REW and acoustics - help needed please

[andyr]

8 minutes ago, gonefishing999 said:

@andyr Andy smoothing should be at least 1/12 . 1/3 will hide a lot of peaks and troughs as shown in Wfall.

 

Peter

 

Thanks, Peter.  I read a comment on the "Intro to REW" manual which mentioned people often use 1/3rd or 1/6th octave smoothing - so I chose 1/3rd.

 

Andy

 

[Jventer]

@andyr

I was referred to this article - very good

https://redspade-audio.blogspot.com.au/2011/05/rew-understanding-decay-and-waterfall.html

 

[Jventer]

Sorry, forgot to ask, what smoothing is best?

 

Edited March 14, 2017 by Jventer

[almikel]

22 minutes ago, andyr said:

 

Thanks, Mike,

 

There is definitely no mains hum to be heard when I listen to music.  I will connect up my computer + external sound card tonight, to listen for mains hum with these devices connected.  (I wasn't aware of any, before.)

 

I'm afraid you've lost me when you say "there's no decay in the 50Hz resonance". 

 

The top of the section of the waterfall plot you reproduced shows 5 vertical lines; these are 40 / 50 / 60 / 70 / 80hz:

• at the bottom of the graph, there is 'white space' around the 40hz line.
• is this showing 'decay'?

yes the white space indicates that at around 40Hz the sound has decayed to below 45dB (your bottom limit on the graph) by approx 180ms

 

30 minutes ago, andyr said:

 

• but there is no other white space showing between 41hz and 68hz ... not just at 50hz?
• so what does this signify?

 

Thanks,

Andy

 

Ideally you want consistent decay across all frequencies - and no tails marching forward - impossible to achieve but that's the goal.

To quote from Paul's bass integration guide - 20dB in 1st 150ms from 40-300Hz is a target as good as any.

Looks like you have a mode (or 2) at 60Hz which is supported by your room measurements.

I'd try some PEQ cut on that.

Even that tail right down near 300ms is still decaying - different from the tail at 50Hz.

 

 

The 50Hz tail is a straight line (taking into account the perspective) - it doesn't reduce from about 80ms or so.

A room resonance (modal or SBIR) will reduce in magnitude. A 50Hz hum will continue forever at the same volume (like your graph).

Try lengthening the limit out to 1000ms - I bet that 50Hz tail it will march all the way out there without dropping in level.

 

cheers

Mike

 

 

[andyr]

12 minutes ago, Jventer said:

@andyr

I was referred to this article - very good

https://redspade-audio.blogspot.com.au/2011/05/rew-understanding-decay-and-waterfall.html

 

 

Thanks, Jv - will read it tonite.

 

Andy

 

[andyr]

3 minutes ago, almikel said:

yes the white space indicates that at around 40Hz the sound has decayed to below 45dB (your bottom limit on the graph) by approx 180ms

 

 

Ideally you want consistent decay across all frequencies - and no tails marching forward - impossible to achieve but that's the goal.

To quote from Paul's bass integration guide - 20dB in 1st 150ms from 40-300Hz is a target as good as any.

Looks like you have a mode (or 2) at 60Hz which is supported by your room measurements.

I'd try some PEQ cut on that.

Even that tail right down near 300ms is still decaying - different from the tail at 50Hz.

 

 

The 50Hz tail is a straight line (taking into account the perspective) - it doesn't reduce from about 80ms or so.

A room resonance (modal or SBIR) will reduce in magnitude. A 50Hz hum will continue forever at the same volume (like your graph).

Try lengthening the limit out to 1000ms - I bet that 50Hz tail it will march all the way out there without dropping in level.

 

cheers

Mike

 

 

 

Thanks, Mike.  Will do some more playing around.

 

Andy

 

[ghost4man]

1 hour ago, Jventer said:

Sorry, forgot to ask, what smoothing is best?

 

Not 1/3.

When I first started I did 1/12 which gives you a more meaningful graph to work with.

To give you some perspective, take a sweep and try and the different grades of smoothing. It will become self evident why 1/3 will give you a poor representation.

 

Its also advisable to take the room the out of the equation and take a sweep with the speakers outside and then do a sweep. 

Right now what is happening is a combination of in room interference and how the speakers couple to this environment.

 

By taking the speakers outside you have an established reference point. Once you take the speakers back inside you will then get a gauge of what is happening against 

this.

