• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Comparing speakers based on measurements like step response

sweetmusic

Member
Joined
Apr 23, 2022
Messages
54
Likes
30
(Apologies for newbie question!)

I'm researching speakers for a possible purchase, and I trust science more than my ears during a 10 min demo at an audio shop. That's led to me spending a lot of time looking at the measurements on sites like Stereophile and SoundStage Network. I feel like I understand the frequency response and impedance plots, and cabinet resonance, and dispersion. But I'm wondering if it's possible to tell if one speaker ought to actually sound better based on the measurements.

Here's an example:
A. Stereophile measurements of the Dynaudio Focus 200 XD
B. Stereophile measurements of the Dynaudio Special 40

These seem to be very similar speakers, from the same manufacturer, with very similar tweeter and woofer, similar sized cabinets. The Special 40 has 1st order crossovers, and the Focus XD has digital crossovers at 4.2 kHz instead of 2 kHz.

When I look at the step response of the Focus 200, it has a sharper point and smoother decay than the step response of the Special 40. IIUC, the Focus 200 has a textbook shape to its step response and is measurably better.

In the cabinet resonances, the Special 40 has a retro cabinet that's maybe too square, and has cabinet resonances. The Focus, despite jamming in the amplifier and DSP, manages to mute pretty much all of the cabinet resonances. The plot looks so clean that it looks like a mistake, TBH. But based on the plot, the Focus just ought to sound better.

Can someone who understands the science of things explain to me if there's any reason why the Special 40 might sound better than the Focus XD? From the data, it seems like no contest.

This is relevant to me because the Focus costs more at MSRP ($6000), but it's available online for 20% off. The Special 40 is $3600 and requires a separate amp, which brings it to the same price or higher. I get that audio enthusiasts like to buy gear and mix and match. But if I want to decide based on audio science, is the Focus the clear winner in this particular matchup?

I'm not stuck on Dynaudio necessarily. It's just that these are speakers that have gotten good reviews, , are about the size of what I'm looking for, and are an indulgence but not completely impossible to afford if I splurge. I mean, getting the next higher trim on a car costs that much, right, so it's not that crazy ;-?
 
Last edited:

Head_Unit

Addicted to Fun and Learning
Forum Donor
Joined
Aug 27, 2018
Messages
813
Likes
403
I got hooked on speakers as a teenager by reading David Weem's book. His references to Audio Engineering Society Journal articles led me to get an older friend to drive me an hour to a university library and photocopy a literally 4" stack of stuff. Reading that stuff decided me to become a loudspeaker engineer. I think every loudspeaker engineer would agree it is like astronomy-you examine stuff and solve some things and find there are other deeper layers.

We can measure some basic things, of which the easiest is frequency response: on-axis, and then the off-axis responses, totaling the sound power radiated into a room and how evenly. Some folks measure distortion...but we have not much idea how important that is and some like Dr. Earl Geddes content our sensitivity varies with sound pressure and many agree it varies with frequency (and that's not even getting into the various types of distortion). Step responses...those are almost always calculated from the frequency response in some way.

Or to rant a different way, pretty much all measurements are actually made in the steady state due to practicality. But our ears it seems don't really hear like that, and I suspect there are various phenomena that aren't revealed in speakers or amps or whatever. Probably not as important as the "basics" but maybe showing differences.

I haven't heard the Special 40. My friend went down to San Diego to check them out and wasn't blown away; some people love them. If you're running full range, the XD is tuned a chunk lower which is a big advantage in my book. If I couldn't listen myself I'd absolutely spend the money for the XD. Either way I'd want a subwoofer and room correction software to get the best bass.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
11,902
Likes
13,392
Location
Riverview FL
Step responses...those are almost always calculated from the frequency response in some way.

It bugged me that Impulse and Step could be calculated from a smooth sine sweep.

So I made an experiment, and am not bothered any more, though maybe a little astonished.


