• 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.

Passive Loudspeaker Speaker Compression/Dynamic Compression

ROOSKIE

Addicted to Fun and Learning
Joined
Feb 27, 2020
Messages
893
Likes
1,580
Location
Minneapolis
Howdy,
this topic got hot in the JBL 4309 thread.
A lot of very interesting points have been mentioned in that thread.
Let's make the jump to a dedicated thread.
Maybe the mods can move some relevant posts here so there is some context and fuel to work with in this thread.

Things discussed included-
How can we measure and test for loudspeaker compression/dynamic compression?
What are the variables that make it hard to isolate true compression/dynamic compression from other loudspeaker qualities?
How much of a factor is it in practice, how much compression must be present before the sound suffers audibly?
How is it that sometimes compression increases SPL at certain frequencies and other times there is the more likely anticipated reduction in relative SPL?
How does compression/dynamic compression relate to other various distortions (especially HD and IMD.)
& more of course
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
35,320
Likes
134,807
Location
Seattle Area
I think we need to start at high level and define what we want. I think determining some random variable is not going to help us. I push speakers hard in every review and many different manifestations occur. In some cases, no matter how much I crank it up, I don't detect any impairment. In this regard, we want a parameter that would show this. If it shows something else, then it is creating concern where none exists.

In other cases, like amplifier compression, I first hear lack of increase in loudness. This then goes into various types of artifacts such as clicks, pops, static, etc.

In yet other cases, speaker doesn't even play the low frequencies. As such, it gets a free pass even though that may not be correct. Or maybe correct if you intend to use it with a sub or something.

Right now, I find that my ears are super reliable indicators of this. I have not seen any measurements or research that predict what I am hearing. My early tests of a number of speakers also failed in determining such objective limits.
 
OP
R

ROOSKIE

Addicted to Fun and Learning
Joined
Feb 27, 2020
Messages
893
Likes
1,580
Location
Minneapolis
Soundstage!, does their "deviation from linearity" test. It often does not corelate with elevated levels in their HD test.
Here is the KEF LS50 meta & the Revel M106 & Triangle BR03.
These are at 90db SPL @ 2m

LS50meta HD
1634259251741.png

LS50meta deviation from linearity
1634259296358.png


Revel M106 HD
1634259364225.png


Revel M106 deviation from linearity
1634259419218.png


Triangle BR03 HD
1634259577854.png


Triangle BR03 deviation from linearity
1634259617621.png
 

Kvalsvoll

Addicted to Fun and Learning
Joined
Apr 25, 2019
Messages
505
Likes
953
Location
Norway
If it shows something else, then it is creating concern where none exists.
Ideally, there should be a simple metric, like a spl-in-dB number, that accurately describes the usable spl output capacity of a speaker.

Just measure distortion at increasing levels until a defined limit is reached, may seem like the obvious solution. But this gets complicated, when realizing that the nature of this distortion can be very complex, and some distortion types is less harmful than others.

A compression test will show when compression begins. As for audibility - compression of peak transients at 3dB is hard to hear, 6dB is easy to hear when compared in ABX. But 6dB compression in a speaker is a lot, and very audible distortion artifacts can be present well below that.

The measurement must also be easy to understand and replicate, and widely adapted, to be usable for comparing speakers.

What at first glance seems like a simple thing to measure, turns out to be a lot more complicated.
 

rebbiputzmaker

Addicted to Fun and Learning
Joined
Jan 28, 2018
Messages
974
Likes
317
I think we need to start at high level and define what we want. I think determining some random variable is not going to help us. I push speakers hard in every review and many different manifestations occur. In some cases, no matter how much I crank it up, I don't detect any impairment. In this regard, we want a parameter that would show this. If it shows something else, then it is creating concern where none exists.

In other cases, like amplifier compression, I first hear lack of increase in loudness. This then goes into various types of artifacts such as clicks, pops, static, etc.

In yet other cases, speaker doesn't even play the low frequencies. As such, it gets a free pass even though that may not be correct. Or maybe correct if you intend to use it with a sub or something.

