Speaker Resonances - The Good, the Bad and the Ugly

[ctrl]

Since the topic of evaluating resonances has been mentioned several times in @amirm's reviews, here is an attempt at a comparative analysis of different resonances in loudspeakers.
For this purpose the measurements of the following loudspeakers are compared:

focal-aria-906

buchardt-s400

ocean-way-hr5

The related question could, in clear terms, be worded as follows:
Why do you nagging bitches hack at one speaker's resonance for days and then completely ignore a similar resonance on the next speaker?

Before we start, a small frustrated outcry on my part: The constantly changing scaling of the CSD diagrams by the Klippel NFS drives me crazy and unfortunately makes comparability massively difficult.

Let's start with probably the least controversial analysis, that for the Aria 906.

Focal Aria 906
First, a quote from the post in the Aria 906 thread that prompted me to open this thread (thanks @hardisj):

The broad bump in the midrange seems it would be troublesome. With our discussion on the Buchardt S400 high-Q midrange peak getting a lot of attention in your recent review, I would expect this speaker's wide-Q bump to be more noticeable to the ear.

The wide mid-range bump around 800Hz has no resonance as a cause. It is rather the case that the baffle-step correction was deliberately not carried out completely.

The decay behaviour in the CSD is inconspicuous, already after 2.5ms the signal around 800Hz is attenuated by about -18dB. The normalized horizontal and vertical frequency responses are inconspicuous, which means that the sound pressure increase around 800Hz becomes the same under all angles.

Around 1.1kHz you can see the hint of a surround resonance(?).

The broad bump can easily be removed by EQ.



Buchardt S400
There really is a resonance on the frequency response between 500-600Hz and in the S400-review thread it was discussed in great detail.

IMHO this could be caused by a combination of interference/resonances that unfortunately increase.
With the 6-inch aluminum cone bass/midrange driver from SBAcoustics there could be a disturbance/resonance in this frequency range.
A standing wave in the speaker enclosure would also be a possibility and could fall within this frequency range.

However, the interaction between the driver and the passive radiator (PR) probably accounts for the largest share.
For this purpose we look at the axis frequency response, the horizontal frequency response measurement at 140° and the vertical frequency response measurements at 120° and 160°.

It is easy to see how shifting the phase alignment of the driver and PR in relation to each other changes the peak of the frequency response into a dip:

That's not really "bad", but when using it as a studio monitor you should of course avoid that the notes C5 and D5/D#5 are 5dB apart in sound pressure - although I don't know if this is still an NFS evaluation for 0.3m (which amplifies the resonances) or one in 1m distance.

The CSD shows that the resonances between 500-600Hz still decay quickly. After 3.5ms the resonances are attenuated by -17dB. Probably interference shifts the resulting resonance as it decays as you can see in the CSD of the first diagramm.

Because the resonance in the frequency response shifts at different angles in frequency, the spike with subsequent dip cannot be completely removed by EQ. This quickly becomes clear when looking at the normalized horizontal frequency response measurements in the first diagram.



Ocean Way HR5
With the HR5 we examine the resonance around 800Hz. My guess is that the resonance is caused by the diffraction slit in front of the 7'' chassis or that a surround resonance is amplified by it. The resonance is most likely not caused by the BR port, since the resonance hardly changes at different angles.

In the upper diagram (comparison of all three loudspeaker) you can clearly see that with the normalized horizontal representation of the angular frequency responses the resonance has disappeared, which means that it hardly changes at different angles.

If we look at selected angular frequency responses, the difference to the resonance in the Buchardt S400 is quite clear. At +-120° horz and +-80° vert the resonance is still there.

The +-80° horz frequency response is not symmetrical, which should be due to the influence of the asymmetrically arranged BR-port.

Now comes the "ugly" part of the analysis. The decay of the 800Hz resonance.
First of all, it should be noted that the 30dB scaling (which is standard) makes the HR5's decay look worse than that of the S400 or Aria 900 - but this is largely due to the larger scaling.
Nevertheless, at 800Hz the resonance is only attenuated by -17dB after 10ms. As shown in the first diagram above and enlarged here:

My guess is that the resonance is caused by the diffraction slit in front of the 7'' chassis or that a surround resonance is amplified by it. Since the resonance slightly changes the pitch as it decays I would consider this relevant in this case, since the damping after 10ms is just -17dB.

My long todo list includes practical experiments on the audibility of resonances, so a thought experiment must suffice here.
Let's assume a piano key is struck at 800Hz (G5). While the tone (with its overtones) swings out, the 800Hz resonance causes a "parasitic" tone slightly shifted in frequency.
Shouldn't this lead to beating during the decay of the piano tone, since the resonance is also excited again and again during the decay?

The small spike in the frequency response can easily be removed by EQ, especially since it is not very pronounced.
What can be changed only slightly is the decay behavior. This will still be quite bad.

