ProSoundWeb Community

Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => Audio Measurement and Testing => Topic started by: Mike Caldwell on January 26, 2018, 09:37:11 PM

Title: Tell me what you see in these phase response graphs
Post by: Mike Caldwell on January 26, 2018, 09:37:11 PM
I've been working with REW software to get an idea of what a couple speakers of mine look like at the high frequency crossover. Pictured are a couple screen shots of the response of a 12X2 and 15X2 bi-amped speaker. These are after playing with the high frequency driver polarity and time delay on the woofer.
For these test the 12X2 crossover is 1700hz and the 15X2 is 1600hz.
Set up outdoors, measurement mic on axis between woofer and horn about 8 foot from the speaker.
Title: Re: Tell me what you see in these phase response graphs
Post by: Mac Kerr on January 26, 2018, 10:03:43 PM
I've been working with REW software to get an idea of what a couple speakers of mine look like at the high frequency crossover. Pictured are a couple screen shots of the response of a 12X2 and 15X2 bi-amped speaker. These are after playing with the high frequency driver polarity and time delay on the woofer.
For these test the 12X2 crossover is 1700hz and the 15X2 is 1600hz.
Set up outdoors, measurement mic on axis between woofer and horn about 8 foot from the speaker.

It's hard to imagine what you're doing to get 16 phase wraps in 1 octave.

A shot of the screen with full bandwidth might be more enlightening.

Mac
Title: Re: Tell me what you see in these phase response graphs
Post by: Mike Caldwell on January 26, 2018, 11:44:43 PM
Here's a couple shots of the 12X2 that show more bandwidth. No EQ was active.

I got about the same phase response when measuring a small powered speaker a few days ago.

I'm thinking the couple deep notches are from my less than perfect testing location.

I should have saved some of the shots before and during the measurement process.
Title: Re: Tell me what you see in these phase response graphs
Post by: Merlijn van Veen on January 27, 2018, 01:49:39 AM
If this was Smaart I would recommend setting the delay locator which effectively takes out all constant delay caused by latency and / or time of flight.
Title: Re: Tell me what you see in these phase response graphs
Post by: Mike Caldwell on January 27, 2018, 10:12:57 AM
If this was Smaart I would recommend setting the delay locator which effectively takes out all constant delay caused by latency and / or time of flight.

REW does not have set or find delay like Smaart, it does have loop back delay finder or a use acoustic source to set delay, using both showed the same results.

For what it's worth I'm using a small Allen Heath ZED10 mixer as the IO via it's USB connection.

I was primarily interested in the phase response at the crossover and if I'm looking at it right I think it is at least somewhat "flat" through the crossover point and the area around it.

I really should have taken a screen shot of the before adjustment.
Title: Re: Tell me what you see in these phase response graphs
Post by: Mac Kerr on January 27, 2018, 01:42:24 PM
I was primarily interested in the phase response at the crossover and if I'm looking at it right I think it is at least somewhat "flat" through the crossover point and the area around it.

I really should have taken a screen shot of the before adjustment.

You're not looking at it right. That sawtooth display at the bottom is the phase response. It is about as far from "flat" phase response as it is possible to get.

Ideally it should be a flat horizontal trace, not a sawtooth showing dozens of 360º phase wraps in a single octave. As Merlijn said, your reference and test signals are in totally different time zones.

Mac
Title: Re: Tell me what you see in these phase response graphs
Post by: Ivan Beaver on January 27, 2018, 03:36:36 PM
You are going to to have to set the receive delay (or whatever it is called in that software) properly.

There are simply to many phase wraps to get any useful information.

It is possible to have a wrap or two, depending on the speaker.  Better ones will have fewer wraps.

A perfect phase response would be a flat line.  So that is the "goal", but rarely achievable-especially at a decent distance away, due to reflections.
Title: Re: Tell me what you see in these phase response graphs
Post by: Marc Sibilia on January 27, 2018, 04:02:27 PM
It's hard to imagine what you're doing to get 16 phase wraps in 1 octave.

Mike,

If you open up the measurement again, in the upper right, there is a button for Controls.  Open that, and press Estimate IR Delay.  After you adjust the Impulse Response, you will have the phase plot you expect.

What you are looking at is the phase wrap of the delay caused by the travel time through the air between your speaker and microphone (as well as any measurement latency).

Marc
Title: Re: Tell me what you see in these phase response graphs
Post by: Mike Caldwell on January 27, 2018, 04:09:08 PM
I know the bottom trace is the phase, and after looking at is why I wanted some other opinions.