Right now, I find that my ears are super reliable indicators of this. I have not seen any measurements or research that predict what I am hearing. My early tests of a number of speakers also failed in determining such objective limits.
If you want to be fair about this there really is going to be no fully reliable measurements at all because what are you really measuring? Maybe even the term speaker is possibly too generic. Each speaker design will have different properties due to the mechanical/physical interactions with the driver. Ported would have different properties versus sealed for cabinets. Too many factors with no control for valid study. Then you have different types of horns and how they are mechanical amplifiers and also affect the driver due other possible interactions. What you’re doing does make sense, although not totally scientific it is reasonable.
 

ctrl

Major Contributor
Joined
Jan 24, 2020
Messages
1,069
Likes
3,453
Location
.de, DE, DEU
Would like to highlight another reason why it is important to develop a "dynamic range" test protocol to determine when a loudspeaker is "dynamically limited".

Definition: A loudspeaker is "dynamically limited" when, at different sound pressure levels, the deviations in frequency response, compared to a reference sound pressure level, exceed a certain limit.
This may not fully describe the "dynamics" of a driver, but it is obviously a basic requirement.

In the thread "Do high-efficiency speakers really have better 'dynamics'?" the majority would probably answered yes to this question.
The 10'' Ciare CH250 mentioned in the thread is a driver with high sensitivity for open-baffle use.

If only [email protected] sound pressure level in free field (regardless of the losses due to sound cancellation, [email protected] half space) without "dynamic limiting" is to be achieved, then the 10'' driver with +-0.5mm Xmax ([voice coil height - magnetic gap height]/2), may only be used down to 200Hz.
1634293300750.png
Otherwise, below this frequency, at high sound pressure levels, the driver will definitely show deviations to a reference sound pressure level frequency response. However, we do not have this knowledge (TSP + specification) when testing loudspeaker A from some manufacturer.

It goes even more extreme with the 12'' driver B&C 320K/C-A with a sensitivity of [email protected]@1m, but with Xmax <<+-0.5mm. But to achieve only [email protected] sound pressure level in the free field without "dynamic compression", a crossover frequency above 200Hz must also be chosen.
1634295806868.png
These two drivers could not even be used in combination with a standard [email protected] without significantly reducing the "dynamic range" of the whole system..

Similarly with a huge PA speaker with 2x12'' woofer, sensitivity >95dB, that @hardisj tested. Harmonic distortion is good in the low end at 96dB, but at 50Hz there is a significant amount of "dynamic compression" already at 102dB. This speaker must always be driven with a subwoofer.
1634307073270.png 1634307092528.png
Without comparative frequency response measurements at different sound pressure levels, this would not have been noticed (Update: I mean the amount of dynamic compression ;)).
 
Last edited:

tktran303

Senior Member
Joined
Mar 27, 2019
Messages
365
Likes
542
@ctrl

I agree with everything you say. And this is what makes the M2 most impressive, to me. The midwoofer doesn’t have the smoothest frequency response. But as erinsaudiocorner.com has shown, the bass and sub-bass distortion @96dB between 30Hz and 1000Hz is outstanding H2/H3 below 50dB down and H4/H5 below 70dB down), and the compression between 76 dB up to 102dB is almost negligible, for the entire music bandwidth (>40Hz)

Compare this to the Neiman KH80; another “simple boxed 2 way speaker” that has a higher smoother FR and listening window; and higher predicted preference score. But whilst the former sounds “almost live” the latter sounds like “a small speaker”
 
Last edited:

tktran303

Senior Member
Joined
Mar 27, 2019
Messages
365
Likes
542
A lot of people who do drive unit testing have found correlations with what they hear. The objectionable distortion that causes one to hit “stop” in the testing of the DUT is easily determined by ear. But the faint rub and fuzz or whistle-type air noises;sometimes that manifests as higher order distortion, but sometimes it’s not easily visible. I haven’t seen objective information or studies on how to measure that.
This is particularly important for open-baffle speakers; where fuzz/buzz or whooshing noises are particularly objectionable

On the other hand, In my own testing of complete loudspeakers; I can often see distortion that that I can’t hear. Once example is when I see distortion that shouldn’t be there. And then tighten all the screws and it goes away. But I certainly couldn’t those ‘not-fully-tightened’ screws.
 

aac

Active Member
Joined
May 17, 2020
Messages
215
Likes
154
PA applications should know a thing or two. Meyer sound has a special signal for that, called M-noise ("music noise" probably).
 
