AI guided Room Optimization.

crwilli said:
Clarification: I am not using DSP. I used AI to interpret the REW measurements and guide me through various ‘setups’ to find an optimum. No DSP was is being used.

But: After I was done and had the subs integrated, it did offer to suggest a series of Parametic EQ settings to tweak things which I accepted In 2 seconds, it provided 10 PEQ adjustments.

Using ROON’s functionality, I have applied all 10 settings. It’s early days and I will need to go back and forth to see if I accept them or not.
Click to expand...
The AI agent was correct in its assertion that group delay is the best method for subwoofer integration. It is.

Adjusting phase over a relatively small range of frequencies while looking at system amplitude response with pink noise creates the opportunity for a 360 out-of-phase situation to measure great in the frequency domain, (a.k.a pink noise spectra) while performing very poorly in the time domain where it counts most.

Human hearing is very time sensitive. Inter-aural time delay is a main contributor to how we localize sounds. 10mSec @ 100 Hz is one 360 degree period, and with subwoofer DSP and group delay it is easy to get there.
 
Last edited:
Here’s why I won’t bother with Pink Noise measurement. I used frequency sweeps from 15hz to 21,000 hz and fed Chat that data.

✅ TL;DR — Sweeps vs Pink Noise

Use REW sweeps. They are far better.

Why sweeps win:
  • Much more accurate
  • Much higher signal-to-noise ratio (especially in the bass)
  • Give you phase, impulse, group delay, and timing
  • Let you integrate subs and mains properly
  • Work consistently and repeatably
Pink noise is only good for:
  • Quick tonal checks
  • Listening/voicing
  • Very rough level matching
Pink noise cannot measure:
  • Group delay
  • Phase
  • Waterfall/decay
  • Sub alignment
  • Modal behavior
  • Time domain performance
Everything we’ve optimized together requires sweeps. Pink noise can’t do it.
 
W9TR said:
The AI agent was correct in its assertion that group delay is the best method for subwoofer integration. It is.

Adjusting phase over a relatively small range of frequencies while looking at system amplitude response with pink noise creates the opportunity for a 360 out-of-phase situation to measure great in the frequency domain, (a.k.a pink noise spectra) while performing very poorly in the time domain where it counts most.

Human hearing is very time sensitive. Inter-aural time delay is a main contributor to how we localize sounds. 10mSec @ 100 Hz is one 360 degree period, and with subwoofer DSP and group delay it is easy to get there.
Click to expand...
If your 360* out of phase? Arn't you in phase?

I can see how timing is the critical element. For clarity. It seems to me, if the timing is correct, the phase is also correct?
 
Roon users can measure with REW (free) and send the data to a third party to generate convolution filters (reasonably priced). The improvement absolutely dwarfs anything you’ll get from cables, conditioners, streamers, or whatever miracle network switch is trending this week.
 
nicoff said:
Roon users can measure with REW (free) and send the data to a third party to generate convolution filters (reasonably priced). The improvement absolutely dwarfs anything you’ll get from cables, conditioners, streamers, or whatever miracle network switch is trending this week.
Click to expand...
Convolution Filters are super powerful and I have dabbled with them. My experience was they are only as good as the target curve you use and how well that matches your desired SQ.

It’s all part of the hobby for me.
 
Kingrex said:
If your 360* out of phase? Arn't you in phase?

I can see how timing is the critical element. For clarity. It seems to me, if the timing is correct, the phase is also correct?
Click to expand...
Yes, you are in phase, but out time

Let’s say the subwoofer is delayed one full cycle. This is not uncommon. Let’s do 100 Hz to make the math easy. The period of a 100 Hz tone is 10 mSec. The woofer’s output is delayed 10 mSec compared to the mid/tweets. The attack of a kick drum will be heard by the listener at two different times and be softened and blurred.
 
W9TR said:
Yes, you are in phase, but out time

Let’s say the subwoofer is delayed one full cycle. This is not uncommon. Let’s do 100 Hz to make the math easy. The period of a 100 Hz tone is 10 mSec. The woofer’s output is delayed 10 mSec compared to the mid/tweets. The attack of a kick drum will be heard by the listener at two different times and be softened and blurred.
Click to expand...
Yes, but today many DSP filters can correct both frecuency and time domains issues.
 
crwilli said:
Here’s why I won’t bother with Pink Noise measurement. I used frequency sweeps from 15hz to 21,000 hz and fed Chat that data.

✅ TL;DR — Sweeps vs Pink Noise

Use REW sweeps. They are far better.

Why sweeps win:
  • Much more accurate
  • Much higher signal-to-noise ratio (especially in the bass)
  • Give you phase, impulse, group delay, and timing
  • Let you integrate subs and mains properly
  • Work consistently and repeatably
Pink noise is only good for:
  • Quick tonal checks
  • Listening/voicing
  • Very rough level matching
Pink noise cannot measure:
  • Group delay
  • Phase
  • Waterfall/decay
  • Sub alignment
  • Modal behavior
  • Time domain performance
Everything we’ve optimized together requires sweeps. Pink noise can’t do it.
Click to expand...
Its not either or …….

Regards
 
You must log in or register to reply here.
Back
Top