Embarking on REW/Roon DSP exercise

[ahofer]

I’m about to get going on a measurement-to-convolution file exercise, now that I have Roon adequately powered in my home. I would welcome any suggestions, and I will try to post my results ASAP.

I ran a few initial sweeps last night, and they look awful. I can’t yet figure out how to put together the filters, but I am googling about for answers. Starting here:

https://community.roonlabs.com/t/a-...w-and-rephase-using-convolution-filters/90990

My cat has taken the listening position, however:

 

[Bombadil]

I’m about to get going on a measurement-to-convolution file exercise, now that I have Roon adequately powered in my home. I would welcome any suggestions, and I will try to post my results ASAP.

I ran a few initial sweeps last night, and they look awful. I can’t yet figure out how to put together the filters, but I am googling about for answers. Starting here:

https://community.roonlabs.com/t/a-...w-and-rephase-using-convolution-filters/90990

My cat has taken the listening position, however:
View attachment 88156

Love the coffee table with butterfly. walnut? oops, might be off topic again

 

[ahofer]

Love the coffee table with butterfly. walnut? oops, might be off topic again

Yes, recovered Walnut from a guy in Brooklyn. https://www.fallenindustry.com/

 

[ahofer]

Bear with my newbie comments.

First cut after a few sweeps: notching down room modes at 65, 128 Hz., then shelved up bass under 140.

There was another measured bump around 28(the doors behind my LP open wide into another room for a 40+ foot total reach). I’m having trouble believing it, given the limits of the Harbeths (port?). I can program all these into my RME and leave room in the Roon for other things. Current Roon EQ settings below.

The troughs I observed in the higher frequencies have me a bit confused, particularly the 200Hz. Not sure how to set the Q.

A fair amount of comb filtering higher up, but I believe the conventional wisdom is not to treat that with EQ.

This does sound better, but I may have overdone it with the shelf - perhaps a touch less definition, but much clearer low end in strings. I’ll try to post the REW results later.

 

[DWPress]

Posting REW results will get you much more constructive feedback from the forum. Have fun!

 

[JoachimStrobel]

You may want to decide your main operating mode: Curtains open or close, AC on/off, table loaded, potentially placed between the speakers while in listening mood. Seriously, whatever you measure, you need the very same environment every time to compare results.
I see a guitar there? You could play it in your favorite hall. Analyse some notes via REW and do the same in your living room to check on your favorite Room curve - just a thought.

 

[ahofer]

You may want to decide your main operating mode: Curtains open or close, AC on/off, table loaded, potentially placed between the speakers while in listening mood. Seriously, whatever you measure, you need the very same environment every time to compare results.
I see a guitar there? You could play it in your favorite hall. Analyse some notes via REW and do the same in your living room to check on your favorite Room curve - just a thought.

Not to mention the pocket doors behind the LP ....

Funny, everyone mentions the guitar, but not the viola in pro carbon case and antique music stand on the left.

The air conditioner is obviously a seasonal thing.

 

[ahofer]

OK, not that I know exactly what I've got, but I have measurement files based on:

-8 sweeps center LP + average
-4 sweeps pointed right from LP+ average
-4 sweeps pointed left from LP + average
-average all of the above (picture below)

You should be able to download all the mdat files here --> https://drive.google.com/drive/folders/1p1NmYm9HIGOwNcZxupG7UcHWeTFd1gd9?usp=sharing

I'm still a little confused about the volume levels I should be targeting with the sweep, but you can see the room modes I noted in my earlier post.

Higher up around 16khz it goes pretty wonky, but I can't hear up there. I presume the sharp downticks starting around 240 are comb filtering?

 

[ahofer]

What I mean by center and left.

 

[ernestcarl]

I neither use Roon nor convolution. Looking at your measurements, the high frequencies above 1kHz looks severely shelved for whatever reason(s).

But here's some auto-generated PEQs using REW that you could test if you want:

I'd prefer to avoid narrow Q filters myself and use MMM -- REW will generate only four or five simpler filters if 'vary max Q above 200Hz' checkbox is On.

There wasn't much of a big difference in the lower half of your graphs so I just averaged all three.

You can fill-in more of the wider bass nulls using the manual controls if desired.

If you xo with a SUB(s) with optimal placement and phase alignment you will be able to satisfactorily fill-in those nulls below 100Hz with less distortion.

 

[ahofer]

I neither use Roon nor convolution. Looking at your measurements, the high frequencies above 1kHz looks severely shelved for whatever reason(s).

But here's some auto-generated PEQs using REW that you could test if you want:
View attachment 88489
I'd prefer to avoid narrow Q filters myself and use MMM -- REW will generate only four or five simpler filters if 'vary max Q above 200Hz' checkbox is On.

