EQing only for MLP - REW vs MSO vs Dirac?

[ppataki]

assuming you are using min phase XO's

I guess if one (like me) uses linear phase crossover then this is a non-issue, isn't it?

Another question: if we use minimum phase, symmetrical 24dB/octave Linkwitz-Riley crossover, then shouldn't phase be kind of linearized by definition? (at least in the crossover region)
It would be great to shed some light on the details of this topic
Thank you

 

[markstein]

Hi,

I am using MSO for subwoofer optimization (2 mono subs) and additional REW assisted careful manual eq up to approx. 500Hz using the MMM method (all IIR filters).

The result of my manual eq sounds much better - to my ears - than Dirac generated filters I tried with my former miniDSP SHD.

I have also the software Acourate Pro which generates high quality FIR filters, however, relies on single point MLP measurement which I do not prefer.

 

[Keith_W]

I guess if one (like me) uses linear phase crossover then this is a non-issue, isn't it?

Another question: if we use minimum phase, symmetrical 24dB/octave Linkwitz-Riley crossover, then shouldn't phase be kind of linearized by definition? (at least in the crossover region)
It would be great to shed some light on the details of this topic
Thank you

Linear phase XO's do not introduce any ADDITIONAL phase rotation. It is important to remember that drivers themselves are minimum phase, meaning they will rotate phase at the extremes of their bandpass. It is possible to compensate for this phase rotation with a linear phase XO, e.g. the new version of Acourate has a feature that allows you to do just that.

Re your second question I am not sure what you are asking. The final phase rotation is a convolution of the electrical XO AND the driver's own amplitude and phase response. So ... if you take (say) a 24dB/oct LR XO with a corner frequency of 100Hz, and you use (say) a driver which has a LF roll off starting at 100dB of 6dB/oct; the result is a 30dB/oct slope.

Each order of filter, or pole, rotates phase by 90deg. So although a 4th order LR has 360deg phase rotation (meaning it sums to flat at the XO point), a 5th order has a 450deg phase rotation meaning it does not sum to flat. Furthermore the subwoofer is rolling phase at its own rate so it is highly unlikely you will get a symmetrical XO.

I have run some sims to show you.

Here we have a LR4 LPF (red) and LR4 HPF (green) with a corner freq of 100Hz. They LPF and HPF sum to flat (pink).

Now I have created a driver with a LF roll off which I have simulated with a 1st order Butterworth filter with a corner freq of 100Hz (brown). If we convolve it with the LR4 HPF (green), the result is blue.

Because of the additional phase rotation caused by the driver, the LPF and HPF no longer sum to flat.

I have also the software Acourate Pro which generates high quality FIR filters, however, relies on single point MLP measurement which I do not prefer.

Acourate does not force you into one way of working if you don't want to. You could use 5 measurements, 10 measurements, or an MMM as the basis of your correction. I wrote a free Acourate guide that should tell you how to use alternative measurement techniques as the basis for your correction. You will still need one measurement from MLP as the basis for all your timing measurements, but you can make amplitude adjustments using almost any technique that you want. As the Pure Acourate Sound project shows, you could even use a Klippel measurement as the basis for your corrections. I don't think any other software package on the market is as flexible.

 

[markstein]

Hi Keith,

Will look into your guide for Acourate, thank you!

As Acourate uses FIR filters (amplitude and phase correction), per my understanding you need to use one single point measurement, as in averaged multipoint measurement any phase information is lost.

Overall, I do neither like Dirac, nor Acourate full bandwidth correction which sounds very lifeless to my ears. This is my subjective impression - my room is asymmetrical and untreated, indirect sound ist very strong @3,5m listening distance, speakers see signature.

 

[Keith_W]

As Acourate uses FIR filters (amplitude and phase correction), per my understanding you need to use one single point measurement, as in averaged multipoint measurement any phase information is lost.

