Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mojave]

I can't find any specs either. The manual doesn't even list specs.

The outputs are split to a group of 8 and stereo outputs. The stereo outputs are the mains outputs according to the manual which means they are probably ASIO channels 0 and 1. The first two channels of the other 8 outputs could be duplicates of the mains, but I couldn't find any info in the manual. Also, there could be channel gaps between the mains and the other 8 channels. For example, they could be ASIO channels 4-11. This means in JRiver you would need to set the output to 16 channels and reroute to the ASIO channels you are using.

With the current 15% off at Musicians Friend it is $467 which is a nice price for 8-10 channels of D/A conversion.

[mwillems]

Anyone used the Motu UltraLite-mk3 Hybrid? It has Mic ins and 10 x balanced analog out channels & is reasonably priced. Cannot find any technical specs on the unit.

It looks like someone got some very basic tech specs from Motu over on the support board (see third post): http://www.motunation.com/forum/viewtopic.php?f=14&t=49197

That's not a particularly full tech description, but it hits some of the bases. 

[jdubs]

Which all got me thinking: would there be any interest around here in something similar to Mitchco's article, except focused on crossover and speaker optimization that can be accomplished using only JRiver and freeware?  Obviously it won't get you quite as far, but can get you a good distance.  I say that because I'm getting ready to start work on a pair of bi-amped bookshelf speakers that I plan to finish putting together and tuning over the next month or two.  Because I'll be going through the steps anyway, I could document the measurements/process as I go.  

100%, yes!!   

-Jim

[mwillems]

100%, yes!!   

-Jim

My current plan is to try and put something together over the holidays when I have a little time off work.  I'll probably do something general on measurement and speaker/room correction using JRiver and freeware first, and then do something separate on bi-amping and filter design later (because that's more technical and probably of interest to fewer people).

[jdubs]

My current plan is to try and put something together over the holidays when I have a little time off work.  I'll probably do something general on measurement and speaker/room correction using JRiver and freeware first, and then do something separate on bi-amping and filter design later (because that's more technical and probably of interest to fewer people).

Awesome - sounds great!  I have a strong interest in both subjects.

Thanks for taking the time to do this.

-Jim

[pschelbert]

Motu Specs

Hi

I have a Motu828mk3hybrid. There are as well no specs avaialble on the manual or their website.
Ask support for it, I got some specs asking directly. However, its not as complete what I got as the RME specs for example ( I own a UFX), where the specs are in the manual and very thorough.
I am wondering why Motu does not want to publish??

Peter

[eugle]

Hi, very interesting topic. I m gonna try Acourate, but could anybody tell me if my Behringer C1u condenser cardioid mic is suitable for measurement? I know, its not a dedicated measurement mic, but it is still not cheap, and has flat fr.... thanks

[mwillems]

Hi, very interesting topic. I m gonna try Acourate, but could anybody tell me if my Behringer C1u condenser cardioid mic is suitable for measurement? I know, its not a dedicated measurement mic, but it is still not cheap, and has flat fr.... thanks

If you have a calibration for it, it should work fine provided you make sure to point it directly at the speaker when measuring (true cardioid response falls off like crazy off-axis).  If it's not calibrated, you're likely to get some anomalies, and there's no way to predict how large or small they might be.  I've seen reasonably expensive microphones with truly impressive irregularities in the response (even when the "calibration" graph from the technical specs was nice and flat). 

If you're going to invest in something like Acourate, it's probably worth your time to invest in calibrating your mic (if it isn't already calibrated) or to buy a calibrated microphone (parts express sells a usable one for around $50). 

[eugle]

Thanks, unfortunately it isn't calibrated. I need to figure out a way to calibrate it...

[theoctavist]

two excellent primers

http://realtraps.com/art_measuring.htm

http://www.gikacoustics.com/room-eq-wizard-tutorial/

[TonyK]

As a long term user of JRiver and having read Mitchco’s review using Acourate I've decided to give room correction a go. Sorry for the slight thread hijack but with regard to various audio interfaces (Tascam, Steinberg, Focusrite etc.) How important is the software supplied with the various products for routing, or are all the various channel inputs and outputs available directly from within Acourate’s “ASIO LogSweep Recorder” using ASIO control, or do you need to set up channel routing from within the chosen audio interfaces software.

Sorry for the slightly daft question but I am new to this pro audio stuff.

[mattkhan]

Unless you have a 1:1 mapping from channel to speaker in so situations then a mixer app is nigh on essential ime. For instance I use the focusrite app for all measurements so I can use a 4 way wav to drive a 2.1 output (the 2 low passes are mixed to 1 output channel),

[TonyK]

Thanks that makes sense, but can you please confirm whether all the various channel inputs and outputs of say a Scarlett 2i4 are available directly from within Acourate’s “ASIO LogSweep Recorder” using ASIO control. I am just trying to envisage various possibilities and how all this works together.

[mattkhan]

Thanks that makes sense, but can you please confirm whether all the various channel inputs and outputs of say a Scarlett 2i4 are available directly from within Acourate’s “ASIO LogSweep Recorder” using ASIO control. I am just trying to envisage various possibilities and how all this works together.

I have a focusrite saffire pro 24 & acourate sees the asio channels that are backed by physical inputs/outputs IIRC. You are probably best off asking Uli directly via the user group to get the complete answer.

[mojave]

I have never needed to use the mixer app for any convolution or routing setup. It is all done in either Audiolense or JRiver for me. All inputs and outputs should be available from any ASIO software including Acourate. I've used many different multi-channel audio devices (Lynx, Steinberg, Tascam, Solid State Logic, etc.) and they all worked this way with ASIO.

[mattkhan]

I have never needed to use the mixer app for any convolution or routing setup. It is all done in either Audiolense or JRiver for me. All inputs and outputs should be available from any ASIO software including Acourate. I've used many different multi-channel audio devices (Lynx, Steinberg, Tascam, Solid State Logic, etc.) and they all worked this way with ASIO.

I think the way you do sweeps in acourate is perhaps a bit obtuse, things are definitely easier with mixing capability

[TonyK]

Thanks, I'm getting a much better idea how all this hangs together, all I need to now is spend some money.

[chrisr]

Great thread!  I just got Acourate and am waiting for my USB mic to arrive. I use a Playback Designs DAC that has its own ASIO driver (USB from PC to DAC). 

