• 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). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Understanding speaker measurements

mk05

Member
Forum Donor
Joined
Apr 28, 2020
Messages
72
Likes
16
#1
After reading a lot, I understand some parts of speaker measurements. Is there a guide somewhere on ASR that explains how to understand each of Amir's measurements, as in what makes a measurement good, what is bad, what does what - that may be readily understood by mere mortals? Perhaps a layman-friendly guide (or guides in general) would allow more people to step-up from fleeting curiosity to involvement, instead of denying the possibility of objective conclusions because the material is too difficult to grasp (without asking for help).

For example, now I understand how to achieve tone, or a speaker's coloration. However, how do I tell the "speed" of a speaker - how tight the speaker is, ie ported/sealed sub? The accuracy of reproduction? The "holographic" soundstage? Approximation of sweet spot (is this the vertical/horizontal graph)? I would love to understand the impedance/phase graph, spectral decay, waterfall, and a graph that looks like a heartbeat.

Thank you again for entertaining my questions.
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
2,428
Likes
5,191
Location
Land O’ Lakes, Florida
#2
“Speed“ usually isn’t really a big deal, but this would show up in the waterfall plot (faster it goes away the better). However, these are not super accurate, the system can show accurate frequency response as it takes samples from all over and can figure out what influence the room has, but a waterfall plot is basically recording the speaker for a short time, and that is too difficult to do the same process on, so the room is influencing the results. These graphs can also be misleading as the usually are not normalized (initial sound made flat), as a speaker with a dip in the response will show a faster decay, but its not actually decaying faster as the amount of decibels reduced is similar, but the initial decibels level is lower than other frequencies, and vice-versa, a rise in response can make it seem like it takes a while to decay.

Not sure on what makes a holographic soundstage, but the width of the soundstage can be determined by those polar plots you mentioned. There is also image width as well as imaging accuracy, the width can be deduced similarly to soundstage width, but looking at a narrower arc, the accuracy of the imaging is how closely the horizontal off-axis looks in reference to the on-axis.

However, a limitation is that this system is showing the on-axis as how this speaker will be used, whereas some speakers are designed to be listened off-axis (this can be seen by looking at if some degree off axis (10/20/30) is noticeably more neutral than the on-axis.
 

andreasmaaan

Major Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
4,845
Likes
4,947
#3
However, how do I tell the "speed" of a speaker - how tight the speaker is, ie ported/sealed sub?
The simplest way would be by measuring the step response, which essentially shows you the speaker's output over time when fed a step signal (an input that instantaneously goes from 0 to 1).

A theoretical ideal step response (unattainable in real life) would look like this (the x-axis is time, and the y-axis is level):

1590589273875.png


Instantly, we go from zero output to maximum output.

Speakers cannot produce DC (zero Hz), however, so the output must rises and then settle. An ideal speaker with a high-pass filter at some point (even as low as 1Hz, theoretically, but of course always higher in practice) will produce a step response something like this:

1590589492266.png


The output still rises instantaneously, but then settles (with some "overshoot" as a result of the high-pass filter).

Really in real life though, no speaker is this perfect. Loudspeaker enclosures and crossovers significantly affect a loudspeaker's step response, and to a lesser extent the nonlinear behaviour of the drivers themselves.

By way of example, here's the step response of a Revel Ultima Studio:



As you can see, in response to an impulse input, the speaker produces three distinct outputs, separated from each other in time. The reason for this is that each driver is high-pass and/or low-pass filtered by the crossover and/or the enclosure, which causes the output at lower frequencies to be delayed relative to the output at higher frequencies.

The first peak is in fact the tweeter, followed by the midrange driver, followed by the woofers. As you can see, the woofers' peak lags that of the tweeter by nearly 3ms.

3ms might sound like quite a lot, indeed it is about as long as it takes sound to travel around 1 metre. However, perhaps surprisingly, despite many attempts by researchers to demonstrate that such a delay is audible, it has yet to be demonstrated to be so in a loudspeaker with music as the signal.

This is probably why step response is not a measurement that Amir generally takes, i.e. there is no evidence that it is subjectively important (except in extreme cases of course).
 
Last edited:

andreasmaaan

Major Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
4,845
Likes
4,947
#4
Approximation of sweet spot (is this the vertical/horizontal graph)?
The polar response graphs will show you (1) how far off the central axis you can move vertically before the direct sound's tonal balance begins to audibly change and (2) what the tonal balance will be of the sounds reflected off floor, walls, ceilings, etc.

The horizontal graphs of course give you this info for side-to-side movements and wall reflections, while the vertical graphs give you it for up/down movements and floor/ceiling reflections.

As to how to piece this together to determine how large a speaker's sweet spot is, essentially all you need to look at is how wide (or tall) a range of angles the speaker covers before the frequency response significantly changes. If it's wide/tall, the sweet spot will be large.

One possible exception to the above comes in the case of controlled-directivity speakers, that may be set up in such a way that they trade time for intensity. This is a long topic, but is explained in some depth here. There's also a great practical interactive demonstration of how this works here.
 

RayDunzl

Major Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
10,148
Likes
8,727
Location
Riverview FL
#5
The simplest way would be by measuring the step response, which essentially shows you the speaker's output over time when fed an impulse-like signal (a very short click containing all frequencies) as the input.
If viewing output waveforms, as via a recording:

A short click input will yield the impulse response.

