My DIY Multi-Speaker / Multi-Sub auto optimizer

[dziemian]

Measurements Software: https://www.rationalacoustics.com/pages/about-smaart

Room Mode Calculator: https://www.vesalaasanen.com/

'Room simulator for non-shoebox rooms'
https://www.audiosciencereview.com/...s/room-simulator-for-non-shoebox-rooms.69609/

Thank you. Anything else apart from Rew and t-smaart? What about decay time and sti analysis? Btw your measurements looks great. I have seen your post on Dirac Art and I wonder as anybody here whatcwould be the comparison

 

[dziemian]

What smaart version is good enough? The pricing is broad.

 

[umamiaudio]

Hi, I just tested UA Perfect Soup with a 9.8.6 Atmos System. I never heard my room as cohesive and clear. Bass was precise and tight. Below you will find my before and after measurements. I plan to add a hardware block diagram soon.

System Calibrated:
View attachment 533066

System Uncalibrated:
View attachment 533067

System Calibrated with LFE:
View attachment 533068

System uncalibrated with subs:
View attachment 533069

Room Layout:
View attachment 533071

Left Front speaker Calibrated Decay Time:
View attachment 533072

LFE subs calibrated Decay Time:
View attachment 533073

Speech Transmission Index (STI) Analysis:
View attachment 533074

Umami Audio VST Plugin:
View attachment 533075

I want to mention thought that these measurements were done from a calibration with LESS measurement points than normally required, so we can expect an even better result when done properly.

 

[Stingber09]

What smaart version is good enough? The pricing is broad.

That depends on your needs. REW is free and good to begin with as you get familiar with the measurements.

 

[dziemian]

I used to measure with REW a couple of years ago. Before a child happened

 

[anlin]

After running Dirac Live + Bass Control (DLBC) as a VST with my 5.2.2 system for the past three years, I decided to give Umami’s Perfect Soup a try since my existing signal chain was already compatible. Here are my initial impressions and brief comparative measurements

System

• Speakers
  • Mains: Genelec 8361A x3
  • Surrounds: Genelec 4030C x2
  • Front heights: Genelec 8020A x2
  • Subs: custom 21" LLT x2
• Processing/amplification
  • AVR: Yamaha RX-V781
  • Sub amp: Yamaha PC-412D
  • Conversion: Lynx Aurora16 + LT-Dante
  • DSP: server with an Intel i3-13100 and a Dante PCIe interface, running Windows 10 (soft-realtime tuned)
• Measurement gear
  • Microphone: Beyerdynamic MM 1
  • Preamplifier/phantom power: Rolls MP13

Signal chain

                                               |----[Dante]--> PC412D/subs
                                               |        
AVR --[analog 5.1.2]--> Aurora16 --[Dante]--> DSP --[Dante]--> Aurora16 --[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

Measurements
(Stereo only for now due to software and time limitations)

L+R only, no correction

L+R, DLBC

L+R (+ supports), Perfect soup

L+R only, no correction

L+R, DLBC

L+R (+ supports), Perfect soup

Subjective evaluation
DLBC and PS both do a good job of taming the worst room modes, but overall PS sounds better in my subjective opinion. I'll spare you the audiophile nonsense

Notes

• PS's latency is low at 10 ms
• The VST runs well with minimal buffers (64 samples @ 96K)
• The calibration process is time consuming, but the results speak for themselves. I ended up doing 30 MLP measurements with 12 secondary listening position measurements
• It's nice to be able to adjust the target curve without having to redo the calibration like with DLBC
• The support is outstanding. A VST channel mapping bug took less than a day to resolve

Conclusion
With my setup the Perfect Soup delivers superior results both subjectively and objectively. I intend to keep running it as my main room correction software for the foreseeable future. I recommend giving it a try if your hardware can run it.

