How I managed to 'open' the soundfield with mid/side processing

[ppataki]

I have posted here on ASR previously about mid/side processing, that time for solving some bass issues in my odd-shaped room

This time I would like to share my findings about using it for (for a lack of a better term) 'opening' the soundfield
Let me explain that a little bit:
I have full-range (one-way) front speakers (assisted with a sub) in both my systems. I use Dirac Live 3 as a DRC. I have noticed many times also in the past that after optimizing with Dirac Live (or simply with REW for that matter) the top end tends to 'collapse' along with the stage - or at least this is what I hear. It sounds like the stage shrinks and sounds a bit dull. I have heard the same from multiple friends also using Dirac or REW with full-range speakers.

I was trying to fix this by applying additional EQ (high-shelves or bell filters) but the result was always too harsh, especially on the vocals
The solution I found is to apply that EQ using mid/side processing and only on the side channel

This is an example of IK Multimedia EQP-1A (basically a Pultec EQ) - but of course any other plugins would be fine provided they supported M/S processing
I switch it to M/S mode then select the Side channel and apply some boosting at 10kHz with a wide bandwidth

The result is that the stage opens up again, it no longer sounds dull and the vocals and other instruments in the phantom center are not impacted - and the instruments coming from the sides actually won't sound harsh at all but nice and clean

Again, just wanted to share my experience, hopefully others might find it useful too

 

[napfkuchen]

Never heard about this but it seems to be a studio tool. Why not limit Dirac to about 300 Hz so that there is no need to repair s.th. which was caused by the full-range-correction? Works fine with my speakers/subwoofers and DLBC.

 

[ppataki]

Never heard about this but it seems to be a studio tool. Why not limit Dirac to about 300 Hz so that there is no need to repair s.th. which was caused by the full-range-correction? Works fine with my speakers/subwoofers and DLBC.

That is a great idea generally speaking, however the particular issue with full range drivers is that their frequency response is very far from being flat throughout the whole frequency range
Therefore one needs to optimize on the full spectrum

 

[goat76]

That is a great idea generally speaking, however the particular issue with full range drivers is that their frequency response is very far from being flat throughout the whole frequency range
Therefore one needs to optimize on the full spectrum

The automatic full-range adjustments done with a program like Dirac are likely not the correct way to solve your issues, as the adjustments made by Dirac are likely the cause of the problems you experience.

You should instead look at the anechoic measurements of your full-range driver and make the corrections based on that, or gate the measurements you have done in REW so that you see the actual direct response of your speakers and make the adjustments based on that above say 300-500 Hz.
As @napfkuchen suggests, try limiting the automatic adjustments by Dirac to the bass range leaving the higher frequencies untouched, or manually adjust it according to the anechoic or gated measurements (if the dispersion of your speakers allows it).

I'm sorry to say this, but all automatic adjustments for the higher frequency range made by Dirac and similar programs will either be a hit or a miss. You can be lucky and it takes you somewhat in the right direction, or it will destroy the sound quality as it seems to be in your case with a "shrunken" soundstage and an overall "dull" sound (or any other type of unsatisfying sound): a hit or a miss.

Stop messing around with the Mid/Side processing. That may solve the problems created by the automatic adjustments, but the automatic adjustments are likely the cause of the problem.

 

[ppataki]

The automatic full-range adjustments done with a program like Dirac are likely not the correct way to solve your issues, as the adjustments made by Dirac are likely the cause of the problems you experience.

You should instead look at the anechoic measurements of your full-range driver and make the corrections based on that, or gate the measurements you have done in REW so that you see the actual direct response of your speakers and make the adjustments based on that above say 300-500 Hz.
As @napfkuchen suggests, try limiting the automatic adjustments by Dirac to the bass range leaving the higher frequencies untouched, or manually adjust it according to the anechoic or gated measurements (if the dispersion of your speakers allows it).

