combo is. The coolest thing about REW is the democratisation of what is required to find out.
If this thread can help people take a simple frequency response, then use the RTA to position their speakers, then check what their chosen roomEQ has done we will have provided a very valuable service.
The ETC/waterfall/whatever stuff is a distraction.Edited by Steve Dodds - 1/31/13 at 3:58am