Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Frank Koenig]

I've measured a few (horn-loaded, compression driver) speakers that exhibit abrupt phase shifts in the top octave while others do no such thing. In speakers where I see this the frequency at which the phenomenon occurs varies with angle and is excess phase (phase minus minimum-phase derived from log-magnitude by Hilbert transform). Speakers that don't do it on-axis appear not to do it at any angle. Speakers that do it sometimes do it at more than one frequency. (These are all windowed impulse response pseudo-anechoic measurements.)

When I started seeing this I wrote it off as some unexplained artifact of my measurement or analysis technique (failed phase unwrapping or some-such) but now I'm convinced it is real. For a given measurement setup it shows up in Smaart, ARTA, and my own analysis code that uses ARTA-captured impulse responses.

The examples are all from ARTA. Speaker A has it and speaker B does not.

At the risk of supplying insufficient information, I don't want to get into makes and models right now. I'm just wondering if any of you who have measured far more speakers than I recognize this pattern and understand its cause. Is it of perceptual significance? Is the driver or the horn or the combo responsible? It's hard to conclude anything from listening since it's always apples to oranges.

Thanks and Happy New Year.

--Frank

[Chris Tsanjoures]

looks like some sort of comb filtering, perhaps caused by horn design, or lack of DSP to correct for horn related anomalies/driver interactions?

actually, knowing what the gear is that you are using and the dsp applied to it is part of the how we would go about coming up with some sort of opinion as to why this is happening. For example, I would expect to see behavior like this with a system comprised of many smaller tweeters arranged together but not close enough to couple at those frequencies.

[Merlijn van Veen]

Is a horn a reflector or an absorber?

I have come to suspect reflections within the horn. At 14 kHz you're dealing with wavelengths of less than one inch. That's pretty small compared to the size of your typical horn and there are a lot of ways it could bounce along the way.

And Happy New Year to you Frank 

[Frank Koenig]

Is a horn a reflector or an absorber?

Hi Merlijn,

I believe horns are intended to be reflectors, although I have seen bits of absorber stuck on them here and there to solve problems

I agree that this looks like some kind of reflection, especially since the frequency of the flip varies with angle. Maybe we're seeing different path lengths for the reflection as we are subject to different aspects of the horn's geometry. If it were, say, a non-piston mode in the diaphragm, I would expect the frequency to be independent of the angle, but I'm totally prepared to be wrong about that (or any of this). And yes, for the first zero to be at 13 kHz the path length difference would have to be about .5 in. Getting down there in scale.

On the other hand, shouldn't a simple reflection (comb filter) be minimum phase?

  H(s) = 1 + Ae^(-st) where -1 <= A <= 1

Neither of the measurement examples has any DSP in the chain, although there is a bit of ASP in each (speaker-level analog crossovers). In both cases the phenomenon of interest is far enough above the crossover frequency that the crossover is likely irrelevant. Speaker B is a BMS 4592ND 2" coaxial driver using BMS's crossover on a BMS 60 x 40 horn.

Thanks for the replies.

--Frank

[Ivan Beaver]

It looks like a reflection.

Have you tried measuring the driver WITHOUT the horn?

My first guess is that it is in the phase plug. 

Which would explain the on axis and off axis "agreeing" with each other.

[Art Welter]

It looks like a reflection.

Have you tried measuring the driver WITHOUT the horn?

My first guess is that it is in the phase plug. 

Which would explain the on axis and off axis "agreeing" with each other.

I'd agree. I have measured a lot of different drivers on the same conical horn, some have smooth phase (and FR) all the way out, others show the same type of phase shift as in the OP, associated with either the phase plug or VHF resonance/breakup in the driver (or both).

[John L Nobile]

How is that ARTA app? I've been looking for something more affordable than Smaart and this hits the right point.

Sorry to go off topic

[Frank Koenig]

Have you tried measuring the driver WITHOUT the horn?
Which would explain the on axis and off axis "agreeing" with each other.

In this case the existence and the frequency of the abrupt phase shifts appear to vary with angle. I took measurements at roughly 10 deg increments and 10 ft. distance -- optimistically in the far field for this horn. In the image below I overlayed phase plots for 6 measurements, individually adjusting the delay to compensate for the small differences in measurement distance. (I could have overlayed the plots a little better but .01 ms is the delay increment in ARTA, which is a whopping 36 deg at 10 kHz.)

The big question is, can anyone hear it? My guess is no, since from what I've read about perception of phase (in isolation, not talking about coupling of multiple sources), any effect is confined to lower frequencies.

The good news is that since this comes and goes with angle there is nothing to do about it on the processing side. Flatten the average magnitude, play nice music, and be happy.

On measuring drivers without a horn, not that I'm likely to try this anytime soon, do I need a plane-wave tube setup to terminate the driver? I believe that is one way it has been done. It sounds like Art has a "reference horn", which is good for driver comparisons.

Thank you all for your insights.

Best,

--Frank

[Frank Koenig]

How is that ARTA app? I've been looking for something more affordable than Smaart and this hits the right point.
Sorry to go off topic

ARTA is awesome, but it is not a replacement for Smaart. ARTA is not "real time". It's great for "lab work" where you can take your time and make systematic measurements. It has a lot of great analysis tools, both for speaker and room work. It also has really nice documentation that explains exactly what it does, unlike Smaart which is a (somewhat irritating) black box. But if you try to use it to align subs 30 minutes from downbeat you'll go nuts. You need Smaart, or something like it, for that.

