I have been doing some verification of the claims about linear phase EQ made by various manufacturers, and the results for Neon HR are rather strange. I'm hoping someone can shed some light on them.

For the tests I generated a stereo white noise file with identical channels. Then I applied EQ to the left channel and displayed the results in the Adobe Audition Phase Analysis Phase Wheel. The Phase Wheel plots phase by frequency with the lower frequencies starting at the center. Each EQ curve created had two nodes: -12db, Q3 at 1kHz and +12db Q3 at 3 kHz.

For most of the parametric EQ's the display showed very similar phase deviations at the two frequency points like this one for Ableton Live EQ 8:



The linear phase EQ's displayed only minor deviation like this one for the Cakewalk LP64 EQ:



However, applying the PSP Neon HR EQ created a totally random distribution including out-of-phase points:



One obvious conclusion would be that PSP is taking the easy way out and generating a random distribution instead of actually trying to minimize phase deviations. However, I will give PSP the benefit of the doubt and assume they have more integrity than that. But I still am puzzled about the outcome.

Danika,

There's definitely something wrong with the testing procedure. I am confident, because I am able to test the phase linearity using another method and get perfect results (send the impulse through the eq and check the resulting response for symmetry and phase response).
Your results are very strange indeed - like the l/r signals are totally out of phase.
Are you sure the right (dry) channel is properly aligned? I will do some testing to make sure it is delayed by the right number of samples (phase-inverting one of two identical channels and running such signal through the eq)...
Maybe you could conduct similar tests as well - reversing the phase of on of the noise channels and running it through the eq to check if they are properly aligned.
Also (I know this is obvious, but I wanna be sure) - because Neon introduces huge latency - make sure the output files you're examining have sufficient length so that you're not checking the phase relationship of the ... silence (or rather some kind of low level dither / antidenormal noise)...

Since I hadn't received an answer on this for awhile, I put it aside. However, I did get back to this project recently and noticed your response. And the latency is Neon HR was in fact the issue. I guess because the delay in Neon HR is so large, the Audition delay compensation doesn't work. So when I manually added a delay of 11712 samples to the white noise, the phase wheel lined up. To answer your other question, I was using only a portion of the file for the phase analysis, so no I wasn't including the silence at the beginning in the measurement.

Danika,

I'm very happy to see this plot! The introduced delay is indeed very long. It couldn't be any shorter, though, unless we want to compromise the frequency resolution - as you're surely aware the law of physics are implacable here Thanks for your testing