REW - Am I doing this right? Need a few tips

[johnk]

Hi,
I have been learning how to get the most out of REW and have watched many videos and read many tutorials, but I am still a bit unclear on a few specific questions.
The results are really good so far, but maybe I can do better with a little help in creating the best filters!
Approach: I measure L and R speakers individually using a moving microphone with a UMIK-1. I use the 90 degree calibration file. I have a house curve file loaded which I found in this forum and I add a touch more bass using the "Add room curve" in the UI (I guess I am in the preference group with untrained listeners).
For target settings I have "Full range speaker" set (the speakers is a 2.5-way Wharfedale Diamond 230 which reach down to ~35Hz). The LF slope is 24 dB/octave.
Q1. I use "Calculate target level from response" to set the target level. In my room the right speaker is in a corner, so if I calculate the target level from the right speaker the level is much higher (0.4dB). I currently use the target level from the quieter, left hand speaker. Is this right or should I average the target levels of both speakers?
Q2. I can see the bass response falls off a cliff around 38Hz so I set the "LF Cutoff" to be 38Hz so the curves between the measurement and the target pretty much align. Does this sound about right? (looks pretty bad at this point, but it is a cheap speaker )

FYI This graph has 'Psycoacoustic' smoothing on. Blue is the target curve and green is the actual measurement.
Previously, I was using the "Match response to target", but have now started creating my own filters and I think I am getting better results when A/B-ing between them.
Q3. When creating my own filters I generally only cut the only exception being a boost for the dip from 1-3.5kHz. I now create a set of filters to try to fix things.

Red is with my filters applied.

The right hand speaker is much more of a problem due to its corner placement. I guess I need some form of a bass trap, but I am trying to compensate using large negative high Q filters. That peak at 103Hz is a bit of a disaster. The red line shows the predicted curve after my filters (which include a -25dB at 103Hz!)

I then load these onto my Raspberry Pi running Volumio.

I just did a quick measure to test how these work and that leaves more questions.

Left speaker. Dotted line - original, dashed line - predicted, solid line measured. I think I need to reduce the filters around 200-400Hz!

Right speaker. Dotted line - original, dashed line - predicted, solid line - measured.

All a bit of a mess, but lots of fun.
Any thoughts would be appreciated.
Thanks,
John

 

[DJBonoBobo]

A few thoughts:
- imagine the correction of room modes in the bass region as one thing and influencing the rest of the FR as a different project.
- for corrections of room modes in bass better use var smoothing or no smoothing.
- Try measuring both channels together in bass, too. below 80 Hz most signals are mono and bass becomes more or less one.
- corrections above the Schroeder frequency (around 200-250 Hz) can be harmful, and measurements in a room can be misleading. Psychoacoustic smoothing and using the MMM technique is good for this, but i would only make broad, low-Q adjustments to taste there. To get a better response in this region, it is better using good speakers, better positions and so on.
- If you have anechoic measurements of your speaker, you could correct flaws of the speaker using it (loke Amir describes in his reviews), but you can't use your in-room measurements for doing this precisely.

 

[johnk]

Thanks for all of the advice.

- imagine the correction of room modes in the bass region as one thing and influencing the rest of the FR as a different project.

I will try to work on room modes first before trying to correct other frequencies.

- for corrections of room modes in bass better use var smoothing or no smoothing.

Good tip!

- Try measuring both channels together in bass, too. below 80 Hz most signals are mono and bass becomes more or less one.

I know the placement of the right speaker is a challenge, but moving walls is difficult. I will add he combined LR measurement to my next set of tests.

- corrections above the Schroeder frequency (around 200-250 Hz) can be harmful, and measurements in a room can be misleading. Psychoacoustic smoothing and using the MMM technique is good for this, but i would only make broad, low-Q adjustments to taste there. To get a better response in this region, it is better using good speakers, better positions and so on.

I will go and read up more about this. The pre-correction vs post-correction graphs are wildly different in this region.
In terms of eq, these speakers have a 'showroom' treble boost which needs a broad low-Q filter to reduce.
I have been looking at replacing the speakers and I am leaning towards the Wharfedale Linton 85th although I am a little concerned that the low directivity will make them difficult to eq. I spent some time in a listening room with them and they did sound rather good to me. I would prefer not having to deal with integration with a sub as it sounds like a whole new project. I was also considering the Elac DFR52 given the excellent review of the Elac DBR62. I am very interested in other suggestions in a similar price range (although I live in Australia so many US speakers are not competitively priced here).

- If you have anechoic measurements of your speaker, you could correct flaws of the speaker using it (loke Amir describes in his reviews), but you can't use your in-room measurements for doing this precisely.