I think there have been a number of threads on this topic.

Best,

--Frank

[John L Nobile]

ARTA is awesome, but it is not a replacement for Smaart. ARTA is not "real time". It's great for "lab work" where you can take your time and make systematic measurements. It has a lot of great analysis tools, both for speaker and room work. It also has really nice documentation that explains exactly what it does, unlike Smaart which is a (somewhat irritating) black box. But if you try to use it to align subs 30 minutes from downbeat you'll go nuts. You need Smaart, or something like it, for that.

I think there have been a number of threads on this topic.

Best,

--Frank

Thanks Frank. This is the first I've heard of ARTA. Guess I'll have to save up a little longer for Smaart

[Ivan Beaver]

In this case the existence and the frequency of the abrupt phase shifts appear to vary with angle. I took measurements at roughly 10 deg increments and 10 ft. distance -- optimistically in the far field for this horn. In the image below I overlayed phase plots for 6 measurements, individually adjusting the delay to compensate for the small differences in measurement distance. (I could have overlayed the plots a little better but .01 ms is the delay increment in ARTA, which is a whopping 36 deg at 10 kHz.)

The big question is, can anyone hear it? My guess is no, since from what I've read about perception of phase (in isolation, not talking about coupling of multiple sources), any effect is confined to lower frequencies.

The good news is that since this comes and goes with angle there is nothing to do about it on the processing side. Flatten the average magnitude, play nice music, and be happy.

On measuring drivers without a horn, not that I'm likely to try this anytime soon, do I need a plane-wave tube setup to terminate the driver? I believe that is one way it has been done. It sounds like Art has a "reference horn", which is good for driver comparisons.

Thank you all for your insights.

Best,

--Frank

If it was a reflection in the phase plug-it could show up differently at different angles

I would measure without any sort of termination-just the raw driver and see what you have.

[Art Welter]

In this case the existence and the frequency of the abrupt phase shifts appear to vary with angle. I took measurements at roughly 10 deg increments and 10 ft. distance -- optimistically in the far field for this horn.

The big question is, can anyone hear it? My guess is no, since from what I've read about perception of phase (in isolation, not talking about coupling of multiple sources), any effect is confined to lower frequencies.

On measuring drivers without a horn, not that I'm likely to try this anytime soon, do I need a plane-wave tube setup to terminate the driver? I believe that is one way it has been done. It sounds like Art has a "reference horn", which is good for driver comparisons.

Frank,

Personally, I can't hear the phase differences, and the frequency response deviations where the problem lies are so high that it would take an A/B test (for me) to identify the problem compared to a smoother unit. Some folks are blessed (or cursed) with hearing more sensitive to ragged FR.

Testing the driver raw would show the problem, though diffraction around the driver body can make identification difficult- I have used a 1/8" sheet of plastic (or Masonite) with a driver hole cut in, and the hole rounded to make a relative "half space" comparison. This can be useful to see how wide the phase plug will spread VHF, and expose driver problems, as opposed to horn problems.

The Maltese horns I use for comparative testing are very smooth above 1kHz, only about 1/4 dB of "ripple", with only 13 x 13 degree conical expansion, so any driver will "fill" the horn, with the proper throat adapter. It loads to around 200 Hz, so the LF response of drivers are also easily compared.

Below you can see the HF phase/FR problem with a B&C DE82 compared to a BMS 4552, a trade of cleaner LF for HF.

Art

[Ivan Beaver]

Frank,

Personally, I can't hear the phase differences, and the frequency response deviations where the problem lies are so high that it would take an A/B test (for me) to identify the problem compared to a smoother unit. Some folks are blessed (or cursed) with hearing more sensitive to ragged FR.

Testing the driver raw would show the problem, though diffraction around the driver body can make identification difficult- I have used a 1/8" sheet of plastic (or Masonite) with a driver hole cut in, and the hole rounded to make a relative "half space" comparison. This can be useful to see how wide the phase plug will spread VHF, and expose driver problems, as opposed to horn problems.

The Maltese horns I use for comparative testing are very smooth above 1kHz, only about 1/4 dB of "ripple", with only 13 x 13 degree conical expansion, so any driver will "fill" the horn, with the proper throat adapter. It loads to around 200 Hz, so the LF response of drivers are also easily compared.

Below you can see the HF phase/FR problem with a B&C DE82 compared to a BMS 4552, a trade of cleaner LF for HF.

Art

DAMN TRADEOFFS-------------

Why can't anybody make perfect drivers??
Yeah it is what we "wrestle with" all the time.  Yes we have to give up one area of performance for another.

It comes down to "What is most important about where/how I am going to be using it".

It may not make sense to somebody else who is looking at a different part of the performance.

But, yet another reason loudspeakers are by far the worst part of the whole reproduction chain.

[Frank Koenig]

I have used a 1/8" sheet of plastic (or Masonite) with a driver hole cut in, and the hole rounded to make a relative "half space" comparison. This can be useful to see how wide the phase plug will spread VHF, and expose driver problems, as opposed to horn problems.

Art, Ivan, I'm more of a curious consumer of loudspeaker systems than a developer, except perhaps with respect to the measurement representation and DSP/ASP side of things, which is purely a hobby at this point. But it's extremely helpful to know that I'm running into real phenomena that others have seen and not just chasing artifacts. That purple trace looks familiar. I'll try the infinite baffle trick someday -- thank you Art for that.

Best,

--Frank

[ProSoundWeb Community]