I got to dig into the software deeper and see what's going on with the delay offset.

Thanks
Mike C.
Title: Re: Tell me what you see in these phase response graphs
Post by: Mike Caldwell on January 27, 2018, 04:11:02 PM
Mike,

If you open up the measurement again, in the upper right, there is a button for Controls.  Open that, and press Estimate IR Delay.  After you adjust the Impulse Response, you will have the phase plot you expect.

What you are looking at is the phase wrap of the delay caused by the travel time through the air between your speaker and microphone (as well as any measurement latency).

Marc


Thanks for that info!
Title: Re: Tell me what you see in these phase response graphs
Post by: Mike Caldwell on January 27, 2018, 08:52:40 PM
Mike,

If you open up the measurement again, in the upper right, there is a button for Controls.  Open that, and press Estimate IR Delay.  After you adjust the Impulse Response, you will have the phase plot you expect.

What you are looking at is the phase wrap of the delay caused by the travel time through the air between your speaker and microphone (as well as any measurement latency).

Marc

That got me something that looks more like it.
I'll take some fresh measurements and post a couple a screen shots.

Is that something I need to apply after each measurement or is there a way to set it to automatically apply the delay at each measurement?

Thanks
Mike C.
Title: Re: Tell me what you see in these phase response graphs
Post by: Erik Jerde on January 28, 2018, 12:51:51 AM
That got me something that looks more like it.
I'll take some fresh measurements and post a couple a screen shots.

Is that something I need to apply after each measurement or is there a way to set it to automatically apply the delay at each measurement?

Thanks
Mike C.

Others more knowledgeable than me may correct this, but you need to find and set the delay each time.  Physically moving a component (even a little) will change it.  If you’re working with a DSP that has fixed latency for each processing block then changes there won’t effect the measurement latency.  Of course this doesn’t hold true for a delay block!  As to messing with analog circuits there I’m unsure.  A little empirical testing should lead you in the right direction.
Title: Re: Tell me what you see in these phase response graphs
Post by: Mike Caldwell on January 28, 2018, 01:10:24 AM
Others more knowledgeable than me may correct this, but you need to find and set the delay each time.  Physically moving a component (even a little) will change it.  If you’re working with a DSP that has fixed latency for each processing block then changes there won’t effect the measurement latency.  Of course this doesn’t hold true for a delay block!  As to messing with analog circuits there I’m unsure.  A little empirical testing should lead you in the right direction.

After the initial set up nothing moves or changes, other than adjustments made in the DSP.

I was wondering if I need to manually apply the delay offset each time I take a measurement or if there is a way to have it automatically applied with each measurement.

REW will not display phase response in real time, you take a measurement, adjust DSP, take another measurement and so on.
Title: Re: Tell me what you see in these phase response graphs
Post by: Ivan Beaver on January 28, 2018, 09:16:51 AM
That got me something that looks more like it.
I'll take some fresh measurements and post a couple a screen shots.

Is that something I need to apply after each measurement or is there a way to set it to automatically apply the delay at each measurement?

Thanks
Mike C.
ANYTHING that changes the phase (time) will require a new setting.

Moving the mic or the speaker is obvious.  Not so obvious is adjust the bandpass filters, eqs etc.

As a "general rule" you want the phase trace to "fall" at the same point the HF response is falling.

This is not always practical, or actually done.  But it should be the goal.

A "perfect phase trace" for a loudspeaker would be sloping down on the low freq, flat throughout most of the passband, and then falling on the top end.
Title: Re: Tell me what you see in these phase response graphs
Post by: Marc Sibilia on January 28, 2018, 10:19:47 AM
Is that something I need to apply after each measurement or is there a way to set it to automatically apply the delay at each measurement?

You can do this on every measurement, but you don't really want to.  Under <Preferences><Analysis>, if you turn off Use Loopback as a Timing Reference, and check <set t=0 at IR peak>, this will be done automatically on every measurement.  Using Estimate IR delay works the same way using the peak of the impulse response to set the delay.

Moving the mic or the speaker is obvious.  Not so obvious is adjust the bandpass filters, eqs etc.

Changing the EQ can change the timing of the peak of the impulse, so you don't really want to use the automatic delay setting, because the phase measurements won't be relative to the same point in time on different plots.  You can manually adjust for the time of flight by going from the <SPL & Phase> to <Impulse>, then under Controls, manually add the same offset to all your measurements <t=0 offset>. 

This is especially important when you are setting up the crossovers and delays and measuring different drivers on separate graphs to try to decide on adjustments to delay.  Without the Loopback timing, there is nothing to relate one graph to another.

