REW Measurements for Room Correction Systems (YPAO, ARC, Audyssey etc)

Very interesting topic. Measured in the same space? Need to knock the gain on the sub down 5 or 10 dBA on the Yammy. ARC looks to do a bit better, but that roll-off starts pretty low, I'd rather see 5 kHz and above boosted.

What is your set up for measuring with REW and your rig?

I'm getting ready to tweak my new room for acoustics and then move on to RTA the system. The setup is a new 7.1 screening room and 2.0 edit suite. I'm running a MacPro + MOTU HD Express to send HDMI video & 8 channels of audio to a Yamaha RX-A1020, feeding (7) JBL Control 5 monitors and a BIC PL-200. My audio inputs are thru a Focusrite Saffire 40.

I'm still trying to figure out if I should run REW on a 2nd computer so I can eventually route the REW signal to any monitor in the room thru the Saffire Mix Controls, mainly because REW can not see or access any of my outboard I/Os..

I have been temporarily running with YPAO calibration in the untweaked room while I have been getting gear on line. Then once I have all acoustic treatments dialed in I figured I would start over and use REW to manually set up the surround .

My 2.0 mix is on a set of Yamaha NS-10m monitors and I won't doing anything with their signal chain, just straight out of the Mac and into a Hafler amp. Only 2.0 consideration is acoustical panels to controlling some early reflection points.

I'd love any advice for how to best use REW with my workflow/set up.

Here's some data for Audyssey MultEQ XT32:
http://mehlau.net/audio/multisub_multeq_xt32/

mine are buried somewhere in the audyssey pro thread, i'll see what i can dig up...

How do your measurements compare to the Anthem Quick Measure ones?

I did but I'm not sure what you want to show?

Here'sre preamp out measurements showing the magnitude response of two exemplary MultEQ XT32 filters (L and sub):


Edited by markus767 - 10/23/12 at 1:52am

Looks like the YPAO hardly does anything

Before



Audyssey XT32



Hand dialed parametric EQ



I used to be really anti Audyssey but the XT32 is 'ok'. It still doesn't deal with modal ringing as well as a hand dialed parametric EQ that is precisely set to match the room mode's Q. Done properly this approach kills the modal ringing. BIG problem with Audyssey IMO is that it messes up what your speakers are doing by trying to apply an arbitrary target curve to the measurement above the modal region. Really in this region (>250Hz) we are correcting the speaker's response NOT the rooms. And far better here to measure the speaker nearfield ('anechoic') as then one can not only accurately correct frequency response issues but also phase and group delay issues. If your speaker is corrected nice and flat once measured nearfield and the response looks horrible at the listening position then you have a badly designed speaker with poor off axis response consistency and the only real solution is to passively treat the room with acoustic treatment.
Edited by Nyal Mellor - 10/23/12 at 8:03pm

+1 a manual calibration is still best if you have a knowledgeable calibrator. None of the auto EQs do a better job with tuning a room. The only exception to this is Trinnov's remapping feature not its EQ. Calibrator's can't remap but if the system is set up well it is not needed.

A lot of assumptions on how we perceive direct and reflected sound. There's simply not enough conclusive psychoacoustic data available to support authorative statements as to what is best. There are also uncertainties arising from the lack of proper standards in music production.

+1

My opinion is of course based on my knowledge and experience.

IMO correction in the non modal region CAN work if one knows the anechioc speaker response, both direct, early reflection and sound power. Harman use such an approach in their JBL Synthesis line. But they have measured each speaker in the line and when when the correction runs it takes into account the speakers anechioc measurements when generating the correction filters...I also believe that correction can work if you have speakers with constant or at least constantly increasing directivity.

Even that is just a crude approach. Steady-state measurements are blind to the important properties of the sound field in acoustically small rooms, i.e. its reflection pattern. Reflection angle, level, spectrum, delay is more or less ignored. But even if we were able to measure this in a meaningful way, what would we do with that information? There's simply not enough psychoacoustic data to provide a good answer.

David Rich on the Secrets site is an exception.

Quick Measure is indended for nothing more than assisting in speaker positioning, especially the sub. Multiple measurement positions are required for effective room correction even if there's only one listening position, otherwise peaks and nulls in that position will affect results adversely.

I don't know whether you're comparing smoothed graphs to unsmoothed but Quick Measure displays an unsmoothed graph from one measurement position whereas the final ARC graphs are smoothed, and use the average as measured from from five to ten positions.

By design, ARC doesn't waste resources trying to fix narrow troughs that are inevitable in-room and likely to be inaudible, therefore the displayed final graphs - the green curves - are indeed representative of what'll be heard. If someone sends us an ARC file we can view the unsmoothed curves for each position from 5 Hz to 24 kHz but this info is not very useful compared to what ARC already shows and could very well be misinterpreted (smoothing helps retrieve information and reject noise but that's another story).

