What is the audible effect of 180 polarity change between HF and mids?

[Jay Barracato]

I appreciate the conversation as I try to move from technician "rule of thumb plus experience" to engineer "theory based decisions supported by measurements". Most of my experience with setting crossovers involves making sure manufacturers recommendations are correctly followed.

This question came up on the other boards, where I really expected that if I was off base, I would hear about it. Distilling the whole thread about the K10 HF came down to "it is always better to match the polarity of the two sources because it gives better transient response". Based on this argument, the other person was amazed that I wasn't willing to give up a whopping chunk of on axis amplitude response, in favor of slightly better response extremely off axis plus this "improved" transient response that showed no where in the data.

Given that the fastest notes in music ever played still consist of over 100 waveforms at a 2000 hz crossover, I have a hard time believing that the initial movement of the vibration source is audible. I can buy that the direction of the cone makes a difference at 80 hz but not in the HF.

Any thoughts?

 

[John Roberts]

Re: What is the audible effect of 180 polarity change between HF and mids?

Absolute polarity is audible for certain asymmetrical waveforms but subtle and generally less audible than a crossover transition with a huge suckout or bump in the summed response. So the absolute polarity if swapped is considered the lesser evil of the two.

Absolute polarity is generally not considered a transient phenomenon but more audible on certain complex waveforms (like brass and vocals). If the dominant components of that complex waveform all fall in the same bandpass, absolute polarity of another bandpass is not significant, so perhaps absolute polarity of the midband (vocal range) bandpass matters even if other bandpasses are flipped.

JR

 

[Langston Holland]

Re: What is the audible effect of 180 polarity change between HF and mids?

Hi Jay:

I'm sure you're assuming proper crossover summation in either case. Obviously if you have a well designed crossover where adjacent passbands are inverted in polarity relative to one another, it'll sound terrible if you simply flip the leads on one of the drivers so that both are in normal polarity.

Study this: http://www.livesoundint.com/archives/2003/july/align/align.php

Our five senses are more aware of change than value. The typical phase trace of a loudspeaker system slopes downward indicating increasingly later arrivals of the beginning of the high frequency waveforms relative to the beginning of low frequency waveforms. This time/frequency modification of the electrical signal that went into the loudspeaker is called phase distortion. The steeper the downward slope is, the greater the time-smearing that has been added and the greater the chance that we'll hear it as a distortion. Additional slope changes over small portions of the phase trace present more opportunities for the ear to know something's wrong. In my experience, phase distortion results in removing depth from the stereo image, smearing of backing vocals such that it's hard to hear individual intonations, muddying of complex instrument waveforms such as piano and acoustic guitar, and a general "this sounds like a PA instead of a band" kind of thing.

EAW's KF730 inverts the polarity of the low and mid/high passbands relative to each other with standard processing (MX8750). The net result of this with this box using the chosen filters and frequencies is less overall phase rotation (change) over the combined output of both passbands. It's the big picture that matters.

With the UX8800 processor, the low and mid/high passbands are still inverted in polarity relative to each other, but the phase trace is horizontal from about 300Hz to 12kHz indicating no phase distortion. How is this possible? To be continued. Maybe.

* Edit to correct my explanation what a downward sloping phase trace means in reference to the arrivals of the high and low frequencies.

 

[Bennett Prescott]

I will also always pick a polarity inversion over additional phase distortion, if inverting a passband saves me a few ms I'm gonna do it. I've never studied the audibility of polarity, but it's nothing compared to another phase wrap. Of course two devices or bandpasses out of polarity are plainly audible...

Langston is incorrect about his interpretation of the phase slope. Delay accumulates in the LF, and must because the same phase shift is a much longer wavelength (time) down there.

 

[Langston Holland]

Re: What is the audible effect of 180 polarity change between HF and mids?

Good point, though I'd like to offer an explanation more in keeping with the train of thought of my post. My poor wording "indicating increasingly later arrivals of the high frequencies relative to the low" should have read "indicating increasingly later arrivals of the beginning of the high frequency waveforms relative to the beginning of the low frequency waveforms".

