PINK noise level bug , high cut bug

[akl]

custom filtered pink noise

(there is no problem with pseudo periodic noise)

DISabling low pass ( = HIGHcut) gives too high output level

max. lowpass freq setting is 20'000 Hz (NOT i.e. 24'000)
setting low pass (the filter at the HIGHend side) to 20'000 Hz generates files with correct output RMS level.

2 ( or even more ) custom settings would be a nice feature.

I switch from 40 --125 to 125 -- max
and something like 125 - 6k3
( though i might use pseudo periodic noise for a steady level)

Thanks

 

[John Mulcahy]

Can you give more details on what problem you are seeing with the high cut? On a loopback test I get exactly the same rms level (to 1 decimal place) with or without the high cut.

 

[akl]

First, I too had a little problem with the crestfactor. Fully random pink noise UNfiltered at 48 kHz sampling, max Level is -12 dB giving a TRUEpeak Level of -0.8 dB FS.

The HIGHcut: if I just set a low cut I get a wrong level of the pink noise.
The max. frequency for HIGHcut is 20'000, NOT higher !

The Pseudo random Noise works till 24'000 Hz.
----
Controlling your remark, everything worked fine

Sorry, I have to be very very precise:
a different machine, a different REW version.
win 7 REW 5.18 :
pink noise 'custom' HIGHcut DISabled
the level depends on the number in the disabled box !
the noise is generated till half sampling ( 24'000 Hz).
100 gives higher level than 250 and higher than 20000. 20000 may be o.k. .

Sorry a seldom and strange bug.

same problem using 5.19 beta 9.

Thanks for reacting such fast.
REW is very nice and comfortable tool.
-akl-

 

[akl]

First, I too had a little problem with the crestfactor. Fully random pink noise UNfiltered at 48 kHz sampling, max Level is -12 dB giving a TRUEpeak Level of -0.8 dB FS.

The HIGHcut: if I just set a low cut I get a wrong level of the pink noise.
The max. frequency for HIGHcut is 20'000, NOT higher !

The Pseudo random Noise works till 24'000 Hz.
----
Controlling your remark, everything worked fine

Sorry, I have to be very very precise:
a different machine, a different REW version.
win 7 REW 5.18 :
pink noise 'custom' HIGHcut DISabled
the level depends on the number in the disabled box !
the noise is generated till half sampling ( 24'000 Hz).
100 gives higher level than 250 and higher than 20000. 20000 may be o.k. .

Sorry a seldom and strange bug.

same problem using 5.19 beta 9.

Thanks for reacting such fast.
REW is very nice and comfortable tool.
-akl-

----

I should probably add some explanation to the files I did upload.
The tool shows the TRUE intersample peaks (TPL), so POSITIVE levels ABOVE full scale may occur !
The tool uses a RMS scale most of us may NOT use. I use a 'sinus peak level equivalent RMS value' giving the same number measuring sinus for peak dBFS and RMS dbFS . The tool uses (which may be physically precise) 0 dBFS for the max level of a square wave. Please add 3 dB (substract 3 dB from the absolute value).

The high rise at the bass end is due to a scale compression (NOT exactly log.). The REW pink noise shows a steady falling line which is perfect.
The -12 dB pink noise measures -14.6 --> -11.6 dB RMS, which is o.k. THE peak level shows a headroom of 0.8 dB. any higher level tahn -11.2 dB will show compression.

The 'black' pic shows a 1.4 dB overload and a wrong RMS level. This is the BUG. level should be -10 dB ( musiscope should show -13 dB FS RMS ) but is -7.9 , more than 5 dB too high.

I use a bandpass filtered pink noise to improve the quality of the loudspeaker sensitivity figure.
As of now I experiment with 125 Hz to 6k3 at different filter slopes and using periodic noise.
1 kHz sinus gives quite bad figures. The REW 500--2k Noise is a BIG improvement. But I like a little more influence of the tweeter and teh woofer, but NOT the UPPER and LOWER END.

The 'musiscope' by www.xivero.com analyses loudness according to the new standards.
-akl-

 

[John Mulcahy]

OK, think I understand now. There is a bug that the level of custom filtered pink noise is affected by the low and high cut filter frequencies even if the filters are disabled. I have fixed that for the next beta release.

The crest factor of the random pink noise is around 12 dB, but will vary depending on the particular segment analysed.

I suggest you use Pink PN, that is synthesised to contain only the frequency range selected and have a crest factor of 6 dB.