Jump to content

Subconscious Auditory Effects of Quasi-Linear Phase Loudspeakers


Chris A

Recommended Posts

Here are the settings that I use for spectrogram view (the sub-menu "Appearance settings" menu is also shown, but is only seen when you press its button on the Controls menu--shown at the top left of the screen shot below):

 

image.png.5e791c7bb968bca775c6056d128b713a.png

 

Chris

Link to comment
Share on other sites

On 2/18/2021 at 9:00 AM, Tarheel TJ said:

Second, It seems that in all of these measurements, phase grows considerably below 100hz, regardless of crossover settings.  Why is this?  Is there something about bass frequencies or drivers that fundamentally creates phase growth?  Is there anything that can be done to reduce phase growth at lower frequencies?

Okay, so I haven't really addressed this question yet.  It's a really, really important one, I've found.  Let's start with an excerpt from Toole's book (Loudspeakers and Rooms, 1st Ed., pg. 420):

 

Quote

18.6.2 Phase Response—The Low Bass

In the recording and reproduction of bass frequencies, there is an accumulation of phase shift at low frequencies that arises whenever a high-pass filter characteristic is inserted into the signal path...Finally...is the loudspeaker, which cannot respond to DC and must be limited in its downward-frequency extension...

 

Fincham (1985) reported that the contribution of the loudspeaker alone could be heard with specially recorded music and a contrived signal, but that it was “quite subtle.” The author heard this demonstration and can concur. Craven and Gerzon (1992) stated that the phase distortion caused by the high-pass response is audible, even if the cutoff frequency is reduced to 5 Hz. They say it causes the bass to lack “tightness” and become “woolly.” Phase equalization of the bass...subjectively extends the effective bass response by the order of half an octave...

So it's the high pass characteristics of the woofers/horn-loaded bass bins which causes this increase of phase with frequency (the higher its cutoff frequency, the steeper the phase curve below 100 Hz).  Since this phase increase is generally not linearly increasing with decreasing frequency, the group delay also begins to take off.

 

Let's examine a few bass bins for their low frequency phase and group delay response. First the phase response (legend is below the plot):

 

 

651172701_LowBassPhaseResponseofSeveralBassBins.jpg.369746960b71dcf37e05068abc019fec.jpg

 

And then group delay:

 

1664943078_LowBassGroupDelayResponseofSeveralBassBins.jpg.5fbf31eec6b462cef316869d05a95975.jpg

 

 

Some observations:

 

1) The lower the cutoff frequency of the bass bin (the TH-SPUD having the lowest at 14.6 Hz, the La Scala having the highest at 60+ Hz), the tendency is a shallower rise in phase response, and the more interesting the group delay response This turns out to be more complicated.  Tapped horns have a phase characteristic that grows with decreasing frequency, because the acoustic center between the two sides of the woofer(s) in a tapped horn lengthens from the mouth of the horn as frequency decreases. 

 

2) You can't see the rise in the group delay response of the Cornwall bass bin (bass reflex) at the port frequency--about 20-25 Hz.  I attribute this design characteristic for the general acceptance of the Cornwall as a loudspeaker not typically sounding like a "bass reflex box".  But the higher levels of modulation distortion from the bass reflex box (only viewable using dual tone signals) is much higher than the horn-loaded bass bins.

 

3) The tapped horn subwoofer exhibits the steepest phase and group delay curves, even within its narrow useful passband (18 Hz to 40 Hz).  This is clearly attributable to the acoustic center of the horn moving progressively backward toward the longest path length portion of the horn, with the front side of the woofers being very close to the horn's mouth.  As the frequency decreases, the contribution of the reverse side of the tapped horn woofers through the long horn path length (~21-23 feet, in this case) increasingly contribute more and more to the summed output, thus moving the acoustic center progressively backward. [I would not expect to see this same sort of characteristic with conventional front-loaded horn subwoofers.]  As the frequency rises above some breakpoint frequency of the TH subwoofer, again the apparent acoustic center appears to move backward again, probably due to backwave vs. frontwave phase cancellations shifting more and more toward cancelling the summed output at higher frequencies.