If you wish to correct for a listening area (as opposed to listening position), the simplest way is to use phase information from a single point measurement and combine it with an amplitude measurement averaged across as many positions as you want as a basis of correction. But there is nothing to stop you from doing something more complex if you wished, the only limitation is coming up with the workflow. Acourate is a toolbox, it has recommendations and guidelines that you do things in a certain way, but it won't stop you from doing anything you want no matter how silly. I have had my share of silly ideas.

Overall, I do neither like Dirac, nor Acourate full bandwidth correction which sounds very lifeless to my ears. This is my subjective impression - my room is asymmetrical and untreated, indirect sound ist very strong @3,5m listening distance, speakers see signature.

I believe you. Full range correction has a real potential to degrade the sound if you take an improper measurement or use the wrong correction strategy. Again, you do not need to do full range correction with Acourate - you can partially correct the frequency response. Again, the guide tells you how.

 

[Ultrasonic]

Dirac is an automated DRC that equalises the entire frequency response to a target curve. I do not think there is a way to avoid the full range correction, at least not with the last version of Dirac I looked at which was a few years ago. Dirac is also notoriously buggy and there are many complaints about failing to correct more than one subwoofer in a multi-sub system, missing bass, and sending corrections the wrong way. Sadly the DSP market is very small and I do not think there is a better option for beginners.

This is incorrect in two important ways. First you absolutely can define the frequency range that Dirac Live is applied over, on the current version and on every previous version I've used for several years. I routinely only use it up to about 200 Hz myself and did so with a new calibration at the weekend.

Secondly, I would describe it as semi-automated. As well as the frequency range choice, the target curve is also fully user adjustable and is intended to be adjusted to taste by the user. Anyone who does one set of measurements, applies filters based on the default target curve and for example concludes Dirac Live is rubbish because it lacks bass is not using it properly.

For context for the rest of the discussion, I use a miniDSP SHD to integrate a single subwoofer. I use REW measurements to optimise speaker positions and make a choice about crossover filters for both main speakers and sub. Further REW measurements are then used to calculate the time delay to apply to the main speakers to align best with the sub. I then use a 9-point Dirac Live measurement around the MLP to optimise the combined bass response. (This is for music. For movies I measure more points spread over a larger area for multiple listeners.)

Full frequency range Dirac Live in my experience always gives a sort of 'closed in' result that I don't like, and even with multiple point calibrations I struggle conceptually with how it's expected to 'work' due to how our ears handle reflections differently to direct sound from the speakers.

For me the main advantage of using Dirac Live over just using REW is that it makes basing EQ on multiple measurements quicker and easier. I previously used to apply bass EQ using single point measurements in REW and I do thing multiple measurement points is the better way to go.

 

[Ultrasonic]

Oh, and if the initial question was meant to imply EQ based on measurements at just a single point I think this is a bad idea. For starters both ears can never be at the exact same position anyway.

 

[Keith_W]

This is incorrect in two important ways. First you absolutely can define the frequency range that Dirac Live is applied over, on the current version and on every previous version I've used for several years. I routinely only use it up to about 200 Hz myself and did so with a new calibration at the weekend.

Secondly, I would describe it as semi-automated. As well as the frequency range choice, the target curve is also fully user adjustable and is intended to be adjusted to taste by the user. Anyone who does one set of measurements, applies filters based on the default target curve and for example concludes Dirac Live is rubbish because it lacks bass is not using it properly.

Thank you for the correction. I have not tried Dirac for many years so it appears that my recollection might be wrong.

 

[dweeeeb2]

Thank you for the correction. I have not tried Dirac for many years so it appears that my recollection might be wrong.

I’d say the strength in Dirac is how easy you can simply change the curve / range, then hit update.

 

[Old_School_Brad]

With practice I've gotten better/faster with these techniques, enough so to obsessively repeat again and again and again and...... Well, faster at least. Anyway, I'm really only interested in EQing for MLP in a 2.2 audio setup. I'm gradually coming to find that my best results are with REW EQing of subs and mains and followed by setting delay on subs and then level matching subs and mains. MSO doesn't seem to do better and is a lot more work. Dirac seems worst of all with compromised imaging and soundstage. To be clear, it's easy and better than no EQing at all, but certainly not as nice as REW. Room is W12xL22xH7ft and moderately treated with ASC bass traps. I'm using MiniDSP studio, Maggie MGIIIa's with old school ML amp and cheapo subs. Thoughts or comments? Thanks and cheers,

My experience has been the opposite of yours.

