New REW measurements, hope knowledgeable people will help me analyse them.

[gorman]

After what's been for me a long journey, I've now more or less settled on what sounds good to me. I'm fully aware that sound can always be better, as such I've taken measurements as best as I could, trying to follow the guidelines kindly provided here by @Keith_W . Measurements were taken at MLP with sweeps, Umik-1 pointed at the ceiling, using the 90 degrees calibration file.

5.3.4 system. Equipment is AVR Denon X3800H, with Dirac Live DLBC license. KEF R7 Meta powered by Audiophonics MPA-S250NC XLR amplifier (rest of the speakers are powered by the AVR), KEF R6 Meta for center speaker, Dynaudio Audience 42W for surrounds and Polk Atrium 5 for Atmos effects, ceiling mounted (I've not measured Atmos, but I'm satisfied by the performance and "holography" of Atmos content in my system). Subwoofers are two frontal (see picture) SVS SB1000 and one side (left side) SVS SB2000.

I repost some info and a picture about my room, previously written for a different thread.

Room is 4.85m x 3.72m (15.9 x 12.2 feet), ceiling at 2.9m (9.5 feet) .
Ceiling and walls are acoustically treated with panels covering a considerable portion of the space (but not all, I did not want a "dead" room. I've also got three bass traps, placed in two corners and one wall (had to work around "normal life" concerns, much as this room is more or less dedicated), they're pretty small by the standards I see around here.

This is how the front looks like:

The third subwoofer (SVS SB2000) is placed along the left wall, facing the sofa where I sit. Behind the sofa there's a bookshelf full of books. The illustrations you see on the wall are acoustic panels too. There's others on other walls.

Front speakers are 36cm / 14" from the back acoustic panels, measured from the end of the cabinet. They are toed in 15° but are well off axis from MLP still (they'd need to be approximately 30° to be on axis).
Center speaker is really close to back wall (but it's sealed, not ported), pretty much dead pointed to the MLP, completely on axis, as stated.
I could angle it down if that is desirable. I cannot aim it higher, though.

Distance from front speakers to MLP is 220cm / 7'2.6" (Dirac "sees" them exactly equidistant from MLP), distance between the two front speakers is 230cm / 7'6.5" (measured tweeter to tweeter). Distance from centre speaker tweeter to MLP is 210cm / 6'10.6".

Target curve in Dirac is this:

I process full range, as I have speakers with different responses that I need to harmonize.

A couple of graphs, just as general reference. I'm not capable of interpreting basically anything except for FR. These are for left front speaker, measurement file contains measurements for F, R, C, SR, SL and LFE.

FR, Var smoothed

Waterfall, 1/6 smoothed

RT60 (not sure if this is useful or if the way I measured makes it useful, TBH)

Spectrogram (another graph I have a very foggy idea of how to interpret)

I attach a zip file with the full measurements. The only idea I get, from looking at the graphs, is that maybe I could change the acoustic panel absorbers behind the front speakers with proper bass trap panels (10cm or thicker).

 

[Keith_W]

I had a look at all your measurements. Looks great to me. About the only problem are all the early reflections in the left speaker. You can see it in the Energy-Time Curve, see this post on how to interpret the ETC. Given that the left speaker is pushed so close to the side wall, it's not surprising.

That RT60 looks to be low-ish, about 200ms or so. It might be appropriate depending on room volume. Do you have a lot of room treatment?

I would not be changing anything about this setup, except to suggest you might benefit from a more symmetrical layout for your system, i.e. L/R speakers equidistant to the walls. Oh, and that Kandinsky should be hung at eye level

 

[alex-z]

First off, it looks like your subwoofers are phase aligned with the mains but the impulse peak is 10ms late. Worth trying to adjust that with distance in the AVR settings. Should bring down your group delay and improve bass quality around the crossover region.

Your acoustic treatment is too thin, I wouldn't even call those bass traps because they won't be doing anything below 100Hz. For your size of room I would actually reduce the number of panels and make them 2-3x thicker.

That deep shelving unit is causing a strong reflection for your centre channel just 1ms after the initial impulse. Pulling the speaker forward may help dialogue quality. Alternatively, covering the area in front of the speaker with 2-3" of absorption. It looks like you have some thin felt currently.

 

[ppataki]

These are great measurements indeed!!

Just a few comments from my end:
- the LFE channel has a really massive distortion (13-35%). I see the SPL is also pretty high though so that might be the cause of that. I would perform a measurement in the MLP at about 75-80dB to see how the distortion looks like (fronts + sub with the crossover and any low shelf filters engaged)
- agree with @alex-z that your GD is also massive - again, a front + sub measurement with the crossover engaged might show a more real-life picture
- the same goes for your Wavelet, there is a 75ms delay on your fronts below 200Hz - I wonder how this would look like with the fronts+sub

 

[gorman]

Thanks to all for your input.