Impulse:

Step:
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
11,902
Likes
13,392
Location
Riverview FL
When I look at the step response of the Focus 200, it has a sharper point and smoother decay than the step response of the Special 40. IIUC, the Focus 200 has a textbook shape to its step response and is measurably better.

One speaker is passive, the other active, with (presumably) some "correction" applied in its software.

Special 40 Passive Step

1650694576827.png


XD Active Step

1650694711069.png




You can apply "room correction" to any speaker to modify its step response. Red without, Black with.


JBL LSR 308, measured at 10 feet (the listening position here)

1650694812905.png



Martin Logan reQuest, at 10 feet

1650694905117.png



Correction via miniDSP Open DRC-DI and Acourate DRC software to generate the filters.

The correction is accomplished by modifying the frequency response and phase of the signal sent to the DAC/Amplifier

Frequency and Phase of the uncrrected signal is a pair of flat lines... as measured at the preamp output:

1650695441076.png


The "correction" changes that. Red is frequency response, Green is Phase

For the left channel of the JBL (green FR, black phase)

1650695609126.png



For the left channel of the Martin Logan - red frequency response, green phase

1650695661688.png
 
Last edited:

ppataki

Senior Member
Joined
Aug 7, 2019
Messages
419
Likes
430
Location
Budapest
Hi @RayDunzl
Is the pre-ringing audible?
Especially for the Martin Logan I mean
(On the Step Response curve it stands out pretty much, hence I am asking if it is audible or not)
Thank you
 

alex-z

Senior Member
Joined
Feb 19, 2021
Messages
479
Likes
815
Location
Canada
Comparing step response of speakers is a fruitless task. Even if one speaker was slightly better than another, by the time you introduce room acoustics and psycho-acoustic masking the difference won't be audible.

The speaker preference rating setup developed by Sean Olive and Floyd Toole isn't perfect but they nailed one thing, frequency response is the most important factor in sound quality. If you can eliminate resonances, provide bass extension, and get a neutral listening window, you are on your way to success. The Dynaudio Focus 200 XD should almost certainly sound better owing to its smoother mid-range and slightly better bass extension.

They are still terrible value though. That kind of money gets you pair of KEF R3, pair of good quality subwoofers, external DSP, high-end amplifier, and enough money leftover to put acoustic treatment in a mid-sized room.
 

Eetu

Addicted to Fun and Learning
Forum Donor
Joined
Mar 11, 2020
Messages
602
Likes
815
Location
Helsinki
the Focus XD has digital crossovers at 4.2 kHz instead of 2 kHz.
I'm pretty sure the xo is at 2.5k.

I agree re: measurements, without hearing them the Focus XD should be the better speaker. I think $6k for a 2-way speaker is terrible value though. I would go for something like the Buchardt A500 instead. And isn't the 200 XD discontinued 20 XD being the current model..?
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
11,902
Likes
13,392
Location
Riverview FL
Hi @RayDunzl
Is the pre-ringing audible?
Especially for the Martin Logan I mean
(On the Step Response curve it stands out pretty much, hence I am asking if it is audible or not)
Thank you

Looking at the output Impulse Response of the DAC...

(without correction)

1650713542287.png


If it is audible, I can't get rid of it, so...

The ringing occurs over a small part of a millisecond....

That makes the frequency very high, at 1/2 the sample frequency (I think)...

Which I can't hear...

So, I'll guess "not audible" for my case.

---

The Impulse Response at the preamp with correction applied gets changed:

Red - ML, Blue JBL

1650713758948.png



---

If watching music play, you never (?) see such squiggles, but then there may not be any Impulses (isolated full scale sample) in music to excite them,
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
2,085
Likes
1,479
Location
Canada
@RayDunzl

If you don’t mind, can you post the before and after mdat for just the ML? Either left or right would be fine. I’m also curious to see what other differences there else may be in the full-range response.
 
OP
S

sweetmusic

Member
Joined
Apr 23, 2022
Messages
54
Likes
30
Thanks, @Head_Unit and @RayDunzl .

