EQ help needed :( - keep a broad 700 Hz cut?

[Opalius]

Hi all
I’m trying to decide whether I should keep a broad 700 Hz cut in my living room setup.

The filter in question is:
700 Hz / -4 dB / Q 1.0

It was used in setup 1 and but not in Setup 2.

The speakers are DALI Oberon 3 in a normal stereo triangle in a living room, and I also have one subwoofer in the system.

The main speakers were not moved between setups. What I’m mostly trying to evaluate is the voicing through the midrange and whether this 700 Hz cut is actually improving the balance.

These are the filters I’m currently using:

1. PK 6.00 kHz, -1.5 dB, Q 1.00
2. PK 45.0 Hz, -5.0 dB, Q 4.00
3. PK 90.0 Hz, -3.0 dB, Q 3.00
4. LS 150.0 Hz, -5.0 dB, Q 0.71
5. PK 700.0 Hz, -4.0 dB, Q 1.00
6. HS 3.30 kHz, -2.0 dB, Q 0.71
7. PK 9.00 kHz, -3.0 dB, Q 3.00
8. HS 14.00 kHz, +2.5 dB, Q 0.71
9. PK 80.0 Hz, +1.0 dB, Q 2.50
10. LS 80.0 Hz, +4.0 dB

I’m attaching screenshots (1/6 smoothing) of the 1 m raw measurement, Setup 1 and 2....

1m mesasurement:

Setup one with filter at 700.....:

Setup withput filter at 700

What I’m trying to understand is whether the 700 Hz cut is actually helping the tonal balance, or whether I’m correcting something that is more room/reflection-related and should probably be left alone?

Any input is welcome.

 

[CB89]

My first question is how the measurement was performed. Is it a single measurement, a multi-point measurement, or a "Moving Microphone Measurement" (MMM)? I would rule out MMM, as the lines in the graph are very smooth.

For corrections above 300 Hz, I wouldn't trust a single measurement. If it's an averaged multi-point measurement, then yes. However, an MMM would be better.

The second question is whether the speaker has a uniform directivity pattern and therefore responds well to an EQ. Otherwise, it's a gamble whether the adjustment will produce the desired effect. (I would still try it, though.)

Due to the ripple in the high frequencies and the roll-off in the high frequencies (in the near field), I assume a single measurement (L+R). A single measurement for L+R is particularly disadvantageous here due to interference.

This is what it looks like for me, for example, in the near field.

 

[levimax]

Common advice is to not correct above Schroeder Frequency which is around 250 Hz depending on room size. A low Q - 4 dB filter is also going to be very audible. I would try it with and without and see which you prefer and maybe try a - 2 dB filter as well and see how it sounds.

 

[Soniclife]

Probably keep it, but you will have to ultimately decide by listening, which will take time unless you are a trained listener. Try it one way for a week, then the other, listen to a wide variety of music, then consider a 2db cut for a play off.

 

[Opalius]

My first question is how the measurement was performed. Is it a single measurement, a multi-point measurement, or a "Moving Microphone Measurement" (MMM)? I would rule out MMM, as the lines in the graph are very smooth.

For corrections above 300 Hz, I wouldn't trust a single measurement. If it's an averaged multi-point measurement, then yes. However, an MMM would be better.

The second question is whether the speaker has a uniform directivity pattern and therefore responds well to an EQ. Otherwise, it's a gamble whether the adjustment will produce the desired effect. (I would still try it, though.)

Due to the ripple in the high frequencies and the roll-off in the high frequencies (in the near field), I assume a single measurement (L+R). A single measurement for L+R is particularly disadvantageous here due to interference.

This is what it looks like for me, for example, in the near field.
View attachment 521171

Fair questions. To clarify: these are Left speaker only measurements. I didn't want to deal with L+R interference while trying to hunt down this specific 700 Hz ghost.

It is a single-point measurement, but I've verified the 'hump' is real. I moved the mic around -+30 cm from the listening position and the 700 Hz peak shows up in every variation, though the exact shape shifts slightly. It’s definitely a persistent room/reflection issue rather than a one-spot fluke.

