EQ, Is too much of a good thing bad?

[Ron Texas]

The above REW screenshots are of my LS50's crossed over to a single sub at 80-ish hz. The first is 1/6th octave smoothing which is how most of these graphs are presented. Next is psycho acoustic smoothing which is supposed to be what we hear. Finally is variable smoothing which is what the author of REW recommends we use.

The house curve is another variable. This is the one @mitchco described elsewhere, +2 db @ 20 hz, -8 db @ 20 khz. Use your imagination for B&K, Dirac or whatever. I can't tell you what's best, but if it doesn't slope down in the highs it will sound bright. The LS50's are 30 degrees off axis, by the way.

If I lower the target level from 91.5 to 91 REW reports no filters are needed for a 2 db target from 20 to 250 hz. That's three more variables, target level, target tolerance, and EQ range. Also hidden are individual maximum boost (8 db) and overall maximum boost (2 db).

There's a lot going on. Many say don't EQ beyond Schroder, which is about 200 hz in typical homes. Higher in a Japanese rabbit hutch and lower in McMansions. I have seen comments by @Floyd Toole that you can go as high as 400 or even 500.

The smoothing makes a gigantic difference in the number of filters generated. Anywhere from none to 12 when all other things are equal. I can tell you what I can hear. Reducing the bump at 68 hz reduces subjective quantity bass, but not the quality. Taking out 2db of the sag at 160 hz changes the timber of Leonard Cohen's voice to one which sounds more natural to me, especially on his later recordings. Attempting to take out the twin peaks on either side of 550 hz gives me the impression of increased harshness, the opposite of what I predicted.

As noted above the author of REW says use variable smoothing for EQ. Others around here say don't go any smoother than 1/12th. In one pro audio forum I found comments like, "don't knock yourself out, use psy or 1/6th.

Nobody here can really tell anyone else exactly which filters to use in their system. The best anyone can do is provide a systematic method. That's because the right answer involves personal taste and individual quirks.

Anyone here want to talk about EQ?

 

[amirm]

As noted above the author of REW says use variable smoothing for EQ. Others around here say don't go any smoother than 1/12th.

Variable smooth is useful to determine your target curve. EQ there is basically a tone control above a few hundred hertz.

Below a few hundred hertz, you do want resolution as low as 1/12 octave.

This is what John must be talking about which seems to correct to me.

Alternatively you can just sweep to say 300 Hz, filter to 1/12 and perform your corrections there. Then, run the variable one to determine the overall slope of the curve (target curve).

 

[amirm]

Also, you don't need many filters for bass. Two to three peaks knocked down is all you need. Mechanically generating them can overproduce filters and in the few tests I have done, it doesn't sound as good to my ears.

 

[Ron Texas]

Also, you don't need many filters for bass. Two to three peaks knocked down is all you need. Mechanically generating them can overproduce filters and in the few tests I have done, it doesn't sound as good to my ears.

This is what I am finding.

 

[ernestcarl]

I don't use auto EQ... however, it's A LOT better if you tried doing a before and after EQ shot using the "moving microphone method" rather than just single measurements.

It is way clearer what effects your EQ has since it averages over a much wider listening area rather than a single spot.

 

[Ron Texas]

I use a few different positions for the mike. It helps.

 

[oldsysop]

Mechanically generating them can overproduce filters and in the few tests I have done, it doesn't sound as good to my ears.

Here we handle measurements, we don't care about your ears.

 

[ernestcarl]

Originally I was leery about adding any more PEQs above 200-300Hz (different measurement points gave radically differing results above the Transition Frequency) but with spatial averaging, it gave me confidence with adding just a few more -- knowing full well it really wouldn't have any detrimental effect in the overall response.

miniDSP Input stage EQ:
80Hz -2dB, q=6
88Hz +6dB, q=6
115Hz +6dB, q=1.5
250Hz +6dB, q=6

miniDSP Output stage EQ:
(Left Speaker)
500Hz +3dB, q=3

(Right Speaker)
450Hz +3dB, q=3

(Subwoofer)
30Hz -3dB, q=4
33Hz -5dB, q=1.8

My previous method of taking 13 measurements for each speaker was rather exhausting. Using the MMM with RTA and pink noise yielded much faster comparisons.

 

[Ron Texas]

My previous method of taking 13 measurements for each speaker was rather exhausting. Using the MMM with RTA and pink noise yielded much faster comparisons.

How does one do MMM with RTA? I would love better, faster and simpler.

 

[ernestcarl]

I just copied Krunok's RTA settings from here: https://www.audiosciencereview.com/...aging-of-measurements-for-room-eq.8148/page-2

Default settings gave me bad results...

And here's a rather lengthy (but very helpful) video where Charles Sprinke effectively shows it (thank god, instead of following text-only instructions!) :

 

[Blumlein 88]

Despite claims about not doing EQ above Schroeder I find it usually helps quite noticeably. Most of that experience is on speakers not at all hewing to the Harman guidelines like my panel ESLs. With Harman products they aren't asking for as much correction and it makes less difference. If a speaker is unbalanced both on and off axis I've found balancing it helps. It still may not match a superb speaker not needing the correction, but the correction has been judged by myself and others a benefit. Plus Sean Olive tested some room EQ and has proof it helped in the scores given by listeners at Harman in some cases. One of those being the algorithm I've used in the past. Dirac seems an improvement over that older algorithm.

 

[JoachimStrobel]

View attachment 37656View attachment 37657View attachment 37658

The above REW screenshots are of my LS50's crossed over to a single sub at 80-ish hz. The first is 1/6th octave smoothing which is how most of these graphs are presented. Next is psycho acoustic smoothing which is supposed to be what we hear. Finally is variable smoothing which is what the author of REW recommends we use.

