Jump to content

Subconscious Auditory Effects of Quasi-Linear Phase Loudspeakers


Chris A

Recommended Posts

So the above requirements for both active and passive crossover filters can be accomplished without using "downward ramping" electrical filters which cause phase and group delay shifts.  The process first looks at the natural response of the crossover ways (drivers/horns) of the loudspeaker and a center crossover frequency is chosen that naturally provides a narrower interference band and does not require boosting of output on one driver channel vs. frequency in the crossover interference bands or excessive attenuation of one driver channel to meet the SPL output of the crossing driver/horn that may have much lower sensitivity (efficiency). 

 

The way this is done is to first measure the natural response of each crossover driver/horn "way", plot the SPL vs. frequency (and phase vs. frequency) to see where the natural crossover frequency is.  Then use attenuating PEQs (also called notch filters) to further attenuate the extensions and peaks of response outside the desired pass band, such as seen in the following measurement:

 

K-402-MEH woofers only (No EQ).jpg

 

The natural crossover frequency of 500-550 Hz is seen from both the SPL vs. frequency and phase vs. frequency curves.  Two more peaks in response above the natural 550 Hz crossover frequency, one at ~1300 Hz and another at ~2500 Hz, will need further attenuating PEQs to push down those response peaks above the woofer crossover frequency.  No other low pass filters are needed on the woofer circuit in order to create a ~24 dB low pass or more crossover between the woofers and the next higher frequency drivers.  This is the technique that is used to create "zeroth order" crossover filters having no induced phase shifts.  This can be done either actively or passively, but note that the effort to successfully do this passively and still meet the above crossover requirements is much, much more difficult than using active PEQs, as found in a DSP crossover.  Additionally, the proper channel delay (usually on the higher frequency channel) in a DSP crossover can be applied at this point to compensate for the physical time misalignments of the drivers/horns within the loudspeaker.

 

Chris

  • Like 1
Link to comment
Share on other sites

Chris, thank you for the explanation.  It all makes sense from what little I know about filters etc. In my case, my LF and HF have a broad region of overlap, so it will be unlikely that I use the natural rolloff of both LF bass bin and the Oris horn since they are not coincidental. In such a case, once I select a crossover frequency, can you show on a diagram or xover display how that should be set up?  My LF bass bin un-restricted, can run from 20hz to about 1200hz IIRC.  My bass bin port resonance lies at 515hz or so. The Oris horn roll-off occurs around 150-200hz. The meat of the usable overlap occurs somewhere around 300-400hz. 

 

I really like the idea for using this technique.  Many Oris owners do not use any crossover on the Oris horn and run them full range, depending on the horn's natural roll-off.

 

The problem I am having trouble visualizing this for the bass bin is that those drivers can respond well above where I want to use the bass bin.  The Kappalite 3015LF drivers don't roll off until about 2kHz, depending on the enclosure.  I want to make sure the PEQ used does not allow the higher frequencies to 'resurrect' on the other side of the PEQ. That is, if I understand the technique correctly.

Link to comment
Share on other sites

It's pretty easy to set one or two PEQs above the crossover frequency to deeply attenuate the added frequency response of the bass bins that isn't needed, still without creating a phase shift between low and higher frequency drivers/horns.  The technique I've described is fairly robust, even when considering drivers whose frequency response has a one or two octave overlap, especially when you've got a crossover with enough PEQs to handle it--like the Xilicas.

Link to comment
Share on other sites

2 minutes ago, Chris A said:

It's pretty easy to set one or two PEQs above the crossover frequency to deeply attenuate the added frequency response of the bass bins that's not needed, still without creating a phase shift between low and higher frequency drivers/horns.  The technique I've described is fairly robust, even when considering drivers whose frequency response has a one or two octave overlap.

How would you set up two PEQs to maximize the attenuation?

