INTERACT FORUM

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1]   Go Down

Author Topic: convolution vs MC19's DSP  (Read 2969 times)

jacqlan111

  • Junior Woodchuck
  • **
  • Posts: 95
convolution vs MC19's DSP
« on: December 04, 2013, 07:07:08 am »

Hi to all
I've been working with MC's capacity for all my DSP needs for a couple of years now.  With the help of this forum's experts I was able to fix crossovers for a 3 way multichannel and multi-poweramp (6 ch + subwoofer = 7.1 settings).

In conjunction with REW measurements Tools, MC'DSP allowed me to mixed all channels to the right destination and choose the best filter slopes at XO freq. cuts (lowpass and highpass) and every other DSP correction filters.

I'm questioning the need to invest in any other software to manage my system.  I like to make my own correction filters one at the time and only where I judge that I must do this particular EQ correction.  Some EQ auto-corrections software produces wrong corrections in places for many reasons (room modes and HF narrow dips)  and in my opinion they should be managed with great care.
It seems that everybody or almost are using convolution engines.  Are they just a command list doing the same thing that we can do manually with MC19 Tools?

Maybe the filter types are limited here but is it that important?

I hope that guys around here that are way ahead of me will clarified these simple questions.

Thank you for your attention.

jacques

Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5234
  • "Linux Merit Badge" Recipient
Re: convolution vs MC19's DSP
« Reply #1 on: December 04, 2013, 10:19:57 am »

I run a pair of active bi-amped speakers, and like to do a lot of my customization manually as well.  I use JRiver's DSP for most of my EQ, but I use a free program called RePhase to do convolution in my system http://sourceforge.net/projects/rephase/ .  RePhase allows you to build custom convolution filters more or less the same way you currently can in JRiver (it's less flexible in some ways, but more flexible in others).  You can import a measurement from another program (Holm or REW) and use that as a starting point, or just start from scratch.  It has no automated features, everything must be configured by hand, which is the way I prefer it  :)

Most of the "flagship" convolution systems rely on automated measurement to do correction (although I think most of them also support manual correction).  But if you're looking for a primarily manual solution RePhase is pretty good, and, well, free  ;D

There's an ongoing thread about some of the automated solutions over here: http://yabb.jriver.com/interact/index.php?topic=85631.msg585671#msg585671

And a really excellent series of guides to one of the automated pay-systems (Acourate) over here (by forum user Mitchco): http://www.computeraudiophile.com/content/556-advanced-acourate-digital-xo-time-alignment-driver-linearization-walkthrough/
Logged

jacqlan111

  • Junior Woodchuck
  • **
  • Posts: 95
Re: convolution vs MC19's DSP
« Reply #2 on: December 04, 2013, 11:03:18 am »

I run a pair of active bi-amped speakers, and like to do a lot of my customization manually as well.  I use JRiver's DSP for most of my EQ, but I use a free program called RePhase to do convolution in my system http://sourceforge.net/projects/rephase/ .  RePhase allows you to build custom convolution filters more or less the same way you currently can in JRiver (it's less flexible in some ways, but more flexible in others).  You can import a measurement from another program (Holm or REW) and use that as a starting point, or just start from scratch.  It has no automated features, everything must be configured by hand, which is the way I prefer it  :)

Most of the "flagship" convolution systems rely on automated measurement to do correction (although I think most of them also support manual correction).  But if you're looking for a primarily manual solution RePhase is pretty good, and, well, free  ;D

There's an ongoing thread about some of the automated solutions over here: http://yabb.jriver.com/interact/index.php?topic=85631.msg585671#msg585671

And a really excellent series of guides to one of the automated pay-systems (Acourate) over here (by forum user Mitchco): http://www.computeraudiophile.com/content/556-advanced-acourate-digital-xo-time-alignment-driver-linearization-walkthrough/

Well, thanks, I feel less alone preferring manual work in this automatized room correction soft fashion?