An illustration that uses a square wave for reference:

"If phase distortion is introduced by a system with non-ideal phase response, all of the harmonics are no longer in phase with the fundamental. In other words, at the fundamentals positive going zero crossing some of the harmonics will no longer be at their positive going zero crossing. As an example, given a 1 kHz square wave, if there is -90 degrees of phase shift at 5 kHz (relative to 1 kHz) then the 5th harmonic that makes up the square wave will have a minima, not a positive going zero crossing, at the reference point in the cycle of the fundamental."

 

[Jay Barracato]

Re: What is the audible effect of 180 polarity change between HF and mids?

Of course two devices or bandpasses out of polarity are plainly audible...

Do you mean two bandpasses producing the same frequency at close to the same level or do you mean you can hear if 500 hz is + and 15kHz is -. That statement doesn't seem to match the detail in the rest of your (highly informative) post.

The crux of the matter is that the phase goes into the crossover region at 0 and comes out at 180 but are basically flat above and below the crossover region. Given that the slope is the same, I would interpret that to mean that the arrival time of each frequency is the same as the frequencies that surround it. Phil's thought problem measurement showed the same type of jump from 90 to -90

 

[Bennett Prescott]

Sorry, Jay, should have been clearer. I am only talking about the magnitude domain effects from two signals in the same frequency range of similar amplitude.

 

[Jay Barracato]

Re: What is the audible effect of 180 polarity change between HF and mids?

That is what I thought. Let me take a shot at summarizing:


The flip in absolute polarity at the crossover may cause cancellation around the crossover if two signals of the same frequency are similar in magnitude but opposite polarity, the flip in absolute polarity at the crossover is probably not noticable when comparing different frequencies within the same passband, and the flip in absolute polarity may cause some distortion if the fundamental is in one passband and a higher harmonic is in the other passband and the opposite polarity.

 

[Bennett Prescott]

Re: What is the audible effect of 180 polarity change between HF and mids?

Jay,Once we start integrating frequency into our understanding of polarity it's easier to just think in phase. With phase all that matters is slope, and having the most consistent slope across the loudspeaker is the goal... or my goal, at least.

Someone else may have more subtlety to offer but I tend to just get 'er done.

 

[Ivan Beaver]

Re: What is the audible effect of 180 polarity change between HF and mids?

Jay,Once we start integrating frequency into our understanding of polarity it's easier to just think in phase. With phase all that matters is slope, and having the most consistent slope across the loudspeaker is the goal... or my goal, at least.

Someone else may have more subtlety to offer but I tend to just get 'er done.

Agreed

There is no "standard" as to whether to flip the polarity or not. As usual "it depends".

And what it depends on is phase. You want to have both pass bands in phase around crossover. This may or may not require the polarities to be the same. It depends on many different factors-which include (but are not limited to) signal arrivals fro both devices-type of crossover filters used, slope of crossover filters used-eq points with an octave or so of crossover (in band and out of band eq).

With our products it is all about the phase. The best phase wins the crossover argument. Sometimes adjacent bands are in polarity-sometimes not. And with 3 way cabinets, they can bee all over the place (in terms of polarity).

As usual this is not an easy question to answer-and one size does not fit all.

Now if all you are concerned with with is the electrical signal out of the crossover-then that is a lot easier. But we don't listen to the electrical signal-we listen with real loudspeakers attached to it, which introduce their own phase shifts, and therefore skew the phase of the electrical signal.

It is the combination of BOTH the electrical and the loudspeaker that give us the final result.

My opinion why phase is not talked aobut very often is that historically it has been much harder to measure and a lot harder to understand than simple amplitude.

I believe our ears are more sensitive to phase than amplitude-but it is much easier to get you head around amplitude-so that is what most people are concerned with.

But loudspeakers can have identical amplitude responses and very different phase responses and sound totally different. Not even close. Yes I have done some testing like that.