Link to comment
Share on other sites

I use REW's EQ facility to help the visualization of what filters to use and the parameters of the filters to reduce the out-of-band response to acceptable levels.  It only takes a few seconds to put in the PEQs in REW (and fewer seconds to put them into the Xilica) needed to accomplish the needed attenuation using PEQs.

Link to comment
Share on other sites

I am using the downloadable Ashly Protea software to play with possible PEQ combinations as compared to a standard crossover. The downside I see to this is the use of at least 2 PEQ positions, which reduce the available PEQ's to work on flattening response.

 

PEQ.png

Link to comment
Share on other sites

  • 3 weeks later...
On 3/20/2019 at 9:02 AM, Arcgotic said:

Regarding the pre-rise of the step response seen with FIR filters usage, what is your observation? Do you have some best practices tips for obtaining a small pre-rise? Number of taps maybe (more or less ?, if the hardware can take it), windowing?

 

I do my filters with rePhase and clearly hear that using FIR filters the sound is not so 'clean'/'crisp' than using simple IIR filters in EqualizerAPO engine. But then again, the phase is more linear using linear-phase FIR filters. 

I recently read a thread (somewhere that I can't remember presently) that basically was saying the same thing as you have above--where they were finding some happy medium somewhere in between the symmetrically shaped impulses that occur with FIR filtering and the one-sided impulse response pulses as seen with IIR filtering.  I do wish that I could find that thread again (but note, the conditions of the subjective testing of this approach did not include full-range directivity of loudspeakers in-room, nor nearfield absorption just around the loudspeakers in order to increase the direct-to-reflected acoustic impulses).

 

This is what I've noticed about my phase flattening exercises and subjective listening relative to your comments above:

  1. Apparently it's the higher frequencies arriving before the lower frequencies that introduces the problems of harshness of string orchestra violin sections and decreased subjective bass response, etc.  This phenomenon has been addressed by Griesinger:

    Griesinger spoken word clarity with without randomized phase of harmonics.GIF
     
  2. Most phase distortion in loudspeakers and other components seem to be low frequency lagging the higher frequencies, which would also be more audible from point "1" above.  When you correct the HF leading the LF, then the subjective listening benefits described in this thread seem to appear.
     
  3. I've noticed that I can get away with a slight pre-rise on the low frequency channel (from slightly overcompensating for LF channel lag via digital delay of the HF channel).  I've found that I can't detect the pre-rise at lower frequencies nearly as acutely as HF pre-rise.
     
  4. When I finally found and corrected a 1-cycle lead of the BMS 4592ND driver HF diaphragm (which I use in the center K-402-MEH loudspeaker) relative to the 4592's MF diaphragm (crossing over at 5.5 kHz) , I found that the center channel K-402-MEH completely disappeared between the Jubs on either side.  Before this change, I had a flat group delay curve at HF where I couldn't see that 1-cycle HF lead of the HF diaphragm, but the phase curve showed an abnormally sloped phase curve toward higher frequencies.  This all occurs above 5 kHz, so the effect that I heard was supposedly out of the most sensitive range of the human hearing curve (especially when factoring in presbycusis), so the effect that I'm describing is actually quite audible even at tweeter frequencies. 
     
  5. I'm guessing that the real-world use of FIR filtering to flatten phase response and decrease the width of the crossover interference bands must be balanced against pre-ringing effects, especially with the HF spectrum leading the LF spectrum.  This isn't terribly surprising in that there is no such thing as a "free lunch" when considering any form of signal processing.  When the signal processing corrections are relatively mild, it seems that the resulting sound is very natural sounding, and conversely unnatural sounding when the signal processing is more extreme.