 

4) All the other horn-loaded bass bins use exponential volume expansion in each case, except the K-402-MEH, which is mostly straight-sided (more than conical expansion--and then room coupling expansion like all the others below ~100-120 Hz), and seem to behave very much the same and exhibit low phase growth relative to bass reflex (Cornwall) even below their cutoff frequencies.

 

5) The KPT-1802-HLS subwoofer behaves more like a bass reflex bass bin in terms of phase and group delay response than it does a front-loaded horn, or I should say, "intermediate" between the other horn-loaded bass bins and the Cornwall, except that it is voiced about 10-15 Hz lower (-3 dB cutoff point) than the Cornwall bass bin.

 

6) The bass phase and group delay response of the K-402-MEH is measured while the loudspeaker is mounted mid-wall in an elevated position (about 1.2 m off the floor to the horn centerline, i.e., nominally "half space").  The Cornwall is measured with its back against a wall on the floor (i.e., nominally "quarter space"), the Jubilee and TH-SPUD are measured fully in a room corner (nominally "eighth space"), and the KPT-1802-HLS is mounted recessed into the front wall on the floor (nominally "quarter space").  Note that at the frequency where the closest walls and ceiling couple to the bass bin acoustically is ~1/4 wavelength.  Below is a table for frequency vs. 1/4 wavelength length/distance.  By that rule of thumb, in the cases above, all bass bins are in eighth space loading by 10 Hz, most by 30 Hz, and only the corner-loaded bass bins at 40 Hz.

 

Frequency (Hz)   1/4 wavelength distance (ft)    1/4 wavelength distance (m)

10                                          28.3                                               8.62

20                                          14.2                                               4.31

30                                            9.4                                               2.88

40                                            7.0                                               2.16

 

7) All of these measurements have the calibrated microphone well inside 1/4 wavelength distance--indeed, well within 1/8th wavelength, so near-field effects are present (just like they are in virtually all home hi-fi listening rooms at these frequencies).  All measurements start at 10 Hz (except for the Cornwall, which starts at 15 Hz) using REW's own sweep generator, which ramps up to reference voltage level at a controlled rate and then ramps again at much higher frequencies (well above those plotted above). 

 

Chris

Link to comment
Share on other sites

By the way, the relative vertical axis position of each phase trace (above) doesn't mean anything--only the relative slope of the phase curves.

 

Chris

Link to comment
Share on other sites

2 hours ago, Tarheel TJ said:

Ok.  I think I got all that right.  Take a look and let me know how you interpret this.  Thanks

 

Spectogram.2.18.21.jpg

That looks very good.  I don't see anything from this view that I can comment on.

 

Chris

  • Thanks 1
Link to comment
Share on other sites

So for those that followed the above conversation on phase flattening below 100 Hz, they might ask, "can you actually hear it?".  For me, that question has a strongly affirmative "yes". You can hear it.

 

So how much?  Here's my answer: I had to re-EQ by demastered stereo tracks (all of them) to cut down on the degree of bass vs. treble response in them .  How much?  Here's the typical correction demastering curve that I first apply when going back to the tracks that I've demastered over the past 6 years:

 

image.png.758d54efcea8b769a459c37289edfec6.png

 

That's about 6 dB of total end-to-end correction that you see in that curve (I show a zoomed-in plot above for clarity).  That doesn't sound like a lot, but that's showing that I'm hearing about double the amplitude of the bass bins below ~50-100 Hz than I did before, and the Jubilee bass bins were not boosted in SPL vs. frequency, but rather just their phase was flattened relative to the compression driver pass band through the crossover interference band.

 

😮

 

Chris

Link to comment
Share on other sites

I assume phase flattening below 100 Hz mentioned above does not eliminate the need to restore the "Missing Octave(s)"?  

 

Was your re-EQ work that provided 6 dB of total end-to-end correction done to all your music files saved on disk?  If so, is it possible to re-EQ the original music once to correct recording mistakes and save those changed files as the new "master" copies?  Then if any further tuning needs to be done to the music due to a DSP change, room treatment, new driver, or a new amp, use DSP to make the adjustment once for all music being played?

