Calibrating my music listening setup - what should I be aiming for?

[Marcus Aseth]

Hello everyone, this is my first post in this forum, I have a pair of kali lp6-v2 and an SVS SB1000 subwoofer hooked to my pc, which I use for gaming, movies and music listening. I am not a professional, and so I come here seeking for guidance.
I want to setup my audio properly for great audio listening experience, and so I've bought a umik-1 and took some measurement using REW (image below):

Treating my room further is not an option, so that's what I have to work with and hope EQ is enough to give me what I want.
But I have few questions about this:
First, what line should I strive to achieve? Is it a flat line?
Second, up to what frequency should I be equalizing, judging from the measurement above? Would you equalize that thing around 750Hz?
Third, in a previous attempt at EQ the end resould is always a weavy line, is that normal? What is the biggest acceptable difference in dB between two frequencies near to each other?
Fourth, given the graph above, where would you place the flat target line?
Fifth, should I only work by decreasing the gain or is ok to increase it as well?

That's all for now, hope I can get it right this time with your help

 

[JStewart]

First, what line should I strive to achieve? Is it a flat line?

Typically its a line that tilts downward from left to right (as your graph does). This is because this is the natural tendency of speakers in rooms.
Ultimately its going to be what sounds best to you with your equipment in your room because there are far more factors in play than just a frequency response graph.

Second, up to what frequency should I be equalizing, judging from the measurement above? Would you equalize that thing around 750Hz?

Not enough info to answer the 750Hz thing. A good starting place, however, will be to remove the response peaks only because we are more sensitive to peaks than dips. I'd personally start with the peaks at 48Hz, 180Hz, and 14000Hz. It also looks like above 5000Hz the response rolls off at a much faster rate than I'd expect given the known measurements of this excellent powered monitor speaker. Some possible reasons for this to consider... Was the UMIK calibration file loaded in REW when the measurement was taken? Where the speakers pointed at the microphone or off axis? Do you get the same result if measuring each speaker separately vs both play together? Are there any controls on the Kali that are engaged that would cause this?

Third, in a previous attempt at EQ the end resould is always a weavy line, is that normal? What is the biggest acceptable difference in dB between two frequencies near to each other?

a) What is a weavy line? b) I doubt there's an answer to that question.

Here's a couple more bits of tid you may find useful:

Erin's Klippel measurements for your speakers:

Kali Audio LP-6v2 (Second Wave) 2-Way Studio Monitor Review

Kali Audio LP-6v2 2-Way Studio Monitor Review

www.erinsaudiocorner.com

If you're unfamiliar with the interpretation of the measurements follow the link to Erins YouTube video on the page above. And of course there's the preeminent text by Dr. Floyd Toole who, along with a few others, pioneered the research in loudspeaker behavior in rooms and human preferences.

Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms (Audio Engineering Society Presents): Toole, Floyd E.: 9781138921368: Books

Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms (Audio Engineering Society Presents) [Toole, Floyd E.] on Amazon.com. *FREE* shipping on qualifying offers. Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms (Audio Engineering Society...

www.amazon.com

You might also have a look at this:

A collection of loudspeakers measurements

Type Kali in the searchbox and you'll find eq settings for your LP6 v2 to correct the anechoic reponse. You are fortunate to have this information. In your shoes I would use this Eq information, then measure with it applied, then cut the remaining low frequency peaks caused by room interaction. In theory, that's the best its going to sound. Edit: Using PEQ. There's always DSP and convolution....

Good luck in your audio journey

 

[Marcus Aseth]

I'd personally start with the peaks at 48Hz, 180Hz, and 14000Hz.

You mean 1400, right?

It also looks like above 5000Hz the response rolls off at a much faster rate than I'd expect given the known measurements of this excellent powered monitor speaker. Some possible reasons for this to consider... Was the UMIK calibration file loaded in REW when the measurement was taken? Where the speakers pointed at the microphone or off axis? Do you get the same result if measuring each speaker separately vs both play together? Are there any controls on the Kali that are engaged that would cause this?

Yes, I had the 90° cal file loaded, the mic was pointing up and I took a single point measurement of the whole system. When I measure only 1 speaker alone, that line becomes relatively flat without any downward slope. No EQ applied from the back of the kali.

The problem is that I don't know what measurement to trust. So for example, thinking that downward slope would disappear with the moving of the mic of even few centimeters, I've attempted a moving mic measurement (swapped to the 0° cal file and pointing the mic horizontally), but I get bogus results, for instance in the image below from the MMM not only there are a lot of up and downs that are not present in the single point measurement image shown above, but also the levels slope downward like crazy from 250Hz and above... I simply don't know what measurement to believe and what's the point of me randomly EQing on some measurement that's wrong to begin with... x_x
So there is clearly some measurement problem here, until that gets fixed I cannot proceed x_x Help please

 

[JStewart]

You mean 1400, right

Yes.