Chris

Link to comment
Share on other sites

This morning I looked at the phase response of typical tube ("valve" in the UK) amplifiers.  What I found might partially explain why so many people like them with their horn-loaded loudspeakers.  First, I'll post the phase and SPL response of a 1979 Cornwall:

 

627044990_CornwallSPLandPhaseResponse(nocorrection).thumb.jpg.5d96ff1e17ae0595466b08eb73764aa0.jpg

 

Note the leading phase of the higher frequencies (particularly the tweeter frequencies above 5 kHz) in the green trace.  Now an SPL and phase plot of a Mullard ECL86 amplifier (3 w/channel, apparently single-ended, documented in 1962):

 

art003gb.thumb.jpg.90157aae5b012201432c40870d3285b8.jpg

 

Note the phase lead at low frequencies (thus partially offsetting the low frequency lag of the Cornwall's K-33 woofer), and the phase lag at higher frequencies (thus partially offsetting the phase lead of the tweeter of the Cornwall).  The phase curve of the Mullard amplifier would not come anywhere close to offsetting the phase leads and lags of the Cornwall, but my subjective listening experience with phase flattening using DSP crossovers is that any improvement in the phase curve flatness vs. frequency at all is immediately audible. The more phase curve correction toward flatness, the better the resulting sound quality.

 

Perhaps the mysteries of what tube amplifiers are doing to horn-loaded loudspeakers isn't quite as mysterious now that this aspect of phase response and subjective sound quality are examined.  Note that the phase corrections that I have documented in this thread with the Jubilee, K-402-MEH, and surround AMT-1/Belle bass bins are much more effective at flattening phase than what a Mullard tube amplifier would otherwise provide, and the difference in sound quality in accomplishing the phase corrections through crossover filter revamping is therefore much more apparent and effective than using a tube amplifier.  Gone are the harshness of the high frequencies, the soundstage is spatially improved, and the perceived bass response improved substantially, exactly what the adherents of tube amplifiers are reporting vs. SS amplifiers.

 

Chris

  • Like 2
Link to comment
Share on other sites

I'm listening to a CD (demastered by myself 4 years ago) of the Los Angeles Guitar Quartet (Delos 3132) with John Eargle, recording engineer:

 

DE-3132-2.jpg

 

I don't remember the sound quality being what it is right now--extremely rich guitar harmonics with a very pleasant engaging presentation, and sparklingly clear, much like a DVD-A or SACD recording without any further mastering applied.  The only thing that changed from last year is the quasi-linear phase Jubilees from reworking the crossover filters in the Xilica.  The listening is so subjectively inviting that I really can't tear myself away to listen to other Christmas music selections--of which this CD is only part of that genre (the Nutcracker selections). 

 

I continue to be amazed in unpredictable ways with these flat-phase, full-range directivity loudspeakers (relative to the configuration before 1 May of this year)--in a listening room that has only been treated at the very front of the room for early reflections.  This includes nine 2'x2'x1" Auralex Sonofiber squares on each side of the room, arranged adjacent to the mouth exits of the K-402s and Jubilee bass bins (and ~1.5 panels placed across the Jubilee bass bin cabinet front panels). The Jubilees are crossed nominally at 550 Hz.  There is also one double thickness Owens-Corning 703 panel bass trap on top of each horn-loaded sub directly behind each Jubilee, that are positioned across the front room corners to form a closed-end cavity to trap lower frequencies, particularly in the 100-200 Hz band.

 

Chris

  • Like 2
Link to comment
Share on other sites

  • 1 year later...

Chris,

    I read this thread a while back, but I was still trying to figure out how to use DSP to optimize SPL response, so I made a mental note to return to it.  I came across it again over the weekend, and I decided to do an experiment.  I simply changed my crossovers in my DSP from LR24 to BW6.  Still using named crossovers, but dropping from 24dB/octave down to 6dB/octave.  This made a HUGE improvement in clarity and intelligibility.  Not subtle at all.  To say the least, my curiosity was aroused.

