REW - Can anyone make sense of my measurements?

[Sparky]

Hi All.

I'll be honest from the start, although I've been taking measurements from the start of my obsession with audio, I've never truly understood what I'm actually doing.
My set up is very simple, a pair of Genelec 8341's with a GLM kit and a pair of Arendal 1723 1S subwoofers running in dual mono.
I run everything through a MiniDSP SHD

I have recently been able to have some time measuring my room (which is rare these days) and have decided to measure across the listening position as opposed to just one MLP measurement (which I always used to do) as I feel multiple positions will allow for a more realistic portrayal of what is actually going on.
Doing this has presented me with a problem however, I have no idea how to combine the measurements correctly and, after doing that, then time align the subs together individually with positive summation as one big sub. Then obviously, time align the subs to the monitors.

I have taken measurements of both my subs at 5 positions across the couch full range (0 to 20,000hz) with no crossover as I don't know which crossover would be more beneficial in relation to how they behave with my monitors.
My mains have been measured across the same 5 positions using GLM full range (without extended phase linearity.
I also measured my mains using various crossovers using the same GLM calibration but didn't run new sweeps which resulted in a minor drop in SPL for those additional measurements for some reason?

Anyway, I have attached a link to the MDAT files if anyone on here has the time, inclination and experience to make sense of it all.

Thank you.


Google Drive REW

 

[Sparky]

The measurements for the mains have been taken at the MLP only as those measurements have already had a multi-seat GLM calibration applied to them.

 

[Keith_W]

Firstly, you have included too many measurements. I have culled them to these four:

Main speakers full range left and right, and subs left and right, measured from the MLP. All were taken with an acoustic timing reference. I note there are no measurements of your subs together with your speakers. You also posted your main speakers with various different HPF's and I don't know which one you are using, so I simply took the full range measurement.

To see whether there are any bass problems, we want to see your main speakers playing together. So I vector summed L and R (yellow), and I staggered the result for clarity. You can see an obvious problem, there is a huge hole in the frequency response between 70-140Hz. This dip is not visible in the individual measurements, meaning it is phase cancellation. Otherwise, the performance of your speakers alone with no sub is pretty impressive.

One option to fill that hole is to give your subs a very high LPF, you will need them to go up to 150Hz at minimum. But this is very bad, it would make the location of your sub audible for sure.

It is far better to try to get rid of that dip in the first place by finding a better position for the speaker or for the listener. You did not post a picture of your room, so I don't know how you have things set up.

... but if you ask me to guess, I am going to say that you have your speakers on a study desk or some kind of cabinet. Look at that huge forest of early and loud reflections. Some of those reflections are coming from 15cm away from the speaker. You want all those reflections to be >-15dB to the main impulse.

That step response gives us a hint as to why you have that huge dip. There should be no peak in the tail of the step response that is larger in amplitude than the main impulse. You have many. All these are low frequency reflections (you can tell because of how wide they are) and they are louder than the direct sound of the loudspeaker! You can also see that the L/R response of the tail are dissimilar, meaning that your speakers are not symmetrically placed in your room. You can even see that at 20ms, the speakers are apparently out-of-phase with each other!

Anyway all the evidence so far supports the conclusion that your speakers are poorly placed. This is what you need to address FIRST before worrying about anything else. As a priority, you need to:

1. Identify what is causing all those early reflections and get rid of it.
2. Find out why you have that dip between 70Hz - 140Hz when L+R mains are playing together.

Look at the eBook in my sig. There is a procedure in there to diagnose dips in the frequency response. I suggest moving the speakers closer to the wall - if it is SBIR, it will push the SBIR frequency up. Or you could try pulling them out further, it will drop that dip down towards a range which you can address with your subwoofers. Where it is right now, it is very challenging to address.

There are other problems with your measurements, e.g. the subs are not time aligned. But we won't worry about that for now, fix the problems one at a time.

 

[Sparky]

Hi Keith. I'm so glad you too this on as I know you know exactly what you're talking about.
We can't really do anything with placement right now so this is what we have to work with I'm afraid.
We do plan to install a media wall in the near future which will allow for the speakers and subs to be positioned much more symmetrically (subs too).
These measurements are from a typical UK lounge so small room and non-symmetrical.

This is why I included some measurements at different crossover frequencies as I really wasn't sure which was the best crossover to use considering the huge hole between 70-140. It really is a massive hole too.

I'm not really sure how to move forward really and select the best x-over for my subs and speakers until we can get the media wall built.

 

[Sparky]

I did download your e-book and had a read a couple of weeks ago. Really informative but I'm so inexperienced that it's like learning a new language!
I found the shortened version a bit more to my understand but even that is pushing it!