Regarding directivity, they are Dali Oberon 3s. Generally, they have decent horizontal dispersion, but they are known for that 'warm' Dali voicing.

My struggle is exactly what you hinted at: is this a 'gamble' where I'm fixing the graph but breaking the power response? Without the filter, the vocalist sounds like they’ve gained 20kg—it loses that light, nimble edge. With the filter, the 'weight' is gone, but I’m worried I’m just sanitizing the speaker’s soul.

Given it’s a living room (where acoustic treatment = divorce), would you trust a -4 dB cut based on a consistent 30 cm cluster, or am I over-correcting for a reflection that my brain would naturally tune out?

 

[CB89]

Only the left speaker. I see, so the Dali speakers are slightly attenuated in the treble. But it looks like a nice slope.

The -4dB is quite significant. I would agree with the other comments and suggest trying -2dB instead. For evaluating tonal balance (as a human), I tend to use psychoacoustic smoothing to avoid overemphasizing narrow peaks. When applying correction, I use variable smoothing.

 

[Opalius]

Only the left speaker. I see, so the Dali speakers are slightly attenuated in the treble. But it looks like a nice slope.

The -4dB is quite significant. I would agree with the other comments and suggest trying -2dB instead. For evaluating tonal balance (as a human), I tend to use psychoacoustic smoothing to avoid overemphasizing narrow peaks. When applying correction, I use variable smoothing.

Thanks for the suggestions. I’ve been experimenting with a middle-of-the-road approach: 700 Hz / -2 dB / Q 2.0.

The tonality feels pretty spot on so far—it doesn't sound "wrong" or overly sanitized. It actually gives the vocals a bit more desirable weight compared to the -4 dB cut, while still cleaning up that persistent hump.

My main uncertainty now is the Q Is Q 2 getting too narrow for a correction up around 700 Hz?

 

[kemmler3D]

Thanks for the suggestions. I’ve been experimenting with a middle-of-the-road approach: 700 Hz / -2 dB / Q 2.0.

The tonality feels pretty spot on so far—it doesn't sound "wrong" or overly sanitized. It actually gives the vocals a bit more desirable weight compared to the -4 dB cut, while still cleaning up that persistent hump.

My main uncertainty now is the Q Is Q 2 getting too narrow for a correction up around 700 Hz?
View attachment 521180

I would say based on personal opinion that Q=2 is too narrow for a correction that high up.

 

[levimax]

My main uncertainty now is the Q Is Q 2 getting too narrow for a correction up around 700 Hz?

Q2 is not particularly narrow and should not create any noticeable time domain issues. There are no hard and fast rules for EQ like this unless you go crazy with ultra narrow filters on every single little peak. For fine tuning like this it often takes time to evaluate and just using one or two songs for a quick check may not give you the best results. As mentioned I would spend time listening with a couple of filters you think are "close" for a couple of days each with a variety of music at a variety of listening levels.

 

[Keith_W]

Guys, read closely. He is taking a 1m measurement. The intention of a 1m measurement is to capture the response of the loudspeaker, free of room influence, for loudspeaker correction. NOT ROOM CORRECTION.

Whether there is any point to this depends on the quality of the measurement. In this case, there are multiple unknowns since he didn't post the MDAT. THE biggest unknown is whether that 700Hz bump is real, or if it's a measurement artefact! If it is a really good quality measurement, then YES he can correct every little bump in the frequency response if he likes. I would make sure that left and right speaker are the same before applying all those PEQ's though. Firstly as a sanity check and secondly to make sure that he is actually correcting the loudspeaker response alone.

 

[dfuller]

The main speakers were not moved between setups. What I’m mostly trying to evaluate is the voicing through the midrange and whether this 700 Hz cut is actually improving the balance.

This is a case where you'll need to listen, I'm afraid.

