Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Donations FAQ Calendar Mark Forums Read

PC Based Computer music servers, crossovers, and equalization

Impulse response, FFTs, deconvolution
Impulse response, FFTs, deconvolution
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools
Old 29th November 2011, 05:01 AM  
theSuede is offline theSuede  Sweden
diyAudio Member
Join Date: Aug 2006
One of the very basic points of 'room correction' is that it's really a three part problem. Both from a measurement and a "music listening" PoV.

You seem to be at home with the math involved, so that base is covered.

I'll just tell you about some of my personal experiences, since starting out with digital direct/reverbant sound correction some ten years ago (my - how time flies....). Split into the three separate points, exkluding the gory innards of the works:

1) Loudspeaker direct sound
This has to be considered the defining base point for everything from here on. Get this wrong, and it doesn't matter what you do in the room-correction - it's still a crappy speaker! I allways do this in a separated step (since the convolutions later on is easily overlayed on to the base point correction) - My reasoning, and my own understanding of this point is that:
Most of the sound from the lower midrange and up is psycho-acoustically dominated by the non-delayed direct radiation sound, the sound arriving at your ears without bouncing off anything - just a straight line through the air between the driver and the ear/mesuring point.

It's very important to "voice" the speaker in this stage, do the intial eq and corrections. I go very easy on the phase corrections, since this has already been partly taken care of in the digital crossover (6x 8192 step FIR convolutions). I seem to prefer a correction with very little pre-ringing present, some small time/amount is ok - but not like in a totally phase linear FFT filter. I have no scientific or other reason for this, it's just something I've gradually have grown to know.

I do this with the impulse average of three 2m distance measurements at (approximately) 0-15-30 degrees. As reflection free as possible - I gate the measurement, and limit the correction amount applied to the region below 300Hz to a very small amount. The correction for the region below 300Hz I get by close-mic measurements, that I multiply in on top of the farfield correction (that should tend towards zero correction below 300Hz). A nice, even frequency response here is just as important as a good room correction. Already after this point, you should notice a very marked improvement of the soundstage, otherwise something is very wrong. The only tricky point here is to do the switch from near-field to far-field correction correctly, and that's more often than not a manual labor (automating say a 12dB/oct correction switchover often leaves a level-difference between the two spliced parts, akin to a "shelf" EQ step - you have to get the curves to mix nicely, with constant sound pressure as a result).

(I'll split this to make it easier to read...) >>>
Old 29th November 2011, 05:18 AM  
theSuede is offline theSuede  Sweden
diyAudio Member
Join Date: Aug 2006
Part 2:

The actual physical placement of the speakers. You just can't get away with doing something stupid here... It's better to get as little NEED for correction as possible. The ways to acheive this are very personal and practical, the do of course vary with application - but please refrain from pointing your speakers straight into a concrete wall or something like that. Avoid early "hard" reflections, try to get a subwoofer region / room interaction that works as well as possible. There's a limit to what levels of correction you can ask from the next stage:

3) The actual room correction!
Since you allready have a well corrected SPEAKER, and a reasonably set up ROOM, the need for correction should be as low as possible.
You should by now allready be playing your measured test signals through the loudspeaker correction convolution (I export the test signals and play them via Foobar+convolution...).

The best measurement signal that I've found is the one you're allready using, the swept sine + deconvolution.

Now, mr Sbagrion's software gets to work again. I usually use a very soft correction setting, very close to the standard "soft" setup with lin.phase response. Since the actual speaker correction is allready made there (should) be very little, if any, pre-ringing in the correction signal. I've never found a "perfect for everything" setting here, even though I've helped quite a few friends to get started by now.

Allways save your base point loudspeaker correction! Don't overwrite & forget - or you'll have to do the near-field measurements all over again...

Convolve the base point impulse with the room correction impulse, and save the result as your intended correction signal...

Then listen to the setup...!


I have some small technical knowledge about the programming and the math works behind it, but I am in no way any kind of expert in the area. But if you have any specific questions, feel free to ask. I just can't promise to be able to help you...
Old 4th December 2011, 09:34 PM  
CopperTop is offline CopperTop  United Kingdom
diyAudio Member
Join Date: Feb 2009
Location: UK

Many thanks for that invaluable information. Are you using off the shelf tools entirely, or is some of it your own software?
Old 5th December 2011, 11:40 AM  
theSuede is offline theSuede  Sweden
diyAudio Member
Join Date: Aug 2006
Almost only OTS software, the only part that I've not assembled from existing parts is a slight rewrite of the sliding window filter in the room-correction (I use less of the direct signal, and concentrate on the room reverbation).

