Really basic REW question

[VDS]

Hi, learning to use REW now that my MEH 402’ are built. Testing my modified Cornwall 1’s and got 35-20,000hz all peaks and valleys within a 10 dB range.  Is this “generally” all you can hope for with passive crossovers?

I don’t really know where to go from here. Any advice on other things to check next? 
waterfall plot to check room reflections? 
I’m just trying to learn the basics so I can start to “see” what I’m hearing.

thanks, Ted

[VDS]

I should say I applied the Var smoothing, as that was a recommendation for novice users.

[pbphoto]

What kind of microphone do you use?  Did you measure close to the speaker or close to your listening position?  Did you apply any smoothing to the data?

 

If you want to measure the speakers, position the mic within 1-2 meters and point it directly at the center of each speaker individually and play around with 1/6 or 1/12th or var and/or psy smoothing.  If you want to measure the speakers in the room, then put it close to your listening position facing upwards at a ~45 degree angle and run some sweeps using var or psychoacoustic smoothing.  If you have a umik-1 or similar calibrated mic, it should come with two calibration files - one for pointing directly at the speaker and one for pointing upwards.

[pbphoto]

My preference is to make as many corrections as you can in the analog domain first via speaker positioning and pre-amp tone controls.  Then look to corrections in the digital domain if needed - less is more IMHO.  Unless of course you decide you want to go with digital crossovers and then there are tons of experts that can help you.

 

Below, the red line is my LS2+sub uncorrected.  Green is corrected.  Measured from my listening position with psy smoothing.

 

[VDS]

I should have said, using Dayton audio Umik 6 mic, I think I applied the correct calibration file (that was confusing, I’ll have to double check). Mic was 1 meter from front of speaker, level, (horizontal) pointing at point between woofer and mid horn. Like I said I was applied Var smoothing which showed a 10 dB range over the entire frequency range, of course fell of cliff at 38 hz.

[Chris A]

Here is an on-axis measurement taken at 1 m that I did on a 1979 Cornwall (w/Crites CT125 tweeter) against the side wall in my listening room, with sufficient absorption on the floor between the microphone and the loudspeaker to absorb floor bounce (1/6th octave smoothing):

 

 

Chris

[VDS]

Here's what I got.  No idea how to read spectogram,  I am researching to understand the data the spectogram gives.  

 

 

[VDS]

the above is 1/6th smoothing

[Chris A]

8 hours ago, VDS said:

I am researching to understand the data the spectogram gives. 

 

Here is the spectrogram view for the SPL response measurement that I posted above, with the display settings menu shown so you can duplicate them:

 

 

Some things to notice:

 

1) Note the bottom axis here is time (the independent variable), in milliseconds, and the vertical axis is frequency (Hz), with the SPL or loudness represented by the scale on the right (dB) in color.  The top of the SPL scale is automatically scaled to the maximum SPL measured during the sweep.

 

2) Note the tweeter's output in the top left of the plot as a vertical bar with a fair amount of decay trail within the first 1.5 ms.  Note that the midrange K-55/K-600 driver/horn lags the tweeter's output by about 0.75 ms.  This is the time misalignment of the tweeter and midrange, and it amounts to about 3.7 full wavelengths of sound at the crossover frequency of ~5000 Hz (i.e., 0.00075 seconds * 5000 Hz).  That's a lot of time misalignment that is audible.  The tweeter needs to either be moved toward the rear of the cabinet by about 10 inches (13584 inches/second speed of sound * 0.00075 seconds), or the tweeter needs to be separately amplified with its own amplifier channel and its output delayed by 0.75 milliseconds.  This will result in a greatly increased sense of soundstage of the higher treble frequencies. 