 

[napilopez]

Thanks! I suspected the big bump in the Aria was not really a cabinet or driver resonance given its behavior doesn't seem to follow the typical resonance pattern. Looking at the baffle step correction curve might make it more clear what's going on:

(For those unaware, a "6.5-inch" woofer typically has a pistonic diameter of about 5 inches, so that's not a typo).

 

[tuga]

Thanks! I suspected the big bump in the Aria was not really a cabinet or driver resonance given its behavior doesn't seem to follow the typical resonance pattern. Looking at the baffle step correction curve might make it more clear what's going on:

View attachment 69100
(For those unaware, a "6.5-inch" woofer typically has a pistonic diameter of about 5 inches, so that's not a typo).

It looks like baffle-step in Soundstage's measurements too:

https://www.soundstagenetwork.com/i...&catid=77:loudspeaker-measurements&Itemid=153

 

[ctrl]

How To interpret CSD

This question or something similar appears regularly in the speaker reviews:

Please help me understand the waterfall, specifically as to why most speakers have longer decay in the bass region. Is it because speaker cabinets in general resonate in those frequencies no matter how well made/braced they are? Or are ports to blame?

The most important thing is that the influence of the measuring room is faded out (via a gate) or calculated out - otherwise you will inevitably draw wrong conclusions.

Since Amir's CSD measurements are time-based, when comparing the resonances one must first convert them into oscillation periods.
The diagram below shows the CSD of the Revel M105. Unfortunately the scaling is limited to 4ms.

a) At 65Hz the time for a complete oscillation is 15ms. So the 4ms shown in the diagram are not even sufficient to represent a full oscillation. Therefore the decay behaviour at low frequencies cannot be assessed.

b) The resonance around 500Hz could for example be caused by the port. For a complete oscillation 2ms are needed. We can see from the CSD that after 4ms, i.e. two complete oscillations, the signal has weakened by about -13dB.
It is generally assumed that the influence of the resonance is no longer relevant with a reduction of -30dB (but I don't know any scientific paper on this subject). Therefore the waterfall diagrams are usually scaled to -30dB.

c) The resonance at 2500Hz requires 0.4ms for one oscillation period.
This means that the signal was only attenuated by about -26dB after about 10 oscillations. However, the signal initially decays quite quickly and then remains at a low level.

d) The Ocean Way HR5 shows a resonance at 750/800Hz caused by the BR-port or the diffraction slit in front of the woofer.
For a complete oscillation 1.3ms are needed. On the diagram you can see that after about 8 oscillations the signal was attenuated by only -17dB.

In cases c) and d) the slow decay is certainly not desired. The influence this has on the sound of a piece of music cannot, as so often, be clearly assessed.
In order to be able to judge the oscillation at low frequencies, the diagram must have a corresponding time scale due to the large oscillation periods.

 

[Curvature]

Amir's CSD are not anechoic. Keep this in mind when making comparisons. They may include vibrations from the NFS corpus and low frequency sound from the environment.

 

[ctrl]

Amir's CSD are not anechoic. Keep this in mind when making comparisons. They may include vibrations from the NFS corpus and low frequency sound from the environment.

AFAIK the CSD is based on NFS data, so it is anechoic. Only the distortion measurements @amirm shows are simple near field measurements.

 

[Curvature]

AFAIK the CSD is based on NFS data, so it is anechoic. Only the distortion measurements @amirm shows are simple near field measurements.

According to some older posts by Amir, the distortion measurements at some point became anechoic. He has never said CSD is anechoic.

 

[ctrl]

He has never said CSD is anechoic.

Since the NFS provides an anechoic transfer function of a loudspeaker, it should not be a problem to create a CSD from it. It is simply a different representation of the already existing data.

 

[Curvature]

Since the NFS provides an anechoic transfer function of a loudspeaker, it should not be a problem to create a CSD from it. It is simply a different representation of the already existing data.

That's logical, but every single CSD has the same 60/120Hz resonance (which tells me environmental noise) and another few around 2.5kHz (which tells me NFS fixture). The inconsistency of Amir's CSD presentation makes it hard to establish this.

 

[ctrl]

That's logical, but every single CSD has the same 60/120Hz resonance (which tells me environmental noise) and another few around 2.5kHz (which tells me NFS fixture). The inconsistency of Amir's CSD presentation makes it hard to establish this.

Yes, it is a pity that CSD representations are so different.

The NFS Mic mount has caused some reflections in the past, but this has been improved.
It's hard to say whether room reflections have crept in in the bass range. However, this would have to be seen in the FR measurements too.

In the bass range below 100Hz, the oscillation period is over 10ms. Since Amir shows the CSD results mostly with a time window of 4-10ms, it is hardly possible to draw conclusions whether room influences are noticeable in certain frequency ranges.
If, regardless of the low frequency concept (CB, BR, TL...), the decay time in the 60 and 120 Hz range is similar for all LS, then this would be a very strong indication for room influences.
But as I said, the time window in the CSD diagrams is too small for that.

 

[dfuller]

I for one would love to see it standardized, perhaps to the same standard that S&R does.