 

The art in all of this - and I certainly dont claim any level expertise, although I have done hundreds upon hundreds of sweeps - is to get a meaningful distribution which

you then interpret  and if necessary influence via PEQs. 

The suggestion that you minimise the number of PEQs goes against what Minidsp is designed for and is to my mind counter intuitive to the spirit of the software.

 

I doubt very much that Andy has done anything in the way of AutoEq but that is another realm altogether.

 

My issue at the moment - which is very typical of anyone who has magnepan speakers - are amplifiers which are not up to the task, as well as dodgy power supply.

[Jventer]

@ghost4man

The speaker weighs 91kg's a bit of work, but will get to it.

 

[davewantsmoore]

7 hours ago, andyr said:

1. the waterfall plot which Jv showed earlier (Mar 6) had a pronounced 'mohawk' at 50hz ... 8-10dB above its surrounding frequencies.  So yes, he probably hears mains hum though his system.

 

Notice that the level of the 50hz tone in both your charts is the same (~70dB).

 

His levels through out his measurements are much lower than yours  (his are too low).

 

7 hours ago, andyr said:

1. however, my waterfall plot actually has a tiny dip at 50hz?  So 50hz is certainly not louder than adjoining frequencies.

 

Sure... and it would not be expected to.     Think about 'typical' mains hum that we have all heard through a system.    It is rarely (if ever) louder than the music... if it were louder than the music, it would be very very loud.

 

As mentioned.  Yours is -15dB below the rest of the signal.  (ie. about 17% distortion).

 

7 hours ago, andyr said:

1. the graph does show a slight rise at 60hz

 

Yes, it appears there are a couple of modes close to 60Hz.

 

6 hours ago, gonefishing999 said:

@andyr Andy smoothing should be at least 1/12 . 1/3 will hide a lot of peaks and troughs as shown in Wfall.

 

 

It doesn't appear there is any smoothing applied to the waterfall.... but the frequency response indicates 1/3.

 

6 hours ago, andyr said:

I'm afraid you've lost me when you say "there's no decay in the 50Hz resonance". 

 

You can see in the chart that 50Hz does not reduce in level  (the X axis), as time passes (the Z axis).    It remains at a constant level out to 300ms.     Looking out to a longer time period, I would expect that it just keeps going (doesn't decay).

 

6 hours ago, andyr said:

Thanks, Peter.  I read a comment on the "Intro to REW" manual which mentioned people often use 1/3rd or 1/6th octave smoothing - so I chose 1/3rd.

 

It really depends on what you are trying to communicate with the chart.    If you are unsure (and asking others), then use 1/12 or 1/24 .... more smoothing will hide too much.

 

1 hour ago, Jventer said:

@ghost4man

The speaker weighs 91kg's a bit of work, but will get to it.

 

 

Unless you are intending to make changes to the speaker based on what you see (ie. you are designing a speaker) .... then I don't think it's very necessary to take the speaker outside.

 

It is safe (and accurate enough) to make the assumption that everything you see on a waterfall or ETC type chart, above ~ 200Hz is the speaker .... and everything you see below there is the room.

[davewantsmoore]

3 hours ago, ghost4man said:

The suggestion that you minimise the number of PEQs goes against what Minidsp is designed for and is to my mind counter intuitive to the spirit of the software.

 

As long as the EQ is warranted (it is often not), then there is no danger of using "too many".

[ghost4man]

Just now, davewantsmoore said:

 

As long as the EQ is warranted (it is often not), then there is no danger of using "too many".

Absolutely Dave.

Wholeheartedly agree. It should be self evident that if they provide you with it then you should use it to resolve an issue

that you identify.

I think 1/12 would be the preferred option. I base that on my own personal experience. 1/6 is to my mind a smoothing  option

which wont yield honest information.

 

I would prefer if minidsp had more PEQ opportunities than the current 6 per driver.

[andyr]

1 hour ago, davewantsmoore said:

 

As long as the EQ is warranted (it is often not), then there is no danger of using "too many".

 

 

I was told, re PEQ, that there is no 'free lunch'.  IOW, any PEQ introduces a phase shift ... which is deleterious to SQ.  If this is true - then, the fewer the better.