I read above that Acourate can only deal with 1 ASIO driver.  So it sounds like you're limited to a single device that has an A/D (for the mic) and a D/A for the output?  Is this correct?  In other words, I can't use my 1) USB mic that does its own A/D to provide a digital input to the PC, and 2) my external PD dac that has its own ASIO driver?  I would need something like a sound card that has an A/D and a D/A all in one?

[mattkhan]

Great thread!  I just got Acourate and am waiting for my USB mic to arrive. I use a Playback Designs DAC that has its own ASIO driver (USB from PC to DAC). 

I read above that Acourate can only deal with 1 ASIO driver.  So it sounds like you're limited to a single device that has an A/D (for the mic) and a D/A for the output?  Is this correct?  In other words, I can't use my 1) USB mic that does its own A/D to provide a digital input to the PC, and 2) my external PD dac that has its own ASIO driver?  I would need something like a sound card that has an A/D and a D/A all in one?

I believe it is possible, the method is briefly described earlier in the thread - http://yabb.jriver.com/interact/index.php?topic=85631.msg585663#msg585663

I think this conversation on the user group is the one that describes it more precisely - https://groups.yahoo.com/neo/groups/acourate/conversations/topics/5628

[chrisr]

Thanks Matt.  I'll check it out!

[Flak]

..............................
Wrt to measuring multiple listening positions, I did try that in other DRC software and found, at least in my setup, that taking measurements in multiple listening positions did not produce a better corrected response.  The best results I got was to triangulate on the listening position/speakers and take one measurement and let the speakers natural polar response deal with various other listening positions.

Hello Mitch,
I just stepped into this discussion and I'd like to comment from my admittedly biased point of view (Dirac Research)

I do not think that a single "right" positioning of the microphone, for one measurement only, does exist.
The reasoning is that measurements are different in different positions... even a few centimeters make a difference up to the point that what we measure at one ear is different from the other one.
So it's reasonable to correct the common behaviour of the different curves from different measurement positions, unless we want to listen with one ear only in a rigidly fixed single point in space

Here we see the eighteen curves of the nine measurements for left and right channels in the nine points that define the listening area together with the two curves (the lighter ones) that represent the averages of the respective L & R channels:

BEFORE CORRECTION:


AFTER CORRECTION:


b.t.w. some may wonder how is it possible that the averages are the ones shown in the images, seeing that the peaks are only a few dBs high while the dips are much lower... those are logarithmic scales (70 dBs plus 70 dBs add to 73 dBs and not 140)

Ciao, Flavio

[JimH]

Welcome, Flavio.  Nice to have you here.

[Flak]

Welcome, Flavio.  Nice to have you here.

Thanks Jim

I'm grateful for the opportunity of posting here
Flavio

[6233638]

Very interesting post Flavio, thanks.
And welcome to the forum!

[dtb300]

Great thread!  I just got Acourate and am waiting for my USB mic to arrive. I use a Playback Designs DAC that has its own ASIO driver (USB from PC to DAC). 

I read above that Acourate can only deal with 1 ASIO driver.  So it sounds like you're limited to a single device that has an A/D (for the mic) and a D/A for the output?  Is this correct?  In other words, I can't use my 1) USB mic that does its own A/D to provide a digital input to the PC, and 2) my external PD dac that has its own ASIO driver?  I would need something like a sound card that has an A/D and a D/A all in one?

Or you can use the Focusrite Scarlett to give you a single point/clock for input/output.  This device uses normal MIC input connections and not USB.

[mattkhan]

I do not think that a single "right" positioning of the microphone, for one measurement only, does exist.
The reasoning is that measurements are different in different positions... even a few centimeters make a difference up to the point that what we measure at one ear is different from the other one.
So it's reasonable to correct the common behaviour of the different curves from different measurement positions, unless we want to listen with one ear only in a rigidly fixed single point in space

fwiw the software mitch (and I) uses allows you to correct the speaker itself separately to the effect of the room so it can still involve multiple measurements, just not multiple seating positions. It also uses a frequency dependent windowing based technique to appropriate smooth the measured response before fitting that to your target curve. i.e. I think it is aiming to see through the position specific information so as to deliver a more robust correction. The windowing parameters are themselves tunable so you can choose to include more or less of the room response in your correction.

I don't know much about exactly how you use dirac or how it works. One thought that springs to mind is how you guard against the natural off axis response of the speaker skewing what you try to correct? are there some limits to the area you should measure so as to avoid this possibility for example?

Here we see the eighteen curves of the nine measurements for left and right channels in the nine points that define the listening area together with the two curves (the lighter ones) that represent the averages of the respective L & R channels:

are those the unsmoothed responses that are then (spatially?) averaged to get the composite view? and is the after view the theoretical after or results of measuring the outcome? what class of problems do you aim to fix above 500Hz or so?

[Mitchco]

Hello Mitch,
I just stepped into this discussion and I'd like to comment from my admittedly biased point of view (Dirac Research)

I do not think that a single "right" positioning of the microphone, for one measurement only, does exist.
The reasoning is that measurements are different in different positions... even a few centimeters make a difference up to the point that what we measure at one ear is different from the other one.
So it's reasonable to correct the common behaviour of the different curves from different measurement positions, unless we want to listen with one ear only in a rigidly fixed single point in space
...
Ciao, Flavio

Aloha Flavio!

Well, I have to respectfully disagree and have measurement proof :-)  I used REW, with the signal routed through JRiver/Convolution engine with the FIR filter engaged, to measure the frequency response of my right speaker at 6 different locations across a 6' x 2' grid at the listening position some 9ft away.  Basically the couch area:



The most ragged curve is measured with the mic the furthest away from the right speaker, i.e. back, left on my couch.  But adding the left speaker in makes up for it (i.e. averages it out).

From a time coherence perspective, looking at the step response of my right speaker, shows everything arriving at the same time measured over a 6' x 2' grid:



Again, both of these charts are measured results, not simulations.

For folks interested in what a good/bad step response looks like, a good explanation can be found here:
http://www.stereophile.com/content/measuring-loudspeakers-part-two-page-2
and
http://www.stereophile.com/content/measuring-loudspeakers-part-two-page-3
Note the reason for using a step response when looking at speakers time coherence versus an impulse response.  Short answer, a step response is more evenly weighted across the frequency range, whereas an impulse response is heavily weighted by high frequencies, usually the tweeter.