To @Keith_W : room dimensions are in the OP, for room volume. It should be 52.3 cubic meters / 1847 cubic feet.
Acoustic treatment is ten absorbing panels on ceiling (not bass traps), size 120x60cm (about 4 feet by 2 feet).
On the walls there are two absorbing panels behind the front speakers, again 120x60 cm each. Plus another four panels (one is the "Kandinsky"), sizes are 100x100cm, 170x90cm, 80x140cm and 90x170cm.

Total surface for these is 13.82 square meters / 148.7 square feet.

I then have a tuned (70Hz) bass trap on the low left wall (you see it in the picture, below the lighthouse absorbing panel), a corner bass trap in the upper back right corner and a very small bass trap on the left wall (the only space I had available, between two windows.

Unfortunately I cannot have the speakers equidistant from the side walls, due to a desk I need to have in the room.

To @alex-z I know those are not bass traps. The only bass traps in the room are the ones I described above in this post. As mentioned, I could substitute the absorbing panels behind the front speakers with thick GIK Monster bass traps, same surface size.
I had thought about the surface in front of the center speaker. The cloth I have placed there is an indication of that. I could have something heavier there, for sure.

To you and to @ppataki I ask:
Measurements were obviously taken with Dirac active, I am unsure what REW does when outputting to LFE, because, for instance, the volume level of LFE seems far higher than what settings in the AVR would suggest. Are your observations derived from the LFE measurement or are they relevant for all measurements?
I will anyway try measuring LFE after recalibrating volume. I didn't do it because I thought the AVR leveled channels anyway (and it does for all other channels, hence why I'm perplexed).

All this considered, I don't know how I would go on measuring front + sub as ppataki is suggesting. The AVR already is doing that automatically through Dirac DLBC, unless I'm misunderstanding what the suggestion is.

 

[gorman]

First off, it looks like your subwoofers are phase aligned with the mains but the impulse peak is 10ms late.

This is also making me doubtful of the LFE measurement. There are three subs, but only one measurement. How this influences results and interacts with Dirac I honestly don't know.

 

[ppataki]

All this considered, I don't know how I would go on measuring front + sub as ppataki is suggesting. The AVR already is doing that automatically through Dirac DLBC, unless I'm misunderstanding what the suggestion is.

Maybe I misunderstood your measurements
Measurement 'L Feb 8' - is it merely the Front Left channel or does it contain already all your subs integrated with it?
I was under the impression of the first, not the latter

If that measurement contains your subs then my comments about massive GD and the massive delay below 200Hz (the Wavelet diagram) still hold true
My comment about the massive distortion is obsolete in that case, the L Feb 8 has great distortion measurements

 

[ppataki]

If you use DLBC then the massive GD and bass delay can be considered 'normal', please see this thread about it:

Excess group delay and mismatched phase response

Hello, I have moved my system to the longwall + 15inch Subwoofer and found out that major issues in the bass region have been solved. To get some perspective here is my old thread. In this new setup, I am trying to figure out measurements such as excess group delay (GD) and minimum phase...

audiosciencereview.com

(there are 7 pages but worth reading!)

 

[gorman]

Maybe I misunderstood your measurements
Measurement 'L Feb 8' - is it merely the Front Left channel or does it contain already all your subs integrated with it?

It's the latter. Considering I'm always using Dirac, I did not see the point in measuring without the subs integrated (and I might well be wrong, I'm a total newbie when compared to all of you).

I will read the discussion you linked to, as it appears to apply to my case.

 

[gorman]

there are 7 pages but worth reading!

I've read them all. Understood... Very little. Could you be so kind as to instruct on how to generate graphs to better illustrate the "normal" problem?

I would like to write to Dirac, in order to get an explanation. I thought that timing was one of Dirac's strong points.

 

[ppataki]

I've read them all. Understood... Very little. Could you be so kind as to instruct on how to generate graphs to better illustrate the "normal" problem?

I would like to write to Dirac, in order to get an explanation. I thought that timing was one of Dirac's strong points.

There are two graphs (among others) that can represent temporal information very well

1. Group delay:
You simply go to the GD tab in REW, use 1/12 smoothing and here you go. It will show you how many ms each frequency is delayed. You can see room related issues as well, for example below in my case between 100 and 200Hz. But between 70 and 25Hz this curve looks pretty good.

