Best approach to EQing and Dirac DSP?

[klettermann]

As I keep falling down the DSP rabbit hole I learn more and more how little I know about all this. I've spent many hours trying to optimize positioning, followed by many more trying to tweak Dirac. I'm running the MiniDSP SHD Studio. Somehow, only now, did I figure out how to get REW sweeps for respective L and R channels (it was a Windows setting issue). And for some reason I never bothered to play with the MiniDSP's parametric equalizer. And so my question:

In seeking some optimum EQ does it make sense to first do some broad EQing using the MiniDSP EQ for gross issues? These would be wild low end swings and maybe the general FR slope. And then follow that with Dirac for further refinement ? It seems like that might give Dirac better raw material so to speak. The number of approaches seems limitless. Comments greatly appreciated, thanks in advance. Happy Thanksgiving and cheers,

 

[staticV3]

In seeking some optimum EQ does it make sense to first do some broad EQing using the MiniDSP EQ for gross issues?

I don't think so. Dirac can handle those gross issues just as well.

 

[klettermann]

I don't think so. Dirac can handle those gross issues just as well.

Hmmmm.... OK, but let me get more specific. Please see below....

Purple: Uncorrected room FR
Red: Default Dirac correction, no tweaking
Blue: Tweaked version of the red Dirac-corrected scan.
Light green: Room noise with HVAC running. I'll work on that at some point.
Light blue: Room noise as-quiet-as-I-can-make-it. No idea what the 4kHz bump is, I'll try to find that too.

The red "raw Dirac" plot is still pretty bumpy between about 90-200Hz and there are some substantial hills and valleys beyond that. The blue was the result of further manual smoothing within Dirac to make the whole FR slope nicer. The point is, it seems like there were further improvements to be made by tweaking the Dirac plot. Or, perhaps, they weren't actually improvements? Anyway, it's pretty clear that some simple manual EQing from 60-200Hz would bring a significant improvement, even without Dirac. Same goes for improving the general slope from 1kHz and above. Or, maybe the red Dirac correction is good enough? To be honest the red and blue both sounded pretty good. Maybe I'm overthinking it all. Thanks and cheers,

 

[tw99]

The blue line looks excellent and I wouldn't mess about with it any more. Even the red one is not that bad

 

[klettermann]

The blue line looks excellent and I wouldn't mess about with it any more. Even the red one is not that bad

That's great news, thanks muchly! Obviously I don't have the experience to really understand what's good. However, the blue line does have another problem that's not apparent from the SPL plot. There's a very sharp, 25dB dip at 94Hz.

This region can't really be EQ'd well. Attempts to do so makes the the ringing was far worse after further tweaking the native Dirac correction. Paradoxically, EQing a massively deep slot at 94Hz minimizes the ringing. Since there's already the dip at 94Hz maybe this OK since the slot is already there anyway? Hope this makes sense. Cheers,

Un-EQ'd Dirac plot
x

 

[Keith_W]

Sum your left and right sweeps and see what happens to the 94Hz dip.

 

[Paffi]

Looks like sbir - try to change the distance to the front wall and measure again. If the dip moves, you have the cause.

 

[Old_School_Brad]

Hmmmm.... OK, but let me get more specific. Please see below....

Purple: Uncorrected room FR
Red: Default Dirac correction, no tweaking
Blue: Tweaked version of the red Dirac-corrected scan.
Light green: Room noise with HVAC running. I'll work on that at some point.
Light blue: Room noise as-quiet-as-I-can-make-it. No idea what the 4kHz bump is, I'll try to find that too.
View attachment 410176

The red "raw Dirac" plot is still pretty bumpy between about 90-200Hz and there are some substantial hills and valleys beyond that. The blue was the result of further manual smoothing within Dirac to make the whole FR slope nicer. The point is, it seems like there were further improvements to be made by tweaking the Dirac plot. Or, perhaps, they weren't actually improvements? Anyway, it's pretty clear that some simple manual EQing from 60-200Hz would bring a significant improvement, even without Dirac. Same goes for improving the general slope from 1kHz and above. Or, maybe the red Dirac correction is good enough? To be honest the red and blue both sounded pretty good. Maybe I'm overthinking it all. Thanks and cheers,

What do you mean by "tweaking the Dirac plot". -By creating your own curve within Dirac versus using the stock curve?

