MQA, DSP and "sound quality"

[Burning Sounds]

Yes, Acourate does partial correction: https://www.computeraudiophile.com/...re-walkthrough/?do=findComment&comment=232841
I am updating my book on a variety of DSP topics, including this one, but if you need the procedure sooner than later, Uli is very responsive on the Acourate forum.

Audiolense also does user configurable partial correction, being able to adjust the frequency and time domains independently. Audiolense also has a user configurable mixed phase target design, so this is not unique to Dirac...

Thanks for the link - I think I can work it out from that image. Yes, the Acourate forum has been very useful - found a fellow LX521 owner in Switzerland who helped me with the complex eq/xovers - I think Uli helped him initially.

I just want to try below Schroeder only out of curiosity as much as anything as I think I get more benefit out of Acourate in the midrange as much as anywhere.

 

[dallasjustice]

I’ve compared the predicted step response in Acourate and Audiolense to REW loopback. It’s always exactly the same. Dirac is a little harder to make the comparison because it uses the (useless) impulse response. Maybe @amirm is talking about the predicted frequency response? Each DSP software uses its own special way of filtering prior to inversion. The filtering is based on psychoacoustics. Therefore, each software reflects that psychoacoustic filter in their predicted response. However, Acourate does allow the user to choose different smoothing for the filter and consequently the predicted response. When you loopback using REW, REW doesn’t use the same psychoacoustic smoothing as the DSP software. So things will look a little different. IME, if you stick with 1/12 per octave smoothing in REW, things will line up very well with Acourate or Audiolense predicted result.

If anyone is attempting to integrate multiple subs with Dirac, you are wasting your time. It’s never a good idea to integrate subs without proper crossovers and delays. Would you buy a pair of speakers that didn’t have crossovers and weren’t time coherent between the drivers?(other than planar). This is especially true if you have ported R/L speakers.

If multi subs are properly integrated and well setup, I find it very hard to believe anyone would prefer 2Ch R/L setup over a multi sub setup. It would have to be a truly exceptional room; very large and very well designed. Even in that case, low frequency distortion will be much lower with multi subs.

 

[RayDunzl]

Noticed this listening position measurement a few days ago. Thought it was pretty.

dallasJustice's M2 setup, with 1-octave smoothing, left and right...

 

[Wombat]

And so it goes.

 

[Sal1950]

And so it goes.

First your money and then your clothes.

 

[hvbias]

Thanks for the link - I think I can work it out from that image. Yes, the Acourate forum has been very useful - found a fellow LX521 owner in Switzerland who helped me with the complex eq/xovers - I think Uli helped him initially.

I just want to try below Schroeder only out of curiosity as much as anything as I think I get more benefit out of Acourate in the midrange as much as anywhere.

Do you have any more information how Uli generated the crossovers? I am going to be using a large format constant directivity mid/treble horn with an active DSP crossover to a bass horn. The EQ needed for the CD horn should be fairly straight forward since the designer already includes settings for a hardware based DSP crossover but there are multiple ways to do the crossover for the midbass to CD horn. I haven't posted anything to the Acourate group yet... just lurked and done searches (and of course read Mitch's exceptional ebook ); seems like some people have taken to doing measurements outdoors for active crossovers.

If multi subs are properly integrated and well setup, I find it very hard to believe anyone would prefer 2Ch R/L setup over a multi sub setup. It would have to be a truly exceptional room; very large and very well designed. Even in that case, low frequency distortion will be much lower with multi subs.

Yeah it's really not about the speakers (traditional, not ones that can fix bass issues via DSP) being deficient in bass when audiophiles talk about disliking subwoofers; it's about non-venue sized rooms being deficient.

 

[Rodney Gold]

I was getting more of a drier pitch defined and more slammy and room pressurising bass with my current speakers without subs...which I preferred...and I listen loud ...full orchestra in the living room and all that...and am a bit of a basshead...

 

[Burning Sounds]

Do you have any more information how Uli generated the crossovers? I am going to be using a large format constant directivity mid/treble horn with an active DSP crossover to a bass horn. The EQ needed for the CD horn should be fairly straight forward since the designer already includes settings for a hardware based DSP crossover but there are multiple ways to do the crossover for the midbass to CD horn. I haven't posted anything to the Acourate group yet... just lurked and done searches (and of course read Mitch's exceptional ebook ); seems like some people have taken to doing measurements outdoors for active crossovers.