Audiolense also has multi-seat measurement/correction capabilities that I tried, but again, ended up that one measurement sufficed and no audible difference, to my ears, when I compared a single measurement correction versus a multi-seat correction when I AB the correction filters in real time through JRiver.  So having multi-seat correction, for me, is a non-feature.

To be fair, 500 Hz on up in my system uses constant directivity horns which are designed for even frequency response coverage across a 90 x 40 degree pattern, so this could be a contributing factor to why only one measurement suffices in my system.

However, for me, having the capability to create digital XO's with a wide variety of slopes, time aligning and linearizing individual drivers, in addition to digital room correction, are must haves.  There is a major audible difference between just room correction and room correction with digital XO, driver time alignment/linearization as measured and described in my article: http://www.computeraudiophile.com/content/556-advanced-acourate-digital-xo-time-alignment-driver-linearization-walkthrough/

Best regards, Mitch

[Flak]

Aloha Mitch  

Well, I have to respectfully disagree and have measurement proof :-)

Proof of what exactly?

From a time coherence perspective, looking at the step response of my right speaker, shows everything arriving at the same time measured over a 6' x 2' grid

Yes, they are similar for the first 1ms (the direct wave, as normally expected), after that they start to be different (the room)

Doing a single measurement compensation will give the best possible result in that point, but the result is unknown in any other points.
Doing a multi-point compensation will result in a compensation that is better on average in the measured points (you add more information). In your case where you fairly explained that your speakers dispersion characteristics are such that the direct wave response does not vary much with position, and the room response is similar for all positions, one measurement can be enough as it will be representative for a larger area... but this is not normally the case.

If one or both of these are not true, as usually happens, more measurements will be needed to avoid over compensating for the behaviour in that one measured point.

Jakob Agren explained that as follows:
"Mathematically it is possible to correct a room, perfectly, in a single point. For this case, only a single measurement in that single point is needed, and also any additional measurements will indeed not contribute at all, and if used, they will ruin the result in that one point.

However, it is important to note the conditions for this to be true. A single point indeed means just that, a single point, with point meaning a spot with no width nor height.

Now this is the theory, in practice we don't sit still, we move about, even if just small distances. Also we got two ears, located at different places, on average about 23 cm between them, so our ears are clearly in distinct positions. On top of this our speakers are placed in a room, introducing a multitude of reflections. The reflected sound will arrive in different points in time, with different amplitude, to each ear, so at any moment in time the sound will not be the same in these two positions. In the attached image two different measurements are shown, with the distance between them being 30cm. only. Now, which one to choose if I can have only one?



Any solution trying to address these issues is a compromise. Completely predicting the reflection patterns is impossible with a small number of measurements, so an educated guess will have to do. The information used come from measurements in and around the measurement position. If you only have a single point you are effectively guessing what the result will be at your two ears. More often than not, your guess will be incorrect because the wave pattern, especially at mid and high frequencies, is fundamentally impossible to predict.

When we take several measurements, our guesses about what's going on in between these positions, become better and better. As a rule of thumb, it becomes easier to predict the behavior in nearby positions the lower the frequency is, and vice versa. That's why a subwoofer correction can work OK in most cases even if just based on a single measurement, whereas a high-performance full-bandwidth optimization require more measurements in order to guarantee that we don't end up mistakenly making things worse at some frequencies. The real problem in room correction is to make the best possible estimate of what can be corrected and what can't. You need several measurements to do a good job at that."

I conclude by saying that you have done a great job and I imagine that the measurements in your room were better than average even before correction, not only because of the dispersion characteristics of your drivers, but also because I read that you "have added bass traps behind the speakers, thick carpet/underlay from speakers to listening position, broadband absorbers on the ceiling, and back wall to help reduce early reflections" and furthermore "that the stereo is offset in the room" to avoid that resonances double their impact.

In other words I do not see any contradiction between your measurements and what has been explained about the usefulness of multiple measurements.

Ciao, Flavio

[Mitchco]

Aloha Mitch  
Proof of what exactly?

Proof that multi-seat measurement/correction is a red herring. 

Doing a single measurement compensation will give the best possible result in that point, but the result is unknown in any other points.
Doing a multi-point compensation will result in a compensation that is better on average in the measured points (you add more information). In your case where you fairly explained that your speakers dispersion characteristics are such that the direct wave response does not vary much with position, and the room response is similar for all positions, one measurement can be enough as it will be representative for a larger area... but this is not normally the case.

If one or both of these are not true, as usually happens, more measurements will be needed to avoid over compensating for the behaviour in that one measured point.

Jakob Agren explained that as follows:
"Mathematically it is possible to correct a room, perfectly, in a single point. For this case, only a single measurement in that single point is needed, and also any additional measurements will indeed not contribute at all, and if used, they will ruin the result in that one point.

However, it is important to note the conditions for this to be true. A single point indeed means just that, a single point, with point meaning a spot with no width nor height.

Now this is the theory, in practice we don't sit still, we move about, even if just small distances. Also we got two ears, located at different places, on average about 23 cm between them, so our ears are clearly in distinct positions. On top of this our speakers are placed in a room, introducing a multitude of reflections. The reflected sound will arrive in different points in time, with different amplitude, to each ear, so at any moment in time the sound will not be the same in these two positions. In the attached image two different measurements are shown, with the distance between them being 30cm. only. Now, which one to choose if I can have only one?



Any solution trying to address these issues is a compromise. Completely predicting the reflection patterns is impossible with a small number of measurements, so an educated guess will have to do. The information used come from measurements in and around the measurement position. If you only have a single point you are effectively guessing what the result will be at your two ears. More often than not, your guess will be incorrect because the wave pattern, especially at mid and high frequencies, is fundamentally impossible to predict.

When we take several measurements, our guesses about what's going on in between these positions, become better and better. As a rule of thumb, it becomes easier to predict the behavior in nearby positions the lower the frequency is, and vice versa. That's why a subwoofer correction can work OK in most cases even if just based on a single measurement, whereas a high-performance full-bandwidth optimization require more measurements in order to guarantee that we don't end up mistakenly making things worse at some frequencies. The real problem in room correction is to make the best possible estimate of what can be corrected and what can't. You need several measurements to do a good job at that."

Rubbish, this myth needs to be busted.  You are showing full resolution frequency response graphs.  What you are not telling the good readers of JRiver is that our ears do not hear at this full visual display resolution. 