 

[klettermann]

Thanks very much for the comments! What's going on is now clearer, even if the reasons are not - or at least not to me. Here's some more info.

Sum your left and right sweeps and see what happens to the 94Hz dip.

I ran every combination of (1) Dirac on/off and (2) left, right and L+R. The dip was there in all cases, it didn't go away. However, it's always significantly worse in the left channel by ~5dB. Regardless of differences in the ringing was more or less the same regardless.

Looks like sbir - try to change the distance to the front wall and measure again. If the dip moves, you have the cause.

After doing the work above I took a quick look at positioning. The speaker's center was 48" from front wall and they're toed in. I moved the speakers back towards the front wall by about 5". This DID make a difference. Not enormous, but an improvement nevertheless. From 20-200Hz SPL was a little flatter and the dip wasn't quite as deep. I have little grasp of SBIR and info with planar dipoles is thin, to say the least. Below is the effect of the 5" change with no smoothing. Does this qualify as an SBIR issue? Regardless. I'm going to revisit placements. Maybe there's more improvement to be gained.

What do you mean by "tweaking the Dirac plot". -By creating your own curve within Dirac versus using the stock curve?

Yeah, I think that's how you'd describe what I'm doing, along the lines of the Dirac pic below.

Thanks again to all and Happy Thanksgiving!

 

[Keith_W]

The SBIR wavelength is 4x the distance of your speaker to the wall. If it's 48", the cancellation should occur at 70Hz (use a frequency-wavelength calculator). If you move it 5" closer to the wall, it should shift the SBIR frequency upwards to 78.5Hz.

 

[Old_School_Brad]

Yeah, I think that's how you'd describe what I'm doing, along the lines of the Dirac pic below.

View attachment 410250

Thanks again to all and Happy Thanksgiving!

Ah, this approach. Looking at the behavior at 200 Hz, I’d be concerned about the apparent +15 dB adjustment, considering potential issues like signal clipping and similar effects. Have you tested with the miniDSP interface and 0 dB test tones in that range to ensure there’s no clipping?
I've had great results keeping the target curve below or near below the dips to avoid this. But this is only viable if you're able to raise the gain enough with down stream components such as the pre amplifier or power amplifier and still keeping noise inaudible.

 

[kyuu]

DSP is an amazing tool, but it's easy to get a bit over obsessed with chasing the flat line to the point where you're tilting at windmills. Or potentially negatively affecting the sound.

For that deep, high-Q dip at 94Hz, as you've found out trying to address that with EQ that is a bad idea. Playing with the positioning of your speakers might help a bit. But ultimately, is that even an audible concern? That dip doesn't appear nearly as bad when you're applying smoothing, and human hearing does not generally pick up on high-Q dips like that. Might be best not to worry about it.

Have you done any A/B testing between the original Dirac filter and your tweaked one? Does your tweaked one sound any better?

 

[klettermann]

Thanks very much to all. I'm going back to the drawing board and start it all over from scratch with all this in mind, as well as rethinking MLP. I think I'm now in a far better position than when I started out. It's been a steeper learning curve than I ever imagined. My partner thinks I've lost my mind, and she's right! More below.....

The SBIR wavelength is 4x the distance of your speaker to the wall. If it's 48", the cancellation should occur at 70Hz (use a frequency-wavelength calculator). If you move it 5" closer to the wall, it should shift the SBIR frequency upwards to 78.5Hz.

By this measure the 94Hz notch is something other than SBIR. `And anyway, what does it even mean with 6' x 2' dipoles? More on the notch problem below.