And when you remove all the eq that was used to make them flat-the same tonal differences you heard-are still there. Due to the major phase response (and other issues).

It is pretty easy to get a loudspeaker with flat amplitude- and quite another to get a good phase response.

 

[Chuck Simon]

Re: What is the audible effect of 180 polarity change between HF and mids?

Interesting question. As an experiment I switched the polatity between the 12" speaker and the high freq driver in a JBL MP 412 and I could not hear any audible difference. Maybe way off-axis there may have been a difference, but within the usual listening area of a floor monitor, I could not tell the difference.

 

[Ivan Beaver]

Re: What is the audible effect of 180 polarity change between HF and mids?

Interesting question. As an experiment I switched the polatity between the 12" speaker and the high freq driver in a JBL MP 412 and I could not hear any audible difference. Maybe way off-axis there may have been a difference, but within the usual listening area of a floor monitor, I could not tell the difference.

On some designs there is little change in the amplitude-but the phase will change.

On other designs there is a huge hole created when the polarity is swapped. Sometimes it is wide-sometimes it is very narrow.

Depending on what you were listening to (pink noise would be best) there may or may not have been some material content around the freq involved.

Also with a floor wedge, you could have been getting some reflections from the floor that could have masked the particular area of interest.

Did you measure it-and not with an RTA? That would be the real test as to whether or not it made a difference.

 

[Jay Barracato]

Re: What is the audible effect of 180 polarity change between HF and mids?

On some designs there is little change in the amplitude-but the phase will change.

On other designs there is a huge hole created when the polarity is swapped. Sometimes it is wide-sometimes it is very narrow.

I think this is the crux of the problem with the K10 and trying to reverse engineer what is happening in the crossover from a summed response.


Looking at the phase well above the crossover and saying the HF is out of polarity because it is 180 degrees from the mids really doesn't say anything about what is happening in the crossover region.

Flipping the polarity of the HF creates such a big suckout in the crossover region, that I believe that flipping the polarity of the HF to match the mids by looking at the higher frequencies is actually causing it to be out of polarity in the crossover region causing the cancellation.

 

[John Roberts]

Re: What is the audible effect of 180 polarity change between HF and mids?

As I suggested in my earlier post... the wrong absolute polarity between passbands is the lesser evil than an amplitude bump or suck out in the transition region.

Note: Ignoring driver phase shift, transition region interference and relative polarity is more of an issue for even order filters than odd, while the ever popular L-R uses an even order filter alignment and same polarity. When the driver phase shift and amplitude loss is already significant they must be added to the filter response.

I believe our ears are more sensitive to phase than amplitude-but it is much easier to get you head around amplitude-so that is what most people are concerned with.

This is contrary to my understanding and experience. I will remind all that the knobs on equalizers are calibrated in amplitude of boost/cut, while EQ tweaks do alter the phase response too.

Phase errors are most audible when they cause constructive or destructive interference that changes amplitude.

That said it all matters, I would just not discount the importance of amplitude response, while it matters how measured. Impulse or transient (amplitude) response could be quite different than steady state sine wave measurements, where 360' or 720' of phase shift are indistinguishable from 0'.

JR

PS: Our brains are pretty sensitive to arrival time of transient events, so this is another possible variable in the mix.

 

[drew gandy]

Re: What is the audible effect of 180 polarity change between HF and mids?

But loudspeakers can have identical amplitude responses and very different phase responses and sound totally different. Not even close. Yes I have done some testing like that.

Hi Ivan,

I've been meaning to ask you about this (as you've mentioned this testing before). Can you elaborate on the speakers in question? I'm trying to figure out how you eliminated the other variables to arrive at the conclusion that it was the phase alone that made the difference. For instance, I think it could be possible to EQ a Magnepan to measure like an SH95 in terms of amplitude (and I'm pretty sure the phase would not look the same) but there are some other huge differences between those 2 speakers that might account for a difference in sound.