I typically work by measuring the HF driver first, using the Impulse Response to set the fixed offset for time of flight, and then using that same offset for midbass measurements and sub measurements.

I don't know of a way to automatically apply that same offset automatically to all subsequent measurements.  But hey, it is an amazingly powerful program worth so much more than we paid for it (free)!

Marc

Title: Re: Tell me what you see in these phase response graphs. Updated measurments
Post by: Mike Caldwell on January 28, 2018, 08:19:55 PM
Ok it was a nice day here in Ohio this afternoon to do some speaker measurements.

Things are looking better with the delay applied to the proper setting!

All EQ was by passed for measurements, the lower frequency stuff
I'm sure was affected by my outdoor surroundings.

Let me know
Thanks
Mike C.

Title: Re: Tell me what you see in these phase response graphs. Updated measurments
Post by: Art Welter on February 06, 2018, 07:40:14 PM
All EQ was by passed for measurements, the lower frequency stuff
I'm sure was affected by my outdoor surroundings.

Let me know
Mike,

The phase traces look more "normal" now, typical of phase wrap due to crossover filters, but the deep dip in combined response SPL around the crossover range could still be indicative that reversing the polarity, different crossover slopes, or delay time, or a combination may be desirable to flatten both phase and amplitude response.

The combined, non-equalized response of the LF and HF ideally should be greater than either LF or HF measured separately, and can be checked against the response when either are muted (while everything else remains the same) to verify that proper on axis summation is observed.

Cheers,
Art
Title: Re: Tell me what you see in these phase response graphs. Updated measurments
Post by: Mike Caldwell on February 06, 2018, 11:45:30 PM
Mike,

The phase traces look more "normal" now, typical of phase wrap due to crossover filters, but the deep dip in combined response SPL around the crossover range could still be indicative that reversing the polarity, different crossover slopes, or delay time, or a combination may be desirable to flatten both phase and amplitude response.

The combined, non-equalized response of the LF and HF ideally should be greater than either LF or HF measured separately, and can be checked against the response when either are muted (while everything else remains the same) to verify that proper on axis summation is observed.

Cheers,
Art

Thanks Art

Here's a few I did a couple days after the others I posted, my fingers were frozen numb after I was done!

I normally flip the polarity on the high freq. and tweak the mid/woofer time delay to find the deepest null point.
After I flip the high freq back to normal I go back and forth a little with the mid/woofer delay watching the amplitude and phase response.
Next time out I'll see what happens trying some different crossover slopes and even some asymmetrical slopes.

Posted is a screen shot of the 15X2 flipped and normal and also shot of a double 12X2 box.
Title: Re: Tell me what you see in these phase response graphs
Post by: Xandy Whitesel on February 07, 2018, 02:33:01 AM

A "perfect phase trace" for a loudspeaker would be sloping down on the low freq, flat throughout most of the passband, and then falling on the top end.

I'd love to hear a more in depth explanation of this statement.  By my understanding both electronic and acoustical manipulations will cause a delay of low end, represented by an upward sloping phase plot towards the low frequency.  How does one achieve a downward slope on the low freq?  Please set me straight.
Title: Re: Tell me what you see in these phase response graphs
Post by: Phil Graham on February 07, 2018, 10:57:45 AM
I'd love to hear a more in depth explanation of this statement.  By my understanding both electronic and acoustical manipulations will cause a delay of low end, represented by an upward sloping phase plot towards the low frequency.  How does one achieve a downward slope on the low freq?  Please set me straight.

Xandy,

You and Ivan are saying the same thing. He's starting at frequency zero and looking to the "right," while you are starting from higher frequencies and looking "left." Because the group delay is defined as -d(theta)/d(frequency), the slope will be down for causal systems (i.e. output proceeds after input) when starting at low frequencies and moving towards higher.

A loudspeaker is inherently a bandpass device (i.e. high pass and low pass in series). So the best case scenario for a loudspeaker, before any additional phase manipulation before the box in a dsp, will be the idealized phase response of an ideal bandpass filter.

Title: Re: Tell me what you see in these phase response graphs. Updated measurments
Post by: Art Welter on February 07, 2018, 02:28:45 PM
Thanks Art

Here's a few I did a couple days after the others I posted, my fingers were frozen numb after I was done!

I normally flip the polarity on the high freq. and tweak the mid/woofer time delay to find the deepest null point.
After I flip the high freq back to normal I go back and forth a little with the mid/woofer delay watching the amplitude and phase response.
Next time out I'll see what happens trying some different crossover slopes and even some asymmetrical slopes.
Nothing quite like outdoor testing in cold weather!