 

[StabMe]

@umamiaudio

Since my system is fully active and i do crossovers and room correction inside Reaper, there is no way i can run calibration of Perfect Soup using microphone and my system using direct access via ASIO, since one can only access separate "ways" (e.g. subs, woofers, tweeters) directly via ASIO. Developer said that using workarounds like creating virtual ASIO devices is a bad idea due to possible drift issues so i decided to try a different workaround.

I ran calibration by exposing 2 subwoofers and 2 woofers directly in Calibration Wizard with the hopes that Perfect Soup will take care of the low end part and i will then add tweeters in Reaper, cross them over to calibration woofers+subs, time align them and then add crude (or maybe not so crude) correction of the resulting response using MMM or maybe more sophisticated methods.

So calibration went with no errors, 30 measurements were recorded an i then got my calibration available. I was planning to start messing with adding tweeters but then i decided to run a quic sweep measurement with REW and the response was not very flat (very ragged).

In order to setup routing of inputs/outputs for the Perfect Soup plugin, i send stereo signal to the input and set up the plugin to have 4 outputs and routing was similar to that which was used during calibration: first two output channels are woofers and third and fourth channels are subs.

Could i be getting very ragged response due to trying to calibrate basically a tweeterless system?

 

[umamiaudio]

It's not designed to manage crossovers within the same speaker. That needs a different set of assumptions than the one Perfect Soup was designed on. In this case I think (I'm pretty sure but not 100% sure) a virtual ASIO device would be okay because it's still the same physical device. Also several people had good results using two devices anyway even though it's unsupported and there's currently nothing to mitigate a drift (if any). You can also use loopback as well if your interface has enough channels. In this case I would try both calibrating as a 2.0 system and a 2.2 system where the tweeter and woofers make a 2 way bookshelf + subs.

7 , 8 ==> loopback into reaper x-over===> 1 , 2 ==> tweeters
                                          3 , 4 ==> mids
                                          5 , 6 ==> Subs
OR

7 , 8 ==> loopback into reaper x-over ===> 1 , 2 ==> tweeters
                                           3 , 4 ==> mids
9 , 10 ==> loopback into reaper =========> 5 , 6 ==> Subs

The next big release will come with drift timing correction as well as support to allow different input and output devices.

 

[john61ct]

Please do BNC word clocking rather than slaving to an input signal

ASRC buffering as a last resort, only if things go wrong with hardware clocking

 

[umamiaudio]

A result from a user on discord.

 

[umamiaudio]

Please do BNC word clocking rather than slaving to an input signal

ASRC buffering as a last resort, only if things go wrong with hardware clocking

BNC fixes drift but doesn't fix driver initialization time so a loopback of some sort is still needed.

 

[soundflo_w]

Hi everyone,
what would be the cheapest way to use Perfect Soup for a typical living room home theater setup (5 channels + 3 subwoofers)?
Ideally, I am looking for a complete "set it and forget it" solution. It needs to be fully automated so that the calibration loader activates instantly when the home theater system is turned on—or at the very worst, by flipping one single additional power switch.
And it needs to be way cheaper than an Denon X3800H + Dirac ART.

I already have a spare Micro ATX Windows PC.

Any advice would be highly appreciated!

---

 

[90sNintendoKid]

I did this with my older Marantz SR6012 which doesnt have Dirac.

AVR pre outs into Behringer UMC 1820 plus Behringer ADA 8200 connected to a Windows PC.

 

[umamiaudio]

Hi everyone,
what would be the cheapest way to use Perfect Soup for a typical living room home theater setup (5 channels + 3 subwoofers)?
Ideally, I am looking for a complete "set it and forget it" solution. It needs to be fully automated so that the calibration loader activates instantly when the home theater system is turned on—or at the very worst, by flipping one single additional power switch.
And it needs to be way cheaper than an Denon X3800H + Dirac ART.

I already have a spare Micro ATX Windows PC.

Any advice would be highly appreciated!

---

Platinum Studio license ($599) + 1 extra sub ($59 -- It already lets you use two subs out of the box), with the current 30% discount (you'll get an email shortly after signup) = $460.6, taxes not included.