 

3) Note that the midrange horn/driver SPL output continues after the tweeters output, way above the crossover frequency of about 5000 Hz,  except 0.7 milliseconds later.  This is due to the absence of a low pass electrical filter in the midrange passive crossover to attenuate the K-55/K-600 assembly above 5 kHz.  This separation in time alignment of their two outputs (the tweeter, then the midrange) is not desirable.  The addition of a low pass electrical filter to the passive crossover would significantly reduce this "interference band" between the tweeter and midrange, but that electrical filter would add another phase delay on the midrange driver/horn crossover--at least 45 degrees of phase lag for every "order" of the electrical crossover filter.  This added delay of that filter needs to be added to the tweeter channel to maintain a time alignment with the midrange.

 

3) Note the unevenness of the midrange SPL response, showing that peaking response around 2000 Hz.  This needs to be EQed flatter.

 

4) Note the apparent disruption or null in response at ~1200 Hz and ~480 Hz.  This is due to the combined effects of a time and polarity misalignment of the woofer to the midrange at the crossover point, and the vertical separation of the midrange horn mouth from the center of the woofer.  This needs to be smoothed out by the DSP crossover using a little time delay on the woofer channel (if amplified separately from the midrange horn/driver) and the woofer's polarity rechecked after delay is added.

 

In order to see the time delay and polarity effects of the crossover region, another plot should be used: excess group delay.  More on that to come.

 

Chris

[Langston]

1 hour ago, Chris A said:

..excess group delay.  More on that to come.

 

Also interesting is that the Spectrogram plot is the standard (not excess) group delay plot flipped backward and turned 90˚ with the addition of SPL.

 

Source.

 

God bless you and your precious family - Langston

 

 

The thing we call time in audio measurements and the thing we call frequency are different coordinates for describing precisely the same signal. - Richard C. Heyser

[Chris A]

There are two parts to the so-called "transfer function" response of the loudspeaker: SPL response (that I first showed above) and phase response.  Most people don't talk about loudspeaker phase response because loudspeakers typically have terrible phase responses.  When the subject comes up in audio conversations, the immediate response by the deniers is that "you can't hear phase". 

 

You can hear phase response...especially in transients and in the higher harmonics (timbre) of instruments and voices. It's just that a lot of people don't listen very carefully to their systems, and therefore choose to ignore phase response.  Those that love the sound of well-dialed-in multiple entry horns (MEHs) apparently can hear phase response--because this is a majority of the difference that you hear with MEHs--along with directivity control.

__________________________________________________________

 

So what does the phase response of a Cornwall look like (i.e., phase vs. frequency)?

 

 

Note that I can't back the vertical scale (phase) off far enough to show the entire phase curve.  We're up to 4200 degrees of phase growth, shown above, and we still can't see the ends of the phase curve.  So we do something else--we "differentiate" the phase curve with respect to frequency--and we get group delay, which we can look look at. So group delay (the slope of the phase curve--derived from the mathematical operation known as "differentiation", a.k.a., take the "derivative"--from Newton's calculus) is shown below:

 

 

Here, psychoacoustic smoothing is used due to the inherently noisy nature of the group delay curve when measured in-room with its attendant early reflections.  So this plot represents the slope of the phase curve vs. frequency (the orange curve). 

 

When another mathematical operation is used, taking the so-called "Hilbert Transform" of the group delay curve, we get the "minimum group delay" curve.  If you subtract the minimum group delay curve from the total group delay curve, shown above, we get the "excess group delay" curve (shown in white trace color), which tells us how much "excess" time delay growth there is in the loudspeaker and driving electronics--from high frequency to low frequency.  The goal is to have a flat, smooth excess group delay curve.  It's not smooth here, and it isn't flat.

 

In fact,the excess group delay curve is disrupted in three major areas (about 475, 1200 and 4800 Hz), corresponding to the two crossovers interference bands of the Cornwall.   The crossover band between the woofer and midrange at ~600-800 Hz produces the bottom two disruption areas at 475 and 1200 Hz on either side of the actual crossover frequency, and the crossover from the midrange to the tweeter at ~5 kHz kHz produces the third excess group delay disruption area at 4.8 kHz.

 

Now, notice how the excess group delay curve rises off the horizontal zero group delay axis when the frequency decreases from the tweeter to the midrange.  We can use this information to directly read how much DSP time delay to apply to the higher frequency driver channels to time align the drivers. 