-drew

 

[Ivan Beaver]

Re: What is the audible effect of 180 polarity change between HF and mids?

Hi Ivan,

I've been meaning to ask you about this (as you've mentioned this testing before). Can you elaborate on the speakers in question? I'm trying to figure out how you eliminated the other variables to arrive at the conclusion that it was the phase alone that made the difference. For instance, I think it could be possible to EQ a Magnepan to measure like an SH95 in terms of amplitude (and I'm pretty sure the phase would not look the same) but there are some other huge differences between those 2 speakers that might account for a difference in sound.

-drew

While it may have seemed like I said the phase was the only thing that was different (and accounitng for the difference in sound), that is not totally what I meant.

Now this was a number of years ago-so some of the details are a bit "foggy".

Both of the loudspeakers used were products that I had no affiliation with-only that we were dealers and had installed quite a number of each over the years-both with good results for the customer.

The simple answer to what I meant was that given two loudspeakers (at least the ones I was playing with)(by respected deisgners -not pyle or Behringer type stuff-these were in the $3000ish-retail- each price range) that have similar price points-similar coverage patterns-similar sized horns (mids and highs)-front loaded woofers-similar sensitivity and similar overall physical size-same size exit HF driver and similar sized mids-the woofers are quite different (both in size and number)-sound quite different-even though the amplitude response was (for all intents and purposes) identical. I would say it +-1/2dB (or less) across the entire freq response.

I did not listen at loud levels (just "healthy)-so both products were well within linear operation.

Here was my basic setup. This was in a fairly large space-so reflections (while still there) were not a big influence on the sound). Both speakers side by side with the mic in the middle about 20' away. I used the one with the least amount of bass (they were similar to begin with) as the reference on the low end and the same thing on the high end.

So I set high/low pass filters so they both had the same bandwidth. I then applied eq (basically until I ran out of filters-I used something like 60 filters??) to the different boxes. Each was driven by the same model amplifier.

If all you looked at was the amplitude-they looked identical. The phase response was quite different-but that was not the intent of the test-only to see what would happen with boxes that had identical amplitude response.

There was quite a bit of difference in the sound quality of the boxes. Not a subtle thing at all. It was not "let me hear the other one again" type of thing-but rather "I can easily pick out which one is which".

So I removed all the eq and listened again. The "sound" difference was still easily picked out-and it was the same "sound" as with all the eq in.

The same thing (kinda) happened recently at the shop. A guy for a demo brought in a loudspeaker to be compared to. So we set it up with a product that had a similar response and sensitivity. We did not go beyound that-no eq-bandwidth matching etc.

There was a distinct sonic difference between the cabinets that was easily picked out.

So basically all I am saying is that there is A LOT more to the sound of a loudspeaker than just amplitude response. Phase is one of those. But there are "other" factors that come into play.

And in general listening-the loudspeakers have a good phase response tend to "sound better-clearer etc" to me (even if the amplitude is a bit all over the place). Amplitude is easily corrected (at least at one spot-depending on the loudspeaker-but the other factors are not so easily corrected.

Even with the best DSP processing available-you can only "correct" fundamental physical errors so much.

Kinda like water filtration. You can take the nasty creek water and make it drinkable-but it will never taste like fresh mountain spring water (assuming the deer haven't pissed in it).

But that is my opinion and my experiences from real world testing-so others may have other opinions.

 

[John Roberts]

Re: What is the audible effect of 180 polarity change between HF and mids?

I think the point of disagreement is reaching for simple cause or explanation for what we hear. No doubt simple EQ can never correct for say time offsets between drivers or other error sources that will degrade the sound reproduction, even if we EQ these to measure flat for steady state sine wave testing. As I suggested before steady state response will not be the same as transient response, because 360' and 720' of phase shift look indistinguishable from 0' in steady state tests but not for transients.