Last edited:

alex-z

Active Member
Joined
Feb 19, 2021
Messages
253
Likes
368
Location
Canada
This sort of testing would take a lot of reviewer time because no single metric covers it.

First you have inter-modulation distortion testing. Many 2 way speakers can play loud in a sine sweep, but do they handle strong bass with vocals on top? What frequencies do you select for testing?

You need at repeat the on-axis measurement many times to check for instantaneous and thermal compression.

Plus if you want to deep dive any problems, you start getting into the realm of accelerometer on the cabinet, anemometer on the port, etc.

For the sake of keeping amir sane I would be happy with just instantaneous compression at 3-4 SPL levels. You can spot a lot of basic things like port chuffing or woofer xmax that way.
 

pogo

Active Member
Joined
Sep 4, 2020
Messages
264
Likes
64
What frequencies do you select for testing?
As testsignale half sine oscillations of different frequencies at the same time could be used, see also here:
Link

A change in amplitude could also be helpful here to show the limits.
 

ctrl

Major Contributor
Joined
Jan 24, 2020
Messages
1,069
Likes
3,453
Location
.de, DE, DEU
This sort of testing would take a lot of reviewer time because no single metric covers it.
....
For the sake of keeping amir sane I would be happy with just instantaneous compression at 3-4 SPL levels. You can spot a lot of basic things like port chuffing or woofer xmax that way.
Exactly, between "do not measure at all" and "measure perfectly" there is a gray area that should be used.

Just as the two measurements of harmonic distortion (86dB and 96dB) Amir makes do not cover the whole complex of distortion, a measurement of "dynamic limitations" of a loudspeaker does not have to cover every aspect of the subject.

We (I mean Amir ;)) can also make an intermediate conclusion after ten measured speakers and if the measurements turn out to be useless, stop it. It would only be important that the measurements follow a precise test protocol to guarantee comparability.
 

pozz

Data Ordinator
Forum Donor
Editor
Joined
May 21, 2019
Messages
3,936
Likes
6,264
@hardisj @amirm

I'm sure you guys know all this, but here's a test regime to think about.

Compression on the signal side is typically expressed as a ratio, input to output:
  • 1:1 means linear output with level.
  • >1:1 is compression, or proportional decrease in output level relative to input. 2:1 means for every 2dB increase in the input signal, there is only a 1dB increase in output.
  • ∞:1 is limiting, equivalent to hard clipping.
  • <1:1 or 1:>1 means expansion, or output proportionally higher than input.
There are also attack and release times (how fast the compression kicks in or drops off, measured in milliseconds) and finally threshold, which determines level. More complex versions might include multiple triggers (multiple thresholds with different ratios), or detection based on crest factor. Multiband compression (different settings for parts of the frequency range) is pretty common.

So function will depend on spectrum, ratio, attack and release times and threshold. As long as we get some numbers we can look up research to figure out audibility.

Some ideas:
  • Plot ratio vs. frequency for a continuous increase in SPL above a baseline. The signal would be a tone increased in level at 12 points per octave, or something to that effect. Take 66dB SPL, around conversational level, and then increase the output 20dB, 30dB, 40dB.
  • Plot level vs. ratio. Take a low level tone, 46dB SPL, and increase it 60dB to identify the threshold where the ratio kicks in.
  • Try to figure out if the compression kicks in or drops off instantaneously or has some time associated. This would be the time in milliseconds that it takes to get from a 1:1 ratio from some compressed ratio or vice versa, to a compressed ratio from a 1:1 baseline. Not sure how to test this.
Not sure how to control for thermal effects. Maybe this can be done with a broadband signal as well, like M-noise, which will simplify the procedure.