Fascinating that the impulse response can be changed like that with DSP. I had no idea! I wasn't totally clear from the thread above: can your ears hear any difference between the original and corrected impulse response, even if not the pre-ringing specifically? And if you can hear a difference, is it clear which sounds better?

Makes sense that frequency range and response matter more. Also, I agree the Dynaudios are kind of crazy expensive.

@Eetu yes, the Buchardt A500 looks like a better value, as well as the simplest package to get it all, including room correction.

KEF LS50 Meta with the KEF sub is also a much better value than the Dynaudio.

Let me do some research on how to add room correction with digital output to active speakers, with HDMI eARC and digital streaming. I've seen other threads here on that topic :)
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
11,902
Likes
13,392
Location
Riverview FL
If you don’t mind, can you post the before and after mdat

Here's L/R/Both before and after correction for JBL (1-6) then ML, (7-12) link good for 7 days

 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
2,085
Likes
1,479
Location
Canada
Here's L/R/Both before and after correction for JBL (1-6) then ML, (7-12) link good for 7 days


Thanks!

The pre-ringing as a consequence of linearization is only barely visible in the step response.

*edit: one of the measurements posted was L+R and the other was actually Left only (sorry, my mistake)

STEP RESPONSE & MAGNITUDE (left ML+subs)
1650755968146.png 1650755975331.png

However, we can see some rippling in the wavelet which is the result of the rather "forced" phase linearization.


WAVELET SPECTROGRAM 1/2 resolution, 40 dB scale, normalized
1650756080700.png 1650760835933.png

I've only learned how to magnify the view of the ripple-effect via this wavelet method recently.

Now, is this audible? I dunno. There may be specific sections of certain tracks where if you listen and A/B test very closely, you might be able to hear it. However, in many cases, it may well be that the global EQ benefit swamps any negative pre-ringing consequence from the phase linearization.
 
Last edited:

Head_Unit

Addicted to Fun and Learning
Forum Donor
Joined
Aug 27, 2018
Messages
813
Likes
403
can your ears hear any difference between the original and corrected impulse response, even if not the pre-ringing specifically?…@Eetu yes, the Buchardt A500 looks like a better value, as well as the simplest package to get it all, including room correction.
The first question I think is not w’ll researched…plus everyone’s hearing differs so YMMV. My friend bought A500, great tech package, room correction really worked. I’ve noted in another thread somewhere he had various teething problems and does not like the external 3rd party gateway thing which Buchardt doesn’t control. Now things seem more under control? I’d be getting the towers though cuz YOLO
 

ppataki

Senior Member
Joined
Aug 7, 2019
Messages
419
Likes
430
Location
Budapest
Thanks!

The pre-ringing as a consequence of linearization is only barely visible in the step response.

*edit: one of the measurements posted was L+R and the other was actually Left only (sorry, my mistake)

STEP RESPONSE & MAGNITUDE (left ML+subs)
View attachment 201866 View attachment 201867

However, we can see some rippling in the wavelet which is the result of the rather "forced" phase linearization.


WAVELET SPECTROGRAM 1/2 resolution, 40 dB scale, normalized
View attachment 201868 View attachment 201879

I've only learned how to magnify the view of the ripple-effect via this wavelet method recently.

Now, is this audible? I dunno. There may be specific sections of certain tracks where if you listen and A/B test very closely, you might be able to hear it. However, in many cases, it may well be that the global EQ benefit swamps any negative pre-ringing consequence from the phase linearization.

@ernestcarl @RayDunzl

Just a dummy question after checking the measurements:
If we talk about phase linearization, the result of that shouldn't be that the below highlighted part becomes fully linear?

1650780055711.png


To become much like this:

1650780098146.png



Can you please comment on that?

I tried many different DRC solutions but none of them could actually 'clean up' that part (if it needs clean-up at all.....that is rather the question)

Thank you
 

sigbergaudio

Major Contributor
Manufacturer
Forum Donor
Joined
Aug 21, 2020
Messages
1,136
Likes
1,981
Location
Norway
@ppataki The phase will look like that in in-room measurements due to reflections whatever you do, unless you do extreme nearfield measurements of a single speaker. So it's a consequence of the interaction of speakers + room, not the speakers themselves.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
2,085
Likes
1,479
Location
Canada
@ernestcarl @RayDunzl

Just a dummy question after checking the measurements:
If we talk about phase linearization, the result of that shouldn't be that the below highlighted part becomes fully linear?

