[TenTec] ARRL Phase Noise Data Plotted to 1 MHz

[Jim Brown]

On 7/28/2014 9:36 AM, Kim Elmore wrote:
> I've been thinking, too. A square wave has an infinite number I'd odd 
> harmonics, not all harmonics. Regardless, it's an excellent way to 
> look at IMD, far better than two tones.

Yup.
>
> What if we modulated the TX with pink noise? That really does have 
> everything in it. Make the pink noise much wider than the expected 
> audio bandwidth. Ideally, the spectrum should be limited to the TX 
> filtered BW. Wouldn't this be nearly the most severe test?

Actually, it's a near ideal test for performance of the rig on SSB. One 
of the major virtues of pink noise as a test signal is that it roughly 
approximates the spectrum of both music and speech.  In pro audio, we've 
used "shaped" pink noise for years to specify the performance of 
loudspeakers, where the shaping is simply bandpass filtering of pink 
noise for the region where the loudspeakers are intended to operate.  I 
strongly encourage the use of pink noise shaped to a bandwidth of 100 Hz 
to 5 kHz for this purpose.

I've discussed this with Rob Sherwood several times, and taken him to 
task for using white noise for that purpose because it does NOT 
approximate speech. His response is that his intent is to represent 
clipped and distorted audio. My problem with that is that good audio 
signal processing should be tailored to the spectra of speech, so using 
white noise as a test signal will cause that signal processing to 
misbehave (or behave differently from how it would work with speech).

For those who don't know, pink noise is random noise with equal power 
per PERCENTAGE bandwidth -- per octave, per third-octave, per 
sixth-octave, etc. It can be generated by applying a 3dB/octave rolloff 
to white noise.

73, Jim K9YC