 

In this case, as the excess group delay curve increases above 1200 Hz, there is an average of about 0.6 ms.  That's the amount of time delay to add to the higher frequency channel (in this case) to bring the tweeters into time alignment with the midrange.  This technique seems to work better than using the spectrogram view, in that it removes the "minimum group delay" from the plot so we can see the excess group delay only, and correct for that. 

 

In this case, the excess group delay curve around the woofer--midrange crossover shows that it's actually in time alignment. But something's still not right here--in terms of polarity and/or vertical spacing from each other--which also produces lobing vs. frequency. Vertical spacing of horns/drivers is apparently that which is responsible for the excess group delay disruptions above and below the center crossover frequency of about 600 Hz. If you flop the polarity of the woofer, you can see if it's a polarity issue, by taking another measurement and looking at the excess group delay and SPL response curves, it can be verified whether or not a polarity reversal is needed.  It may be that a slight amount of woofer channel time delay is needed to settle the excess group delay spikes at 475 and 1200 Hz, because those spikes are audible...

 

(Taken from here)

 

Chris

[VDS]

On 7/2/2021 at 5:50 AM, Chris A said:

 

Here is the spectrogram view for the SPL response measurement that I posted above, with the display settings menu shown so you can duplicate them:

 

 

Some things to notice:

 

1) Note the bottom axis here is time (the independent variable), in milliseconds, and the vertical axis is frequency (Hz), with the SPL or loudness represented by the scale on the right (dB) in color.  The top of the SPL scale is automatically scaled to the maximum SPL measured during the sweep.

 

2) Note the tweeter's output in the top left of the plot as a vertical bar with a fair amount of decay trail within the first 1.5 ms.  Note that the midrange K-55/K-600 driver/horn lags the tweeter's output by about 0.75 ms.  This is the time misalignment of the tweeter and midrange, and it amounts to about 3.7 full wavelengths of sound at the crossover frequency of ~5000 Hz (i.e., 0.00075 seconds * 5000 Hz).  That's a lot of time misalignment that is audible.  The tweeter needs to either be moved toward the rear of the cabinet by about 10 inches (13584 inches/second speed of sound * 0.00075 seconds), or the tweeter needs to be separately amplified with its own amplifier channel and its output delayed by 0.75 milliseconds.  This will result in a greatly increased sense of soundstage of the higher treble frequencies. 

Basic question.  In spectagram is the blue/green “fluff or clouds” on the right side reflected sound that the room is adding? Do we have want to see the colors taper quickly as they move to the right?   

[Chris A]

7 hours ago, VDS said:

 In spectagram is the blue/green “fluff or clouds” on the right side reflected sound that the room is adding?

Yes.  If we increase the time horizontal limit of the plot to 100 ms (from 10 ms shown above), you can see more room reflections:

 

 

That grouping of low level reflections around 20-30 ms are actually the axial room mode reflections from the opposite side wall. My listening room is 15.5 feet wide, and the microphone is about 3 feet from the front of the Cornwall, so the round-trip distance is about 27 feet, and the round trip time is about 23 ms (given that the speed of sound is about 1132 feet/second).  [The metric equivalences to these linear measurements are: room is 4.7 metres wide, microphone is at 1 m from the loudspeaker, and round-trip distance is 8.2 metres, and speed of sound is 345 m/s.]

 

You can also see long-period reflections at 240 Hz and at 80 Hz and below.  The energy below 80 Hz is due to standing room resonances (i.e., half and full wavelength resonances at the room width, height, and depth: 125, 73, and 28 Hz full-wavelength resonances, and 62, 36, and 14 Hz half-wavelength resonances).  Since the Cornwall doesn't put out much energy below 40 Hz, the lower room modes are not seen in the above spectrogram plot.

 

7 hours ago, VDS said:

Do we have want to see the colors taper quickly as they move to the right?   

Yes, but not too quickly.  We want those longer wavelength and lower frequency reflections at longer time delays to give a feeling of space in the room.  It would sound like an anechoic chamber if all reflections were absorbed--which is a terrible place to listen, in my experience. 

