Not All Pink Noise is the Same!

[Dyno]

I did a little experiment that y'all might find interesting...

Background
I have a modest Yamaha/Klipsch 5.1 HT system that I've been tuning with REW. I used its generator to play random pink noise and did my best to match a house curve. My room is carpeted, but not otherwise treated. I have a CD from Sheffield Labs that has a pink noise track. I wondered how the Sheffield CD pink noise would compare to REW pink noise. I also wondered if the format (CD, streaming, laptop) would make a difference.

Procedure
Using REW, I recorded "pink noise" from four sources;
1. From the REW Generator to the receiver via cable and the Aux input
2. REW pink noise from a file on a USB stick
3. Pink noise from a Sheffield Labs CD
4. Sheffield Labs Pink noise streamed from Spotify (premium)

I left the equalizer and volume settings constant for all four recordings. I adjusted the levels in the chart manually the converge at 1000Hz.

Results

Conclusions

• Wow! I didn't expect that much spread!
• The Sheffield noise has less low-frequency content than the REW noise
• Some small difference between CD and Spotify, but probably insignificant
• The choice of pink noise source will influence tuning of an equalizer

Recommendation

• Choose your pink noise source to match your listening format when tuning

 

[John Mulcahy]

Instead of capturing the system output play the noise sources directly into a line input. If they are really pink they will be flat on the RTA, after enough averaging.

 

[John Mulcahy]

One other consideration: if you played the signal on both speakers the result will depend on whether the same signal goes to both channels (correlated noise) or whether each channel gets different noise signals (uncorrelated noise).

 

[Dyno]

One other consideration: if you played the signal on both speakers the result will depend on whether the same signal goes to both channels (correlated noise) or whether each channel gets different noise signals (uncorrelated noise).

The Sheffield pink noise is uncorrelated. Does REW generate correlated or uncorrelated noise?

 

[John Mulcahy]

In 5.31 and earlier the noise is correlated, in the V5.40 beta builds it can be either.

 

[Dyno]

Indeed, there is a difference in the RTA of Correlated vs. Uncorrelated pink noise.

Red is Correlated
Green is Uncorrelated

In most cases, the uncorrelated response is lower, so I guess there is some cancelation happening between the two channels.

 

[John Mulcahy]

Correlated noise will add coherently at low frequencies, uncorrelated will not, though how well centred the mic is between the speakers and how close their in-room responses are will also affect the summation.