Sure, happy to help if I can. I have a Word file somewhere with the instructions for the LX521 EQ/x-over settings, so I'll need to take out the specific LX521 information and also make sure it makes sense as the original was in German. I'll PM you in a couple of days if that's OK.

 

[pirad]

I revisited Dirac after more than a year and dug up these:

Dirac manual page 17
Using the microphone positions indicated by the Dirac Live Calibration Tool™ will generally give you consistent results.
However, the microphone positions are not required to be in exactly these positions - if your listening environment looks different then
you may use a different set of microphone positions. Taking all measurements close to the sweet spot will generally not give optimum
performance because the microphone positions need some spread in order to acquire enough acoustical information about the room.

Elsewhere there is information that even the number of nine mic positions is "typical" or "approximate" (marked with tilde) . Dirac demands
only one must-do: the central position first measurement.


Dirac FAQ
3.6. Why is there no validation feature?
The predicted response may look too good to be true, but it is important to interpret the curves in the right way; the prediction curves show the response in exactly the point each curve is based on.
In theory the curve can be totally different just a small distance away from it, but in practice this is rarely the case. In order to avoid optimizing the filter for a really small area at the expense of the actual listening area, multiple measurements are used, taken in different points. Their average is the basis for the filter calculations. Together with the spatial robustness of the Dirac Live room correction technology this will result in a filter that sounds good in a large area. However, for a measurement to produce the predicted curve the microphone need to be in the exactly same position, otherwise the result will just be close to the prediction. It is also important to compare a validation measurement to the curve that corresponds to that measurement position and not the average curve.
If you want to measure the result of the filter you can either use some application such as HOLMImpulse, or you can select the Dirac Audio Processor as the output device on the sound setup page. If this method is selected it is very important to note that the Dirac Live Calibration Tool will disable filtering in the Dirac Audio Processor before each measurement, and you will have to enable it manually real quick. The reason the Dirac Live Calibration Tool disables filtering is that most measurements will not be for validation but for room analysis, and in this case any filtering has to be disabled.

It's not hard to see some wishy-washy talk and contradictions in the above. The conclusion (in blue) is that you can measure the calibrated curve, even with Dirac -- "if you want".
My take is Dirac hopes to avoid the situation where amateur users (vast majority of applications) accuse them of misrepresenting the filter effects shown in the simulated curve.
It would not be difficult to introduce simple "filter validation" function. Instead they direct to outside apps or a workaround which is not in the manual. These solutions are
definitely not for amateurs, and the assumption is the advanced users know what they are doing and understand how it all works.

 

[Rodney Gold]

Does it really matter ? At the end of it all , as all these packages have a subjective target curve applied , you just do what sounds best for you.
I found dirac to have a more diffuse wall of sound compared to trinnov which has uncanny 3d holographic imaging .. I could more readily suspend belief I was listening to hifi with it. It did not "shrink" the performance , if anything it gave much better and a more realistic width and depth of soundstage
The measurements are good to reveal room or other issues , but the measurements are somewhat useless to predict whether the correction is good to your taste .. maybe getting that "perfect" graph is more comforting to some .. who knows
As I use dirac in a minidsp hardware box , measuring the output is a little easier ..but I never bothered to do so
I did however find that not following the dirac order of subsidary measurements (ie doing it random by myself) gave me a result that skewed sound towards one channel for some reason or another..not sure if it was a user fault or diracs fault.

 

[Wombat]

I agree, choose what pleases you.

In the real world how do you get to audition the multitude of offerings. Retailers offer limited selections and other personal opinions are just that, personal. After all of these years there is no reliable description of variation from transparent performance to enable individual preferences to be predicted prior to mail-order purchase.

 

[pirad]

It matters as far as this forum's name is "audiosciencereview" and not "tellmewhatyoulike". One can have subjective opinions, but not subjective facts. The calibrated curve should be a fact confirmed by other experiments (quote from Paul Dirac ). At the end of the day I don't make graphs (I use them during the day) and rely on two pairs of ears of two Grammy winners in the best classical recordings category . I am fortunate enough to have been able to bring my system to the philharmonics concert hall where they made the awarded recordings. We play them. And they tell me if the sound is right. They made two main suggestions: set the subs LPF at 100Hz and play at lower volumes, the ones of a real unamplified orchestra. The biggest honor of my audio life is that they prefer my modest and affordable system to a world renowned 500k behemoth. And my system is two 2-way unassuming mains and a couple of subs plus an integrated amp. No MSG...err...DSP.
PS. Given that Trinnov is 50-100x Dirac price it better perform better
PSPS. Real Dirac quotes--really good!
3996-Paul_Dirac

 

[Wombat]

Of course the Dirac quotes are mainly subjective. Open to cherry-picking like most others, even the Einstein one in my signature.