 

Mark

Link to comment
Share on other sites

25 minutes ago, Marks said:

I assume phase flattening below 100 Hz mentioned above does not eliminate the need to restore the "Missing Octave(s)"?  

Yes.  There are many, many stereo albums released before 1991 (and some after 1991) that need restoration of bass below ~100-200 Hz.

 

26 minutes ago, Marks said:

Was your re-EQ work that provided 6 dB of total end-to-end correction done to all your music files saved on disk?

No, generally I have to restore each track separately, unfortunately, depending on the instrumentation and genre/mix of each track. 

 

For some albums, I can run a macro (in Audacity, they're called "chains") do batch process an entire album, once I find the right tweaked demastering curve.  Notably, this applies to classical and jazz tracks.  Rock tracks seem to take the most effort to get right.

 

30 minutes ago, Marks said:

If so, is it possible to re-EQ the original music once to correct recording mistakes and save those changed files as the new "master" copies?

I write over the old tracks after verifying their sound quality.  Some albums it's been easier to go back to the original CD and re-rip the tracks, then demaster from scratch (this generally applies to those albums that I found were more difficult to demaster initially).  Also, some of the early albums that I demastered from early 2015 were easier to re-rip and demaster anew.

 

33 minutes ago, Marks said:

Then if any further tuning needs to be done to the music due to a DSP change, room treatment, new driver, or a new amp, use DSP to make the adjustment once for all music being played?

I thought about doing it that way but found that it's easier just to get all the stereo tracks in my ripped/demastered library up to the same standard (including adjusting their average loudness back to zero on the ReplayGain scale) so I don't have to mess with them again. 

 

I really don't expect to have to go back again to update the stereo tracks that I've demastered.  After demastering something like 1500 albums over 6 years (that's well over 20K individual tracks, with many now re-demastered for the second time), I think my learning curve has reached a plateau, including getting the Jubilees dialed in with flat SPL and phase so this doesn't happen again.

 

Chris

Link to comment
Share on other sites

  • 1 month later...

I found another article over at Linkwitz's web site on this subject (flat or linear phase response) that basically says the same things that you find above, but with more detail and in other areas I haven't discussed (perhaps explaining more on the "subjective" portion of this subject):

 

http://www.linkwitzlab.com/Attributes_Of_Linear_Phase_Loudspeakers.pdf

 

Chris

Link to comment
Share on other sites

  • 3 months later...
On 6/5/2019 at 11:30 AM, Chris A said:

all you have to do is put the HF and LF drivers together without phase shifts:

  1. Don't use the "crossover filters" that come with DSP crossovers--clear any crossover filters if they're set.
  2. Set the HF or LF channel delay to get perfect impulse response in the time domain--as seen in the spectrogram view.
  3. Flatten each driver's SPL response within their pass bands.
  4. Match the channel gains between flattened phase drivers.
  5. Use output channel PEQs to trim off response on each end of the bass and high frequency drivers until you've got overall flat SPL across the crossover interference band and smooth handover of SPL vs. frequency.  The drivers themselves will tell you where that transition/crossover should occur.  [If you're using MEHs, you'll have to use multiple PEQs to attenuate the bass bin peaks in response above the first notch frequency.]
  6. Use the input channel PEQs to further flatten the overall response within the interference band to correct any dips or peaks in response within that band.

Voila!  Flat phase.  It's really that easy.

 

Chris

Chris,

I think I understand 1/2 of this. 
instead of using Xilica supplied crossovers, we use natural fall off of drivers, possibly tweaking with PEQ, to create a natural crossover point? 
Can you supply more details, dumbed down as much as possible?

Ted

Link to comment
Share on other sites

12 hours ago, VDS said:

...instead of using Xilica supplied crossovers, we use natural fall off of drivers, possibly tweaking with PEQ, to create a natural crossover point?