Generating sound fields in real rooms is far more complex than simple amplitude response and EQ as a tool has limitations. If the raw material is seriously flawed, you are just putting lipstick on a pig. Even if we constructed a phase shift equalizer and dialed in the steady state phase response too, we could still have the same difference between steady state and transient response. Complete phase correction really needs to be time correction, and for multi-driver loudspeakers even this can only be correct for one listening axis or vector.

This seems like one of the attractive things about different drivers summing together in one horn, if I understand the technology correctly, and if they are properly dialed in, but I am not the speaker guy here.

JR

Note: all things equal, EQ for flat steady state response is better than not, since most music is made up of relatively persistent tones, not just transient bursts, but this is only one aspect of the sound quality so making this perfect only improves but does not fix basic flaws.

 

[Jay Barracato]

Re: What is the audible effect of 180 polarity change between HF and mids?

While music is definitely more transient in nature than a steady state test signal, I can't help but wondering if any meaningful definition of "transient response" hasn't been destroyed by marketing programs.

Here is a "Fermi" numbers estimation for the number of waveforms in a guitar note tone at 2000Hz:

The fastest guitar player out their playing at 220 bpm in 16th notes would play 14.7 notes per second or 0.068 sec per note

At 2000 hz each waveform is 0.0005 sec

Therefore, each note consists of 136 waveforms

For the same guitar player at 500 hz, there are still 34 waveforms per note.

I don't believe I have ever seen any research into "how fast can we hear" or what is the minimum number of vibrations needed for us to recognize it as sound.

 

[John Roberts]

Re: What is the audible effect of 180 polarity change between HF and mids?

You can educate yourself about the audibility of stuff like this with tone bursts. Back in the '70's when I was designing dynamic processors like companding noise reductions and compressor/limiters, I built my own tone burst generator. Basically a glorified on/off (actually variable amount of off) gate that started and stopped at zero crossings for adjustable on and off times. I even made sure it presented even numbers of positive/negative cycles. (if you don't gate at zero crossings you will hear HF clicks. If you don't present full cycles you will get DC content that will cause cap coupled paths to bounce up and down. )

As you progressively shorten the number of cycles presented of a sine wave burst, the sound character becomes less dominated by the fundamental pitch of the sine wave, and more influenced by the characteristic frequency of step function defining the on/off gate or burst time. This is kind of predicted by mathematical analysis of the waveforms and the step function it is multiplied by, but simply put a short one or two cycle long tone burst sounds more like a click or thump than a tone.

The majority of sound energy we receive is tone like, and steady state EQ matters for that. But there is a lot of information in the transients that is not tone like and not so easily corrected.

I suspect listening to tone bursts in loudspeakers could be revealing, while not immediately obvious which one is more accurate when two speakers sound different. Maybe develop a baseline for what the short burst should sound like first by spending some time with high quality headphones listening to bursts.

I found the completely adjustable on time, rep rate, and amount of attenuation when gated off, helped me parse out quirky behavior in dynamic circuits that might reveal themselves on only a single note in a musical passage.

JR

PS: My old tone burst gate, could also be used to gate musical inputs so I could artificially increase the dynamics or peak to average (crest factor) of pre-recorded music to better stress and isolate transient response flaws in dynamic processors. I don't expect loudspeakers to have the same kind of quirks related to dynamic level changes. I suspect speakers to instead have more static dead spots or unwanted honks on spot frequency content (like the length of a burst in combination with a sine wave carrier pitch). Of course this is mostly speculation about speakers on my part. I am not a speaker guy. (They're too hard

 

[Ivan Beaver]

Re: What is the audible effect of 180 polarity change between HF and mids?

As usual JR is correct.

i remember the first time I hear short bursts of HF sinewaves.

The freq that we "heard" was nowhere near the freq of the sinewave bursts. It "appeared" to be several octaves lower-and sounded more like a thud/cllick than a high freq tone.

The origional was around 5-8Kish if I remember correctly.

And nowhere as loud as you would think. A quick easy way to tear up some drivers if you try to get it loud.