The prevailing science, that anyone can look up and educate themselves on, http://lmgtfy.com/?q=ears+critical+bandwidth+1%2F6+octave is that our ears critical bandwidth, while frequency dependent, is typically 1/3 to 1/6 octaves.

One link from the list above is from the Handbook for Sound Engineers and is representative of the science on the psychoacoustics of the critical bandwidths of the human ear: http://books.google.com/books?id=S4nBNZ_EJwwC&lpg=PA49&ots=eLJGYvbWyH&dq=ears%20critical%20bandwidth%201%2F6%20octave&pg=PA49#v=onepage&q=ears%20critical%20bandwidth%201/6%20octave&f=false

If one wants to get a proper "visual" of what our ears are actually hearing, then applying 1/6 octave smoothing to a full resolution frequency response display is a much more representative view.

Here is a full resolution frequency response measurement of my right speaker:



By applying 1/6 octave smoothing to the same measurement:



Is more visually representative of what our ears hear based on science.  The 6 x 1/6 octave frequency response measures I made across a 6' x 2' grid at the couch area in my previous post, based on one analysis measurement/correction, is practical proof of this science.

Taking your full resolution frequency response graphs above and applying 1/6 octave smoothing will show that that the two measures, relative to how our ears hear, are virtually identical. That's why multi-seat measurement/correction is a red herring. 

Either you don't know the science or you choose this way to try and differentiate your product.  Either way, the discussion is over for me.

[Flak]

I ignore Mitch's offenses and I will answer for this Forum's readers...

If one wants to get a proper "visual" of what our ears are actually hearing, then applying 1/6 octave smoothing to a full resolution frequency response display is a much more representative view.

Here is a picture showing 9 measurements smoothed at 1/6 octave (blue traces) and the average in all of these points (the black trace)
These measurements are 3 feet apart from the central measurement from a sofa in a standard room, distance to the speakers is about 9 feet or so.



Now, when you can choose only one of the blue ones, which one is it?
Is this blue trace representative for all of the other ones?
I think it should be fairly obvious that for this room in these positions there is not a magical one point that can be measured that will be a good fit for all of them.

Even at a very heavy handed 1/3 octave smoothing it is not hard to find frequencies where the points are more than 10dB apart, here it is:



Flavio

[JimH]

Keep it civil.

[mwillems]

At the risk of sticking my neck out:

@Mitchco- I absolutely agree with your statement of the scientific consensus regarding the resolution of our ears.  The hitch (from my perspective) is that my own in-room measurements, even with 1/3 or 1/6 octave smoothing applied, can vary quite a lot based on relatively small changes in position (a few feet).  Your measurements show strong consistency over a wide area, but (as you note) having a constant directivity horn and a well-treated room may be contributing to that consistency.

Most speakers that I've measured in untreated rooms were prone to significant positional variations in frequency response even with 1/6 or 1/3 octave smoothing applied.  I can observe (at certain frequencies) a six dB or more variation between two locations three feet apart in my living room with 1/6 octave smoothing applied.  And my living room setup has some features that should "tighten" variation: I have some small room treatments in place and I have a horn-loaded HF stage that controls directivity moderately well up to around 12KHz.  By contrast, when I measure the little cone speakers I have in my kitchen (with no acoustical treatment and laminate flooring), I can see nearly 10 dB variations at some frequencies with 1/6 octave smoothing.  

If I had based my living-room room-correction on a single position, I would have gotten poor results; the way I know that is because I tried it and the resulting correction measured very poorly (and sounded pretty bad) on the other end of the couch.  It's kind of a catch-22: a given speaker/room combination may not need multiple measurements or multiple-listening-position correction, but the only way to be sure whether your setup is one of them is to... take multiple measurements at different listening positions (as you did).    

I have a great deal of respect for you (you've been very helpful to me), and I always enjoy reading your posts. I just think it's a little strong to say that positional variation in frequency response is a myth or red herring. I agree 100% that measurements at different positions may not be relevant for every room, but they've certainly been relevant for me.  

P.S.-  I actually had been meaning to thank you; I was shopping for horn throat adapters a few months back and found a review you wrote of one on Parts Express.  I grabbed the adapter based on your recommendation and it was exactly what I needed!  It's a small world in online audio  

[mattkhan]

I read the argument not as *for* HF correction based on multiple measurement positions but rather as *against* HF correction full stop. Correcting based on a hypothetical average seems to reduce to applying random noise unless the intent is to shape the overall contour of the signal to a preference curve (in which case you could reasonably argue you should buy different speakers). The question then becomes what secret sauce is dirac deploying that is beyond simply correcting an average?

[Flak]

.....
Correcting based on a hypothetical average seems to reduce to applying random noise unless the intent is to shape the overall contour of the signal to a preference curve (in which case you could reasonably argue you should buy different speakers). The question then becomes what secret sauce is dirac deploying that is beyond simply correcting an average?

Hello Mattkhan,

I quote you when you say that correcting based on an hypothetical average is not necessarily a good idea, and that is not what Dirac Live does.
Even if advances have been made since then you may be interested in this 2009 document (eventually from page 7 "FAITHFUL STEREO REPRODUCTION") which had been partially presented at the 123rd AES convention:
http://www.dirac.se/media/12044/on_room_correction.pdf

Also thanks to the thread OP (6233638) for his welcome
Ciao, Flavio

[mwillems]

I read the argument not as *for* HF correction based on multiple measurement positions but rather as *against* HF correction full stop. Correcting based on a hypothetical average seems to reduce to applying random noise unless the intent is to shape the overall contour of the signal to a preference curve (in which case you could reasonably argue you should buy different speakers).

Whose argument do you mean (Flavio's or Mitch's)?  I didn't get that out of either argument, but I'll concede I may just have misunderstood  

But, to your point, assuming we're talking about room correction exclusively (vice speaker correction, where HF correction can be very useful), an average can be significantly better than "applying random noise" because the room may be exerting uniform effects at HF that can be usefully corrected, and that uniformity would be revealed by multiple measurements.  

An example, I have two speakers that are toed in and centered on the center seat of my couch, which means that my left seat is about 13 or 14 feet from the right speaker and about 9 feet from the left speaker, and the right seat is about 9 feet from the right speaker and 13 or 14 feet from the left speaker (i.e. they are symmetrical distance-wise).  The two speakers measure almost identically when close-miced on axis.  