2. Wavelet, this shows you the Peak Energy Delay focusing on the low frequencies. So in other words, how much those are lagging behind the rest
Go to Spectrogram and configure it like this:

It shall look something like this:

Mine is around 7.6ms (I have two subs integrated manually and then Dirac Live is applied - not DLBC but the 'standard' one; this way it does not 'see' the subs since they are already integrated into the fronts properly)
Btw. a rule of thumb is that a value up to 20ms is great, and a value between 20 and 40 is still acceptable
With DLBC people usually get values well above 40ms

 

[gorman]

Btw. a rule of thumb is that a value up to 20ms is great, and a value between 20 and 40 is still acceptable
With DLBC people usually get values well above 40ms

Thanks for the "how to", I now see what you mean. I will most likely write to Dirac tomorrow to try and have their version of what's going on. I've read the seven pages of the discussion you linked hoping to come to a "Dirac explained it's like this because..." point. But it never came.

As such, since this seems to be widespread, I think customers are owed an explanation. In practice, what should I be experiencing? As I think I have stated, the system sounds good to me, so I'm not sure I want to understand but... who am I kidding? I'm here because I do want to understand. So any specific "telltale sign" I could look for, while listening to music? Or is that delay something more "general" that has to do with the "feel" of the music? Sorry for the ignorance of my words, I really am navigating blind in this sea...

EDIT: I also wish to thank you all. With these few messages and a little bit of experimenting I've now understood more of what REW allows to analyze.

 

[ppataki]

I also wish to thank you all.

You are welcome! If you are happy with the answers you can use the Like button

It is difficult to describe how this delay sounds but generally speaking the less delay you have (on the Wavelet diagram) the more precise and impactful your bass will become

And I say it is difficult because as you said it too, you already like your system as-is (which is a good thing). But it is very probable that if you heard how it sounded with much less group delay you would never want to come back to the earlier state

You can try to do an experiment with using one or two subs to manually integrate them and that would hopefully bring way better results. If you want to go that route I am happy to help with the steps

 

[ZolaIII]

@gorman switch the speaker, subwoofer position on the left or far both L & R if it looks better like that to you. In order to move speaker from left wall and improve early refractions as Keith suggested you already. Sub's won't mind and you don't need any additional space to do it. I have mixed fielding about Dirac Live and more like Audiseey for its equal loudness compensation support tho it won't work very good with regular low crossover as it is. If you don't want to do it (because price you paid for half Dirac license) at least use target for uper 70 dB SPL or how you prefer to listen to (at which SPL), don't use flat 88 dB one without equal loudness whatsoever.

 

[gorman]

You can try to do an experiment with using one or two subs to manually integrate them and that would hopefully bring way better results. If you want to go that route I am happy to help with the steps

I don't usually listen to music through a device that could handle separate subs integration. There's an old Slimdevices's Squeezebox, a CoreELEC machine... and, sure, an HTPC, but I use that pretty much only for gaming nowadays. As such, I think I'm stuck with whatever Dirac can provide by itself. Unless I buy extra hardware. Probably a simple MiniDSP? Not sure about what product would be the right one for this. As I think, if it should be a stable solution, it should be capable of "fixing" all three subwoofers.

Spectrogram (another graph I have a very foggy idea of how to interpret)

Just to check that I am understanding this correctly, @ppataki , am I right in believing I could have noticed the problem from this spectrogram I initially posted? Below about 80Hz (which is, not by chance I guess, the crossover point) the spectrogram changes noticeably. Or is this showing decay and not delay?

switch the speaker, subwoofer position on the left or far both L & R if it looks better like that to you. In order to move speaker from left wall and improve early refractions as Keith suggested you already. Sub's won't mind and you don't need any additional space to do it. I have mixed fielding about Dirac Live and more like Audiseey for its equal loudness compensation support tho it won't work very good with regular low crossover as it is. If you don't want to do it (because price you paid for half Dirac license) at least use target for uper 70 dB SPL or how you prefer to listen to (at which SPL), don't use flat 88 dB one without equal loudness whatsoever.

That's how subs and speakers were set initially. Result was significantly worse for bass. I moved the subs to the other possible position and FR was better, smoother. No doubts about this. Early refractions from left speaker are quite likely due not much to the wall but to the window that's just beyond the curtain you see on the left. I use the curtains only when I watch stuff during the day. They probably would improve refractions from the window (I suppose they would for sure, actually).

I wonder if I should do a whole different calibration in Dirac for when I have the curtains drawn. I probably should.
For sure I'm going to do something for early refractions from central speaker.

In regard to loudness compensation, I tested solutions in Dirac (there's a guy that sells a guide to try and emulate loudness compensation through custom curves). What I discovered was that, at volume levels I'm likely to listen to... not much of a difference. It's easier to set a target curve I'm happy with at the volume I usually listen to. I checked tens of times with an SPL meter and the volume I like to listen to always hovers around 75/80db (which is fine, as far as I understand regulations about noise exposure). As such, I've no need for loudness compensation.