   Now I am trying to go all the way and eliminate named crossover slopes entirely, using your "fractional order" method.  I think I understand how it works, but I have one question.  Is there a risk to HF compression drivers by not using a crossover?  My CD/horn combo drops like a stone below ~390hz.  Assuming I throw a few PEQs cuts below this frequency, I can quickly kill output below my 400hz crossover frequency.  However, below these PEQs (in the <150hz region) the input signal goes back up, as it is now outside of the PEQ's effective region.  Does this signal represent a danger to my compression drivers?  Am I going to shatter my CDs by hitting them with bass signals?  Obviously they have no output down there, but does the signal move the voice coil anyway, or do the electrical properties of the voice coil reject signals at that low of a frequency?

 

Thanks

  • Like 1
Link to comment
Share on other sites

2 hours ago, Tarheel TJ said:

Is there a risk to HF compression drivers by not using a crossover?... However, below these PEQs (in the <150hz region) the input signal goes back up, as it is now outside of the PEQ's effective region.  Does this signal represent a danger to my compression drivers? Am I going to shatter my CDs by hitting them with bass signals?

 

How much power is in your HF amplifier channels?  If you're using something that can exceed the input power limits of the compression drivers (greater than 150 w/channel), then you could conceivably create an issue with the driver.  If you're running amplifiers having 100s of watts output per channel and cranking it up using beryllium diaphragms, you're more than tempting fate.  I wouldn't recommend beryllium diaphragm drivers for commercial duty (outside PA, very large cinemas, traveling rock band setups, etc.).

 

But I've found that these fears are way, way overblown.  I've run into perhaps one or two guys that I've helped on the forum (out of the many dozens that I've helped) that are clearly playing their HF channels well above 105 dB, probably around 110-115 dB in a home environment, that have complained about noises in one driver.  In those cases, simply adding the extra 45 degrees or 90 degrees of phase delay (i.e., first or second order high pass filters) will cure the problem.  But I've never heard of anyone blowing a 2" compression driver diaphragm in home hi-fi duty.  Perhaps if you're using them outside and pumping 100s of watts into them, there may be some concern.

 

Chris

  • Thanks 1
Link to comment
Share on other sites

2 minutes ago, PrestonTom said:

Not necessarily a danger, but you will pick up distortion.

...at very high SPL, I might add.  The lower frequency drivers usually roll off by themselves (just like Klipsch Heritage used up until the past 20 years or so--no low pass filters on the midrange K-55 driver). 

 

Chris

Link to comment
Share on other sites

  • Klipsch Employees
6 hours ago, Chris A said:

 

How much power is in your HF amplifier channels?  If you're using something that can exceed the input power limits of the compression drivers (greater than 150 w/channel), then you could conceivably create an issue with the driver.  If you're running amplifiers having 100s of watts output per channel and cranking it up using beryllium diaphragms, you're more than tempting fate.  I wouldn't recommend beryllium diaphragm drivers for commercial duty (outside PA, very large cinemas, traveling rock band setups, etc.).

 

But I've found that these fears are way, way overblown.  I've run into perhaps one or two guys that I've helped on the forum (out of the many dozens that I've helped) that are clearly playing their HF channels well above 105 dB, probably around 110-115 dB in a home environment, that have complained about noises in one driver.  In those cases, simply adding the extra 45 degrees or 90 degrees of phase delay (i.e., first or second order high pass filters) will cure the problem.  But I've never heard of anyone blowing a 2" compression driver diaphragm in home hi-fi duty.  Perhaps if you're using them outside and pumping 100s of watts into them, there may be some concern.

 

Chris

Overblown?  How?  Pun intended. 

Link to comment
Share on other sites

I should add that the advice I gave above on adding a "half first order" crossover filter (i.e., high pass only) is the same type of filter configuration in DSP as the passive crossover filters used in analog, and was used while PWK was still at the helm of the company. I think I remember someone saying that PWK mentioned they  "sounded better".  An example of a type "B" crossover for the Cornwall:

 

CornwallTypeB.jpg

 

Note the absence of a low pass inductor on the midrange, and the single capacitor on the tweeter with a "T2A" autoformer to pad down the tweeter's  and midrange's output.  What PWK didn't have was the ability to directly add delay to the tweeter and woofer via DSP (in the Cornwall's case) to align them in time with the midrange, which had a long full exponential horn that created the time delay.