Looks like the polarity was correct, as well as slopes you might try overlapping the crossover points, that is bringing the LF point higher or HF lower to compensate for the acoustical roll-offs. That said, 1600 Hz is already pretty high for a 15"...
Title: Re: Tell me what you see in these phase response graphs. Updated measurments
Post by: Mike Caldwell on February 07, 2018, 04:43:13 PM
Nothing quite like outdoor testing in cold weather!

Looks like the polarity was correct, as well as slopes you might try overlapping the crossover points, that is bringing the LF point higher or HF lower to compensate for the acoustical roll-offs. That said, 1600 Hz is already pretty high for a 15"...

Yea I kind of scooted the crossover up a little higher for a little more driver protection for when they get pushed sometimes.
To be honest I need to build another preset for "less demanding" use with a lower crossover.
Title: Re: Tell me what you see in these phase response graphs
Post by: fedele de marco on February 11, 2018, 04:41:35 PM
(Why not? Yes...you can
Also check your measurement setup and signal path.


REW does not have set or find delay like Smaart, it does have loop back delay finder or a use acoustic source to set delay, using both showed the same results.
Title: Re: Tell me what you see in these phase response graphs. Updated measurments
Post by: Jean-Pierre Coetzee on March 24, 2018, 04:36:43 AM
Yea I kind of scooted the crossover up a little higher for a little more driver protection for when they get pushed sometimes.
To be honest I need to build another preset for "less demanding" use with a lower crossover.

I'm not sure what DSP you have but you should be able to get a little better phase response if you have the DSP power available. Also don't just check the response on axis, if you plan to use the box with say a 60 degree horn then check it at points in between as well up to 60 degrees, you may find that crossing over a 15 that high will have detrimental effects as you move further out of the coverage pattern. designing a loudspeaker is about more than just on axis response as I'm sure that Ivan can tell you.
Title: Re: Tell me what you see in these phase response graphs
Post by: Merlijn van Veen on March 24, 2018, 10:50:18 AM
I'd love to hear a more in depth explanation of this statement.  By my understanding both electronic and acoustical manipulations will cause a delay of low end, represented by an upward sloping phase plot towards the low frequency.  How does one achieve a downward slope on the low freq?  Please set me straight.

Correct me if I’m wrong, but if there were to be such a thing as a “perfect” phase trace for loudspeakers, I imagine it would be flat. Like a microphone cable. Preserving the impulse response. Evidently, trying to achieve this, comes with its own set of tradeoffs.
Title: Re: Tell me what you see in these phase response graphs. Updated measurments
Post by: Mike Caldwell on March 25, 2018, 08:10:33 PM
I'm not sure what DSP you have but you should be able to get a little better phase response if you have the DSP power available. Also don't just check the response on axis, if you plan to use the box with say a 60 degree horn then check it at points in between as well up to 60 degrees, you may find that crossing over a 15 that high will have detrimental effects as you move further out of the coverage pattern. designing a loudspeaker is about more than just on axis response as I'm sure that Ivan can tell you.

Now that it's warming up a little I'm going to do some more measurements and test. The last time out my fingers got so numb I couldn't feel the keys on my laptop.

My DSP's are DBX Venue 360's
Title: Re: Tell me what you see in these phase response graphs
Post by: Ivan Beaver on March 26, 2018, 08:15:54 PM
Correct me if I’m wrong, but if there were to be such a thing as a “perfect” phase trace for loudspeakers, I imagine it would be flat. Like a microphone cable. Preserving the impulse response. Evidently, trying to achieve this, comes with its own set of tradeoffs.
I have always believed that the "perfect phase response" would be as follows.

It would "fall" below the lower knee of the response, be flat across the operating passband of the loudspeaker, then fall as the HF response falls.

So it would kind of look like a "S" that was turned 90* to the left. Kinda
Title: Re: Tell me what you see in these phase response graphs
Post by: Chris Grimshaw on March 27, 2018, 04:46:44 AM
I have always believed that the "perfect phase response" would be as follows.

It would "fall" below the lower knee of the response, be flat across the operating passband of the loudspeaker, then fall as the HF response falls.

So it would kind of look like a "S" that was turned 90* to the left. Kinda

I believe that's only a function of the rising/falling frequency response at each end of the passband of the speaker.
If we're talking about ideal speakers, then it'd be good for 1Hz-50+kHz anyway, putting those phase shifts well outside of the audible range.

Chris