Ah, this approach. Looking at the behavior at 200 Hz, I’d be concerned about the apparent +15 dB adjustment, considering potential issues like signal clipping and similar effects. Have you tested with the miniDSP interface and 0 dB test tones in that range to ensure there’s no clipping?
I've had great results keeping the target curve below or near below the dips to avoid this. But this is only viable if you're able to raise the gain enough with down stream components such as the pre amplifier or power amplifier and still keeping noise inaudible.

Yeah, good advice, which I'll follow in the next EQing/Dirac series. As for clipping, I've never heard it with this system. The Maggies are inefficient, to be sure, but present a nice, stable 4ohm load to the amp. And the amp is a Mark Levinson No. 332 which puts out a clean 400 wpc into 4ohms. That's probably what allowed me escape clipping when EQing a dip higher. I think my gain is fine to do this.

DSP is an amazing tool, but it's easy to get a bit over obsessed with chasing the flat line to the point where you're tilting at windmills. Or potentially negatively affecting the sound.

For that deep, high-Q dip at 94Hz, as you've found out trying to address that with EQ that is a bad idea. Playing with the positioning of your speakers might help a bit. But ultimately, is that even an audible concern? That dip doesn't appear nearly as bad when you're applying smoothing, and human hearing does not generally pick up on high-Q dips like that. Might be best not to worry about it.

Have you done any A/B testing between the original Dirac filter and your tweaked one? Does your tweaked one sound any better?

Yes, amazing! Maybe too amazing. For sure, I've fallen very deep into windmill tilting. As for the dip, I suspect it isn't actually audible. Moreover, I found that EQing the dip down actually has the effect of reducing the ringing. I'm going to try that approach. After all, it's just making and existing and hopefully inaudible dip deeper.

I've tried so many filters at this point that I've lost track. The whole process really demands a disciplined, systematic approach to saving filters, file labeling, documenting results etc. Winging it just doesn't work, at least not for me. That's another reason to start fresh now that I have at least some idea of how the pieces work and fit together.

I'll post whatever I come up with. Thanks again and cheers,

 

[Keith_W]

By this measure the 94Hz notch is something other than SBIR. `And anyway, what does it even mean with 6' x 2' dipoles? More on the notch problem below.

You just made me stop and think The first SBIR frequency is a cancellation, but this is with a monopole design! With a monopole, the back wave is in phase with the forward wave. The reflection from the front wall is in-phase with the incident wave. By the time the reflection travels back to the speaker, half a period has passed, so it is now out-of-phase which results in a cancellation. With a dipole, the back wave is out-of-phase with the forward wave. So the first SBIR frequency should be reinforcement. You should be seeing a peak at 70Hz if you have your dipole placed 48" from the wall. Then again, I don't know enough about dipoles to know if dipoles behave as dipoles at long wavelengths, or whether they are omnis. As you can gather, what you see (if it is SBIR) depends on whether the rear wave is in-phase or out-of-phase with the forward wave.

I will need to defer to somebody more knowledgeable than me about this and humbly warn you that I might be in error. I am confident that my understanding of SBIR with monopoles is correct, not so confident about SBIR with dipoles. So i'll shout out to @NTK and @Duke.

Regardless, there is no practical benefit of knowing whether that cancellation at 94Hz is due to SBIR or a room mode, because either way the solution is the same. You can choose to ignore it (which is what I would do, because it is so narrow) or you can use another speaker or subwoofer to fill it. I for one, acknowledge that tilting at windmills is fun and you learn a heck of a lot in the process. So get yourself a horse and a lance and go for it!

 

[DWPress]

Just to chime in with my own experience - use the MMM method to correlate your static sweeps and make sure you're not correcting something that only exists in one place in time/space. I would use the miniDSP EQ to then get rid of the worst offenders up to about 400-500Hz as I find presenting Dirac with something halfway decent allows better correction filters in the end.