 

[dasherzx]

Common advice is to not correct above Schroeder Frequency which is around 250 Hz depending on room size. A low Q - 4 dB filter is also going to be very audible. I would try it with and without and see which you prefer and maybe try a - 2 dB filter as well and see how it sounds.

i still go full range correction if mids and highs are very uneven (after var smoothing) in rew. then i do a final measurement of the corrected result.

 

[Sal1950]

Firstly as a sanity check and secondly to make sure that he is actually correcting the loudspeaker response alone.

Now hold on just one minute Keith my friend,

 

[Opalius]

Guys, read closely. He is taking a 1m measurement. The intention of a 1m measurement is to capture the response of the loudspeaker, free of room influence, for loudspeaker correction. NOT ROOM CORRECTION.

Whether there is any point to this depends on the quality of the measurement. In this case, there are multiple unknowns since he didn't post the MDAT. THE biggest unknown is whether that 700Hz bump is real, or if it's a measurement artefact! If it is a really good quality measurement, then YES he can correct every little bump in the frequency response if he likes. I would make sure that left and right speaker are the same before applying all those PEQ's though. Firstly as a sanity check and secondly to make sure that he is actually correcting the loudspeaker response alone.

Thanks for the catch—you’re absolutely right that the 1m measurement was intended to isolate the speaker's response as much as possible.

I apologize for not posting the MDAT.
To be honest, I was a bit hesitant because my signal chain involves an interface between the PC and the mic, and I was worried that might invite some scrutiny or "measurement-shaming!"

However, I am 100% sure the 700 Hz hump is a real characteristic of the room output and not a measurement artifact. When I measure the speaker at the 1m BUT close to a sidewall wall, it’s there.! Measure away from sidewall, ( 1.5 m ) or at a angle from side wall, and it's gone. Also when measuring the speaker from the position they are normally in ( 80 cm from sidewall) and at a distance of 50 cm, 1 m, 2 m and so forth from speaker, the hump will show op diffrent in size and slightly in place -+ 100 hz and +- 2 db.

 

[Keith_W]

To be honest, I was a bit hesitant because my signal chain involves an interface between the PC and the mic, and I was worried that might invite some scrutiny or "measurement-shaming!"

Nobody knows how to take a proper measurement unless you specifically learn how to do so. Amir wasn't born with a Klippel in his mouth, he had to learn how to use it like everyone else There is no shame in not knowing how to take a measurement. Take a look at the eBook in my sig - there is a section in there about how to measure loudspeakers.

However, I am 100% sure the 700 Hz hump is a real characteristic of the room output and not a measurement artifact. When I measure the speaker at the 1m BUT close to a sidewall wall, it’s there.! Measure away from sidewall, ( 1.5 m ) or at a angle from side wall, and it's gone. Also when measuring the speaker from the position they are normally in ( 80 cm from sidewall) and at a distance of 50 cm, 1 m, 2 m and so forth from speaker, the hump will show op diffrent in size and slightly in place -+ 100 hz and +- 2 db.

Your description 100% tells me it is an artefact of microphone positioning - i.e. it's a measurement artefact. What's the point of correcting it if it disappears when you move the microphone somewhere else? You're only introducing spectral distortion and messing up the performance of your loudspeakers. Depending on the size of your room, 700Hz may be within your transition zone. Don't try to correct too finely in the transition zone.

I hope you understand what I am trying to say. Above the transition zone: obtain a quasi-anechoic measurement of your loudspeaker. Mic 1m away, speaker elevated and preferably outdoors. Examine your measurement to see the time arrival of the first reflection (impulse response or ETC) and window it out. Once you have this measurement, then by all means you can correct every bump in the freq response. Remember that obtaining a quasi-anechoic measurement of a loudspeaker down to long wavelengths is very difficult, so you have to stop correction somewhere otherwise you are correcting ghosts.