On and off, posters in the Anthem threads have questioned why we don't include a verification phase where the correction is on while the speakers are swept for a second round to see whether the final response matches the predicted correction. For one, we know that it does match and for another it would lead to some people becoming obsessive about small displayed differences which make no real-world difference while listening. In other words, it would be a waste of time.

To qualify that statement for readers I dug up some old ARC v2 results sent by jayray, a regular poster on the Anthem thread who has run practically every ARC beta from the last few years. In this beta version he stumbled on a bug where the correction was left on during sweeps. The result is shown in the second set of graphs below, the ones without green curves. The two sets of measurements were taken nine days apart so temperature, humidity, and pressure might account for minor deviations but note how closely the red measured graph in the second set matches the target and the green corrected graph in the first set. Clearly, ARC is doing everything that's intended of it here, and the intentions were determined through many years of research using many test subjects in blind listening tests, bad speakers, good speakers, bad rooms, good rooms, at Canada's NRC, in-house, and in the field. The conclusion was that this type of correction benefitted all scenarios.

As a reminder, these are ARC v2 results whereas we are on v3, the recommended correction range is still 5 kHz since the "omni" mic becomes progressively directional above that, and jayray has since moved his sub by about eight feet resulting in its in-room corrected response being ruler flat to 100 Hz. Results are posted by him somewhere in the Anthem D2v thread in case anyone is interested. The dip in response shown here at around 115 Hz in the center channel isn't fully corrected because by design there's a 6 dB boost limit at any one frequency - ARC doesn't stress speakers by trying to make them play something they can't or shouldn't play.





Edited by Nick @ Anthem - 10/25/12 at 12:53pm

Waterfalls are pretty but often overrated because if the frequency response graph already shows a spike, it's already showing that there's a resonance.

I 'kind of' agree. You are right in what you say.

However...

If you know the anechoic response of the speaker - on axis, early reflections and sound power and you know what the response looks like at the listening position then you can make a pretty good guess at where to put correction filters. I agree it is not perfect but it MAY sound better than not using correction at all. Likewise if the speaker is well behaved off axis then correction MAY well improve things. Not everyone has great speakers and a well treated room.

There are many full range room correction systems. Some are more clever than others. Of course if you have very well designed speakers with nice off axis response, an appropriately treated room, then you may be better off with no correction.

Personally at this time I do not use or advocate use of full range correction. When I set up client's systems and they have Audyssey on their receivers I always turn it off. Anthem's ARC does good things in the bass but has some peculiarities in how it manages sub to mains crossover.

I think we will see some real advances in this area in the next few years.

First we would have to get a better understanding how humans perceive sound in acoustically small rooms, especially when that sound is radiated by loudspeakers.

You probably know Blackbird studios. One of their rooms is more or less one big diffusor. One of the effects is that first reflections are 30dB below the direct sound. From there it decays very smoothly. A common notion in acoustics is to reduce reflections by 25dB to make them inaudible. This is supported by quite a few psychoacoustic studies. Nonetheless they now use absorbers at first reflection points at Blackbird. How can that be?

D'Antonio asks "Why doesn’t the room sound anechoic with room reflections 30 dB below the direct sound?" in http://www.rpginc.com/docs%5CTechnology%5CPresentations%5CStudio%20Design%20From%20Mono2Surround.pdf
While his explanation sounds reasonable it's nothing more than conjecture. We obviously need to learn much more.

I agree with your points, though at some point one has to stop being theoretical and get practical. Not knowing enough about the psychoacoustics of reflections as you say does not stop people building studios that integrate different approaches of dealing with reflections. Nor I think should our lack of a complete understanding of psychoacoustics stop us from experimenting with different approaches to DRC. In the end the only real way to be sure about anything is to physically implement whatever the theories are and see if they work in practice. It is not as if the consequences of implementing DRC in a system are life and death (though hanging around on the audio forums you might consider it is judging by some people's attitudes!). I mean even with clinical drugs no one really knows what will happen the first time a compound is put into a human. Got to try it to see.

Some would argue that those advances are here now in the form of Trinnov and DiraLive.

Do you have any experience with either?

Trinnov is very pricey but there's a standalone 8-ch Dirac box for ~$2k that's supposed to be out right about now.

Have used Trinnov and at some point will likely become a dealer...their stuff definitely works...got to get my demo room up and running first...when it's done you should visit Noah!

Dirac have not had a chance yet to use in anger...afaik it's only in Datasat and Theta processors?

Also check out Realsoundlab CONEQ...a very interesting approach...sound power correction