I'm sorry to say this, but all automatic adjustments for the higher frequency range made by Dirac and similar programs will either be a hit or a miss. You can be lucky and it takes you somewhat in the right direction, or it will destroy the sound quality as it seems to be in your case with a "shrunken" soundstage and an overall "dull" sound (or any other type of unsatisfying sound): a hit or a miss.

Stop messing around with the Mid/Side processing. That may solve the problems created by the automatic adjustments, but the automatic adjustments are likely the cause of the problem.

Leaving it uncorrected above 300-500Hz is not an option for the reason I have mentioned above (I have tried it already with all my systems and it sounds bad)
A gated measurement is a great idea indeed, I will give that a try and report back

 

[napfkuchen]

Leaving it uncorrected above 300-500Hz is not an option for the reason I have mentioned above (I have tried it already with all my systems and it sounds bad)
A gated measurement is a great idea indeed, I will give that a try and report back

I have little experience with Dirac calibration when it comes to using different speakers. What I can say, however, is that with my old speakers the result was unbearable without Dirac and unconvincing with Dirac activated (regardless of the settings). According to the Spinorama tests, my current speakers don't actually need any EQ. In my living room it also shows that, depending on the speaker/position, the in-room response before EQ follows the desired slope above 275 to 350 Hz, as hoped. Even if I select the full-range EQ now, the sound does not change noticeably.
As @goat76 already wrote, the behavior of the automatic calibration systems above 300-500 Hz is unfortunately not as reliable as hoped. The EQ curves after calibration presented are not a measurement but rather the desired result (unfortunately). Maybe measurements with REW and manual corrections would be a sensible option?

 

[kemmler3D]

Therefore one needs to optimize on the full spectrum

I would say you should correct the upper frequencies (especially mids) by hand based on nearfield measurements, auto-correction by Dirac or REW above a few hundred Hz is expected to cause problems.

Mid / side processing will give you an artificially exaggerated soundstage but I wouldn't call that a form of correction so much as enhancement.

On my own system I only auto-correct up to about 300hz and I do a few small corrections above that by hand / ear and looking at Klippel scans of my speakers.

 

[ppataki]

Yes, that is exactly what I will try using gated measurement as suggested by @goat76
I will leave Dirac up to 300Hz then make manual corrections based on the gated measurement in the MLP between 300 and 20.000 Hz

 

[ppataki]

Mid / side processing will give you an artificially exaggerated soundstage but I wouldn't call that a form of correction so much as enhancement.

I agree, I also consider this as rather an enhancement (that also happened to 'fix' my issue) and I really do like the sound it produces
I am really curious how the manual correction based on the gated measurement will sound - I have tried manual correction in the past based on non-gated measurement and it sounded utter rubbish (=even more dull and collapsed sound)

 

[LIΟN]

I have tried manual correction in the past based on non-gated measurement and it sounded utter rubbish (=even more dull and collapsed sound)

It would be much better to observe what is happening with progressive gating and apply EQ accordingly. - Of course, you can't control all the reflections in your room with EQ. But even so, it's much better to gate (window) so you know exactly what you're experiencing.
What are your speakers?

 

[ppataki]

It would be much better to observe what is happening with progressive gating and apply EQ accordingly. - Of course, you can't control all the reflections in your room with EQ. But even so, it's much better to gate (window) so you know exactly what you're experiencing.
What are your speakers?

I have these in the living room and these in my home office

 

[Galliardist]

That is a great idea generally speaking, however the particular issue with full range drivers is that their frequency response is very far from being flat throughout the whole frequency range
Therefore one needs to optimize on the full spectrum

So why hobble your system with them in the first place? Presumably, you see some advantage?

 

[ppataki]

So why hobble your system with them in the first place? Presumably, you see some advantage?