There wasn't much of a big difference in the lower half of your graphs so I just averaged all three.
View attachment 88490

View attachment 88491
You can fill-in more of the wider bass nulls using the manual controls if desired.

If you xo with a SUB(s) with optimal placement and phase alignment you will be able to satisfactorily fill-in those nulls below 100Hz with less distortion.

I have to learn how to auto-generate. I wondered as well about that shelf. Not clear why.

 

[ernestcarl]

I have to learn how to auto-generate. I wondered as well about that shelf. Not clear why.

Presuming 'center' (teal? colored line) is a sum L+R measurement, then it could be partly just some measurement artifact from phase cancellation. As for the left and right by themselves, could be the crossovers going bad? Like this:

Just one possibility.

 

[ernestcarl]

Ah.... I missed reading your post#9

Don't know how many bands Roon's PEQ option allows, but looking at the vector avg of your summed LCR measurements closer in the 'wavelet view' of REW's spectogram tab, there are only four main modal peaks that badly need to be EQ'd. Thus, we can simplify and minimize the number of PEQ bands required:

The high frequencies above 2-3kHz looks like its almost completely gone in the above view. There is something seriously wrong here. Could be the tweeters, crossovers, or the amplifiers at fault -- the speakers look pretty vintage so... a closer check-up might be necessary.

 

[ahofer]

Many thanks for your generous help with this!! Especially @ernestcarl

I've done a bunch more measurements this morning, both with the UMIK and a hand-held spl meter at various distances and both using and not using the UMIK calibration file. The results are inconsistent. One problem is clear - my first batch of measurements were done with only one speaker playing (not L+R in the settings). But even with both speakers I have measurements that have the shelf and measurements that don't. so...I have to work now, but I'll get back to it.

the handheld SPL with test tones indicated results from the LP at the +/-2 dbA at levels on either side of the 'shelf' so did not produce evidence of the phenomena in REW.

Suddenly I was getting low level warnings as well. At pretty healthy volumes.

UPDATE: I added some normal center readings from this morning to this, one using mic calibration and one without:

https://drive.google.com/drive/folders/1p1NmYm9HIGOwNcZxupG7UcHWeTFd1gd9?usp=sharing

But I also got some even worse results, which I'm not posting yet. Is it possible I'm triggering some kind of protection?

 

[ahofer]

OK, As I do my reading, I see what I’ve been doing wrong. I need to measure one speaker at a time. When I get a chance I’ll re-do. The room modes, which I’ve now canceled out, should be similar, but the rest very different.

Apologies.

 

[ahofer]

So, just for comparison, I did a bunch of measurements of a second system, with KEF LS50W. They are behind me with a bed in between, and since I am now a standing-desk person, I am way above the driver axis. I added some close-up measurements for comparison.

They are all saved in one mdat workbook - KEF LS50NYC.
https://drive.google.com/drive/folders/1p1NmYm9HIGOwNcZxupG7UcHWeTFd1gd9?usp=sharing

Showing in the capture are the L, R, and both channel measurements from my desk (blue, light green, orange and brown=avg), a close L measurement (purple), and a 5ft measurement Left in dark red. Also a KEF EQ'd measurement (Avg) in dark green.

Regardless of position, it looks to me like there is still a shelf at 1Khz., and a weird series of dips in higher frequencies.

I'm confused by the high frequency dips/peaks, and the overall lower level above 1Khz.

 

[RayDunzl]

I'm confused by the high frequency dips/drops, and the overall lower level above 1Khz.

Odd...

The spacing of the "waves" in the trebles response.

A thousand cycles, another thousand, then two...

 

[ahofer]

I tried elevating the speakers. Here's the result - same pattern.

REW keeps telling me the level is low, but this was really loud, and there was like 40-50db of headroom. I don’t dare go louder.

And the room as I did it:

 

[ahofer]

New measurements from the living room, all from LP. Along with suggested overdone EQ.

https://drive.google.com/drive/folders/1p1NmYm9HIGOwNcZxupG7UcHWeTFd1gd9?usp=sharing

Update: I programmed all these in for fun. You wouldn’t believe what crap it sounds like. Or maybe you would.

 

[ernestcarl]

Occasionally, REW will tell me the levels are off as well while it not being so... make sure the microphone cable and connections are good -- unplug and replug, or try another cable... -- as well as try restarting the program.

Just looking at the per channel measurements the treble is louder now. Can you try measuring each tweeter in the nearfield at half a meter or less? Tweeter axis is fine... Since both look similarly problematic in the high frequencies either left or right channel is fine -- both speakers if you can.

It looks a bit like combing filter, but one should not see this much in the nearfield esp. when only measuring a single channel exclusively.