 

[Sparky]

Looking a bit further into it, the only way to get rid of the suck-out is to apply either the 90 or 100hz x-over onto the speakers.
It's a shame to have to do it really as they play really well down low.
I'm not sure you'll agree though?

 

[Sparky]

Attached are a series of vector averaged sweeps. These 3 VA's seem to allow me to avoid the nasty dip and fill it in with the subs.
I tried a 70hz VA too but that reintroduced the start of the dip so left it out.

 

[Keith_W]

Would you be able to show us a picture of the room? Or better still, a diagram with dimensions drawn in?

Re: those vector averaged sweeps. You left crucial information out. What are you vector averaging?

BTW the reason I participate in these threads is because I learn from your measurements as well. The learning goes two ways, sometimes I see something really baffling. I live for those moments

 

[Sparky]

Would you be able to show us a picture of the room? Or better still, a diagram with dimensions drawn in?

Re: those vector averaged sweeps. You left crucial information out. What are you vector averaging?

BTW the reason I participate in these threads is because I learn from your measurements as well. The learning goes two ways, sometimes I see something really baffling. I live for those moments

I'll draw out a quick sketch in a minute.

The vector averages are from measurements I took across the couch. I took 5 measurements of Left and Right mains, I then cross corr aligned them and produced a vector average of each. Is that correct?

Well I'm more than happy to oblige you Keith!

 

[Sparky]

Hope this helps mate. If you need more detail, I'll do my very best for you.

 

[Sparky]

Sorry. I made a mistake. the vector averages are the mains combined at certain x-overs in order to find the optimal crossover to get rid of the dip.
I do apologise if I've caused any confusion here.

 

[Sparky]

So just to be clear. The MDAT files in Google Drive are as follows:

Subwoofer Left and Right measured in 5 locations across the couch.
Mains measurements Left and Right from MLP only as GLM has already EQ's for the 5 locations across the couch and so measuring again across the couch would carry no benefit.

I hope that makes sense.

 

[Keith_W]

I'll draw out a quick sketch in a minute.

The vector averages are from measurements I took across the couch. I took 5 measurements of Left and Right mains, I then cross corr aligned them and produced a vector average of each. Is that correct?

To be clear, that vector average are measurements of the L and R speakers PLAYING TOGETHER across 5 positions in your listening area?

You did not take the most useful measurement, which would be the full range measurement. I mean, I said that the frequency range of concern is 70Hz - 140Hz.

Look at your measurements. The roll-off starts at 150Hz. I am willing to wager that that 70 - 140Hz dip is still there. You just failed to measure it.

As for your diagram, can you please state whether the distance from your head to the rear wall equals the distance of the loudspeaker to the front wall? It certainly looks it from that diagram. And would you be able to measure that distance?

And can you clarify what we are working with here. Have those measurements already been equalised with DSP, or are they raw measurements?

 

[Sparky]

As per previous post:

So just to be clear. The MDAT files in Google Drive are as follows:

Subwoofer Left and Right measured in 5 locations across the couch as follows:
Left - Left Mid - MLP - Right Mid - Right all measured full range.
Mains measurements Left and Right from MLP as above only as GLM has already EQ's for the 5 locations across the couch and so measuring again across the couch would carry no benefit.

The crossovers for the mains are applied within the GLM software and so are limited to how the GLM software applies the crossovers.
I have no control over this it seems as applying a crossover in the miniDSP doesn't seem to work as the DAC in the speakers has the final say.
This is one aspect of the Genelecs which frustrate me. You don't have full control in how crossovers are applied as they are designed to work with other Genelec products.

We/I sit on the couch which is 3.14 metres from the drivers if sat in the MLP.
The wall behind isn't actually a wall but a window which is approx 0.5 metres from the back of my head when sat down.

I hope that makes sense.

 

[Sparky]

Genelec seems to be applying a crossover from 100hz with varying degrees of filter type?

 

[Keith_W]

As per previous post:

So just to be clear. The MDAT files in Google Drive are as follows:

Subwoofer Left and Right measured in 5 locations across the couch as follows:
Left - Left Mid - MLP - Right Mid - Right all measured full range.
Mains measurements Left and Right from MLP as above only as GLM has already EQ's for the 5 locations across the couch and so measuring again across the couch would carry no benefit.

That is far from clear. I asked whether DSP has been applied. The answer should be yes or no. And it's not clear to me why you think that measuring again would be of no benefit. If you have applied DSP, then you want to see if it has worked, right? Otherwise, why are you taking these measurements?

The crossovers for the mains are applied within the GLM software and so are limited to how the GLM software applies the crossovers.

I have no control over this it seems as applying a crossover in the miniDSP doesn't seem to work as the DAC in the speakers has the final say.

Well, you seem to be able to measure your mains full range. Which means you can use your MiniDSP to apply crossovers.

