REW convert frequency spectrum to fractional octave - Home Theater Forum and Systems - HomeTheaterShack.com

Old 08-23-15, 06:38 PM Thread Starter
New Member

Join Date: Mar 2015
Posts: 4
REW convert frequency spectrum to fractional octave

When taking a RTA measurement in REW you have the choice of taking spectrum measurements (power/hz) or various fractional octave measurements (ie 1/12 RTA). It is impossible to ever compare the graphs of the two due to the different units, though it is mathematically possible to convert at least in one direction; spectrum to fractional octave.

Does REW have the ability to take a saved spectrum measurement and convert it to a fractional octave? I ask since it appears that the sweep measurements are spectrum in nature, which make them difficult to use for EQ in the higher frequencies.
bflataudio is offline

Old 08-24-15, 01:45 PM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,705
Re: REW convert frequency spectrum to fractional octave

In the RTA window you can convert between spectrum and the various octave fractions by simply changing the graph selection, REW retains the underlying raw data to allow that.

It is not really correct to compare an RTA plot with a system measurement. The RTA is showing the spectral content of what is being captured, the sweep measurement is showing how the audio system alters signals passed through it (it is a transfer function measurement). However, if the RTA plot is the result of playing Pink Periodic Noise through the system you will find the saved measurement is a very close match to a sweep. Here is an example I have just captured on my desk, the difference between the two is more down to the background noise than the measurement methods:

JohnM is offline
Old 08-24-15, 02:14 PM Thread Starter
New Member

Join Date: Mar 2015
Posts: 4
Re: REW convert frequency spectrum to fractional octave

I guess I was trying to combine two questions into one.
1. When plotting a sweep and pink noise RTA on the same SPL graph, my lines are not nearly as close as yours, so I assumed it was due to the frequency vs fractional octave binning.
2. I also mistakenly took a bunch of pink noise measurements with the spectrum setting instead of RTA and I hoped to be able to convert them so that I don't have to re-do everything.

When switching between the RTA and spectrum on the live screen, even in loop back) the difference between fractional octave (apparent power) and frequency spectrum (linear power) is really obvious. The graph drops off significantly and steadily as frequency increases as the octave bins sum more frequency ranges.

Am I either doing something wrong, or misunderstanding the situation? I'm new to measurements and REW so I'm still learning.
bflataudio is offline

Old 08-26-15, 02:36 PM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,705
Re: REW convert frequency spectrum to fractional octave

The match is good at higher frequncies, which suggests the difference may have been due to the type of pink noise source you used. Random pink noise needs to be averaged over a very long interval for the low frequencies to approach a pink frequency distribution. Pink periodic noise, however, does not need any averaging. RTA settings for use with REW's pink periodic noise (pink PN) are shown below.

JohnM is offline
Old 08-26-15, 03:01 PM Thread Starter
New Member

Join Date: Mar 2015
Posts: 4
Re: REW convert frequency spectrum to fractional octave

I tested with REW's Pink Noise generator, NOT the Pink PN...is that the mistake? If I am reading the help file properly, the PN noise eliminates the requirement for averaging in the RTA, but should yield the same results as the standard Pink with RTA averaging.

I tested with the "Adjust RTA levels" box checked, 4 averages, Rectangular window, and 75% overlap for both spectrum and RTA measurements.

What difference does the 93.75% overlap setting make?

I took another set of measurements in a quiet room to easier illustrate the difference I'm seeing. I had to adjust the offset of the spectrum measurement to line up with the RTA, but you can still see that the spectrum diverges nearly linearly from the RTA above 100hz.

Spectrum:
RTA:
Compared:
bflataudio is offline
Old 08-27-15, 01:02 PM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,705
Re: REW convert frequency spectrum to fractional octave

Quote:
bflataudio wrote: View Post
I tested with REW's Pink Noise generator, NOT the Pink PN...is that the mistake?
Yes

Quote:
I took another set of measurements in a quiet room
That results in you measuring the differences in noise floor of the two techniques, which really isn't relevant. As you will have noticed from your previous measurements the agreement at higher frequencies when there is something there to be measured is very good even with random pink noise.
JohnM is offline
Old 08-27-15, 01:15 PM Thread Starter
New Member

Join Date: Mar 2015
Posts: 4
Re: REW convert frequency spectrum to fractional octave

Quote:
JohnM wrote: View Post
Yes
Aha! It was my error. How many averages should one use if measuring pink noise (not PN) or from a CD or other source to converge?

Quote:
JohnM wrote: View Post
That results in you measuring the differences in noise floor of the two techniques, which really isn't relevant. As you will have noticed from your previous measurements the agreement at higher frequencies when there is something there to be measured is very good even with random pink noise.
I thought noise floor was a factor of the microphone and recording system. How does the noise floor differ based on the measurement or calculations performed?

Also, other than increased processor load (as indicated in the help file) what is the result of increased overlap percentages? Does it help accuracy?
bflataudio is offline

 Bookmarks

 Tags convert , fractional , frequency , octave , rew , spectrum

Message:
Options

## Register Now

Random Question
Random Question #2

User Name:
OR