 

The reflections at less than 4-6 ms are the early reflections that strongly affect soundstage imaging, so we want to channel the loudspeaker's output (via directivity) and absorb much of the remaining in-room reflections within 4-6 feet of the loudspeakers (horizontally and vertically) to preserve that soundstage imaging, emphasizing the direct arrivals from the loudspeakers.  Jubilees and K-402-MEHs have directivity control that makes that job much easier to accomplish, without having to pull the loudspeakers away from the walls or use dipole loudspeakers (like Magnepans or electrostatic panel loudspeakers that present "head-in-a-vise" listening position constraints). 

 

Chris

[VDS]

So I’m starting to get a feel for using Xilica to flatten frequency response.  I’m trying to listen and see if I hear any unpleasant frequencies and trying to adjust. Early days though!

im trying to read spectagrams and see how room treatments affect the 4-6ms reverberation. 
wonder what else I can measure with REW and possible correct with Xilica? Using 402 MEH, so time alignment seems good.

thanks, Ted

[Chris A]

On 7/8/2021 at 7:41 PM, VDS said:

wonder what else I can measure with REW and possible correct with Xilica?

The Xilica does EQ, delays, and channel gains, as well as fast and slow (integrated) limiters for those applications where extreme loudspeaker SPL is regularly encountered in operation. 

 

REW can tell you a lot more than what the DSP crossover can correct in terms of the performance drivers/horns themselves, such as harmonic distortion, as well as crossover filter performance and room acoustic performance.  REW can also give you many insights as to where to place your loudspeakers in-room, and how that will affect their low frequency performance (i.e., via the "Room Sim" facility).

 

Chris

[VDS]

I thought might measure a lot more than DSP can effect.Im working on seeing the effects of PEQ by taking a lot of sweeps to start to get a feel for what the changes in settings actually do. Lots of “unintended consequences” that surprise me when I take measurements!

Now I’m trying to see how the room is effecting the sound.  I’m starting to look at Spectagram (kind of starting to understand).  Wondering how to interpret Impulse and Decay, RTA(?).  Thinking a fast drop off in first 10ms is desirable for sound clarity.

[Chris A]

15 minutes ago, VDS said:

Thinking a fast drop off in first 10ms is desirable for sound clarity.

Actually, I would put that need to damp acoustic output to be within the first 4-6 ms.  Ten milliseconds is a bit too long.  There's surprisingly little in the psychocacoustic literature about this (but a lot more information on the precedence or Haas effect, which extends out to 20--30 milliseconds from the direct arrivals).  The type of effect here is not precedence effect but rather the effect of hearing apparent soundstage width, depth, and clarity.

 

This is one of two basic effects of using fully horn-loaded loudspeakers.  The other effect is the dramatic increase in efficiency of the drivers due to the acoustic transformer action of the horn--the much better impedance matching with the air, which in turn translates to a virtual elimination of modulation distortion relative to using the drivers in direct radiating mode.  "Anything that moves--distorts".  What is typically forgotten here is that the type of distortion is modulation distortion--and this is a major reason why horn loudspeakers sound different--the absence of modulation distortion.

 

Chris

[Langston]

39 minutes ago, VDS said:

Lots of “unintended consequences” that surprise me when I take measurements!

 

The most exciting phrase to hear in science, the one that heralds new discoveries, is not 'Eureka!' but 'That's funny...' - Isaac Asimov

[VDS]

Yes! Now I get to ask, why? And explore some more

[Edgar]

2 hours ago, Chris A said:

There's surprisingly little in the psychocacoustic literature about this (but a lot more information on the precedence or Haas effect, which extends out to 20--30 milliseconds from the direct arrivals).  The type of effect here is not precedence effect but rather the effect of hearing apparent soundstage width, depth, and clarity.

 

This is where I learned about psychoacoustics. Doesn't directly address the situation that you describe above, but the info is probably there if you dig deeply enough. Chapter 4, especially.