The HF measurements at the three seats all show FR/phase nonlinearities that aren't present when the speakers are close-miced.  They agree with each other in many particulars, but disagree in some particulars.  Correction of some of the nonlinearities that agree among the seats has produced better (read flatter and more on target) results in all seats, but correction of the areas of disagreement based on a single measurement did not produce good results for me.  

Essentially, my room appears to introduce some HF effects that are uniform across my couch region, and some HF effects which are not uniform across the same region.  It may have something to do with the fact that my room is asymmetrical and my couch has it's back against a bay window about as wide as the couch (which, even with curtains, is pretty reflective).  

Bottom line: from where I sit (quite literally) a room can introduce effects at HF that are consistent across a given region of the room and which can (sometimes) be usefully corrected.  An average can be one way to cook out commonalities, although just looking at the multiple measurements can often tell you what you need to know.  

I do agree that HF room correction is significantly less useful/relevant than low frequency room correction, I just think it can have a place if you can find agreements across the region of interest in your room.

[Mitchco]

I ignore Mitch's offenses and I will answer for this Forum's readers...

Here is a picture showing 9 measurements smoothed at 1/6 octave (blue traces) and the average in all of these points (the black trace)
These measurements are 3 feet apart from the central measurement from a sofa in a standard room, distance to the speakers is about 9 feet or so.



Now, when you can choose only one of the blue ones, which one is it?
Is this blue trace representative for all of the other ones?
I think it should be fairly obvious that for this room in these positions there is not a magical one point that can be measured that will be a good fit for all of them.

Even at a very heavy handed 1/3 octave smoothing it is not hard to find frequencies where the points are more than 10dB apart, here it is:



Flavio

Again, not comparing apples to apples.  Your graphs are showing the uncorrected frequency response whereas my graphs are with the DRC correction filter in the signal path.  Point being, if the correction filter is doing its job, then there should not be large variations in frequency response across a couch listening area for example.   This should be especially true for Dirac since the claim is that multi-seat correction does matter.   Therefore, Dirac's measured results (with the correction filter in the signal path) should be as good if not better than what I show below, given my single measurement point to generate the correction filters.  So let’s see the “measured corrected" responses at multiple locations around the LP.

2nd, I don’t know what kind of smoothing Dirac uses, but it appears to be very different than REW (and other measurement software I have used).  Here are my six measures (with the correction filter in the signal path) at full resolution:



With 1/6 octave smoothing:



With 1/3 octave smoothing:



To compare apples to apples, let's use a vendor neutral solution.  Given that REW is free, runs on Windows or Mac, and has been around for almost 10 years and earned the respect of audio community, let’s use REW to compare apples to apples, with the correction filter in the signal path.  Therefore, when posting frequency response graphs (vertical range 45 to 105 dB and horizontal range 20 to 20 kHz with 1/6 octave smoothing and standard 500 ms window), the comparison will be a valid one.  Further, given the measurement files (.mdat) are portable, these can be exchanged and overlays can be produced for direct comparison.  Are you (and other members of the forum) willing to compare apples to apples?

[mattkhan]

My comment was just based on the graphs presented by Flavio. In isolation, they suggest to me that an average based correction would be wrong for all seats hence not v useful. i.e. the example might argue *against* the efficacy of a single measurement but it doesn't seem to do much to argue *for* multiple measurements.

The whole discussion seems a bit of a red herring tbh anyway given that Dirac doesn't correct based on that average either & so it comes back, for me, to my earlier comment of which class of problems does dirac seek to fix? It's latency budget seems relatively small next to other solutions (perhaps not much more than a typical minimum phase IIR solution if we're talking <20ms) so how much can it do in the time available?The linked paper is an interesting one that I've read before but doesn't really say much about Dirac itself other than it's not just a "simple" minimum phase system. Obviously I don't expect that to come on here and say exactly what they do, just would be interesting to read some more detail.

Fair point re your room, do you know the source of those consistent non linearities?

[mattkhan]

Are you (and other members of the forum) willing to compare apples to apples?

do you propose a particular set of mic positions or should I just take them across the seating area as I see fit?

[mwillems]

To compare apples to apples, let's use a vendor neutral solution.  Given that REW is free, runs on Windows or Mac, and has been around for almost 10 years and earned the respect of audio community, let’s use REW to compare apples to apples, with the correction filter in the signal path.  Therefore, when posting frequency response graphs (vertical range 45 to 105 dB and horizontal range 20 to 20 kHz with 1/6 octave smoothing and standard 500 ms window), the comparison will be a valid one.  Further, given the measurement files (.mdat) are portable, these can be exchanged and overlays can be produced for direct comparison.  Are you (and other members of the forum) willing to compare apples to apples?

Mitch, I'm game. I usually take my measurements in Holm, but I'll retake my measurements in REW (it might take me a few days).  

I have one suggestion and one question

The suggestion: a slightly finer vertical scale. With a 60dB scale (i.e. 45 to 105dB) it's very hard to see whether a given variation is say 3dB or 5dB and that's a big difference.  Would you be up for a 40dB scale instead (say 60 to 100)?  With the smoothed graphs everything should still be visible for most folks, right?

EDIT: The question: I ask this knowing that I don't have a good answer, and hoping to hear your thoughts. How will comparing the corrected response give us a truly apples to apples comparison if we're all measuring different speakers in different rooms?  Couldn't anyone's better result be explained by a different room or a speaker with better directivity, rather than by better correction methodology/technique?  I mean if someone has a 40 by 60 foot room with the speakers 15 feet away from the listening position, they're going to probably see much less variation over a six foot area than someone in a 10 by 12 room with the speakers six feet away.  

Not sure what to do about that, and I'm still up to compare notes.  Maybe the way to get at that is to describe the room, the speakers, and the distances of the listening positions to the speakers/walls?  

Regardless, I'm definitely game, and I'll try and get something posted over the weekend.

Fair point re your room, do you know the source of those consistent non linearities?

@ mattkhan- I don't, the main effects I saw are between 2KHz and 3KHz.  My best guess is the bay window, it's literally about 6 inches behind the entire listening position, which is about a wavelength at 2.2KHz.  

[Mitchco]

At the risk of sticking my neck out:

@Mitchco- I absolutely agree with your statement of the scientific consensus regarding the resolution of our ears.  The hitch (from my perspective) is that my own in-room measurements, even with 1/3 or 1/6 octave smoothing applied, can vary quite a lot based on relatively small changes in position (a few feet).  Your measurements show strong consistency over a wide area, but (as you note) having a constant directivity horn and a well-treated room may be contributing to that consistency.

