[ultimatemj]

So, I'm trying to take some simple RTA measurements using a Dayton Umm6, their provided calibration file, and uncorrelated (stereo) pink noise.

My measurements are coming up with more slant than my ears hear. Comparing it to my housecurve you'd think it was super bass heavy (+30db above vocals) and dull up high (-15db from 4k up).

So, I found a quiet place in my house and turned on the RTA to see what the mic picked up....and it seems "slanted"...sorry forgot to smooth it, but you can see 20Hz is about +30db!

Did I miss a step in my setup? Or a preference? Or? :huh:

Thanks in advance for any insights!

 

[JohnM]

You have the RTA in spectrum mode, select an RTA mode for an RTA-style display. When making measurements using pink noise it is better to use Periodic Noise (Pink PN), the REW generator can produce this and save it as a WAV file if you cannot play directly from the generator. With Pink PN you do not need to use any averaging, use these settings:

 

[ultimatemj]

That was it, thanks!

On to my tuning :wave:

 

[ultimatemj]

JohnM,

I can't find anything explaining the differences between uncorrelated pink noise and periodic pink noise.

Are they different? If so, can you explain or point me to some information about their differences?

Thanks,
M.J.

 

[JohnM]

Yes, they are different. Random pink noise is usually produced by applying a pink shaping filter to a uniform (white) random signal. It has the required pink spectrum only on average, over any short period its spectrum can deviate significantly from the ideal especially at low frequencies, so a lot of averaging is required to use it for measurement. Periodic Noise is a signal that is constructed to have the desired spectrum in a period that matches the analyser's FFT length so it does not require any averaging and the analyser display reacts much more rapidly to changes in the system than it would if testing with random Pink noise. Pink PN is ideal for live adjustment of EQ filters. It is also possible to control the crest factor of periodic noise (the ratio between the peak level and the rms level). Random pink noise typically has a crest factor of around 12 dB. The PN sequences REW generates are optimised to have a crest factor that does not exceed 6 dB so higher levels can be used without clipping.

 

[VirtualRain]

Yes, they are different. Random pink noise is usually produced by applying a pink shaping filter to a uniform (white) random signal. It has the required pink spectrum only on average, over any short period its spectrum can deviate significantly from the ideal especially at low frequencies, so a lot of averaging is required to use it for measurement. Periodic Noise is a signal that is constructed to have the desired spectrum in a period that matches the analyser's FFT length so it does not require any averaging and the analyser display reacts much more rapidly to changes in the system than it would if testing with random Pink noise. Pink PN is ideal for live adjustment of EQ filters. It is also possible to control the crest factor of periodic noise (the ratio between the peak level and the rms level). Random pink noise typically has a crest factor of around 12 dB. The PN sequences REW generates are optimised to have a crest factor that does not exceed 6 dB so higher levels can be used without clipping.

I find the RTA incredibly useful in making some adjustments. However, I'm curious why there is no way to band-limit pink PN... For example, the RTA is great for sub placement and sub/mains or sub/center cross over adjustments but having to run it full range is a bit hard on the ears and tweeters. It would be great to be able to generate Pink PN from 15-300Hz for these situations. Is that possible and I've missing it, or maybe something for a future update, or its just not possible?

 

[JohnM]

Yes, the PN could cover any chosen span, I'll put that on the todo list.

 

[ultimatemj]

If I understand this correctly, uncorrellated PN will work for capturing RTA, but periodic PN is better for "live use" of RTA.

I don't totally get the statement "it does not require any averaging". My understanding is that averaging multiple mic positions is always necessary above the Schroeder frequency.

Would you mind expanding on your comment?

Thanks in advance for any clarification!

 

[JohnM]

"Averaging" gets used in many contexts, I was using it to refer to the RTA's Averaging setting, not to whether one should measure in more than one position. To understand why averaging is essential with random pink noise it is worth reflecting on what the frequency content of random pink noise looks like on an RTA. You might be tempted to say that's trivial, it's a flat line, right? Here is an example of the frequency content of about 1.5 seconds of pink noise (one 64k FFT's worth at 44.1 kHz sampling), first the span out to 20 kHz then a closer look at the span to 200 Hz:

Not terribly flat, is it? Any short section of random pink noise has a similar amount of variation. The overall frequency content of random pink noise converges to flat as the time period increases, but to have something that is flat to within a couple of dB typically means averaging about 30s or so:

The idea of playing pink noise and viewing the result on the RTA is that any deviation from flat is due to the overall system's response, but that is only true of you wait long enough for the input to the system (the pink noise signal) to average to something close to flat. That means to get a reasonably accurate (to within a few dB) view of the response at a single position you would need to leave the mic stationary there for about 30s, with the RTA averaging all the while. Random pink noise is pretty useless for measuring responses.

To get around the deficiencies of random pink noise we can use periodic pink noise. The pink PN signal is mathematically constructed so that any segment of the length of the RTA's FFT contains every frequency the FFT can resolve at exactly the right amplitude to have a perfectly pink profile, so it is exactly flat on the RTA. To get an accurate view of the response at a single position with pink PN the mic only needs to be there for one FFT duration, or about 1.5s for the 64k FFT at 44.1 kHz used in the sample plots above. There is no need to do any averaging as every FFT gives exactly the same result (background noise contributions aside). However, if you move the mic away before that time has elapsed you won't see the contributions of all frequencies for that location, and it's impossible to say which frequencies will have been missed. If you move the mic around for your measurement then move it slowly and go back to the same places more than once if you want a reasonable approximation to the effect of averaging measurements made at each position.

 

[ultimatemj]

WOW!

Another reason I seem to have been "chasing my tail"!

Looks like time to save my current settings somewhere and start from scratch!

Just gotta make sure I get the Periodic PN burned at the correct rate (44 vs 48)...

Thanks again JohnM!

 

[JohnM]

I find the RTA incredibly useful in making some adjustments. However, I'm curious why there is no way to band-limit pink PN... For example, the RTA is great for sub placement and sub/mains or sub/center cross over adjustments but having to run it full range is a bit hard on the ears and tweeters. It would be great to be able to generate Pink PN from 15-300Hz for these situations. Is that possible and I've missing it, or maybe something for a future update, or its just not possible?

I've added the ability to custom filter the Pink PN in V5.14 beta 8, it is available from the installers directory.