"Distorted" step response

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
"Distorted" step response

Hi, I'm new in this group.

What should I think about this "distorted" step response graph I took from my 10" subwoofer? Does it indicate something, like broken element, or is it normal? What that "double peak" means it has?

I thought that step response is always smooth like in this one (graph below) I took from woofer of my shelf speaker. Both were taken in near field.


Thanks, Tipo
 

Attachments

  • pic3.jpg
    pic3.jpg
    98.3 KB · Views: 409
It may sound strange at first but the upper step response is nicer than the lower one !

The distortion that you see is either cone resonance or maybe from the reflex tunnel (though the step response reminds me from a cloded box). Has it been taken without crossvover ?
This resonance can be reduced significantly as soon as its source has been determined.

BTW: The reason why I like the upper response more than the lower one (probably reflex, isn't it ?) is because it has less overshoot.

Regards

Charles
 
Attached is freq response and step response curves for both closed and vented configurations.

Box is 40l, tube 340/66mm, element Peerless 10" XLS. I simply closed box by putting foam into tube.

There's practically no difference between closed and vented step responses.

Speakers sounds and "feels" good. I'm just curious why step response is "distorted". I don't have questions about freq resp.

There' s no crossover/filter/EQ in measured system.

Thanks, Tipo
 

Attachments

  • fr_sr closed_vented nf.jpg
    fr_sr closed_vented nf.jpg
    83.9 KB · Views: 309
Speakers sounds and "feels" good. I'm just curious why step response is "distorted". I don't have questions about freq resp.There' s no crossover/filter/EQ in measured system.

1.) Sound and feels good: Everything is OK.

2.) Distorted step response and frequency response: They are depending on each other. The ripples you see on the step response are caused by very the uneven response above 500 Hz.

3.) As soon as you measure this SUB-woofer with a lowpass in front most of the ripple will be gone.

Regards

Charles
 
tipo1000 said:
Attached is freq response and step response curves for both closed and vented configurations.

Box is 40l, tube 340/66mm, element Peerless 10" XLS. I simply closed box by putting foam into tube.

There's practically no difference between closed and vented step responses.

Speakers sounds and "feels" good. I'm just curious why step response is "distorted". I don't have questions about freq resp.

There' s no crossover/filter/EQ in measured system.

Thanks, Tipo

Ah, this is a close field measurement, near the woofer?

That is clearly visible in the frequency response, but not in the step response. The port is tuned to ~21 Hz, and the port contribution is neglected (due to the close field measurement).

Am I right?
 
phase_accurate said:


1.) Sound and feels good: Everything is OK.

2.) Distorted step response and frequency response: They are depending on each other. The ripples you see on the step response are caused by very the uneven response above 500 Hz.

3.) As soon as you measure this SUB-woofer with a lowpass in front most of the ripple will be gone.


And you can even tell it's around 500 Hz by figuring out the period of the step response ripples. They're about 1.8 milliseconds (ms), indicating a problem around 550 Hz (period of 2 ms). Lo and behold, the response graph shows a narrow and deep null there, indicating a high-Q resonance. I wouldn't worry about it, since the subwoofer crossover should be many dB down by then.

Phase_accurate liked the first step response, even though it looked a bit grungy, because the grunge is well outside the passband and the time response is otherwise well damped. The second step response has ringing around 10 ms, hinting at a resonance around 100 Hz. That will be noticeable in a subwoofer.
 
tipo1000 said:
Tube is 340/66mm and in opposite side of box.
Charles: What ripple you mean?

How can you see location and length of tube from these curves?

BR, Tipo

Length of tube is easy: you have a peak at 410 and another at 880. Ducts are open at both ends, so they tend to resonate at multiples of a half-wavelength. Imagine a sine half-wavelength with a node at the middle of the tube and the maximum at either end: that's the first resonance. Now visualize a sine with a maximum at either end and a full wavelength within the tube. That's the second resonance.

Let's look at the second resonance, the full wavelength: 340 meters/second / 880 Hz = 38 cm. That's a bit longer than your duct, so there might be some end effects going on.
 