 

More recent updates of the Heritage line of loudspeakers have seen the replacement of the long exponential midrange horns, K-400, K500, K-600, etc., with a K-510 like midrange (straight-sided/tractrix mouth flare) having a much shallower horn that reduces the time delay of the midrange to the tweeter.  This horn is constructed very much like the following figure from Dr. Bruce Edgar's original 1981 Speaker Builder article on the tractrix horn (approximately the red line):

 

1510934234_TractrixhorncrosssectionconstructionfromBruceEdgar.thumb.GIF.da2e1714c738a4928550b80b18d61127.GIF

 

As you can see, this horn is much shorter--and it's actually about 1/3 the length of the exponential horn it replaced.  That means that instead of the midrange being ~3 wavelengths behind the tweeter at the 4.5-->7 kHz crossover point using the full exponential midrange horn, it's been reduced to about one or two full wavelengths behind the tweeter in the newest versions of the Cornwall and Forte loudspeakers. Vertical separation of drivers from each other also add to the diffraction of the radiated sound field in front of the tweeter+midrange horns/drivers.  I believe that the crossovers used currently are probably using second order filters (Roy hasn't explained that yet), but the second order filters (adding an additional 180 degrees of crossover filter induced phase lag on the lower frequency drivers) only add ~0.68 inches to the physical time misalignment over using first order filters...an added 90 degrees of phase shift over first order that could be avoided.  Given the rave reviews of the Forte and Cornwall updates (and to a lesser degree the Heresy), it seems likely that at least a portion of that increased satisfaction by listeners is due to the reduced time misalignments.

 

Chris

Link to comment
Share on other sites

     It sounds like blowing a compression driver is not much of a concern.  I am running a 30 watt tube amp through compression drivers that are rated for 220 watts continuous, so I guess there is nothing to worry about there.  That said, I'm not sure going to a fractional order crossover really makes much difference in my setup.  I'd be interested to see what others have to say about my results.

    For starters, here is how my system measured about a week ago.  This is using Linkwitz Riley 24dB/octave crossover slopes for every driver and crossover point.  As you can see, the frequency response is relatively flat (other than the sub), but the phase growth is considerable.  Over  700 degrees of phase growth between 90hz and 20khz.

LR24.jpg

Link to comment
Share on other sites

    Next up, I tried keeping the crossovers set where they are, but reducing the slope to a Butterworth 6dB/octave slope.  This made a dramatic difference in phase behavior.  I would agree with Chris that it is a difficult to describe effect, but not a subtle one.  Everything sounds a lot more "together".  The best way I can describe it is that it sounds less like a stereo system and a lot more like real life.  Big step in the right direction.

   The results show a clear difference.  Huge reduction in phase growth, only about 240 degees of phase growth from 90hz-20khz.  A nearly three fold reduction!  This is very audible.  It made me curious to go as far as possible with phase growth reduction.

BW6.jpg

Link to comment
Share on other sites

Next up, I wanted to try Chris' "fractional order crossover" method, to see if I could further reduce phase growth.  Unfortunately, it didn't seem to make much difference for me.  The phase growth stayed about the same as where it was with the BW 6dB/octave setup (~240 degees from 90hz up).  Any thoughts?

 

Fractional.jpg

Link to comment
Share on other sites

3 hours ago, Chris A said:

As you can see, this horn is much shorter--and it's actually about 1/3 the length of the exponential horn it replaced.  That means that instead of the midrange being ~3 wavelengths behind the tweeter at the 4.5-->7 kHz crossover point using the full exponential midrange horn, it's been reduced to about one or two full wavelengths behind the tweeter in the newest versions of the Cornwall and Forte loudspeakers.

 

It also makes it closer in alignment to the woofers in the Forte and Cornwall. That's a win/win...

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...