 

Andy

Edited March 14, 2017 by andyr

[davewantsmoore]

1 hour ago, ghost4man said:

honest information

 

 

It just depends on what you are trying to communicate in your use of the graph.

[davewantsmoore]

6 minutes ago, andyr said:

I was told, re PEQ, that there is no 'free lunch'.  IOW, any PEQ introduces a phase shift ... which is deleterious to SQ.  f this is true - then, the fewer the better.

 

That is correct (EQ normally introduces a phase shift) ... but the conclusion is (wrong) an oversimplification.

 

The response of a speaker is mostly "minimum-phase".    This means that the amplitude and phase are related.    Where there is a wiggle in the amplitude, there is a corresponding wiggle in the phase.

 

Typical EQ (which a typical PEQ is), has the same deal, it is minimum phase.   Amplitude response changes, cause a corresponding change in the phase response.

 

So....  An amplitude error in a speaker (which has a corresponding error in the phase response) .... that is fixed with minimum-phase EQ .... with have the error in both the amplitude and phase fixed.    ie.  if you flatten the frequency response, the phase response will be also flat.

 

So contrary to being a problem - this is actually desirable.   OTOH, true "linear phase" for example, alters only the frequency (and not the phase at all), and so it will correct the amplitude, but if there is a (typically there will be) corresponding phase error - that won't be corrected.

 

 

Errors caused by the room, are not usually minimum-phase .... but as always, you can see the effects through measurement .... so there is no need to use blanket rules.

[Guest]

@davewantsmoore

Do you still have that impulse decay plot you took at the GTG, with all the human bass traps inside the concrete room, might be interesting.

[davewantsmoore]

I've been wondering when you would ask.    It seems I didn't hit save.  >_<

[gonefishing999]

@andyr

21 hours ago, andyr said:

I was told, re PEQ, that there is no 'free lunch'.  IOW, any PEQ introduces a phase shift ... which is deleterious to SQ.  If this is true - then, the fewer the better.

 

Hard to pull off but fewer the better.

 

Peter

[andyr]

2 hours ago, gonefishing999 said:

@andyr

Hard to pull off but fewer the better.

 

Peter

 

Mmmm, well that's what one expert told me ... but another expert (Dave, here) says that you - and my "other expert" - are confused. 

 

Andy

 

[davewantsmoore]

17 hours ago, gonefishing999 said:

but fewer the better.

 

Why?

[davewantsmoore]

15 hours ago, andyr said:

Mmmm, well that's what one expert told me

 

The reasoning that your expert gave (assuming the message hasn't been lost in translation somehow) indicates that he is not an expert.

[almikel]

PEQ can be a very powerful tool when used well

 

On 3/14/2017 at 7:19 PM, davewantsmoore said:

 

As long as the EQ is warranted (it is often not), then there is no danger of using "too many".

^ this

On 3/14/2017 at 9:00 PM, davewantsmoore said:

 

That is correct (EQ normally introduces a phase shift) ... but the conclusion is (wrong) an oversimplification.

 

The response of a speaker is mostly "minimum-phase".    This means that the amplitude and phase are related.    Where there is a wiggle in the amplitude, there is a corresponding wiggle in the phase.

 

Typical EQ (which a typical PEQ is), has the same deal, it is minimum phase.   Amplitude response changes, cause a corresponding change in the phase response.

 

So....  An amplitude error in a speaker (which has a corresponding error in the phase response) .... that is fixed with minimum-phase EQ .... with have the error in both the amplitude and phase fixed.    ie.  if you flatten the frequency response, the phase response will be also flat.

 

So contrary to being a problem - this is actually desirable.   .

^ and this

If you have a wiggle in the amplitude response which is minimum phase, then flattening it with PEQ will fix the amplitude and the phase response.

 

The challenge when using PEQ to fix wiggles in the frequency response is knowing which ones are minimum phase.

Wiggles caused by diffraction off a speaker baffle (as an example), are never minimum phase and should not be corrected with PEQ.

On 3/14/2017 at 9:00 PM, davewantsmoore said:

Errors caused by the room, are not usually minimum-phase .... but as always, you can see the effects through measurement .... so there is no need to use blanket rules.

Room cancellations are never minimum phase, but room peaks can be minimum phase - hence the often provided advice that it's OK to cut peaks with PEQ, as the PEQ is very effective if the peaks are minimum phase.

Peaks due to modal behaviour are usually minimum phase - peaks due to SBIR are not.