In fact, I don't want an alternative to MC19 unless it brings a clear advance in manual correction quality.  If not, I'll wait until JRiver offers new filter types.  Though, I am not sure if other type of filter is better.

You seems to think there is no added value in others?

regards
jacques
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5234
  • "Linux Merit Badge" Recipient
Re: convolution vs MC19's DSP
« Reply #3 on: December 04, 2013, 11:09:52 am »

Well, thanks, I feel less alone preferring manual work in this automatized room correction soft fashion?

In fact, I don't want an alternative to MC19 until it brings a clear advance in manual correction quality.  If not, I'll wait until JRiver offers new filter types.  Though, I am not sure if other type of filter is better.

You seems to think there is no added value in others?

I wouldn't say that, there's definitely value in convolution (no matter whether you get there manually or automatically).  Convolution doesn't have to be a replacement for JRiver's DSP, they can work easily in tandem.  For example, I use both JRiver's built-in EQ and convolution at the same time.  

JRiver's built-in filters cannot change phase without also changing frequency response. There is no way in PEQ to, for example, unwrap the phase shift introduced by a crossover filter (without also undoing the crossover filter). I use RePhase to do phase manipulation that I cannot do in JRiver's PEQ, and apply that phase shift on top of the PEQ filters I use in JRiver.  

So, for example, I build a RePhase filter to fix my crossover phase shift (so phase is now flat through the crossover region) and to correct some of the phase shift near my port tuning frequency, but that's all.  I run that convolution filter in JRiver's convolution module in DSP Studio.  Then I copy and route the channels and apply frequency response correction in JRiver's Parametric EQ like normal.  So the convolution is just an extra layer that I add on top to correct phase issues.

I do it this way because convolution involves latency, and the more you're trying to accomplish in convolution the more latency is required.  By keeping as much as I can in JRiver's low-latency PEQ, I can use a fairly short convolution filter, which allows me to use it with latency sensitive content (video, etc.).
Logged

jacqlan111

  • Junior Woodchuck
  • **
  • Posts: 95
Re: convolution vs MC19's DSP
« Reply #4 on: December 04, 2013, 11:32:03 am »


JRiver's built-in filters cannot change phase without also changing frequency response. There is no way in PEQ to, for example, unwrap the phase shift introduced by a crossover filter (without also undoing the crossover filter). I use RePhase to do phase manipulation that I cannot do in JRiver's PEQ, and apply that phase shift on top of the PEQ filters I use in JRiver.  

So, for example, I build a RePhase filter to fix my crossover phase shift (so phase is now flat through the crossover region) and to correct some of the phase shift near my port tuning frequency, but that's all.  I run that convolution filter in JRiver's convolution module in DSP Studio.  Then I copy and route the channels and apply frequency response correction in JRiver's Parametric EQ like normal.  So the convolution is just an extra layer that I add on top to correct phase issues.

This is highly interesting.
With PEQ I am always fixing phases between drivers everytime I do any small modifications that affect the acoustic crossover region, and I do it every day (mucho hours) for 3 crossovers.  So I'm doing foreward and backward timing (delays) adjustments,  always in the first arrival portion of the signal but giving the phase matching the greater weight.

I have to try your solution. 
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5234
  • "Linux Merit Badge" Recipient
Re: convolution vs MC19's DSP
« Reply #5 on: December 04, 2013, 11:44:51 am »

This is highly interesting.
With PEQ I am always fixing phases between drivers everytime I do any small modifications that affect the acoustic crossover region, and I do it every day (mucho hours) for 3 crossovers.  So I'm doing foreward and backward timing (delays) adjustments,  always in the first arrival portion of the signal but giving the phase matching the greater weight.

I have to try your solution. 

It works best if your drivers are more or less time aligned already, then you can just do fine tweaking to level it out.  Depending on what kind of crossover topography you currently have it will be easier or harder to "linearize" the phase, but give it a shot and let me know if I can be of help.
Logged
Pages: [1]   Go Up