REW Convolution Questions

[mrlawng9]

I've been trying to find info on REW's new'ish convolution generation features but am coming up short. I'm interested in how it compares to Dirac, Audiolense, Accourate, etc... I'd like to learn if it is designed similarly to its automated competitors and whether it's based on similar or different principals. It would also be great to know from people who have used such software, whether REW performs similarly or whether it has any advantages or cons compared to them.

Also, is it only capable of generating convolution files for fully integrated speakers (e.g. store bought 3-way) or can it be used to implement crossovers and TA for say a 2.2 system? If it is possible to do this, I'd be appreciative of a high-level overview.

Many thanks!

 

[RayDunzl]

REW's author (@JohnPM) is a member here.

 

[OCA]

how it compares to Dirac, Audiolense, Accourate

These are very different tools. Dirac would best compare with Audyssey and Acourate and Audiolense would compare with each other. REW is also another different category and would best compare to Smaart.

 

[mrlawng9]

REW's author (@JohnPM) is a member here.

Thank you for the call out to John. It would be great if he commented on this.

 

[mrlawng9]

These are very different tools. DIrac would best compare with Audyssey and Acourate and Audiolense would compare with each other. REW is also another different category and would best compare to Smaart.

Thanks Serkan. I've seen your YT video on REW Convolution. Thanks for the effort you put into that.

How would you recommend augmenting the procedure from 2 loudspeakers to multiple? Obviously, we would have 4 in-room measurements vs 2 for a 2.2 system. My thoughts are to leverage REW's EQ functionality and amend step 16 of the PDF you wrote to add LP and HP filters as appropriate for a 2.2 system. After adding the crossovers in the EQ tab, it would be easy enough to have REW generate 4 separate measurements from target shape but how do I merge them? I'm a bit confused if this would be A+B, A*B, or something else entirely via REW's trace arithmetic functionality?

 

[mrlawng9]

I think I may have gotten myself confused and now see how I confused myself.

After setting up my father's MiniDSP SHD Studio recently, I think I got caught thinking about how to use REW convolution in MiniDSPs quasi-multi-channel world. What I mean is that devices like the SHD Studio allow for 2.2 system architectures but the Dirac function is only a 2-channel correction when the subs and satellites have already been summed and optimized (e.g. EQ, xovers, and delays were already performed prior to running Dirac).

There would be no way to use REW to create 4 discrete convolution channels with integrated xovers to stream from Roon into a MiniDSP SHD Studio unfortunately. However, this could be accomplished with multiple DACs.

In this scenario (the one requiring multiple DACs) are there any additional steps to sum / integrate the 4 discrete convolution files / channels or is the process the same as described in Serkan's PDF (albeit with 2 extra speakers to measure and optimize via the process)?

 

[OCA]

I don't know which pdf you're talking about, I've only posted one for Roon users about two years ago and it must have been quite outdated by now.

IMHO, if you want to get great results from your filters with a minimal learning curve and then forget about DSP and enjoy your music, go for Acourate. If you'll design your own speakers, you might consider Audiolense. If you want to dive into the rabbit hole of digital room correction and get stuck in it, download REW and follow AVNirvana forum (and this one of course).

 

[mrlawng9]

I don't know which pdf you're talking about, I've only posted one for Roon users about two years ago and it must have been quite outdated by now.

IMHO, if you want to get great results from your filters with a minimal learning curve and then forget about DSP and enjoy your music, go for Acourate. If you'll design your own speakers, you might consider Audiolense. If you want to dive into the rabbit hole of digital room correction and get stuck in it, download REW and follow AVNirvana forum (and this one of course).

I meant the google drive document linked to your YT video. I misspoke calling it a PDF.

I'm actually pretty far down the DSP rabbit hole and have a lot of experience with Audiolense, Dirac, and other software.

I'm curious about REW convolution (with and without RePhase) to have MORE control over the process.

Unfortunately, I can't seem to find any substantial reviews of REW convolution, in-depth discussions of how it compares and contrasts to the other named solutions, and info on how to use it for more than 2 channel corrections.

 

[OCA]

I meant the google drive document linked to your YT video. I misspoke calling it a PDF.

I'm actually pretty far down the DSP rabbit hole and have a lot of experience with Audiolense, Dirac, and other software.

I'm curious about REW convolution (with and without RePhase) to have MORE control over the process.

Unfortunately, I can't seem to find any substantial reviews of REW convolution, in-depth discussions of how it compares and contrasts to the other named solutions, and info on how to use it for more than 2 channel corrections.

REW can create bit perfect convolution files for up to 1536kHz resolution in mono or stereo. The filters will be designed per channel at the end of the day. I don't fully understand what you mean by multichannel convolution.

Added: If you mean some Dirac ART like multispeaker solution, all of the software you have listed can create these filters very accurately but there are no multichannel receivers which open up their convolution engines to users. There's kind of a monoploy there reserved for Dirac and Audyssey.

 

[mrlawng9]

REW can create bit perfect convolution files for up to 1536kHz resolution in mono or stereo. The filters will be designed per channel at the end of the day. I don't fully understand what you mean by multichannel convolution.

Let's use a 2.2 system as a 'multichannel' convolution use case. I'd like to start with 4 individual raw speakers - Left and Right mains and two subwoofers. In this scenario, the mains and subs aren't equalized, crossed over (let's assume the mains are 2 way with an internal crossover that I won't touch - I'm only concerned about creating a crossover between the mains and the subs), or time aligned. The goal would be to use REW convolution to create crossovers, embed time alignment parameters (though this is not critical as it can be inserted into roon or HQPlayer manually), and to implement my desired house curve. The output would be 4 discrete convolution files that work seamlessly together - 1 convolution file corresponding to each speaker (and each requiring a dedicated digital channel - e.g. a 2.2 channel system).

Audiolense and Acourate do this and integrate all information into the convolution files. For example, crossover and TA info are embedded in the various convolution files.

Based on the great demo you've provided I have a pretty good idea how to accomplish most of this. What I don't know is how to embed TA info (again, not critical) and whether I have to do anything to make sure all 4 files play well together. I'm a bit stuck on whether there would be any additional trace arithmetic procedures / requirements to make all these files work together seamlessly.

*Added - I'm not targeting a receiver to implement these files. I'd use Roon or HQPlayer to create a 4 channel mapping and would need a multichannel DAC or multiple DACs. I have the hardware and have accomplished this in the past with available software. As I said, I'm curious about REW because it provides a lot of control over the process - even more if I use RePhase in addition. And, over the years I've become pretty savvy with REW. Audiolense and other solutions have STEEP learning curves and support can be tough. Using REW as a one stop shop is very desirable!

 

[AlfaNovember]

@OCA, your convolution YouTube tutorial was hugely beneficial for my system, and really opened my eyes to learning REW. Thank you.