Schroder freq and below: the speaker is corrected together with the room. The mic should be at the listening position, then obtain individual measurements of L, R, each sub if any, along with a timing reference. You should consider taking a measurement over a wider area (i.e. listening area not listening position) with multiple measurements or an MMM. Experiment with speaker/sub positioning until you have as few nulls as possible, then chop off all the peaks with EQ. If necessary, you could apply some "broad, low Q, tone control" equalisation (as per Toole) to taste.

 

[Opalius]

Nobody knows how to take a proper measurement unless you specifically learn how to do so. Amir wasn't born with a Klippel in his mouth, he had to learn how to use it like everyone else There is no shame in not knowing how to take a measurement. Take a look at the eBook in my sig - there is a section in there about how to measure loudspeakers.



Your description 100% tells me it is an artefact of microphone positioning - i.e. it's a measurement artefact. What's the point of correcting it if it disappears when you move the microphone somewhere else? You're only introducing spectral distortion and messing up the performance of your loudspeakers. Depending on the size of your room, 700Hz may be within your transition zone. Don't try to correct too finely in the transition zone.

I hope you understand what I am trying to say. Above the transition zone: obtain a quasi-anechoic measurement of your loudspeaker. Mic 1m away, speaker elevated and preferably outdoors. Examine your measurement to see the time arrival of the first reflection (impulse response or ETC) and window it out. Once you have this measurement, then by all means you can correct every bump in the freq response. Remember that obtaining a quasi-anechoic measurement of a loudspeaker down to long wavelengths is very difficult, so you have to stop correction somewhere otherwise you are correcting ghosts.

Schroder freq and below: the speaker is corrected together with the room. The mic should be at the listening position, then obtain individual measurements of L, R, each sub if any, along with a timing reference. You should consider taking a measurement over a wider area (i.e. listening area not listening position) with multiple measurements or an MMM. Experiment with speaker/sub positioning until you have as few nulls as possible, then chop off all the peaks with EQ. If necessary, you could apply some "broad, low Q, tone control" equalisation (as per Toole) to taste.

Okey.... To be sure, I just took 6 different measurements at seating position - (moving the mic 30 cm left, right, front, and back) across my entire seating area - yes I have a BIG butt and head....
That 700 Hz hump showed up in every single spot. When I apply that broad -4 dB filter and average those 6 measurements together, the hump is gone across the whole area.

Since it’s not just a "fluke" in one position but a consistent extra "energy" in the room, it feels like a real problem that needs fixing rather than a measurement error.

If it stays there no matter where I move the mic (within 30 cm), doesn't that suggest it's a real room/speaker interaction that's worth EQing out? Without it, the "weight" on the vocals is just too much?

 

[Keith_W]

If the red/blue curves are an average over a listening area, then yes it's fair enough to EQ it out. Especially since you are using psychoacoustic smoothing.

 

[90sNintendoKid]

How are you doing the 1m measurement? I use Magic Beans True Target and for nearfield measurements within 1m, it’s a moving mic method approach with periodic pink noise. If you lookup some Magic Beans YouTube videos, you can see the mmm method in action.

 

[Pareto Pragmatic]

I’ve been experimenting with a middle-of-the-road approach: 700 Hz / -2 dB / Q 2.0.

I have issues that are not normal (issues at 700 will be very much from a floor resonance for me at high SPL), but I do find that trying to reduce such issues is better than trying to totally correct them. And about half a full correction is where I often land. And if things sound good, just be happy with that and listen. Measurements matter, but your preferences are what really matters as the primary goal.

One suggestion. Try pulling down 600 and 900 with narrower filters. You may or may not get better results.

 

[CB89]

@Opalius
That looks pretty decent. And it's great that you went to the trouble of multi-point measurement for more meaningful results. I can still highly recommend using MMM for measurement, as it significantly reduces the effort. You only need to set the settings correctly once, and then it takes 30-60 seconds per channel to generate >50 average measurements. And you can watch live as the measurement stabilizes until nothing changes anymore.

The sharp drop above 9kHz is due to the speakers? Well, nothing can be done about it, but the rest looks good.

Have you set a house curve for the low-frequency range?
I would be interested in the final result (with subwoofer) at the listening position.