Joffieb
Member
Thread Starter
- Joined
- Sep 12, 2021
- Posts
- 41
More
- Preamp, Processor or Receiver
- Rotel RC995
- Main Amp
- Rotel RB981
- DAC
- Modi 3
- Streaming Equipment
- Audioengine B-Fi
- Streaming Subscriptions
- Tidal, Spotify
- Front Speakers
- B&W 685s2
- Subwoofers
- AE Aegis Neo V2
- Other Equipment
- MiniDsp nanoDigi 2x8
Hey everyone,
I’ve recently discovered Eric Winer and heard a few podcasts he’s been on. His philosophy is that measurement microphones are far superior to human hearing and that measurements are the only real way to set up and correct a room and system. I decided to lean into that idea and really set my system up using nothing but measurements and Chat GPT to analyse them.
Previously I had used the Sumiko Master Set procedure supplemented by PEQ filters created in REW through a WiiM Pro. It is never an easy process especially with Harbeths (P3ESR) that aren’t bass heavy (or fussy). (I had modified the procedure by having an (almost) willing assistant pull the speakers forward with a towel under the stands while I listened from the listening position for bass changes and with my B&W 685 s2’s which have very punchy bass, the procedure was amazingly definitive and quickish.)
NEW PROCEDURE (ChatGPT, REW AND WiiM Pro)
Speaker Placement
I measured my Harbeths at several distances from the front wall 82 cm, 103 cm, 110 cm, 115 cm and 123 cm capturing SPL, GD, spectrograms, distortion and impulse response at each position.
After going through everything, ChatGPT pointed to 110 cm from the wall as the best location, with the most balanced SPL, lowest distortion, cleanest impulse and tightest bass decay.
Subwoofer Integration
I then worked on the sub. Measurements showed that 180° phase, with reduced gain and crossover lowered, integrated the best. The handoff around 65 Hz smoothed out nicely and group delay was cleanest there too.
Filters I Tested
With the position and sub locked in, I tried three different EQ approaches:
ChatGPT filters (it created these for me):
“NOLS” (no low shelf) → very neutral and tight.
“LS” (with low shelf) → warmer, fuller, but slower bass.
REW filters – generated directly in Room EQ Wizard, aimed for maximum flatness.
WiiM RoomFit filters – my hardware’s built in auto correction.
I then measured the system using each filter and had ChatGpt assess which was best. I have attached the full report as well as my mdat file.
According to CHatGPT the winning combo is:
110 cm speaker distance + 180° sub phase + ChatGPT’s NOLS filters.
To my ears, without any filters the Harbeths and Sub sound fantastic. The P3ESR’s aren’t nearly as fussy as my previous speakers and their tone and revealing nature are quite a joy, and the Sub is turned very low with a 60Hz LPF so there’s not much boom to start. The Harbeths aren’t huge rockers and no at all efficient. I’ve moved to a 200WPC Rotel and can crank them up and the bass heavy music is okay but it’s not my go to music anyway.
Once I tried all the filters, some sounded different to others but all were still good and I’d probably choose them based on the music in question.
Thanks
I’ve recently discovered Eric Winer and heard a few podcasts he’s been on. His philosophy is that measurement microphones are far superior to human hearing and that measurements are the only real way to set up and correct a room and system. I decided to lean into that idea and really set my system up using nothing but measurements and Chat GPT to analyse them.
Previously I had used the Sumiko Master Set procedure supplemented by PEQ filters created in REW through a WiiM Pro. It is never an easy process especially with Harbeths (P3ESR) that aren’t bass heavy (or fussy). (I had modified the procedure by having an (almost) willing assistant pull the speakers forward with a towel under the stands while I listened from the listening position for bass changes and with my B&W 685 s2’s which have very punchy bass, the procedure was amazingly definitive and quickish.)
NEW PROCEDURE (ChatGPT, REW AND WiiM Pro)
Speaker Placement
I measured my Harbeths at several distances from the front wall 82 cm, 103 cm, 110 cm, 115 cm and 123 cm capturing SPL, GD, spectrograms, distortion and impulse response at each position.
After going through everything, ChatGPT pointed to 110 cm from the wall as the best location, with the most balanced SPL, lowest distortion, cleanest impulse and tightest bass decay.
Subwoofer Integration
I then worked on the sub. Measurements showed that 180° phase, with reduced gain and crossover lowered, integrated the best. The handoff around 65 Hz smoothed out nicely and group delay was cleanest there too.
Filters I Tested
With the position and sub locked in, I tried three different EQ approaches:
ChatGPT filters (it created these for me):
“NOLS” (no low shelf) → very neutral and tight.
“LS” (with low shelf) → warmer, fuller, but slower bass.
REW filters – generated directly in Room EQ Wizard, aimed for maximum flatness.
WiiM RoomFit filters – my hardware’s built in auto correction.
I then measured the system using each filter and had ChatGpt assess which was best. I have attached the full report as well as my mdat file.
According to CHatGPT the winning combo is:
110 cm speaker distance + 180° sub phase + ChatGPT’s NOLS filters.
To my ears, without any filters the Harbeths and Sub sound fantastic. The P3ESR’s aren’t nearly as fussy as my previous speakers and their tone and revealing nature are quite a joy, and the Sub is turned very low with a 60Hz LPF so there’s not much boom to start. The Harbeths aren’t huge rockers and no at all efficient. I’ve moved to a 200WPC Rotel and can crank them up and the bass heavy music is okay but it’s not my go to music anyway.
Once I tried all the filters, some sounded different to others but all were still good and I’d probably choose them based on the music in question.
Thanks