The closest model that Amir has reviewed is the Wharfedale Diamond 220. The 230 is a 2.5-way floostanding version of the 220 with larger woofers (165mm over 130mm) so I am not sure how useful the measurements would be for my situation. I haven't found any great measurements of the Diamond 230.
I am having fun trying to get the most from these speakers. At this point I am unsure if more of my problems are caused by the room or the speaker.
Thanks again for all your help.
-John

 

[Ruhled]

I'd like to see your measurements with variable smoothing instead of psycho. That would be a better starting point for advice. But instead of boosting I would lower the target of your preference curve 3-5 db to match your un-eq'ed mid freq response's troughs so you can work those peaks down a bit too. But do this with variable smoothing.

 

[johnk]

Here are a few more graphs based on the same data.
Var smoothing lower frequencies: Left Green, Right Orange
Looking at it like this, I don't understand the 140Hz peak/trough.

Var smoothing full range: Left Green, Right Orange
The higher frequency boost is one that I do want to reduce.

I would lower the target of your preference curve 3-5 db to match your un-eq'ed mid freq response's troughs so you can work those peaks down a bit too.

Thank you I will try that.
Cheers, John

Appendix:
Same graph from a measurement 6 months ago. There have been a few small furniture changes and probably very slight difference in speaker placement (within 10cm). But it is quite different around the 140Hz mark.

and again 6 months earlier ..

I realise comparing these that I can make big improvements to the channel matching by taking a bit more care with positioning. I need to put some markers on the floor.

 

[JeremyFife]

Hi, that 'furniture change ' is really interesting to me. I also have an asymmetric room setup, one speaker in a corner. Small room, limited options.
Do you mind giving more information on what changed?

 

[DJBonoBobo]

I realise comparing these that I can make big improvements to the channel matching by taking a bit more care with positioning. I need to put some markers on the floor.

Yes, careful positioning should always come first, EQ last.

 

[johnk]

Hi, that 'furniture change ' is really interesting to me. I also have an asymmetric room setup, one speaker in a corner. Small room, limited options.
Do you mind giving more information on what changed?

I have been going back through old photos to check, but there are no clear shots. I think the toe in was larger. The rug in front of the has also changed, but it is not much thicker than the old one. Both speakers might have moved ~20cm nearer to the corner. I need to find a few hours spare to play with the placement ..
This is very interesting as I hadn't actually compared old readings before. I will take more notes of the details in the future.

 

[Ruhled]

Those are some pretty major room nodes esp 100-200 hz. where you have 20db of swing. I would definitely work on repositioning if you can before attempting to eq this. The new graphs poorly correlate with the psycho graphs so im not sure what that's about but this is why you shouldn't use psycho smoothing. It hides the real picture in the bass regions.

I would ditch mmm so you can make faster work of finding best positioning vs room nodes and maybe switch back to it later to fine tune. Just to save some time.

If you have a sub turn it down, if you have adjustable mid/tweeter levels try turning them up.

I wouldn't worry about the tiny dip at 3khz either. You have way bigger issues in the low freqs than that broad wiggle.

 

[Miniyouuuu]

I had similar curves with my Lintons and before any EQ I sealed the port of the worst speaker with foam. It’s better than using 15dB EQ filters. Lintons have 2 ports esch speaker. I got better results with just 1 sealed port.
I never use PEQ higher than 250hz or higher than 12dB, as they change a lot the natural sound of the speaker.
You have some frequency cancelations. The represent room mode issues, you should move a bit the speakers.

 

[ozzy9832001]

Here are a few more graphs based on the same data.
Var smoothing lower frequencies: Left Green, Right Orange
Looking at it like this, I don't understand the 140Hz peak/trough.
View attachment 320530

Var smoothing full range: Left Green, Right Orange
The higher frequency boost is one that I do want to reduce.
View attachment 320531


Thank you I will try that.
Cheers, John

Appendix:
Same graph from a measurement 6 months ago. There have been a few small furniture changes and probably very slight difference in speaker placement (within 10cm). But it is quite different around the 140Hz mark.
View attachment 320533
and again 6 months earlier ..
View attachment 320534
I realise comparing these that I can make big improvements to the channel matching by taking a bit more care with positioning. I need to put some markers on the floor.

By changing the speakers position, even as little as few inches you have moved the problem to different parts of the spectrum. It looks SBIR related. It's not present in the right corner speaker because the corner is allowing it to couple better to the room. Distance from a boundary has changed.

 

[johnk]

Thanks for all the pointers. I now have a plan to try to improve things. I might also try plugging the ports to see if that helps (although they are not easy to access). I'll post some follow up measurements in the next week or two once I have done some more testing. I am interested to see how good I can make these sound before I consider upgrading.