When I started out with DRC, I got the same first impression as you. It did sound like crap, most of the time. I guess the "big thing" is to find a solution, a general recipy, that improves everything slightly - without being to intrusive on the original signal. AND to have reasonably good speakers and a good setup as a starting point.
Old 5th December 2011, 06:30 PM  
CopperTop is offline CopperTop  United Kingdom
diyAudio Member
Join Date: Feb 2009
Location: UK

So how do you blend the near and far field speaker correction? Is this a case of mixing (you mention multiplying - is this in the time domain, or convolution?) two wav files, maybe having aligned the phase in the overlapping frequency regions 'by eye'?

I maybe did get a half reasonable correction last week when I measured the IR with a 2 minute sweep on each speaker and used the DRC 'normal' setup. Although I wouldn't leave it on all the time, it didn't offend me, and seemed to improve the 'separation' both between left and right channels, and also 'within the mix'. And I seemed to get deeper, clearer bass notes that really stood out on a couple of tracks. The problems, if there were any, were a cut in the top end brightness (maybe it's now flat!), and a hint of 'smearing' (pre-echo) at the start of certain sounds.

And there's another factor that I'd very much like your opinion on: my understanding of the way room correction works, is that when performing time domain correction, in essence the FIR filter is effectively following the 'dry' signal with a delayed version which cancels out echoes and reverberation. I'm wondering if there is a subtle problem-ette from this in that while the audio that reaches your ears is corrected, there is also a path through the floor which you feel in your feet, or posterior (if sitting down!). Is it possible that the correction may actually increase the vibration level you feel while reducing the audio level you hear, giving rise to a strange sensation of loud music being made quieter?
Old 6th December 2011, 10:23 AM  
theSuede is offline theSuede  Sweden
diyAudio Member
Join Date: Aug 2006
I blend by convolution. That's the rewrite, I make the initial pulse of the measured room response converge into a perfect 1-frame pulse to make the room-correction "let the speaker do its' thing".

The main problem with a "one-step" room correction is that you don't separate the integrated frequency response from the time-response - and they are not entirely dependent on each other. The "room response" sum with a long integration window does not have to be "flat" in any way to make the sound of the speaker balanced. The main coherency and "integrity" of the music/signal comes from direct radiation.

So the ideal is to have a good, flat frequency response within a ( 2/f + 10 ) ms time window - this is where you have the "speaker sound". The "reverbation" or "room response" after this initial pulse should not count very heavily towards frequency response correction, it just has to be dampened.

Basically - the frequency response of the system, in-room, does not have to be flat. It doesn't matter.
If you DO push a correction system into making the (long t) gated room response "flat", then you have probably made the speaker direct sound misbehave. You take a (maybe) perfectly ok speaker, and make it pre-compensate for errors that will happen LATER in the reverbation - to make the sum flat. Psycho-acoustically this is not a very good behaviour, and can make the sound very "wrong" even though general impression is that the sound is "balanced".

The FIR/convolution approach to room correction is just as you say depending on a cancellation effect, and this is also why stronger correction means more "hot-spot" sensitivity. Also because of this I prefer to adjust the ROOM for a good response in the 400-500Hz+ region, and limit the correction to the lower frequecies (that have less directivity and phase sensitivity as compared to real distance relations between origin/receiver).


Impulse response, FFTs, deconvolutionHide this!Advertise here!
Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Response to an impulse of a driver ginetto61 Multi-Way 32 27th December 2009 07:08 PM
Loudspeaker impulse response optimization thadman Multi-Way 7 18th April 2009 05:42 PM
Direct Impulse Response measurements jzagaja Multi-Way 33 4th October 2007 12:33 PM
impulse response vs. enclosure size East Subwoofers 5 23rd February 2006 01:53 PM
Impulse response differences Sjef Multi-Way 6 28th August 2005 09:43 PM

New To Site? Need Help?

All times are GMT. The time now is 10:28 PM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 68.42% / PHP 0.09%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio