My DIY Multi-Speaker / Multi-Sub auto optimizer

[anlin]

A brief update and some more measurements. There have been some changes to the algorithm since the last update, so I thought I'd post some more up-to-date measurements. Hardware changes since the last post are indicated in bold.

System

• Speakers
  • Mains: Genelec 8361A x3
  • Surrounds: Genelec 4030C x2
  • Front heights: Genelec 8020A x2
  • Subs: custom 21" LLT x2
• Processing/amplification
  • Signal source: Turtle AV Dante-USB-C interface / Blustream SW42DA-V2 (soon)
  • Sub amp: Yamaha PC-412D
  • Conversion: Lynx Aurora (n) + LT-Dante
  • DSP: Intel 270K Plus based server with a Dante PCIe interface, running soft-realtime tuned Windows 10
• Measurement gear
  • Microphone: Beyerdynamic MM 1
  • Preamplifier/phantom power: Lynx Aurora (n) PRE4

Signal chain

                               |----[Dante]--> PC412D/subs
                               |       
Turtle/Blustream --[Dante]--> DSP --[Dante]--> Aurora (n) --[AES]--> L/C/R
                                                   |
                                                   |--[Analog]--> LS,RS,LFH,RFH

Listening space
5 x 2.5 x 3.5 m (WxHxD). Some furniture but no specific acoustic treatment. MLP is at 3 m from LCR

Measurements

All channels individually, PS MIMO active, flat target, 1/6 oct smoothing

L+R frequency response and phase, PS MIMO active, flat target, 1/6 oct smoothing

L+R group delay, PS MIMO active, flat target, 1/3 oct smoothing

22 ms at 20 Hz

Subjective evaluation
It still sounds great

Notes

• The overall latency of the signal chain is 12 ms, confirmed with measurements
  • 10 ms / PS filters
  • 1.3 ms / VST host and ASIO buffers (32 + 32 samples in each direction @ 96K)
  • 0.5 ms / 2 hops over the Dante network (0.25 ms in one direction)

 

[90sNintendoKid]

A brief update and some more measurements. There have been some changes to the algorithm since the last update, so I thought I'd post some more up-to-date measurements. Hardware changes since the last post are indicated in bold.

System

• Speakers
  • Mains: Genelec 8361A x3
  • Surrounds: Genelec 4030C x2
  • Front heights: Genelec 8020A x2
  • Subs: custom 21" LLT x2
• Processing/amplification
  • Signal source: Turtle AV Dante-USB-C interface / Blustream SW42DA-V2 (soon)
  • Sub amp: Yamaha PC-412D
  • Conversion: Lynx Aurora (n) + LT-Dante
  • DSP: Intel 270K Plus based server with a Dante PCIe interface, running soft-realtime tuned Windows 10
• Measurement gear
  • Microphone: Beyerdynamic MM 1
  • Preamplifier/phantom power: Lynx Aurora (n) PRE4

Signal chain

                               |----[Dante]--> PC412D/subs
                               |      
Turtle/Blustream --[Dante]--> DSP --[Dante]--> Aurora (n) --[AES]--> L/C/R
                                                   |
                                                   |--[Analog]--> LS,RS,LFH,RFH

Listening space
5 x 2.5 x 3.5 m (WxHxD). Some furniture but no specific acoustic treatment. MLP is at 3 m from LCR

Measurements

All channels individually, PS MIMO active, flat target, 1/6 oct smoothing
View attachment 554059


L+R frequency response and phase, PS MIMO active, flat target, 1/6 oct smoothing

View attachment 554060


L+R group delay, PS MIMO active, flat target, 1/3 oct smoothing

View attachment 554064
22 ms at 20 Hz

Subjective evaluation
It still sounds great

Notes

• The overall latency of the signal chain is 12 ms, confirmed with measurements
  • 10 ms / PS filters
  • 1.3 ms / VST host and ASIO buffers (32 + 32 samples in each direction @ 96K)
  • 0.5 ms / 2 hops over the Dante network (0.25 ms in one direction)

The all Genelec Atmos system made my eyes pop out. Congratulations I suspect that sounds awesome aside from the EQ.