 

On 3/14/2017 at 7:19 PM, davewantsmoore said:

 

As long as the EQ is warranted (it is often not), then there is no danger of using "too many".

^ this again

Poorly applied PEQ can make the situation worse if using PEQ to fix issues that aren't minimum phase.

An example of poorly applied PEQ would be getting a ruler flat on axis speaker response when the dips or peaks were caused by diffraction off the speaker baffle.

Geddes discusses this in his paper on directivity - Geddes uses the term "acoustic" issue when describing a non minimum phase speaker issue.

http://www.gedlee.com/Papers/directivity.pdf

He also describes a resonance (electro-mechanical) that is minimum phase and could be fixed via PEQ.

 

What is minimum phase and what isn't is very hard to get your head around - the simplest explanation is the one on REW, which I've read lots but still struggle with (and way simpler than any other definition I've found)

https://www.roomeqwizard.com/help/help_en-GB/html/minimumphase.html

 

@davewantsmooreif you can describe minimum phase better than the REW page go for it.

 

The best thing is you don't need to understand what minimum phase is, because REW can show you via the "Excess Group Delay" function/graph - apply PEQ in the flat areas only and you should be fine - also useful for showing time alignment issues.

 

On 3/14/2017 at 7:23 PM, ghost4man said:

Absolutely Dave.

Wholeheartedly agree. It should be self evident that if they provide you with it then you should use it to resolve an issue

that you identify.

I think 1/12 would be the preferred option. I base that on my own personal experience. 1/6 is to my mind a smoothing  option

which wont yield honest information.

 

I would prefer if minidsp had more PEQ opportunities than the current 6 per driver.

Hi Ghost

As above - MiniDSP may provide 6 bands of PEQ per driver, but only use them where it assists (ie minimum phase issues).

If PEQ is used on non minimum phase issues (like baffle diffraction) - you'll likely not improve the sound.

 

Particularly relevant for room PEQ - the correction is valid only for that point in space, and for every other point in the room that had the same response - hence why lopping off peaks for modal behaviour works so well, but not so well for SBIR.

Modal behaviour is single frequency (refer Geddes "iso" frequency) and consistent through the room - ie lop the top off a mode and that mode will have less intensity everywhere in the room.

SBIR is position dependant and will change in frequency depending on position (not "iso" frequency, not minimum phase) - EQ may work at a specific position in the room, and likely make the sound much worse elsewhere in the room...

 

...no wonder EQ has such a bad name - it's easy to use it badly and make things worse.

 

IMO when EQ is applied properly, it's indispensable (and I love having EQ tone controls on the remote).

When applied poorly you understand why EQ has such a bad reputation. 

 

cheers

Mike

[ghost4man]

@almikel

 

Mate thanks for the lengthy but clear response.

Have a question for you. Do you think when AutoEq is applied in minidsp that the software takes the points that you make into  account

or does it simply go through the process of levelling at will?

 

As you know AutoEq takes the guess work out for you so I wonder how judicious it is when the software takes over. In other words how do

you know  or rather how does the software - if at all - knows which ones are minimum phase and which ones arent.

 

You say:

 

The best thing is you don't need to understand what minimum phase is, because REW can show you via the "Excess Group Delay" function/graph - apply PEQ in the flat areas only and you should be fine - also useful for showing time alignment issues.

 

Does AutoEq rely upon this and seek the best EQ as a result? I'm just not getting the impression - even though I followed the steps thoroughly - that

it takes into account phase considerations.

 

Do maggies present a different type of problem given the differences with which sound is distributed over say traditional box speakers?

 

 

 

My issue at the moment is poor power supply - already tested and determined - and amplifiers which arent up to scratch which has led me

to cease testing until I sort out this issue. 

One of the many hazards of having maggies.

[andyr]

7 hours ago, almikel said:

PEQ can be a very powerful tool when used well

 

^ this

^ and this

If you have a wiggle in the amplitude response which is minimum phase, then flattening it with PEQ will fix the amplitude and the phase response.

 

The challenge when using PEQ to fix wiggles in the frequency response is knowing which ones are minimum phase.

Wiggles caused by diffraction off a speaker baffle (as an example), are never minimum phase and should not be corrected with PEQ.