 

[Old_School_Brad]

Yeah, good advice, which I'll follow in the next EQing/Dirac series. As for clipping, I've never heard it with this system. The Maggies are inefficient, to be sure, but present a nice, stable 4ohm load to the amp. And the amp is a Mark Levinson No. 332 which puts out a clean 400 wpc into 4ohms. That's probably what allowed me escape clipping when EQing a dip higher. I think my gain is fine to do this.

I wasn’t referring to amplifier clipping due to insufficient power but rather signal clipping.

I always verified using known tracks and test tones in the miniDSP user interface after closing Dirac Live, ensuring the signal wasn’t clipping and that the master volume was set correctly before powering on the amplifier.

 

[markstein]

Hmmmm.... OK, but let me get more specific. Please see below....

Purple: Uncorrected room FR
Red: Default Dirac correction, no tweaking
Blue: Tweaked version of the red Dirac-corrected scan.
Light green: Room noise with HVAC running. I'll work on that at some point.
Light blue: Room noise as-quiet-as-I-can-make-it. No idea what the 4kHz bump is, I'll try to find that too.
View attachment 410176

The red "raw Dirac" plot is still pretty bumpy between about 90-200Hz and there are some substantial hills and valleys beyond that. The blue was the result of further manual smoothing within Dirac to make the whole FR slope nicer. The point is, it seems like there were further improvements to be made by tweaking the Dirac plot. Or, perhaps, they weren't actually improvements? Anyway, it's pretty clear that some simple manual EQing from 60-200Hz would bring a significant improvement, even without Dirac. Same goes for improving the general slope from 1kHz and above. Or, maybe the red Dirac correction is good enough? To be honest the red and blue both sounded pretty good. Maybe I'm overthinking it all. Thanks and cheers,

I personally would stop correcting at about 300-500 Hz upwards and let Dirac handle the bass correction only. Never liked Dirac full correction in my untreated unsymmetrical living room. Btw. I also never liked Acourate full correction. Sounds liveless and boring.

 

[klettermann]

I wasn’t referring to amplifier clipping due to insufficient power but rather signal clipping.

I always verified using known tracks and test tones in the miniDSP user interface after closing Dirac Live, ensuring the signal wasn’t clipping and that the master volume was set correctly before powering on the amplifier.

Not quite following you here. Can you elaborate? Thanks much.

 

[Old_School_Brad]

Not quite following you here. Can you elaborate? Thanks much.

Dirac attempts to boost frequencies where the target curve exceeds the measurement curve. However, this introduces the risk of signal clipping within the miniDSP if the voltage corresponding to the boost surpasses the device’s capability.

What I’m emphasizing is the importance of monitoring the meters in the miniDSP graphical interface while playing various audio content at maximum volume. The output meters should remain below 0 dBFS, as this represents the miniDSP’s maximum output. For example, if you apply a 15 dB boost at 200 Hz, a test tone at 200 Hz will produce a significantly higher output compared to a tone at 300 Hz, due to the applied boost.

 

[markstein]

Dirac attempts to boost frequencies where the target curve exceeds the measurement curve. However, this introduces the risk of signal clipping within the miniDSP if the voltage corresponding to the boost surpasses the device’s capability.

What I’m emphasizing is the importance of monitoring the meters in the miniDSP graphical interface while playing various audio content at maximum volume. The output meters should remain below 0 dBFS, as this represents the miniDSP’s maximum output. For example, if you apply a 15 dB boost at 200 Hz, a test tone at 200 Hz will produce a significantly higher output compared to a tone at 300 Hz, due to the applied boost.

This will likely not be sufficient, as the MiniDSP does not show true peak volume but RMS volume only. There is a high risk of clipping when you are close to 0dBFS.

Better attenuate the master volume by 9dB, as 9dB AFAIK is the max. boost Dirac applies in its MiniDSP implementation.

I asked for a true peak display in the MiniDSP forum but never received a feedback from the company.