I have only been using full range speakers for 4-5 years when I first listened to one (it was an Avantone Mixcubes with Dirac) I was instantly blown away and threw all my gear out the window (it was a nuVero 140 at that time)
The holographic stage presentation and the unbelievably 'live' sound are the things that still totally catch me
All of this is utterly subjective of course
Technically speaking they are inferior and shall not be used; I am well aware of that

 

[kemmler3D]

I agree, I also consider this as rather an enhancement (that also happened to 'fix' my issue) and I really do like the sound it produces
I am really curious how the manual correction based on the gated measurement will sound - I have tried manual correction in the past based on non-gated measurement and it sounded utter rubbish (=even more dull and collapsed sound)

I would suggest doing wide corrections of only 2dB, maybe 3dB for a very wide correction, (Q >1) to keep things sounding natural.

If you take an on-axis measurement and do narrow corrections based on that, you're just squeezing one end of the balloon and puffing up the other, i.e. off-axis gets worse.

This will tend to make reflected sound much worse and that leads to bad imaging.

Narrow corrections are sort of OK in certain situations but I wouldn't want to do them without at least 60 degrees worth of measurements, preferably more.

 

[ppataki]

I had some time today to play around with it
I cannot post the measurement now since I am in the countryside but Sunday evening I will post those too

So, if I just perform a 1-point measurement with Dirac at the MLP then run a verification measurement in REW the frequency response looks jolly good (like unbelievably good) but when listening to it the bass is ringing (due to overcompensating room mode-related dips) and the highs are non-existing, and the stage is collapsed

When I perform a complete 9-point measurement and then run a verification measurement in the MLP the frequency response looks more like POS but the bass no longer rings and the highs have come alive - actually more than they should
So I applied a low Q, low gain correction as suggested by @kemmler3D to manually remove the big and wide peak that Dirac left there at around 10kHz, then applied a high-shelf to boost the overall treble and now the harshness is gone and the stage is wide open too

As mentioned above, I will post the measurements on Sunday + I have another idea to try next week when I will have a few hours to spare

 

[hex168]

Another reason to stick with low Q corrections is that the resolution of your gated measurement will limited by by your gating window.

 

[Keith_W]

Hello Peter. In general, bad results following DSP may be due to:

1. Wrong measurement - e.g. improper placement of microphone, reflections contaminating the measurement, inadequate gating to remove reflections. Get a tape measure and measure the distance to the closest reflecting surface - sofa, wall, floor, coffee table, etc. and look for it on the ETC. Then work out what frequency will be affected.

2. Wrong DSP or inappropriate correction strategy, e.g. improper phase extraction leading to inversion of excess phase, attempting aggressive correction of upper frequencies, etc. Given that Dirac isn't very transparent about what it is doing, I would be making sure that the initial measurement is correct.

 

[goat76]

As Dirac expects the measurements to be made with the microphone at about 9 different positions around the listening position(s), and with the loudspeakers in the position where the user placed them, there will be many reflection points contaminating the measurements and it's impossible for the program to “separate” those from the direct sound from the loudspeakers.

That's the main problem with all these automatic room correction programs, they base the full-range corrections on in-room measurements with many uncertainties of how much the measurements are contaminated by room reflections. The microphone picks up everything as equally as important no matter if it's the direct sound or room reflections, while our hearing (at least according to the science) puts more emphasis on the direct sound and filters out much of the reflections as “secondary” sounds.

 

[ppataki]

So here are the measurements I was referring to in post #15

This is when doing a verification measurement of a 1-point Dirac optimization in the MLP

It looks superb but in reality bass rings like hell due to overcompensation, see the Waterfall graph below

And the highs sound extremely dull and the stage collapsed - I am not sure how to prove that with this measurement but I will attach the mdat file just in case others can see something that I missed


And this is the verification measurement of the 9-point optimization in the MLP:

Bass does not look that good but actually it sounds way better, ringing is much reduced:

But as mentioned in post #15 the highs are now way too exaggerated (that wide peak at 10kHz). I manually removed that with one single low Q bell filter then applied some high shelf to taste and I am happy with the results. Treble is no more harsh and stage is nice and wide open

Basically what I noticed is that Dirac tends to keep the original 'profile' of the speakers (with the multi-point optimization at least), see the original, unprocessed response at MLP below:

That wide peak at 10kHz did not get removed with the 9-point optimization (it nicely filled the dip around 4kHz though)