You can use it with a free host like Cantabile/ equalizer APO / etc that auto starts when your PC starts. If you need help getting setup, please feel free to message me.

 

[anlin]

Hi everyone,
what would be the cheapest way to use Perfect Soup for a typical living room home theater setup (5 channels + 3 subwoofers)?
Ideally, I am looking for a complete "set it and forget it" solution. It needs to be fully automated so that the calibration loader activates instantly when the home theater system is turned on—or at the very worst, by flipping one single additional power switch.
And it needs to be way cheaper than an Denon X3800H + Dirac ART.

I already have a spare Micro ATX Windows PC.

Any advice would be highly appreciated!

---

If you want an appliance type experience you can set Windows up in kiosk mode or modify the registry to replace the default Windows shell (explorer.exe) with your VST host of choice. I'm personally using SAVIHost on a headless server with some custom software to manage core affinities and thread priorities.

 

[umamiaudio]

If you want an appliance type experience you can set Windows up in kiosk mode or modify the registry to replace the default Windows shell (explorer.exe) with your VST host of choice. I'm personally using SAVIHost on a headless server with some custom software to manage core affinities and thread priorities.

GPU / iGPU processing support is coming very very soon! It's just waiting for me to finish up the new wizard so it can be released together. If you use a headless system, it's basically a dedicated processing core.

 

[welwynnick]

Your website says GPU acceleration is used to calculate the filters. Is GPU acceleration also used to run the filters?

Not at the moment.

Hi umamiaudio, you said before that GPU acceleration is used to calculate the filters, but not to run them.

GPU / iGPU processing support is coming very very soon! It's just waiting for me to finish up the new wizard so it can be released together. If you use a headless system, it's basically a dedicated processing core.

To be clear, does this now mean that you will soon be able to run the filters with GPU acceleration? . That would be excellent. Is CUDA needed?
Does it make any difference whether PS is embedded or stand-alone?

 

[umamiaudio]

Hi umamiaudio, you said before that GPU acceleration is used to calculate the filters, but not to run them.

To be clear, does this now mean that you will soon be able to run the filters with GPU acceleration? . That would be excellent. Is CUDA needed?
Does it make any difference whether PS is embedded or stand-alone?

The server uses GPU acceleration to calculate your filters, but the CPU runs them locally on your computer, but now that local part can be offloaded to your GPU / iGPU as well. For people in production this helps conserve precious CPU resources for other plugins. For the HT crowd, if you have a big system with many speakers, you can now run it at smallish buffer sizes and have 100% stability if your system has a dedicated GPU. Basically it's just more options for the user to use their existing hardware and to use resources more effectively. It will run on AMD, Nvidia, Intel, and Apple M series chips. It will run on both discrete and integrated GPU. The only platform where hasn't been tested is Intel Apple.

 

[soundflo_w]

Platinum Studio license ($599) + 1 extra sub ($59 -- It already lets you use two subs out of the box), with the current 30% discount (you'll get an email shortly after signup) = $460.6, taxes not included.

You can use it with a free host like Cantabile/ equalizer APO / etc that auto starts when your PC starts. If you need help getting setup, please feel free to message me.

Didn't the plugin used to allow purchasing a single month's subscription for measurements, and then using it indefinitely until the next time you needed to remeasure?

 

[umamiaudio]

Didn't the plugin used to allow purchasing a single month's subscription for measurements, and then using it indefinitely until the next time you needed to remeasure?

I'm changing things up as I go so a deal might come up and then never again. For example, all the beta testers who provided feedback got free copies. That's never coming back. What you suggested was never a thing though. We do have a rent to own model that is $27.99/month (with discount) for 24 months. With the license for the extra support, that's $69.29 first month and then $27.99 for the remaining months. It looks like it's a lot more if you total it up, but if you factor in the 3 years of calibration it comes with, it's only about $100 more over the course of 3 years.