For reference: https://www.audiosciencereview.com/...ighlight-dynamic-compression-expansion.26548/
 

tuga

Major Contributor
Forum Donor
Joined
Feb 5, 2020
Messages
2,774
Likes
2,807
Location
Oxford, England
Ideally, there should be a simple metric, like a spl-in-dB number, that accurately describes the usable spl output capacity of a speaker.

Just measure distortion at increasing levels until a defined limit is reached, may seem like the obvious solution. But this gets complicated, when realizing that the nature of this distortion can be very complex, and some distortion types is less harmful than others.

A compression test will show when compression begins. As for audibility - compression of peak transients at 3dB is hard to hear, 6dB is easy to hear when compared in ABX. But 6dB compression in a speaker is a lot, and very audible distortion artifacts can be present well below that.

The measurement must also be easy to understand and replicate, and widely adapted, to be usable for comparing speakers.

What at first glance seems like a simple thing to measure, turns out to be a lot more complicated.

What do you think of these two graphs:

5. MIL - Maximum Input Level (intermodulation distortion ≤ 5%)

6. MOL - Maximum Output Level (intermodulation distortion ≤ 5%)

H87LuMP.png
 
Last edited:

Kvalsvoll

Addicted to Fun and Learning
Joined
Apr 25, 2019
Messages
505
Likes
953
Location
Norway
What do you think of these two graphs:

5. MIL - Maximmum Input Level (intermodulation distortion ≤ 5%)

6. MOL - Maximumm Output Level (intermodulation distortion ≤ 5%)

H87LuMP.png
Maximum output level before excessive distortion, viewed across frequency range like this, can be informative. But this graph looks strange - perhaps they were amplifier limited for this measurement.

But for this speaker, I would rather just eyeball it - 15" good woofer with good motor system + HF compression driver in some kind of smaller horn -> enough capacity for some serious listening in all domestic sized rooms. Just make sure to add a bass-system with decent capacity to fill in the lower frequencies.

Experience with different types of speakers often is all that is required - a typical hifi-speaker with 30hz extension and 2x 6" woofers will never play more than background music, while big 15" and horns usually will be able to play loud enough.

The problem with such an approach is when you encounter something different - like the speakers I make. They are very small, yet claim to be able to play loud enough. Then we need measurements, to show that this is the case.

Using intermodulation distortion for limits can perhaps be a good idea. We know that releaving the speaker drivers that reproduce midrange from bass duty, increases usable capacity significantly, even if the bass is still reproduced from small distorting drivers.
 

abdo123

Major Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
4,101
Likes
3,649
Location
Brussels, Belgium
@amirm @hardisj is it possibe to calculate the displacement limits of the speaker using the Klippel LSI module without taking the componoents apart?

I feel that would be the ultimate test of dynamic ability / compression .etc
 

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
1,537
Likes
3,258
Location
Berlin, Germany
I feel a need of clarification when we talk about compression? More often than not, compression refers to a relative slow sensitivity drift when the voice coil has heated up vs cold state (T_vc = T_ambient). With the long time constants involved, this doesn't cause much distortion by itself, with low level steady state signals after a power burst the driver distorts basically the same, no matter whether hot or cold. But the output RMS level is reduced when the VC is hot (unless the amplifier output impedance is amlost a current source)

OTOH, when the driver is overdriven and clipping, not being able to reproduce the signal anymore, we also measure a reduced total RMS level (or fundamental level) and this can be seen as another form of compression which in turn is not affected much by VC temperature.

Both effects do overlay in practice and to measure one of them correctly would require the other being factored out properly, by choice of signals and procedure/preconditioning. Most likely the second kind of compression will be dominating sound quality deterioration but sometimes the first type can be relevant, for example in small AMTs with as little as ~5W of continuous power dissipation and faster time constants can show a short but significant "brownout" of level after/during a high power burst.
 

Flaesh

Member
Joined
Jun 13, 2021
Messages
30
Likes
21
Location
Eburg
Nice tread. Let's increase test signal if compression\limitation seem insufficient :):
1635130411914.png
 
Top Bottom