Most speakers that I've measured in untreated rooms were prone to significant positional variations in frequency response even with 1/6 or 1/3 octave smoothing applied.  I can observe (at certain frequencies) a six dB or more variation between two locations three feet apart in my living room with 1/6 octave smoothing applied.  And my living room setup has some features that should "tighten" variation: I have some small room treatments in place and I have a horn-loaded HF stage that controls directivity moderately well up to around 12KHz.  By contrast, when I measure the little cone speakers I have in my kitchen (with no acoustical treatment and laminate flooring), I can see nearly 10 dB variations at some frequencies with 1/6 octave smoothing.  

If I had based my living-room room-correction on a single position, I would have gotten poor results; the way I know that is because I tried it and the resulting correction measured very poorly (and sounded pretty bad) on the other end of the couch.  It's kind of a catch-22: a given speaker/room combination may not need multiple measurements or multiple-listening-position correction, but the only way to be sure whether your setup is one of them is to... take multiple measurements at different listening positions (as you did).    

I have a great deal of respect for you (you've been very helpful to me), and I always enjoy reading your posts. I just think it's a little strong to say that positional variation in frequency response is a myth or red herring. I agree 100% that measurements at different positions may not be relevant for every room, but they've certainly been relevant for me.  

P.S.-  I actually had been meaning to thank you; I was shopping for horn throat adapters a few months back and found a review you wrote of one on Parts Express.  I grabbed the adapter based on your recommendation and it was exactly what I needed!  It's a small world in online audio  

mwillems, I appreciate your view and the respect is mutual.  The Eminence horn throat adapter is the best I have come across and glad my review was helpful.

You will have to pardon me if I am a bit tough on vendors that make claims, but provide no proof or evidence of said claim.  In this case, Dirac makes the claim that multi-seat correction does matter, but does not back up the claim with any "independent" acoustical measurements of the corrected response.  While I use Acourate to generate the correction filters, I use REW to "independently" verify the correction is indeed doing what it is supposed to be doing.  As you can see, I have posted a response to see if Flavio is willing to compare apples to apples.

My room is less than optimal, has minimal, but carefully placed room treatments, and my speakers are based on a 1958 design.  It is a far cry from the studio/control room environments I have worked in which one can see some of the rooms and acoustical treatments here:  http://www.computeraudiophile.com/blogs/mitchco/hear-music-way-it-was-intended-be-reproduced-%96-part-5-67/

If one can achieve +-5 dB across the listening window, then one has achieved what most control room monitoring achieves for producing the music we all listen to and is good enough.  I achieved that with one measurement used for the correction (as have many others).  In fact, I think Acourate's DRC is so good that I can be near or within that tolerance, even at a full resolution using REW's averaging (see button with red square) the 6 full resolution responses:



I believe there are more important contributors to my results than just room correction alone.  Using Acourate's digital XO is an important factor for the reasons described in this paper: http://files.computeraudiophile.com/2013/1202/XOWhitePaper.pdf and opens the door to driver time alignment and driver linearization.  That's why I posted the step response in an earlier post as I believe time coherence is as important as frequency response.  For example, here is the step response of my Lynx Hilo Ad DA converter:



Virtually text book response and used as a reference point.

Step response of my time aligned (but not linearized) and DRC right speaker (again achieved with a single measurement used for correction) at the LP:



Comparing that to the ideal step response in the Stereophile articles I linked to earlier, it is as good as it gets.

For comparison, here is the same speaker, but the difference is no time alignment or DRC and using 3 way passive XO:



Massive difference, not only measurement wise, but also very audible. For me, this is a much more important factor then quibbling over a few dB from a frequency response perspective and yet another reason why I think multi-seat correction is a red herring.



 

[Mitchco]

Mitch, I'm game. I usually take my measurements in Holm, but I'll retake my measurements in REW (it might take me a few days).  

I have one suggestion and one question

The suggestion: a slightly finer vertical scale. With a 60dB scale (i.e. 45 to 105dB) it's very hard to see whether a given variation is say 3dB or 5dB and that's a big difference.  Would you be up for a 40dB scale instead (say 60 to 100)?  With the smoothed graphs everything should still be visible for most folks, right?

EDIT: The question: I ask this knowing that I don't have a good answer, and hoping to hear your thoughts. How will comparing the corrected response give us a truly apples to apples comparison if we're all measuring different speakers in different rooms?  Couldn't anyone's better result be explained by a different room or a speaker with better directivity, rather than by better correction methodology/technique?  I mean if someone has a 40 by 60 foot room with the speakers 15 feet away from the listening position, they're going to probably see much less variation over a six foot area than someone in a 10 by 12 room with the speakers six feet away.  

Not sure what to do about that, and I'm still up to compare notes.  Maybe the way to get at that is to describe the room, the speakers, and the distances of the listening positions to the speakers/walls?  

Regardless, I'm definitely game, and I'll try and get something posted over the weekend.

@ mattkhan- I don't, the main effects I saw are between 2KHz and 3KHz.  My best guess is the bay window, it's literally about 6 inches behind the entire listening position, which is about a wavelength at 2.2KHz.  

mattkhan and mwillems, I propose that we take 6 measurements over a 6' x 2' grid to cover a typical couch area, whether sitting back in the couch or sitting upright, center, left, and right positions, if you know what I mean.

REW recommends the following for posting measurement graphs: http://www.hometheatershack.com/forums/rew-forum/934-please-read-posting-graph.html which is what I have already recommended.  However, given that the file is portable, we can slice and dice at any vertical and horizontal scales as required.

mwillems, wrt your question on various room sizes/absorption coefficients, speakers directivity indexes, etc, given that we are all correcting to a similar target frequency response, the results actually should be very close.  After all, I would argue that this is the raison d'ętre behind DRC in the first place :-)  And would be a good measure of how effective a particular DRC software is working.  Perhaps we can agree on a target response of say flat to 1 kHz, and using 1 kHz as a hinge point, a straight line to -6 dB at 20 kHz. Or something close enough to this as what we really want to compare is the variability across the listening area that tells us how well the DRC software is working.  Make sense?

[mwillems]

I think I've got it; thanks for the detailed answer (both of them).