I don't know much about GLM. Maybe if I read their manual, I would have some idea what they are trying to do. Some kind of spatially averaged correction, perhaps. In any case, you can choose to use your MiniDSP, or you could use GLM. I personally dislike "black box" DSP solutions - i.e. measurement goes into the black box, correction comes out. Those things tend to take control away from the user. If it was me, I would use the MiniDSP. But really the choice is up to you.

We/I sit on the couch which is 3.14 metres from the drivers if sat in the MLP.
The wall behind isn't actually a wall but a window which is approx 0.5 metres from the back of my head when sat down.

So: speakers are 1m from the front wall, and listener's head is about 0.5m from the rear wall/window.

SBIR cancellation wavelength = 4m. That equals 86Hz.
LBIR cancellation wavelength = 2m. That equals 172Hz.

It is possible that that 70-140Hz dip is a combination of SBIR and LBIR. Can you push your speakers closer to the front wall?

 

[Marc v E]

Imo it would sound instantly better if you'd turn your room layout 90 degrees to the left:

Speakers to the left side,
Couch to the right, at least 50 cm from the wall,
Table where your couch is now.

Much of sound quality is layout ime.

 

[Sparky]

That is far from clear. I asked whether DSP has been applied. The answer should be yes or no. And it's not clear to me why you think that measuring again would be of no benefit. If you have applied DSP, then you want to see if it has worked, right? Otherwise, why are you taking these measurements?



Well, you seem to be able to measure your mains full range. Which means you can use your MiniDSP to apply crossovers.

I don't know much about GLM. Maybe if I read their manual, I would have some idea what they are trying to do. Some kind of spatially averaged correction, perhaps. In any case, you can choose to use your MiniDSP, or you could use GLM. I personally dislike "black box" DSP solutions - i.e. measurement goes into the black box, correction comes out. Those things tend to take control away from the user. If it was me, I would use the MiniDSP. But really the choice is up to you.



So: speakers are 1m from the front wall, and listener's head is about 0.5m from the rear wall/window.

SBIR cancellation wavelength = 4m. That equals 86Hz.
LBIR cancellation wavelength = 2m. That equals 172Hz.

It is possible that that 70-140Hz dip is a combination of SBIR and LBIR. Can you push your speakers closer to the front wall?

My reasoning behind not carrying out multiple sweeps across the couch with the mains is because the GLM software takes care of all of that for me with its own calculations.
I carried out the exact same measurements with the GLM microphone first across the couch (GLM doesn't recognise any third-party mic) as I did with my subs and the GLM software gave me the optimal FR for the MLP as you saw.
The only issue here is the dip as we've discussed but I think that is down to just very poor placement of the mains.

I don't really have much wiggle room when it comes to placement. I might be able to move the speakers back toward the front wall a touch but it won't be much.

I think for now, I just have to work with what I've got.

If you feel I should re-measure, please let me know what you are looking for so I can give you the exact measurements you require.

Let me know Keith. I certainly don't want to waste your time by having to use inappropriate measurements that don't meet your needs.

 

[Sparky]

Imo it would sound instantly better if you'd turn your room layout 90 degrees to the left:

Speakers to the left side,
Couch to the right, at least 50 cm from the wall,
Table where your couch is now.

Much of sound quality is layout ime.

Hi marc,

We are installing a media wall and integrating the mains and subwoofers into it over the summer so the room will be being rotated 90 degrees.
I agree with you, doing this will drastically improve the situation as I'm well aware that the current speaker/sub positions are terrible.

 

[Keith_W]

Look at your measurements. The roll-off starts at 150Hz. I am willing to wager that that 70 - 140Hz dip is still there. You just failed to measure it.

Here, i'll show you. This is a comparison of your vector averaged measurements that you posted here (blue) vs. the sum of L+R main speakers from the earlier measurements in your original post (green).

It is like I said, that 70-140Hz dip is probably still there, but you failed to measure it.

Right now, we do not know whether GLM has done a proper job integrating your subwoofers to your mains. Just to be clear, this is what you need:

1. Mic at MLP. The GLM correction and XO's need to be applied. L and R speaker, and L and R sub individually with a timing reference. Then all speakers and subs playing together. That's five measurements.
2. All speakers and subs playing together at the four remaining positions across the listening area. Do not vector average them, just post the individual measurements.

We also don't know whether you could do a better job with your MiniDSP and manual correction. If you want to find out, this is what you need:

3. Mic at MLP. TURN OFF GLM AND ALL XO's. Full range measurement of L and R speaker, and L and R sub, all with a timing reference. That's four measurements.
4. Once the DSP has been designed, you will need a verification measurement. This will be: all speakers and subs with DSP applied playing together at 5 positions across your listening area. Then you look at the deviance and decide if you want to refine it. Obviously this is not something you are going to do now.