The RTA settings for MMM (top right on last chart) are off. Please try:
Mode: 1/48 octave
Averaging: Forever
Window: Rectangular

Use 90 deg
The signal generator should be set to Pink Noise PN as opposed to just pink noise.

For the original roll off question, please measure each speaker separately and post the REW mdat file. Maybe someone will see something of note.

 

[Marcus Aseth]

Yes.

The RTA settings for MMM (top right on last chart) are off. Please try:
Mode: 1/48 octave
Averaging: Forever
Window: Rectangular

Use 90 deg
The signal generator should be set to Pink Noise PN as opposed to just pink noise.

For the original roll off question, please measure each speaker separately and post the REW mdat file. Maybe someone will see something of note.

thank you for all the help ^_^ Just one clarification: if I use the 90° cal file for the MMM does it mean I need to keep the mic pointed up, as opposed to a youtube video in which they do it with the mic held horizontally?

 

[JStewart]

if I use the 90° cal file for the MMM does it mean I need to keep the mic pointed up

Yup. Take a look at your cal files sometime if you want to have a better idea of what they are doing. If you think you'll be less likely to interfere with the measurement if using the 0 deg file then there's no issue. I assumed you were holding the mic in your hand when taking the measurement.

 

[Marcus Aseth]

The signal generator should be set to Pink Noise PN as opposed to just pink noise.

I cannot find this option you mention, could you share a screenshot of it? ^_^

 

[skid00]

I cannot find this option you mention, could you share a screenshot of it? ^_^

You want the Noise tab, then Pink Periodic in the second row of tabs, IIRC.

I point the mic directly at the speaker I'm testing, NOT at the ceiling (I'm also doing some EQ at higher frequencies).

I slowly move the mic around, about a two-foot diameter circle with it's center where my head would be. There are issues with the chair reflections, when doing this.

Sometimes I use a monopod to get the mic further from my body, to avoid those reflections.

Make sure your sequence lengths are correct (I don't IIRC this either! ).

 

[ddude003]

This could be some fun reading for you @Marcus Aseth... https://www.bksv.com/media/doc/17-197.pdf Especially figure 5 which @DanDan refers to below...

 

[DanDan]

Measuring a single point with two sources can suffer from HF cancellation, which two ears combined do not. So measure L, R, L+R. Regular listener conditions and preferences have been surveyed over the years. They have remained remarkably consistent, e.g. a 6dB Tilt between LF and HF. Check out the Bruel and Kjaer and Harman Target curves.

 

[Marcus Aseth]

Regular listener conditions and preferences have been surveyed over the years. They have remained remarkably consistent, e.g. a 6dB Tilt between LF and HF

@DanDan do you happen to have the recommended curve as a .txt I can import as an EQ target into REW?
Also, is the curve shown in Fig.5 of @ddude003 link the recommended one?

Also, I was taking a look at the correction file computed from ErinsAudioCorner data which @JStewart suggested to use before EQ, this file here -> https://raw.githubusercontent.com/pierreaubert/spinorama/develop/datas/eq/Kali LP-6v2/iir-autoeq.txt
my question is, if I go for a target house curve what does it matters if I apply the correction or not? In the end, I would be still overwriting the correction changes in order to match the curve, right?

 

[JStewart]

I would be still overwriting the correction changes in order to match the curve, right?

Yes you would be.
You might try both methods and see if you have a personal preference.

If you search around you’ll find advocates for both full range correction to a house curve and for correcting only the room interaction below the transition frequency and the anechoic response above or not at all.

Interestingly, while Dr. Toole clearly is not an advocate of full range correction based on science, @Mitchco of

https://accuratesound.ca/services.html

said somewhere that his clients overwhelmingly choose full range correction. I’m not qualified to guess why that might be.

You‘ll be able to derive the B&K curve from this paper:

https://www.bksv.com/media/doc/17-197.pdf

 

[DanDan]

You can Google the B&K or Harmon curves, but I can tell you that I have settled on 0 at 30 +3 at 100 0 at 2K -3 at 10K -6 at 15K. There is the matter of taste and individual speakers and rooms but my goal is professional, translation to the real world from my Studio CR. This works for me, spectacularly well.

 

[Marcus Aseth]

Make sure your sequence lengths are correct

not sure what this means ^_^'

 

[Marcus Aseth]

ok, got around to do the MMM with the recommended settings, here's the result for the full system without EQ and full system with EQ:

NO EQ -

WITH EQ -

I've also took a single point measurement (the image below) of the EQ'd system and it looks a bit different, mainly the downward slope is not noticeable, but I suspect that's because it's a single point measurement and there is a lot more variability in the higher frequencies and the way the two speakers interact with each other, so if I where to take maybe 9 single point measurement and average those, I would end up with the image above from the MMM.