When I watch stuff at night, I use Audyssey anyway, for Dynamic Volume, Dialog Enhancement, Audyssey LFC. And if I want to listen to music, I use either Sennheiser 800 or Focal Clear MG with equalization. As such... no need or interest for loudness compensation. But thanks for the suggestion.

What I think I'm gonna do later this afternoon is try and measure the results of my Audyssey correction. Curious to see what happens with group delay. I'll wait for that data point before writing to Dirac. I think it could be relevant in showing there's a problem (if Audyssey does not mess up too).

 

[ppataki]

I don't usually listen to music through a device that could handle separate subs integration. There's an old Slimdevices's Squeezebox, a CoreELEC machine... and, sure, an HTPC, but I use that pretty much only for gaming nowadays. As such, I think I'm stuck with whatever Dirac can provide by itself. Unless I buy extra hardware. Probably a simple MiniDSP? Not sure about what product would be the right one for this. As I think, if it should be a stable solution, it should be capable of "fixing" all three subwoofers.

You just need the followings:
- a device that can handle different volume and delay per channel + crossover (I guess your AVR should be able to do it)
- a PC running REW with a measurement mic (preferably UMIK-2 since it has its own clock source and hence more precise for time alignment measurements)

You need to align the volume for the subs then you need to perform an acoustic timing measurement with REW. Then you enter the volume and delay settings into your device then you set the crossover (80Hz) and finally play around with sub polarity and measure in REW how it looks the best.

I know it is easier said than done but that is the high-level process.

Just to check that I am understanding this correctly, @ppataki , am I right in believing I could have noticed the problem from this spectrogram I initially posted? Below about 80Hz (which is, not by chance I guess, the crossover point) the spectrogram changes noticeably. Or is this showing decay and not delay?

Yes exactly
However, the Wavelet diagram is much better for this task since it provides you with a more microscopic view

 

[ZolaIII]

@ppataki now I am afraid that the window will ring maybe decoupling from floor with accustic packs or similar could help, left speaker neads to get at least 50~60 cm from side wall if he wants to improve decay time from that side. On 2 m with sealed sub's delay should be under 10 ms (7~8) so not much of a big deal.

 

[gorman]

You just need the followings:
- a device that can handle different volume and delay per channel + crossover (I guess your AVR should be able to do it)
- a PC running REW with a measurement mic (preferably UMIK-2 since it has its own clock source and hence more precise for time alignment measurements)

You need to align the volume for the subs then you need to perform an acoustic timing measurement with REW. Then you enter the volume and delay settings into your device then you set the crossover (80Hz) and finally play around with sub polarity and measure in REW how it looks the best.

I know it is easier said than done but that is the high-level process.

The AVR can do that, sure. But in manual mode. Not with Dirac running. So all this would be... experimental? Just to understand if it could be fixed that way. But then, without Dirac, I'd have a much, much harder time correcting the three subwoofers and, as I said, I don't use my HTPC to listen to music. So... the experiment would be just for science, without a practical outcome for my case. Or am I missing something?

Because when I activate Dirac, as far as I know... hmmm... I need to check something later. If I remember correctly there's an option to NOT set up delay and levels with Dirac (in the PC program, not on the AVR). I could try activating that and see if, then, I can set them on the AVR (normally they're greyed out, with a message stating that Dirac handles them).
I could take note of current settings and copy them all, except for subs delays. Try that, measure, and see if things get better. It's just subs delay that I would need to act upon, right? That was alex-z suggestion, initially. I'm not sure I'm remembering correctly, though.

Yes exactly
However, the Wavelet diagram is much better for this task since it provides you with a more microscopic view

Ok, that's good. It's important, for me, to keep on learning stuff. Otherwise I'd just be wasting time.

 

[gorman]

now I am afraid that the window will ring maybe decoupling from floor with accustic packs or similar could help

If you refer to the left speaker, when you talk about decoupling, the speaker stands on one of these already: Aperta 200.

 

[ZolaIII]

If you refer to the left speaker, when you talk about decoupling, the speaker stands on one of these already: Aperta 200.

I am talking about the sub on that side.
Edit: and packs only on nead for a stand if you really don't want one.
This is just for illustration purpose of medium cary wage ones for reference:

https://www.amazon.com/IsoAcoustics-Orea-Equipment-Isolators-Bronze/dp/B07KGCRZ9V

You can find similar in a package and for less money. Goal is that it transfer it to very low frequency and more contained, plane little better silicone and cuple of coins does the trick.