 

[pirad]

This one dedicated to audiophiles:

It is more important to have beauty in one's equations than to have them fit experiment...

 

[Cosmik]

One can have subjective opinions, but not subjective facts.

Is it a fact that humans distinguish between direct sound and reverberant sound? Or not? In effect that they "stream" the direct and the reverberant separately? Or, putting it another way, that they "dereverberate" the sound by being able to focus their attention on the direct?

If it is a fact that they don't, then it makes sense to have an in-room 'target curve' because the listener hears everything combined, as a wash of 'colour' and you've got to match the colours to what the producers heard.

If it is a fact that humans do stream the direct sound separately, then it makes no sense to have a target curve - because the listener will hear the direct sound EQ'ed arbitrarily to compensate for the room's acoustic properties.

I don't know the answer to the question, but I do know that there is quite a logical case (plus subjective experience) for thinking that humans retain their ability to recognise sounds and their locations in heavily reverberant surroundings, which casts doubt on the validity of full range 'room correction'.

It also looks to me as though 99% of people never ask the question at all, including the purveyors of the software. Even if the software appears to improve the sound (as judged by people salivating at the prospect - not exactly unbiased!) it may be as a result of incidental partial compensation of non-neutral speakers (as well as novelty, artificial enhancement of detail within certain recordings, etc.). The speakers could be corrected more scientifically and methodically.

 

[Cosmik]

This one dedicated to audiophiles:

It is more important to have beauty in one's equations than to have them fit experiment...

Indeed, this sort of thing has occurred to me in the case of audio DSP 'room correction'. Quoting @dallasjustice above:

Each DSP software uses its own special way of filtering prior to inversion. The filtering is based on psychoacoustics. Therefore, each software reflects that psychoacoustic filter in their predicted response.

In other words, they're arbitrary, messy, undefined, subjective.

Compare that with the simplicity and beauty of "Use neutral speakers and no room correction".

(Sure, DSP is needed to get the neutral speaker, but this is a simplification of the system, because uncorrected speakers behave in terribly complicated ways, even if (in fact, because) the hardware is simple).

 

[dallasjustice]

I did a validation of Dirac years ago and posted it on WBF. That was before my number got retired there.
http://www.whatsbestforum.com/showthread.php?15212-Dirac-validation

 

[pirad]

I did a validation of Dirac years ago and posted it on WBF. That was before my number got retired there.
http://www.whatsbestforum.com/showthread.php?15212-Dirac-validation

I started reading your WBF thread and will probably have more questions, but easier for me to ask as I go, if you don't mind.
Post#1.
-I had no problem validating with REW on Mac. I have 1.2.0 version. Did you try later with REW?
-Why 18 measurements, 9 for each speaker? Do you mean measurement points or sweeps?

 

[Burning Sounds]

The weak point of LX521 is the MiniDsp 4x10HD. What's the point of testing MQA decoding DXD 352kHz studio codec and then feed it to MiniDsp mediocre 96khz DSP? Also the output stage is not so good, surprisingly RCA SE better than the balanced XLR.

Yes, I think the MiniDSP4x10HD was really intended as an inexpensive way to get people up and running with their LX521 build and the interface is easy to understand. I don't use MiniDSP4x10HD myself, preferring to do the x-overs/eq in JRiver or in Acourate itself. Your second point was why I asked the question - it sounded like you were still using LX521 so I wondered how you were getting output from a 2 channel DAC into a system that requires 8 channels.

 

[pirad]

Yes, I think the MiniDSP4x10HD was really intended as an inexpensive way to get people up and running with their LX521 build and the interface is easy to understand. I don't use MiniDSP4x10HD myself, preferring to do the x-overs/eq in JRiver or in Acourate itself. Your second point was why I asked the question - it sounded like you were still using LX521 so I wondered how you were getting output from a 2 channel DAC into a system that requires 8 channels.

But the minidsp 4x10hd has both digital and analogue stereo inputs, only
outputs are 8 analogue +2 digital ?? It also works as ADC/ DAC, so and so...