View attachment 201920

To become much like this:

View attachment 201921


Can you please comment on that?

I tried many different DRC solutions but none of them could actually 'clean up' that part (if it needs clean-up at all.....that is rather the question)

Thank you

No…

Frequency dependent windowing algorithms will always be applied prior phase linearization.

I believe some “room correction” software will allow the user to try to reduce reflection effects seen in the room even above the low bass range by correcting or maybe even cancelling it out to various degrees, but this is very, very problematic since relative source, path and direction of these innumerable reflections change significantly with movement.

However, reflection cancellation through specialized FIR DSP can work well in much more fixed, consistent systems like what we see of the inside of certain horn designs e.g. Fulcrum Acoustics coaxial horn-loaded (much smaller waveguides) speakers. Depending on the speaker design philosophy this kind of DSP correction may be acceptable or not…
 
Last edited:

ppataki

Senior Member
Joined
Aug 7, 2019
Messages
419
Likes
430
Location
Budapest
Thank you @ernestcarl

One more question if I may:

I just noticed that if I apply a frequency dependent window in REW, the frequency response will have dips that are normally not there....

Without FDW:

1650897052798.png


With FDW of 9 cycles:

1650897076342.png


The FR response looks worse vs without FDW....

Do you know why that is and what should cause this?
And I am wondering which one is the 'real' response then? ....

Thank you
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
2,085
Likes
1,479
Location
Canada
Thank you @ernestcarl

One more question if I may:

I just noticed that if I apply a frequency dependent window in REW, the frequency response will have dips that are normally not there....

Without FDW:

View attachment 202260

With FDW of 9 cycles:

View attachment 202261

The FR response looks worse vs without FDW....

Do you know why that is and what should cause this?
And I am wondering which one is the 'real' response then? ....

Thank you

FDW will window out, cut out or "throw away" energy that arrives late from the magnitude graph. So the energy is still arriving at the microphone/ears, but only late in time.

Both windowed and the non-windowed responses are 'real'. More advanced room correction algorithms have the ability to psychoacoustically smooth and window out the measured response -- adjustable in varying degrees depending on the software -- that may better represent what we "hear" for the purpose of equalization. REW is limited compared to something like Accourate or Audiolense.

*BTW, I would refer or look at both windowed and unwindowed responses when using REW.
 
Last edited:

ppataki

Senior Member
Joined
Aug 7, 2019
Messages
419
Likes
430
Location
Budapest
Thank you @ernestcarl
Are there any recommendations whether to use windowed or unwindowed for EQ-ing frequency response?
So far I have always used the unwindowed measurement

I guess for correcting phase with RePhase for example one shall use the windowed measurement, right?
Thank you
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
2,085
Likes
1,479
Location
Canada
Are there any recommendations whether to use windowed or unwindowed for EQ-ing frequency response?
So far I have always used the unwindowed measurement

I sort of use both as reference by overlaying the same measurements with different settings and or techniques. e.g. 1) MMM, 2) swept sine smoothed but non-windowed, 3) swept sine smoothed & windowed. In the bass, sometimes I do not bother with smoothing at all.

Now, I can't really be 100% certain whether the filtered theoretical "direct-sound" trace is really more representative to what I truly hear vs the MMM, for example. Rather than pick one and simply be done with it, I obsessively compare what EQ effects have on either filtering methods.

I guess for correcting phase with RePhase for example one shall use the windowed measurement, right?
Thank you

Yep, though, sometimes in rare cases it may almost seem unnecessary for adjustments just of the phase:

1650902875822.png

Directivity of the ML panel is so narrow that the phase response is easily readable even with largest window size setting.
 
Top Bottom