 

[anlin]

Thanks! I'm very happy with the system. The rears and heights are very underpowered compared to the subs and the LCR speakers but enough for my current listening space, though PS would probably work even better with more capable surround/Atmos speakers.

 

[kemmler3D]

There seems to be a lot of disturbance between 300-500hz even though the bass looks amazing. What do you think drives that?

Would love to see a waterfall too if you have the time.

 

[anlin]

I'm not entirely sure what causes the unevenness around and above 300 Hz. I have a some quiet (for enterprise rackmount gear) servers and networking gear in the same room, so a measurement artifact caused by the background noise is not out of the question, although the untreated front and back concrete walls and the large window on the left side of the room are probably to blame.

 

[kemmler3D]

It actually does look like it could be fan noise now that you mention it... The waterfall shows some tones reaching a steady level. And the decay looks very clean above that band. I wonder what it would look like if you redid everything with the fans off. Although I realize that's a lot of work.

 

[anlin]

I will remeasure without the fan noise the next time I have the servers off. The AV system is powered by two Eaton 9PX UPSs so I'll have to bypass and switch them off too to get a truly quiet baseline

 

[umamiaudio]

I will remeasure without the fan noise the next time I have the servers off. The AV system is powered by two Eaton 9PX UPSs so I'll have to bypass and switch them off too to get a truly quiet baseline

If you can, wait until the next algorithm update and do new measurements as well because (unlike what some people think) the noise DOES affect the algorithm because the algorithm doesn't know if it's the room or a fan, so that throws it off as well.

 

[soundflo_w]

I'm considering Perfect Soup for my home studio. 2x Neumann KH120 and 3 DIY subwoofers.
It should work with a Scarlett 18i20 3rd Gen, right?
Is there a way to optimize subwoofer placement for the best calibration result?
Is optimal subwoofer placement for Perfect Soup the same as the general multi-sub wisdom ("subwoofer crawl" etc.)?
Thanks!

 

[umamiaudio]

I'm considering Perfect Soup for my home studio. 2x Neumann KH120 and 3 DIY subwoofers.
It should work with a Scarlett 18i20 3rd Gen, right?
Is there a way to optimize subwoofer placement for the best calibration result?
Is optimal subwoofer placement for Perfect Soup the same as the general multi-sub wisdom ("subwoofer crawl" etc.)?
Thanks!

Hi! Thanks for checking us out. Perfect Soup works with your setup.

The best way to place your subs is to use REW: put the mic at your main listening position, then test different placements. What you’re looking for is the combination that gives the widest “coverage,” meaning the nulls don’t all stack up at the same frequencies.

This method doesn’t guarantee the absolute best sound, but any placement with poor coverage cannot be the best. In other words, the best‑sounding setup will always be one of the placements where the nulls are spread out, not clustered.

 

[soundflo_w]

Thanks for the reply!

Just as a side note, I’ve been wondering about this question:
Would it be correct to say that each additional subwoofer in a multi-sub optimization setup acts like an active bass trap (similar to a PSI Audio AVAA), with the difference that the subwoofers have foreknowledge of the signal and can proactively cancel reflections, rather than just reacting to them?

 

[umamiaudio]

Thanks for the reply!

Just as a side note, I’ve been wondering about this question:
Would it be correct to say that each additional subwoofer in a multi-sub optimization setup acts like an active bass trap (similar to a PSI Audio AVAA), with the difference that the subwoofers have foreknowledge of the signal and can proactively cancel reflections, rather than just reacting to them?

Yes that's a part of how it works, but it doesn't just cancel reflection because it can still act as a source as well.

 

[john61ct]

What is PS?