DSP_Geek said:


And you can even tell it's around 500 Hz by figuring out the period of the step response ripples. They're about 1.8 milliseconds (ms), indicating a problem around 550 Hz (period of 2 ms). Lo and behold, the response graph shows a narrow and deep null there, indicating a high-Q resonance. I wouldn't worry about it, since the subwoofer crossover should be many dB down by then.

Hmm, in the early part if the step response, there is a ringing with a period of ~0.5 ms. This is consistent with the peak at slightly less than 2 kHz.

At the null there is not a resonance, but an anti-resonance, and these typically does not result in oscillations (I might be wrong here), at least not as powerful as a resonance with the same Q.

I would guess that the 2 ms ripple comes from the sharp corner at 550 Hz, not the null at 650 Hz.
 
Svante said:


Hmm, in the early part if the step response, there is a ringing with a period of ~0.5 ms. This is consistent with the peak at slightly less than 2 kHz.

At the null there is not a resonance, but an anti-resonance, and these typically does not result in oscillations (I might be wrong here), at least not as powerful as a resonance with the same Q.

I would guess that the 2 ms ripple comes from the sharp corner at 550 Hz, not the null at 650 Hz.

Actually, anti-resonances will cause ringing in a step response. Here's a hand-waving explanation:

Look at a second-order transfer function equation set for unity gain, where w = 1:

Hunity(s) = (s^2 + s/Q + 1)/(s^2 + s/Q + 1)

The bandpass function is:

Hbandpass(s) = (s/Q)/(s^2 + s/Q + 1)

The bandstop function is:

Hunity(s) = (s^2 + 1)/(s^2 + s/Q + 1)

Note the bandstop is unity - bandpass. Now, set Q quite high. The bandpass function will ring on transients such as a step function. But wait! Since the bandstop is unity - bandpass, that will ring too, but in opposite phase from the bandpass. The ringing is identical for equivalent Q, only the sign changes.
 
DSP_Geek said:


Actually, anti-resonances will cause ringing in a step response. Here's a hand-waving explanation:

Look at a second-order transfer function equation set for unity gain, where w = 1:

Hunity(s) = (s^2 + s/Q + 1)/(s^2 + s/Q + 1)

The bandpass function is:

Hbandpass(s) = (s/Q)/(s^2 + s/Q + 1)

The bandstop function is:

Hunity(s) = (s^2 + 1)/(s^2 + s/Q + 1)

Note the bandstop is unity - bandpass. Now, set Q quite high. The bandpass function will ring on transients such as a step function. But wait! Since the bandstop is unity - bandpass, that will ring too, but in opposite phase from the bandpass. The ringing is identical for equivalent Q, only the sign changes.


Mm, you are forgetting one thing. The bandpass only lets a narrow band through. This means that the signal has little energy.

The bandstop removes very little.

The amplitude if the ringings in these two signals will be the same, as you say.

But here is the twist; if these two signals are normalised, so they have the same amount of energy, the amplitude of the ringing will become larger in the bandpassed signal.

In real life the systems are not as simple as these, but my impression is that a peak of 10 dB causes much more ringing than a dip of 10 dB.
 
Svante said:

Mm, you are forgetting one thing. The bandpass only lets a narrow band through. This means that the signal has little energy.

The bandstop removes very little.

The amplitude if the ringings in these two signals will be the same, as you say.

But here is the twist; if these two signals are normalised, so they have the same amount of energy, the amplitude of the ringing will become larger in the bandpassed signal.

In real life the systems are not as simple as these, but my impression is that a peak of 10 dB causes much more ringing than a dip of 10 dB.

That's the virtue of the step and impulse functions: they contain all frequencies so any resonant modes will get excited, no matter the bandwidth.

Ears are relatively insensitive to time domain phenomena such as ringing; they mostly work in the frequency domain. A 10 dB peak is more obvious because it increases narrow-band energy, which the ear notices more readily than a narrow-band lack of energy.

I still suspect the ringing will be the same, and I'm willing to throw a 'scope on a resonant system to find out, but things are _really_ busy for me right about now. Let's talk about this some time in November.


Cheers,
Francois.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.