RTA vs Measurement difference

[edso_73]

Hi to all,

sorry for the stupid question.
I'm a newbie on REW, and I'm struggling to figure out where I'm going wrong.

My setup is composed of a UMIK-II connected to a laptop, a miniDSP HD, and amplifier and a commercial speaker.
I did several test to check how it works REW, and I noticed a big difference on the SPL I get from the measurement done with the sweep signal and the SPL level returned by the RTA.
As a quick check, it can be noticed also that difference in the level of noise (figure below) got from the measurement (performing the sweep with the volume at 0) and the RTA, which is around 23 dB @ 1kHz.

I'm afraid the RTA needs to be set for the calibration input level, where I was thinking setting the calibration file under the Preferences was enough.

Given your experience, where I'm going wrong?

Thank you
Cheers
Edoardo

 

[John Mulcahy]

the level of noise (figure below) got from the measurement (performing the sweep with the volume at 0)

That is not a reliable way to get a background noise measurement, the sweep decorrelation process will be completely wrong as it has no peak to act as a reference The spectrum of the noise floor is captured before the measurement and can be viewed on the Distortion graph.

Sweep measurements and RTA captures are very different things, even if the results may look similar.

A sweep measurement determines the transfer function of the system, how signal levels are changed by the system. That transfer function is, for convenience and to ease interpretation, drawn at an SPL level corresponding to the nominal level of the sweep. If the nominal level was 75 dB, for example, then a frequency at which the response is 87 dB is showing that the overall system has (87 - 75) = 12 dB of gain at that frequency.

The RTA shows the frequency content of an input signal, breaking it up into bins and showing the level in each bin. In RTA mode those bins are an octave fraction wide, e.g. 1/48 octave. In spectrum mode they are fractions of a Hz wide (sample rate divided by FFT length). For a broadband signal (i.e. not a single tone) the level in any bin will be less than the overall level of the signal because the overall level is the sum of the levels in all the bins. The more bins there are (higher octave fraction for RTA or longer FFT for spectrum), the lower the level in any individual bin. That makes it difficult to compare an RTA periodic pink noise measurement to a sweep measurement, for example. To reduce that problem there is an option in the RTA appearance settings to "Adjust RTA levels". That gives a plot with levels that are closer to those a sweep would produce.

 

[edso_73]

Hi John,

I understand it now.
Thank you very much for your reply. It clarified many doubts I had.