Now, the thing that is giving me concern is the Distortion graph:

I don't really know how to interpret it, but it looks wild... is that normal?! I'll share the Single Point Measurement file for anyone that wants to check more closely (below in the attachments)

 

[DanDan]

I made the same mistake recently. The Distortion percentages are read in the bottom panel when you slide the cursor over that SPL graph.

 

[Marcus Aseth]

I made the same mistake recently. The Distortion percentages are read in the bottom panel when you slide the cursor over that SPL graph.

I'm aware of that, but my issue here is that I have no clue about what to make of those numbers - I was hoping someone that saw several other distortion graphs could tell me how bad that is, if I need to do something about it and what, if it is completely normal and I should ignore it, etc...

 

[John Mulcahy]

It isn't unusual to see high harmonic distortion from subs at low frequencies and high SPL. That measurement is fairly loud at nearly 90 dB, which is getting to the distortion-limited peak SPL of the sub at 20 Hz. If you are boosting the sub's response with EQ at the low end that will likely be making things a bit worse.

 

[JStewart]

Well, John's an expert. I'm not, but having seen a bunch of distortion graphs there's nothing that really jumps out. The 550 and 3kHz peaks are a bit interesting even if not likely audible. @John Mulcahy , could they be caused by something vibrating in sympathy with the speaker, aka resonance?

@Marcus Aseth , a good sanity check might be to measure a single speaker and compare it to Erin's distortion data for your Kali.

 

[Marcus Aseth]

@Marcus Aseth , a good sanity check might be to measure a single speaker and compare it to Erin's distortion data for your Kali.

Thanks, I'll try that, also I don't understand why the measurement is at 90db when I always make sure to set the volume to 75db before any measurement... was a mistake on my part? I doubt it, but I'll try again later (can't right now, everyone's sleeping)

 

[Mitchco]

If you search around you’ll find advocates for both full range correction to a house curve and for correcting only the room interaction below the transition frequency and the anechoic response above or not at all.

Interestingly, while Dr. Toole clearly is not an advocate of full range correction based on science, @Mitchco of
https://accuratesound.ca/services.html said somewhere that his clients overwhelmingly choose full range correction. I’m not qualified to guess why that might be.

Using Frequency Dependent Windowing (FDW) is the key. Using a FDW for a full range frequency correction actually meets the criteria of your top paragraph. One needs to look at the math to understand why:

 

[DanDan]

The Distortion graphs look normal to me. High levels of distortion at LF are fairly inaudible. There are a few possible gain changes so that 90dB may not be accurate.

 

[sam_adams]

@Marcus Aseth, as @John Mulcahy pointed out, what you are seeing at low frequencies is that the sub is distortion limited—most likely as a result of bass boosting at low frequencies pushing the driver out of the linear region of its excursion. Once the driver reaches a value of approximately 10% THD, it usually is beginning to reach its Xmax limit. You can see at what frequencies the distortion is occurring by switching the setting 'Plot harmonics at the harmonic frequency' on and off. (Noise floor in blue and fundamental in black at 100% below.)

'Plot harmonics at the harmonic frequency' OFF:

'Plot harmonics at the harmonic frequency' On:

You can easily see the 2nd and 3rd harmonics (red and gold traces respectively) are exceeding 10% distortion.

 

[Marcus Aseth]

So I'm currently stuck in this eternal cycle of taking single point measurements, equalizing for that, then taking a MMM and now the equalization looks wrong, so I equalize for it, and now the single point measurements look bad... I cannot get the measurements to coincide in order to guide me towards the correct EQ...

The blue line on the bottom is the MMM, the 3 on top are 3 single point measurements, the green is the center (listening position), the blue is 10 cm to the left of it, the orange 10cm to the right.
And so the MMM shows that I lose something like 3db in the region from 30 to 38 Hz, but in the 3 single points measurements I don't have that severe of a problem in there...
I don't know which measurement I'm supposed to believe... x_x

 

[Marcus Aseth]

I did all 4 measurements again, and it feels like I'm being trolled by the results:

If I EQ based on the Single Point Measurements (the purple highlighted is the average of them), then I might try and boost 655Hz a bit, but if I mind the Moving Mic Measurement, then I would remove some gain at that frequence... and that loss of several db around 30-40Hz is still present, which is extremely confusing since I managed to EQ in a way that that problem was fixed in all 3 SPM...
Who is lying to me here?!

EDIT: it is worth of notice that during the MMM I am in the room, and I leave for the SPM, but is my body affecting the result that much?! but I am in the room when I listen to the music, does it means that the MMM is more correct?
Should I take the SPM to the left and right of my listening position when I'm sitting there then?