I'll try and get my ducks lined up

I believe there are more important contributors to my results than just room correction alone.  Using Acourate's digital XO is an important factor for the reasons described in this paper: http://files.computeraudiophile.com/2013/1202/XOWhitePaper.pdf and opens the door to driver time alignment and driver linearization.

On that you are preaching to the choir.  JRiver is my crossover (using rephase to linearize the crossover), and I've definitely seen the benefits of digital XO's and time alignment both in my step response and in improved distortion performance.  Dealing with the crossover made a much larger difference for me (both in terms of improved measurement and perceived audio quality) than room correction.

[Mikkel]

Things to consider:
- Add speaker brand to the post
- Make one close-range measurement of a speaker to get a response that above 300hz is "somewhat" anechoic. This will make it possible for us to evaluate the speaker-room interaction. Or alternatively choose a very short window and post the graph for that.

Best of luck! I will enjoy it from the sideline.


Best regards,
Mikkel

[Flak]

Again, not comparing apples to apples.  Your graphs are showing the uncorrected frequency response whereas my graphs are with the DRC correction filter in the signal path.  Point being, if the correction filter is doing its job, then there should not be large variations in frequency response across a couch listening area for example.   This should be especially true for Dirac since the claim is that multi-seat correction does matter.   Therefore, Dirac's measured results (with the correction filter in the signal path) should be as good if not better than what I show below, given my single measurement point to generate the correction filters.  So let’s see the “measured corrected" responses at multiple locations around the LP.

I submit to you a thought experiment:

Assume you have a single perfect speaker with a single input in a perfect room such that in two points in a couch it measures flat from DC to Nyquist/2.
That is, the speaker is perfectly flat in the two points for all frequencies.
Now we apply a filter before the speaker, let's say the filter applies a 3dB notch at 500Hz.
What do you expect the response to be in the two measured points after applying this filter?
The answer is, both points will be perfectly flat, except for a 3dB notch at 500Hz.

Now assume the original response was flat but for a 3dB notch at 500Hz in point 1 and a 3dB peak at 500Hz in point 2.
We only measure point 1, and design a filter that is flat but for a 3dB peak at 500Hz to cancel out the notch in point 1.
After applying this filter, the response in point 1 will be flat, how will it be in point 2?
Did the filter do what it was supposed to? Maybe.
Did the filter have enough information to fix the issues in point 1 and point 2 at the same time? No.
What would have been the preferred filter when taking both points into account?
Flat if point 1 and point 2 are equally important.
Can a single filter acting on a single speaker behave different in different points? No.

If we agree on the above, then we can also agree that there is no filter acting on a single speaker that can remove variations between different points (short of the special case of the 0 filter).
All you can hope to achieve is a filter that does the best possible job in all points at the same time and accept it is a compromise.
If the room varies a lot, having too few measurement points will give you a filter that is not representative for your listening area since there was not enough information at design time.
If your room has a nice and uniform response this will not be an issue.

Flavio

[mattkhan]

I submit to you a thought experiment:

fwiw I think this link is a nice, simple summary of that experiment

[Mikkel]

Hello Flavio,

I see no point in disagreeing in what is logical. One or multiple filters cannot treat the effect of room response in a differential manner. It can only take into account what is known and then try to avoid doing excessive damage in other listening positions while correcting at the main listening position (of course, the algorithm tries to correct the other positions as well but gives primacy to the main listening position). This is of course a good thing, no doubt. Not taking data from other positions into account there is a chance of making the sound worse at the other positions. If it has any practical relevance is of course another matter (see below).

Anyway, in some situations multi-seat correction may not yield much better results than single-seat correction for the reason as I just mentioned: Mainly because the main listening position has priority in the correction procedure. The corrections in all other places than the main listening seat may thus be minimal (or they may not) because they are constrained by the effects of the correction of the main listening position

I don't doubt what you are saying but it would still be interesting to see the differences in correction between single vs multi-seat correction, since the changes may be subtle (or they may be big depending on the sound at each of the measurement positions).

Btw, I do multi-seat correction. All things being equal, it is the better option - even if doing multi-point correction within a small area similar to the area just around your head. More (good) data --> better results... up to a certain point of course.


BR Mikkel

[mattkhan]

My contribution is here - https://drive.google.com/file/d/0BxdmSMpV-t3GRTFFa18tdWJKME0/edit?usp=sharing

I couldn't manage 6 positions as I couldn't work out a way to get the mic into the 6th position without being in the way of the speakers. Floor plan is attached, naming convention for the measurements is

<mic position>-<speaker>-<DRC status>

so 1-L-On means mic on the far left of the pic, L speaker with DRC (provided by acourate hosted by jriver) on, acourate is a single measurement system and I use position 2 for measurements.

Speakers are MK MP150 Mk2
Attempts at close mic measurement can be found in this post - http://yabb.jriver.com/interact/index.php?topic=88942.msg614015#msg614015 - it's a bit tricky though as there are 3 tweeters that have different crossovers for dispersion control reasons.

My comments;

- room is completely untreated except for normal furnishing, it's a lounge home cinema system I use for music
- the amount of correction for my sub is quite large, the results without correction are therefore quite meaningless so I would ignore <100Hz
- it seems I wasn't super careful with mic orientation so you might want to ignore >15kHz too
- there is certainly evidence of suboptimal correction in the other locations <1kHz
- one might expect the system to sound rather bright without correction as there is consistently a dip in the low kHz range and then it rises up again towards 10kHz before rolling off, in contrast the "on" view consistently rolls off as per the target curve
- subjectively... it sounds vastly superior with correction on, both for stereo and for surround

I've attached a graph (averages.jpg) showing the average of the 5 positions at 1/6 smoothing for the L and R with DRC on and off.
There are also some more graphs (1/6 smoothed) showing the individual collection of L and R measurements with DRC on and off.

FWIW I know the rationale of taking multiple measurements but I did take a conscious decision to optimise for a single seat (mine). Position 3 is where you go for a lie down during a film and is just outside the axis of the L speaker, attempting to account for that seems a waste of time. Position 1 is tucked into a corner and has always been rubbish so if you care about sound you don't sit there. Positions 4 and 5 would involve leaning forward so file under not going to happen. Despite all this, my wife reports that it sounds great from both those positions & much improved over no correction so what do graphs know  

[AudioVero]

I like to step into this discussion and to tell a bit about my point of view:

I have seen Flavio's argumentation about multipoint measurements already at other forums. The arguments for multipoint are repeated again and again like a prayer mill. And it seems to be so very convincing. The pictures look so nice. But IMHO it is nothing else than marketing.