 

[Curvature]

Yes, it is a pity that CSD representations are so different.

The NFS Mic mount has caused some reflections in the past, but this has been improved.
It's hard to say whether room reflections have crept in in the bass range. However, this would have to be seen in the FR measurements too.

In the bass range below 100Hz, the oscillation period is over 10ms. Since Amir shows the CSD results mostly with a time window of 4-10ms, it is hardly possible to draw conclusions whether room influences are noticeable in certain frequency ranges.
If, regardless of the low frequency concept (CB, BR, TL...), the decay time in the 60 and 120 Hz range is similar for all LS, then this would be a very strong indication for room influences.
But as I said, the time window in the CSD diagrams is too small for that.

Let's look at this. Four random speakers of different designs.

• Orange: B&O Beolab 20
• Green: Genelec S360A
• Purple: JBL AC25
• Red: Arendal 1961 Center

On-axis measurements aligned at 2.5kHz.

My tiny yellow dots show some MF resonances, and my yellow lines show HF resonances common across different speakers. The HF ones are consistent across all of them, the MF ones are inconsistent but there.

In all the CSDs (but no CSD was posted for the Genelec) you see the HF resonances are fairly consistent. Has to be the fixture.

In the JBL AC25 and Arendal 1961 Center the LF response decline below 100-200Hz. But somewhere around 2ms for both there's an LF bloom that is disproportionate to the initial FR. Environmental noise, no?

 

[Curvature]

The 4-5kHz range show prominent resonances in many speakers too.

 

[fineMen]

The 4-5kHz range show prominent resonances in many speakers too.

Often the relationship between time and frequency is not really understood. Because the definition of frequency already has a time reference, namely so and so many repetitions per time unit, and because at the same time (pun not intended) the repetition is defined by the amplitude changing in time ... it has probably no sense at all to speak of a temporal development of the amplitude of a frequency.

(** you might want to think this through yourselves. How would you define / identify the repetition, if the strength of a frequency changes? The concept of repetition would be disturbed and by that, logically, the concept of a pure, clean frequency. Terms are decisive! See also: https://en.wikipedia.org/wiki/Logos. One cannot speak logically (sic!) if the meaning or concepts casted into terms aren't or would not remain well defined.)

By the way, everything that has to do with frequency is completely and unambiguously describable by a spectrum of frequencies with respective (never changing!) amplitude and (never changing!) phase - a time element does not come into play.

In summary, the so-called CSD only makes confusion. Based on confusion one can then of course discuss a whole lot quite seriously. Obviously it is necessary not to be clear about the elementary connections. Otherwise the fun is over, and who wants that?

What you want to discuss here are reflections (presumably at the mike's fixture?), not resonances, right? Obviously, I will spare the explanation in order not to bore anyone with self-evident facts, both are completely different physical phenomena. That the CSD does not show this difference shows its deficiency.

To quote the much-quoted author, even if only in spirit if permitted, resonances are bad, reflections don't matter. CSD is void.

 

[KSTR]

Often the relationship between time and frequency is not really understood. Because the definition of frequency already has a time reference, namely so and so many repetitions per time unit, and because at the same time (pun not intended) the repetition is defined by the amplitude changing in time ... it has probably no sense at all to speak of a temporal development of the amplitude of a frequency.

A single frequency has no time aspect by definition, that is true. And in general we have always a trade-off in resolution, one cannot have infinite frequency and time resolution at the same time.
But a narrow-band group of frequencies can have envelope, or in other words, an envelope (=window function) on a single frequency can be represented as a narrow-band group of frequencies, depending on the envelope, when making a FFT of sufficient length, just as the FFT itself does than from any applied windowing.

CSD/Waterfall (a set of stepped FFTs) is just one of many ways to display an impulse response and certainly has good uses. It's very sensitive to processing parameters, though. Best would be to have the IR raw data and visualize yourself.

 

[KSTR]

What you want to discuss here are reflections (presumably at the mike's fixture?), not resonances, right? Obviously, I will spare the explanation in order not to bore anyone with self-evident facts, both are completely different physical phenomena. That the CSD does not show this difference shows its deficiency.

Repeated reflection == standing waves == resonance

 

[fineMen]

Repeated reflection == standing waves == resonance

not quite. Reflection needs the right phase to build up what you might see as a resonance. Very different concepts altogether. You may want to derive a 0-dimensional standard resonator's differential equation from the 1-dimensional wave equation. Let me know if you do.

 

[Curvature]

not quite. Reflection needs the right phase to build up what you might see as a resonance. Very different concepts altogether. You may want to derive a 0-dimensional standard resonator's differential equation from the 1-dimensional wave equation. Let me know if you do.

There are consistent high-Q FR trends across speakers measurements and CSD results.

I think the HF and MF ones are to do with the NFS fixture, either because of mechanical vibrations or reflections, and the LF ones visible in the CSD to be noise, since the CSD is not anechoic.

The device in question.

Near Field Scanner System (NFS)

Near Field Scanning System, NFS

www.klippel.de