A step signal can be created with a very low frequency square wave.

Impulse: https://audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44352

Step: https://audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44440

---

However, a swept sine measurement signal will reveal both after some fancy math is applied, as noted in the links above.
 
Last edited:

andreasmaaan

Major Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
4,845
Likes
4,947
#6
If viewing output waveforms, as via a recording:

A short click input will yield the impulse response.

A step signal can be created with a very low frequency square wave.

Impulse: https://audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44352

Step: https://audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44440

---

However, a swept sine measurement signal will reveal both after some fancy math is applied, as noted in the links above.
Yes thanks, brain fart. Post corrected.
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
2,428
Likes
5,191
Location
Land O’ Lakes, Florida
#7
That is a quasi-anechoic measurement. If you measured truly dead flat speakers in an anechoic chamber they'd be flat. Using a quasi-anechoic method, the result is a downward slope in response being 10 db lower at 20 khz vs 20 hz or so.
You would get that for an in-room measurement at the listening position, not quasi-anechoic. Quasi means they did a measurement and windowed it before the first reflection point, and maybe spliced it with a near-field measurement of the bass. This is what John Atkinson does.
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
9,875
Likes
13,365
#8
You would get that for an in-room measurement at the listening position, not quasi-anechoic. Quasi means they did a measurement and windowed it before the first reflection point, and maybe spliced it with a near-field measurement of the bass. This is what John Atkinson does.
Okay, you are correct. I'll delete the mis-leading post.
 
OP
M

mk05

Member
Forum Donor
Joined
Apr 28, 2020
Messages
72
Likes
16
Thread Starter #9
“Speed“ usually isn’t really a big deal, but this would show up in the waterfall plot (faster it goes away the better)
Thank you! And trying to understand your following notes on decay. Could you possibly show two waterfall plots, what is ideal and what is not? What do you mean by faster it goes away the better? In which way? Is it best to be a smooth decline, or rather just like a wall? Sometimes, I hear people say that some speakers have a nice "decay," like hearing a bell and the note resonates or sings (whatever correct term may be) shorter or longer vs other speakers. I'd like to know what that rate on the measurement is, maybe that will allow me to understand what I like, and look for!

Not sure on what makes a holographic soundstage, but the width of the soundstage can be determined by those polar plots you mentioned. There is also image width as well as imaging accuracy, the width can be deduced similarly to soundstage width, but looking at a narrower arc, the accuracy of the imaging is how closely the horizontal off-axis looks in reference to the on-axis.
Would I be correct that accuracy of imaging is portrayed by the spinorama (0-60deg chart)?
Holographic nature, width of soundstage (vs image width? huh), clarity of imaging (accuracy of reproduction?) - these are the factors that I would like to understand by the measurements tested. Is it possible to point me towards the respective graphs/charts that would imply or demonstrate those traits of a loudspeaker?

Right now, all I know is the frequency response. However, I'd like to know the "how well" part. While measurements are nice, are some of those characteristics above impossible to tell from measurements?

However, a limitation is that this system is showing the on-axis as how this speaker will be used, whereas some speakers are designed to be listened off-axis (this can be seen by looking at if some degree off axis (10/20/30) is noticeably more neutral than the on-axis.
I think I understand this part as the toe-in/out. So this relates to the 0-90deg by 5deg chart? Many times, I see that speakers lose freq response in areas you may not want (like a dip in voice freq), once you start pointing the tweeter away from the ear, in order to calm the treble end. To me that seems like poor speaker design? Does this mean that you guys actively analyze freq response characteristics of 5deg or 10deg to make sure there isn't loss of fidelity (or freq response turns into preferred response) in areas with respective toeing, in order to make the purchase, with said toeing considered? Because that is fascinating, if so.
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
2,428
Likes
5,191
Location
Land O’ Lakes, Florida
#10
Could you possibly show two waterfall plots, what is ideal and what is not?
I don't think Amir's setup can do this accurately, so let's looks at Stereophile's where it's pretty much anechoic down to a certain frequency.

You want a fast decay and you want a clean decay (looks similar as it's decaying).


Very good; if I had to nitpick, there are two small dips in the upper treble that get increasingly more prominent.

Also good. The bass decays a bit longer, and the decay isn't clean between 1kHz-2kHz, the response is dropping but then picks back up.

Now, how audible is all this? I have no clue.

Sometimes, I hear people say that some speakers have a nice "decay," like hearing a bell and the note resonates or sings (whatever correct term may be) shorter or longer
That is silly, all the sonic information is included in the recording, so it already has the ideal decay built-in. Now, sometimes a song is poorly recorded and vocals/instruments sound like they were recorded in such a way that there is almost no reverb, but this is a poor recording, and shouldn't be an excuse to promote speakers with long decay.


Would I be correct that accuracy of imaging is portrayed by the spinorama
It can be deduced, if the listening window and to a degree the early reflections graphs are close in level and shape to the on-axis (for a speaker designed to be listened off-axis), it should have a wide image.

Many times, I see that speakers lose freq response in areas you may not want (like a dip in voice freq), once you start pointing the tweeter away from the ear, in order to calm the treble end. To me that seems like poor speaker design?
Designers need to know how most of their customers will use their speakers, if they know that most will not use toe-in, then they should focus more on making the off-axis perform well in terms of evenness of the frequency response.
 

Similar threads

Top Bottom