After extensively using REW and rePhase, I decided to try Dirac Live. On the very first run, it outperformed my best manual effort -and it only took a fraction of the time.

After further tweaking and getting familiar with how Dirac Live works, I’ve been able to achieve far better sound quality than I could ever manage manually. This has been my experience with both 2.1 and 2.0 setups, using four different sets of speakers in the same room.

 

[Digital_Thor]

I have a quite a bunch of like-wise HIFI-heads, which goes from turntables-lovers to pure streaming worshippers. Some of them are even hardcore digital-fanatics, that praise FIR and extremely steep filters so much, that I find it almost unobtainable - even though their systems sound ok.
I have mixed feelings, since I find that the Toole's theory works quite well, and when I hear a system with auto-tuned FIR, it often seems like you can get a new unique result every time you let a given algorithm do its thing, which is at times more or less kinda muffled or "dead" - IMO
My vague theory is, that often when we get a new tech - we go a bit too far. My first DSP was used as though everything else did not matter. I turned values up and down, but quickly found out, that unless you understood what was possible in the real world regarding gain, power respone, physics, amp power, room-acoustics and x-max, then I would never really make it sound good - just - weird or wrong.
Working the other way - by knowing what is possible in the real world, and then find the tools and means necesarry to reach that goal - seems much better - sound better too - IMO.
So my tough lesson, was that if a speaker did not sound right, I could not just crank my IIR DSP in digital wonderland, and then force it to be awesome in real life analog world.
Now it seems like people sometimes think, that FIR DSP and the right algorithm can take data from the listening position, and turn any system - no matter the original specs or shortcomings - and make a smooth straight frequency response, without changing the speaker physically.
And I can't wrap my head around this approach, when people like Erin and Amirm, put so much effort in measuring the speaker without the room entirely -while clearly saying that measuring at home - will never get precise - because why would you need a klippe scanner?
I actually hear people claim, that power-reponse, speaker type, driver type, dispersion, cabinet construction and shape - all does not matter. You just do some moving-mic-magic or measure a few places, and then FIR straightens out everything - boom - done.

 

[klettermann]

My experience has been the opposite of yours.

After extensively using REW and rePhase, I decided to try Dirac Live. On the very first run, it outperformed my best manual effort -and it only took a fraction of the time.

After further tweaking and getting familiar with how Dirac Live works, I’ve been able to achieve far better sound quality than I could ever manage manually. This has been my experience with both 2.1 and 2.0 setups, using four different sets of speakers in the same room.

Frankly, I'd bet more on your experience than mine. I only jumped on the DSP bandwagon last Nov (apart from Audessy for HT, but I'm totally uncritical of HT sound so that doesn't count). I've spent countless hours just trying to figure it all out, most of which just keeps showing me how much I don't know. My "experience" with Dirac was early on in those dark times where I had little grasp of anything except room modes and FR. I'll come back to it in the near future, doubtless with more questions. Cheers,

 

[klettermann]

This is incorrect in two important ways. First you absolutely can define the frequency range that Dirac Live is applied over, on the current version and on every previous version I've used for several years. I routinely only use it up to about 200 Hz myself and did so with a new calibration at the weekend.

Secondly, I would describe it as semi-automated. As well as the frequency range choice, the target curve is also fully user adjustable and is intended to be adjusted to taste by the user. Anyone who does one set of measurements, applies filters based on the default target curve and for example concludes Dirac Live is rubbish because it lacks bass is not using it properly.

For context for the rest of the discussion, I use a miniDSP SHD to integrate a single subwoofer. I use REW measurements to optimise speaker positions and make a choice about crossover filters for both main speakers and sub. Further REW measurements are then used to calculate the time delay to apply to the main speakers to align best with the sub. I then use a 9-point Dirac Live measurement around the MLP to optimise the combined bass response. (This is for music. For movies I measure more points spread over a larger area for multiple listeners.)