The house curve is another variable. This is the one @mitchco described elsewhere, +2 db @ 20 hz, -8 db @ 20 khz. Use your imagination for B&K, Dirac or whatever. I can't tell you what's best, but if it doesn't slope down in the highs it will sound bright. The LS50's are 30 degrees off axis, by the way.

If I lower the target level from 91.5 to 91 REW reports no filters are needed for a 2 db target from 20 to 250 hz. That's three more variables, target level, target tolerance, and EQ range. Also hidden are individual maximum boost (8 db) and overall maximum boost (2 db).

There's a lot going on. Many say don't EQ beyond Schroder, which is about 200 hz in typical homes. Higher in a Japanese rabbit hutch and lower in McMansions. I have seen comments by @Floyd Toole that you can go as high as 400 or even 500.

The smoothing makes a gigantic difference in the number of filters generated. Anywhere from none to 12 when all other things are equal. I can tell you what I can hear. Reducing the bump at 68 hz reduces subjective quantity bass, but not the quality. Taking out 2db of the sag at 160 hz changes the timber of Leonard Cohen's voice to one which sounds more natural to me, especially on his later recordings. Attempting to take out the twin peaks on either side of 550 hz gives me the impression of increased harshness, the opposite of what I predicted.

As noted above the author of REW says use variable smoothing for EQ. Others around here say don't go any smoother than 1/12th. In one pro audio forum I found comments like, "don't knock yourself out, use psy or 1/6th.

Nobody here can really tell anyone else exactly which filters to use in their system. The best anyone can do is provide a systematic method. That's because the right answer involves personal taste and individual quirks.

Anyone here want to talk about EQ?

I believe people saying that the time domain matters too, or most. Frequency plots for a room response are nice to look at, that may be all. Hence REW is nice for checking this and that, but may not be serious enough for more.

 

[ernestcarl]

I believe people saying that the time domain matters too, or most. Frequency plots for a room response are nice to look at, that may be all. Hence REW is nice for checking this and that, but may not be serious enough for more.

But what is that "more required" by the average audio enthusiast that REW simply cannot function?

 

[RayDunzl]

But what is that "more required" by the average audio enthusiast that REW simply cannot function?

What is an average audio enthusiast?

*note: the average human has one tit and one testicle

 

[digitalfrost]

As noted above the author of REW says use variable smoothing for EQ. Others around here say don't go any smoother than 1/12th. In one pro audio forum I found comments like, "don't knock yourself out, use psy or 1/6th.

From the REW manual:

Variable smoothing applies 1/48 octave below 100 Hz, 1/3 octave above 10 kHz and varies between 1/48 and 1/3 octave from 100 Hz to 10 kHz, reaching 1/6 octave at 1 kHz. Variable smoothing is recommended for responses that are to be equalised.

This is bascially a form of frequency dependent windowing.

https://www.audiovero.de/acourate-w...ionen:td-functions:frequency_dependent_window

I am convinced this is the right way to go with room correction, i.e. more correction in the bass, and less correction as you move higher up in frequency.

 

[ernestcarl]

What is an average audio enthusiast?

*note: the average human has one tit and one testicle

I wouldn't know.

 

[audimus]

My conclusions from considerable experimentation over the last 6 months is that there are no rules to how this should be done. There are a lot of conjectures, opinions (and some very dogmatic ones), thumb rules, etc., which may have worked in some context but don’t apply in general. But it works more often than it does not and in most real world cases is better than not doing it.

This is certainly problematic for anyone that wants to rely on a purely mechanistic process and avoid any reliance on ears.

Either you bring listening into the tweaking loop, and really understand the concepts behind the measurements in REW (which is a monstrous mix of tech features that can get you doing a lot of things that are wrong and not know it) or use a commercial system.

The latter vendors, I suspect have done a lot more experimentation and under controlled conditions with this than any public studies or published papers but that body of knowledge is unavailable to us for obvious commercial reasons. Analyzing some of the results of what these commercial system do, they seem to break every thumb rule out there as long as it works. EQ at higher frequencies? They do it. Boosts of as much as 10db? They do it. High target curve slopes? They do it. All depends on what you are starting with.

I am glad the OP simply used EQ instead of room correction because the very dogmatic opinion that this is only about correcting room characteristics is debatable looking at what any of the commercial systems actually do. I can understand them from a marketing purpose of calling it only room correction publicly rather than something that implies correcting the results of flaws or deficiencies in the equipment chain on the sound output since almost all of them depend on vendor relationships for their business model.

Denon is not going to call Audyssey anything but room correction even if, in practice, it gets rid of some annoying results of their design or makes it sound better to most of their audiences than a perfectly flat curve.

 

[ernestcarl]

But it works more often than it does not and in most real world cases is better than not doing it.

I wonder if folks such as Steve Guttenberg uses EQ and/or DSP... I've heard him talk about using subs with some speakers before.

I'm guessing he still does everything by ear.

 

[Blumlein 88]

What is an average audio enthusiast?

*note: the average human has one tit and one testicle

These days that might turn out to be the average human, or maybe the average over their shifting lifetime.

 

[Sal1950]

Despite claims about not doing EQ above Schroeder I find it usually helps quite noticeably.

Agreed, I've listened to my system with EQ full range and below 500hz. I prefer it full range.

What is an average audio enthusiast?

One that doesn't want to spend half of his life pouring over online info, books, etc to expand his understanding of the technology.
A fairly basic setup satisfies him without trying to learn how to manually set filters and curves in REW or AcourateDRC , etc. One that would rather listen to music than attempt to get a degree in EE, Acoustics, etc.
This was supposed to be fun, not make me pull my hair out. I've been complete bald since my 40's LOL