 

Yes. Once the drivers are individually flattened in their SPL response, I look at the natural crossover point between drivers/ways that's about centered in their downslope on either side of the desired crossover frequency.  Then I add attenuating PEQs to steepen the natural fall-off in SPL to each output channel, just until their natural crossing point is achieved, and the sum of the two driver/horn ways visually match what you'd get if you used regular crossover filters (typically with near-second order slopes on either side of the crossover frequency, like this:

 

Jubilee with Danley-Style Crossovers SPL Response.jpg

 

In this example, I added attenuating PEQs on the TAD (high frequency compression driver, yellow trace, below) channel and a very small amount of attenuation on the bass bin response (red trace), shown here:

 

image.thumb.png.be042c15e377d9caac90b5352274e9fa.png

 

The biggest attenuating PEQ is #4 here:  -10 dB at 400 Hz, with 0.14 octave bandwidth. 

 

The bass bin PEQ is not shown in the table above, but is -3 dB at 488 Hz with 0.25 octave bandwidth.  That's all I had to add to the output channel PEQs to get the two drivers/horns to sum to the yellow trace in the top figure.  That's it. No other filters are used, and more importantly, no phase shifts occur through the crossover region. Below you will see that there is no phase growth around the ~550 Hz crossover point that was naturally chosen--which is the big payoff of using this method:

 

2119949661_JubSPLPhaseusingZerothOrderXOvers.jpg.b9f99ad932f6c9749ed6a6dc93120e02.jpg

 

Anticipating your next question..."how do I choose the crossover point and PEQs?"--I simply look at the flattened responses and choose one frequency, then run full sweep measurements with both drivers/ways to see their summed responses, adding attenuating (and sometimes small boosting) PEQs.  It's not as hard as it looks.

 

In the case of the MEH bass bin, there will be one or two SPL peaks above the first notch frequency that have to be attenuated:

 

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

 

Those peaks at 1300 and 2400 Hz need to be separately attenuated (ref. PEQs #4 and #5, below):

 

image.thumb.png.63c1cd0c7709f7103d42ee3221074864.png

 

The bass bin SPL response of the prototype K-402-MEH is actually flatter than that of the Danley SH-50 bass bin SPL response.

 

Chris

Link to comment
Share on other sites

Chris, I kind of understand your technique , but how are you stopping out of band signals going to the drivers?

 

If I use a 24db LR High Pass on a mid driver then I know that the lower bass signal won't be going through the voice coil of the mid driver. Looking at your Xilica screen shot Dev1 out Ch 2 the yellow trace indicates to me that bass signal is not attenuated and will be sent to the mid driver with associated inter-modulation effects.

 

What am I missing ?

Link to comment
Share on other sites

10 hours ago, Wirrunna said:

What am I missing ?

FM and AM distortion depends upon movement of the driver's diaphragm at upper and lower frequencies, simultaneously. 

 

If there is almost no virtual movement of the diaphragm at lower frequencies (due to the bandpass nature of acoustic drivers used for higher frequency reproduction), there is no modulation distortion.  This is more than 50% of the reason why horn-loaded drivers sound the way that they do (the other reason is largely due to their directivity control relative to direct radiators). 

 

PWK did tests and reported on this subject more than 50 years ago to highlight that horn loaded drivers experience over 25 dB lower modulation distortion than the same drivers used in direct radiating mode.  This is because the diaphragms of horn-loaded drivers typically have to move only ~1/5th the distance of the same drivers used in direct radiating mode to produce the same on-axis SPL. 

 

So the issue remains: is there significant AM distortion of midrange/high frequency drivers that are fully horn loaded when they are fed full-range signals? (Note that the drivers themselves are not in any way damaged by this.) 

 

My answer to this question: if you keep it below 110 dB at 1 metre, you never experience audible AM distortion in midrange and high frequency drivers that are fed full-range signals. If however, you find yourself cranking it up to 110 dB and above levels quite often, then you can re-insert your crossover filters using your DSP crossover (perhaps in a readily available preset that can be quickly selected)...and live with the phase shifts...all the time that you use that preset with crossover filters.