Room cancellations are never minimum phase, but room peaks can be minimum phase - hence the often provided advice that it's OK to cut peaks with PEQ, as the PEQ is very effective if the peaks are minimum phase.

Peaks due to modal behaviour are usually minimum phase - peaks due to SBIR are not.

 

^ this again

Poorly applied PEQ can make the situation worse if using PEQ to fix issues that aren't minimum phase.

An example of poorly applied PEQ would be getting a ruler flat on axis speaker response when the dips or peaks were caused by diffraction off the speaker baffle.

Geddes discusses this in his paper on directivity - Geddes uses the term "acoustic" issue when describing a non minimum phase speaker issue.

http://www.gedlee.com/Papers/directivity.pdf

He also describes a resonance (electro-mechanical) that is minimum phase and could be fixed via PEQ.

 

What is minimum phase and what isn't is very hard to get your head around - the simplest explanation is the one on REW, which I've read lots but still struggle with (and way simpler than any other definition I've found)

https://www.roomeqwizard.com/help/help_en-GB/html/minimumphase.html

 

@davewantsmooreif you can describe minimum phase better than the REW page go for it.

 

The best thing is you don't need to understand what minimum phase is, because REW can show you via the "Excess Group Delay" function/graph - apply PEQ in the flat areas only and you should be fine - also useful for showing time alignment issues.

 

Hi Ghost

As above - MiniDSP may provide 6 bands of PEQ per driver, but only use them where it assists (ie minimum phase issues).

If PEQ is used on non minimum phase issues (like baffle diffraction) - you'll likely not improve the sound.

 

Particularly relevant for room PEQ - the correction is valid only for that point in space, and for every other point in the room that had the same response - hence why lopping off peaks for modal behaviour works so well, but not so well for SBIR.

Modal behaviour is single frequency (refer Geddes "iso" frequency) and consistent through the room - ie lop the top off a mode and that mode will have less intensity everywhere in the room.

SBIR is position dependant and will change in frequency depending on position (not "iso" frequency, not minimum phase) - EQ may work at a specific position in the room, and likely make the sound much worse elsewhere in the room...

 

...no wonder EQ has such a bad name - it's easy to use it badly and make things worse.

 

IMO when EQ is applied properly, it's indispensable (and I love having EQ tone controls on the remote).

When applied poorly you understand why EQ has such a bad reputation. 

 

cheers

Mike

 

Thanks for your explanation, Mike - and the REW link.

 

Andy

 

[davewantsmoore]

8 hours ago, almikel said:

An example of poorly applied PEQ would be getting a ruler flat on axis speaker response when the dips or peaks were caused by diffraction off the speaker baffle.

Geddes discusses this in his paper on directivity - Geddes uses the term "acoustic" issue when describing a non minimum phase speaker issue.

 

Yes.   This is why consider the the speakers response as a full 3d map is so helpful.     As you can immediately see where an error in the response has been corrected on one axis, that has negatively affected the speaker at other radiation angles.

 

It is essentially THE defining performance characteristic of a speaker .... as you correct the directly arriving (aka, the "on-axis sound") sound to have less errors .... does the speaker also have less errors at every other axis?

 

Here are a couple of examples.

 

The two speakers have been corrected so that the directly arriving sound is perfectly flat.    You wouldn't do this in practice .... but what it is attempting to show, is how much the speaker deviates from "corrected" at other angles.

 

This is what you would see with a "perfect" response.

 

 

 

Where the speakers significantly deviate from perfect, you would be able to measure (at some point in space, at least) that the 'excess phase' was >> zero.

 

 

The R16 speaker is a DIY multiway dipole design, by a member here.

 

I would post mine ... but I do not have data which is so detailed  (it was compiled by Princeton university) ... and my data is also not (easily) directly comparable to the others speakers Princeton have worked up data sets for.

 

[davewantsmoore]

3 hours ago, ghost4man said:

Do maggies present a different type of problem given the differences with which sound is distributed over say traditional box speakers?

 

Not really.   The headline issue is still the same.   If you want to make a correction to the response of the speaker ....  then it is essential to know what your correction is doing to the entire "3d soundfield" ...  not just the point you've measured from .... and a room compounds this problem.

 

"room correction" programs, like for example, "Dirac Live" .... try to consider these issues (which is why they have you take more than 1 measurement before calculating their correction) .... but IMV they are still usually quite heavy handed.