Full frequency range Dirac Live in my experience always gives a sort of 'closed in' result that I don't like, and even with multiple point calibrations I struggle conceptually with how it's expected to 'work' due to how our ears handle reflections differently to direct sound from the speakers.

For me the main advantage of using Dirac Live over just using REW is that it makes basing EQ on multiple measurements quicker and easier. I previously used to apply bass EQ using single point measurements in REW and I do thing multiple measurement points is the better way to go.

^^This is the place I'm trying to get to. Maybe I'm thick, but getting to the point of being able to actually to DO this with some facility, and sufficient understanding of the results to do it right, has been a pretty steep learning curve. For a complete beginner it's all unintelligible technobabble gibberish. Well, so is organic chemistry unless you're a chemist. Anyway, conversations like this are enormously helpful. When I really, really feel like I've got placement, timing and XO dialed in I'll come back to Dirac. Thanks to all and cheers,

 

[klettermann]

@klettermann your order could be causing problems. You stated “REW EQing of subs and mains and followed by setting delay on subs and then level matching subs and mains. ”

I find that setting delay impacts FR due to constructive / destructive interference. Also I find that I need to level match L to R before level matching with sub. so the order I use is:
EQ L (using MMM for all FR measurements)
EQ R
LVL L=R
set sub delay to mains
EQ Sub with mains
Then check FR with all playing. Some frequencies might combine better than others, so I apply equal correction to L+R for these new peaks.
bit off topic so feel free to ignore, just a suggestion.

And I’ve had no luck with Dirac. REW is the way!

This makes sense, as @Keith_W agrees. I got 2 subs playing mono (1 virtual sub), so first do timing and then EQ the Summed subs together? Cheers,

 

[dweeeeb2]

This makes sense, as @Keith_W agrees. I got 2 subs playing mono (1 virtual sub), so first do timing and then EQ the Summed subs together? Cheers,

If you can’t vary the timing or eq independently b/w subs then you could benefit from ensuring their placement is “optimized” for the MLP first, ie place one so as it fills in the nodes of the other one, otherwise you could have twice the nodal problems
If you are happy with their position then yes, try to get the best time alignment with your mains, with the XO’s active. Then determine the EQ for your virtual sub by conducting sweeps(or MMM) with the mains playing.

 

[klettermann]

If you can’t vary the timing or eq independently b/w subs then you could benefit from ensuring their placement is “optimized” for the MLP first, ie place one so as it fills in the nodes of the other one, otherwise you could have twice the nodal problems
If you are happy with their position then yes, try to get the best time alignment with your mains, with the XO’s active. Then determine the EQ for your virtual sub by conducting sweeps(or MMM) with the mains playing.

I can vary sub timing and EQ independently. My understanding from comments above, and others, is to: (1) try to find best positioning, (2) optimize timing/phase, and (3) optimizing XO's. And only then EQ the summed subs. The reason being that doing so individually would likely alter the effect of timing, position etc. Hence EQing both at the MLP. I guess I'll try both ways and see what happens. Seems like there are advocates for every possible permutation. Hobby becomes obsession.... Cheers,

 

[dweeeeb2]

I can vary sub timing and EQ independently. My understanding from comments above, and others, is to: (1) try to find best positioning, (2) optimize timing/phase, and (3) optimizing XO's. And only then EQ the summed subs. The reason being that doing so individually would likely alter the effect of timing, position etc. Hence EQing both at the MLP. I guess I'll try both ways and see what happens. Seems like there are advocates for every possible permutation. Hobby becomes obsession.... Cheers,

Sorry I wasn’t clearer but thats what I’m suggesting when I said “eq your virtual sub”. Eq the 2 subs as one. I do it with the mains playing just incase they impact (within crossover region) the bass response at the MLP.
EDIT: My reference to individual sub timing / eq was in reference mainly to the fact that playing with the timing can alleviate some of the positional problems. Feel free imo to use the delays to get position / NON EQ’d response as best you can as part of your point (1). Unfortunately every action has pros and cons.