 

If you are using FIR filtering extensively, then the above comment about phase shifts can be avoided, but at the cost of significantly more processing power and more setup/dial-in time required.

 

Chris

Link to comment
Share on other sites

Followed advice of not using Xilica supplied crossover settings, using natural roll off of drivers to set XO point, Flattened FQ response.  

Now for the super Phase for Dummies question, What does this phase graph tell me? Shows Phase of 136 degrees, are we theoretically looking for lowest possible (zero degrees)?

What is going on at 4-6khz? I think phase is a combination of speaker and room, but honestly I'm in a fog.  Hopefully I'm not the only one struggling to understand the majority of this thread!

 

REW_spl.phase_f3.jpg.f3d7093438093c726f7fbd92f4183110.jpg

Link to comment
Share on other sites

59 minutes ago, VDS said:

What does this phase graph tell me? Shows Phase of 136 degrees, are we theoretically looking for lowest possible (zero degrees)?

No, the exact vertical position of the phase plot means nothing.  It's the change in the phase with frequency that means everything.  So when reading phase, you can actually add or subtract 360 degrees under the Controls menu while in the "SPL & Phase" plot.  We're looking for a horizontal line with no perturbations in the plot (i.e., changes in the slope of the line) as a perfect phase response.  Next best is a straight line vs. frequency that is tilted from horizontal. 

 

The group delay plot is nothing but a plot of the instantaneous slope of the phase curve, so that 3-6 kHz region will also show a spiking rise in the group delay plot.

 

59 minutes ago, VDS said:

What is going on at 4-6khz?

You've got some early reflections around those frequencies.  I'd check the concentricity of the Axi2050 driver on the throat of the K-402 to make sure that it's mounted centered on the K-402 throat.  Next, I'd make sure that there are no acoustically reflective objects around either the horn mouth or near the microphone.  Move any large objects away from the microphone and horn mouth to make sure there are no early reflections.  The wavelength of sound at 4 kHz is 3.4 inches, and the first quarter wavelength is 0.85 inches, which is about the distance from the Axi2050 diaphragm (nominally the acoustic center of the driver) to the K-402 horn throat, so any disruptions around the 4-6 kHz region says that the driver is likely mounted not quite entered on the K-402 horn throat.

 

Additionally, you can use non-hardening modeling clay to smooth any uneven areas where the driver contacts the horn (inside the horn at the throat). 

 

Chris

 

 

 

Link to comment
Share on other sites

By the way, you need to make sure that your plot is centered on the impulse response peak by selecting the "Estimate IR Delay" button under the Controls Menu under the SPL & Phase plot.  When the dialog box appears, select "Shift IR" (the left button), which will recenter the measurement at time zero to coincide with the peak of the impulse response curve.  When you do this, your phase plots will be correct, otherwise, they won't be correct.  You can also select "Generate Minimum Phase" to see the minimum phase curve.

 

Chris

  • Like 1
Link to comment
Share on other sites

Chris, the Shift IR makes a significant difference.

What does the minimum phase curve tell us?

The first image is K-Horn bass bin 20 to 700Hz at 89db after EQ in my 10.8 x 6.5 x 2.6 m room.

The second  image shows the result of Shift IR.

Jul 14 M051R1 Bass 20-700 90db PEQ TL89db 20-500 1db.jpg

Jul 14 M051R1 Bass 20-700 90db PEQ TL89db 20-500 1db after Shift IR.jpg

 

Edit - The flat black line is Minimum Phase, dotted red line is phase and the black line next to it is Excess phase

Link to comment
Share on other sites

Uhmmm, I meant that you should use "Shift IR" for full-range sweeps, not so much using bass bins only.  If the phase shifts look the same when you use a 20-20000 Hz sweep with all driver channels turned on, then they are believable for the bass bin only. 

 

The reason for this is the ambiguity in assigning the peak of the impulse response using a bass bin only--there aren't enough higher frequencies in the IR calculations to pick a good "peak" position for the zeroing of the time data (i.e., the shift of the data to match the peak of the calculated impulse response).

 

Chris

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...