Flavio knows a bit more of course. We can see this by his thought experiment

So to learn more about the truth it would be nice to see only three curves instead of the 18 curves example:
1. the curve with the deepest suckout at e.g. 40 Hz (or any other frequency)
2. the curve with the least suckout at the same frequency
3. the averaged curve

Let's assume some arbitrary values. The average is 0 dB at 40 Hz (we just pick one frequency for discussion). Curve 1 shows -20 dB and curve 2 shows 3 dB.
If the target is 0 dB this means that the correction is also 0 dB as the average is already there. So listening point 1 still has -20 dB after correction, point 2 is still 3 dB.

But how is the average calculated? There are many different algorithms to calculate an average. So another algorithm may lead to an averaged curve with -6 dB at 40 Hz. To reach the 0 dB target the correction thus will boost by 6 dB. So at the listening position 1 we get -14 dB, which is an improvement and at position 2 we get 9 dB , indeed a deterioration.

So we simply learn that one correction will never fit to all positions. And indeed we apply only one correction filter.

Now let's assume we do only one measurement. And by chance we get the result with -20 dB suckout at 40 Hz. Then a target of 0 dB will lead to a boost of 20 dB (in case of primitive correction approach without boost limitation). And we already feel and know that this boost will lead to the worst correction. It may possibly sound perfect at the single nailed position but will cause a room roar elsewhere.

If we look a bit closer to the averaging of the curves in logarithmic domain we will learn that the averaged curve is more located at the upper part of all curves. It's just maths. See the 18 curves example given by Flavio. So the averaging simply spoken is nothing else than a protection against a correction overboost.

But is the averaging improving the listening result? I'm convinced that the averaging of measurements leads to an average listening quality but not top quality.

IMHO there also is an inherent flaw in the design. The example frequency responses show a steady state result (the time information is lost). We reach the example suckout of -20 dB at 40 Hz by playing a continuous sine wave. But music is usually not steady state. We can play the logsweep signal reversed. The measurement will result in the same frequency responses. But obviously it sounds different.

Acourate uses a much different approach. It looks at the sound waves arriving at the measurement position also under time aspects. In our 40 Hz example Acourate investigates in the sound wave arriving at the microphone and checks for time behaviour, transients and steady state. As the correction always is focusing on the direct sound and a short time around the direct sound (keyword: frequency dependent windowing) the steady state is much less important. Think about the sound wave radiated from loudspeakers and arriving at the ears. Is the soundwave arriving and passing at two positions of 30 cm distance really so much different as Flavio tries to convince us?

BTW it is possible to average multiple measurements by Acourate. You can even calculate the upper enevelope on multiple measurements and derive a correction if you like. Of course I have also studied all these methods during the development of Acourate. But to my knowledge up to now no Acourate user has ever been forced to do multipoint measurements and corrections.

My resume: multipoint measurements and averaging simply help to protect against over-correction. Nothing else. It does not necessarily lead to the optimal results, the averaging algorithm may be just arbitrarily selected. It is more important to look what happens in reality. A correction based on steady state signal analysis cannot be the right method.

I hope this helps

Uli

[Mikkel]

Dear Uli,

Thanks for contributing to the discussion. Your points are of high quality but I don't see them contradicting the (theoretical) point made by Flavio.

The important contribution in your post is frequency-dependent windowing. This is used by (at least) Audiolense, your product and Audyssey.
Since you would normally use a bigger window for low-frequency correction (due to the larger wave lengths I presume?), variation (would be my guess) will show up because it is impossible to keep the speaker-room interaction out of the data. Since this part of the frequency band is not directional in our typical small rooms I think it makes good sense to use multiple measurements. As soon as travelling up the frequency band the direct sound is a function of the radiation pattern and hence the quality of the speaker, hence multipoint correction serves no important purpose (since time-domain correction only applies to the main listening position anyway).

I don't know how other products work, but with Audiolense one can set a tolerance as to the variation for the main listening position (I would expect Acourate to provide something similar). As you mention, multipoint measurements serve the purpose of not overcorrecting. It equally serves the purpose of trying to find the best balance between the quality of the main measurement position and the remaining measurement positions. In some cases this may possibly provide better results overall given the tolerance for correction for the main measurement position is large enough to accomodate the correction required elsewhere and small enough not to severely deteriorate the quality at the main measurement position.

As I mentioned in my other post, though, in other situations the variation in frequency response may be too big between various positions, meaning that in real life listening it makes little improvement for the total listening area - possibly a slight decrease in quality at the main listening position. But anything that matters? Only measurements (or simulations) combined with double blind listening tests can tell.

Anyway, I'm not expert in all this but know just enough to sort of understand how it all works (but only just).

My point here is though: Mitcho and Flavio pull the discussion into two corners instead of focusing on the cases in which both positions have their right.


Best regards,
Mikkel

[AudioVero]

Mikkel,

multipoint measurements are quite common today. See Dirac, Audiolense, Lyngdorf, Audyssey.
I do not darn them, anyway they are better than the first approaches of room correction in the old days. Because they help to avoid overcorrection. That's it.
Indeed it is not complicated to code a multipoint measurement. You can erect big buildings around this theory and add weigthing methods and fuzzy logic.

But I definitely contradict to statements that single point measurements are wrong. You have of course to know how to interprete the measurement. IMO good skills in maths is not sufficient. You have to understand a bit more about the nature of things.

I do not think that a single "right" positioning of the microphone, for one measurement only, does exist.
The reasoning is that measurements are different in different positions... even a few centimeters make a difference up to the point that what we measure at one ear is different from the other one.
So it's reasonable to correct the common behaviour of the different curves from different measurement positions, unless we want to listen with one ear only in a rigidly fixed single point in space

Such a statement tells me that someone does not really understand. How much does the sound change when you listen to e.g. a conversation in your room and move your head... even more than just a few centimeters?

BTW frequency dependent windowing is just a part of the process. And of course the speaker-room interaction is part of the game. Otherwise we would just need to correct the speaker and it would play nicely everywhere.
Keep in mind: listening to music is NOT equal to viewing frequency responses ! Neither to a single response nor to a bunch of them. 

Uli