Help me understand Impulse Response, Step Response, and Time Alignment

[Heinrich]

Please show me a scientific study that shows ...

Why?! If there is, an AI will find it today. The problem then is the analysis of the result(s). Without a solid background, or to the contrary, with just the background of long term audio enthusiasm, not too much is gained. Audio is not the right pair of glasses (as we say), to look at the phenomena.

The original problem statement was the complicated integration of SVS woofers and DIY woofers and mains in a room. You were told, that the measurement of the relevant relative phases of all three speakers in room was done wrong. The concept didn't fit. An alternative was given. Did we progress in this regard? What do you need to know.

While that might be true in general terms, and is even a way to cheat on apparent extension for higher frequencies, for the specific example, it's v much audible. Clean 20Hz does not sound like 20Hz with as heavy dose of distortion on top

That's true to some extent, but it wasn't the original question. with four 13" and two 18" (?) plus mains combined there shouldn't be that much of distortion at regular, bearable listening intensity. When I go LOAUD w/ my humble set, I measure about 3% HD2 @25Hz, and no music ever hits that, and 25Hz cannot be perceived as a musical note without overtones aka harmonics, and it is all about wide spectrum noises , bumps, that defunct the idea of HD and that was it maybe. Why don't people just listen and enjoy their costly purchases?

 

[Heinrich]

In a real-world scenario, I couldn't care less, when I am completely satisfied with the sound result based on my own subjective opinion.

 

[levimax]

Why?! If there is, an AI will find it today. The problem then is the analysis of the result(s). Without a solid background, or to the contrary, with just the background of long term audio enthusiasm, not too much is gained. Audio is not the right pair of glasses (as we say), to look at the phenomena.

The original problem statement was the complicated integration of SVS woofers and DIY woofers and mains in a room. You were told, that the measurement of the relevant relative phases of all three speakers in room was done wrong. The concept didn't fit. An alternative was given. Did we progress in this regard? What do you need to know.


That's true to some extent, but it wasn't the original question. with four 13" and two 18" (?) plus mains combined there shouldn't be that much of distortion at regular, bearable listening intensity. When I go LOAUD w/ my humble set, I measure about 3% HD2 @25Hz, and no music ever hits that, and 25Hz cannot be perceived as a musical note without overtones aka harmonics, and it is all about wide spectrum noises , bumps, that defunct the idea of HD and that was it maybe. Why don't people just listen and enjoy their costly purchases?

My original questions was trying to better understand what impulse and step responses represent in the physical world and how it affects sub integration. I never mentioned audibility of time domain issues. The problem is you took something I said about audibility out of context, the audibility I mentioned was about heavy (35% THD) distortion which is very audible. You then built on this to condescendingly make points about my lack of understanding of time domain issues and math. When your misquote was pointed out you then started on a unrelated distortion discussion.

I would be very interested to hear your perspectives on the value or lack thereof of impulse response and step response graphs and if and how you used these to help better integrate different subs and any explanations to help lay people better understand what these represent both their value and limitations.

Yes I have learned a lot from this thread.

 

[audiofooled]

My original questions was trying to better understand what impulse and step responses represent in the physical world and how it affects sub integration. I never mentioned audibility of time domain issues. The problem is you took something I said about audibility out of context, the audibility I mentioned was about heavy (35% THD) distortion which is very audible. You then built on this to condescendingly make points about my lack of understanding of time domain issues and math. When your misquote was pointed out you then started on a unrelated distortion discussion.

I would be very interested to hear your perspectives on the value or lack thereof of impulse response and step response graphs and if and how you used these to help better integrate different subs and any explanations to help lay people better understand what these represent both their value and limitations.

Yes I have learned a lot from this thread.

I've prepared some graphs of my system in the hopes it helps in building some intuition. Beforehand, I'd like to emphasize that the essential part would be determining what you want to achieve with the system and measure accordingly. Correct interpretation of the measurements and of course listen, remeasure, reinterpret, repeat and hold your breath and wait a millisecond or two (pun intended) in case you see something as ugly as this:

In blue you see the step response in the nearfield, sub cone level, close to the ground. Pink would be ear level of the same setup I described earlier, measured at 3,5m distance. What the hell is this and what went wrong?

First, note the highlighted area on the pink graph. Then look at this:

The vertical cursor is a corresponding dip of the highlighted area and three peaks in the vicinity which are misaligned in time at that measurement position. More about that later.

Next, look at this:

It's the other side of the same wavelet so read it backwards, with vertical cursor positioned at where the pink graph is at -100%.

This is another view of the same thing, showing that everything is aligned at 0 (nearfield) and then misaligned as the total summation 265Hz and bellow develops towards the listening position:

This is because at this measurement position, pressure is lower at 0 and summed at 6dB higher at 10-12ms sound has to travel to this measurement position. In reality, it's quite the opposite, nearfield pressure sums a lot higher but is lost with distance.

Considering goals I was trying to achieve, (besides stereo bass and AE), that is floor bounce control where mid bass frequencies sum at the floor, whereas at the listening position at ear level they don't (highlighted area on the pink graph also shows physical spacing between the mains). And wise vice versa and more importantly, low frequencies are very much attenuated on the floor, but sum at ear ear level with a pressure loss as compared to the nearfield. Otherwise, room resonances are much worse and quite destructive to the room structure.

As for "perfect" step response, this is the same measurements as the first picture, but with Multi Time Window (MTW):

Where it is obvious that measurement distance and low frequency extension would greatly affect measured step response.

I hope this helps.

 

[levimax]

I've prepared some graphs of my system in the hopes it helps in building some intuition. Beforehand, I'd like to emphasize that the essential part would be determining what you want to achieve with the system and measure accordingly. Correct interpretation of the measurements and of course listen, remeasure, reinterpret, repeat and hold your breath and wait a millisecond or two (pun intended) in case you see something as ugly as this:

View attachment 498129

In blue you see the step response in the nearfield, sub cone level, close to the ground. Pink would be ear level of the same setup I described earlier, measured at 3,5m distance. What the hell is this and what went wrong?

First, note the highlighted area on the pink graph. Then look at this:

View attachment 498130

The vertical cursor is a corresponding dip of the highlighted area and three peaks in the vicinity which are misaligned in time at that measurement position. More about that later.

Next, look at this:

View attachment 498131

It's the other side of the same wavelet so read it backwards, with vertical cursor positioned at where the pink graph is at -100%.

This is another view of the same thing, showing that everything is aligned at 0 (nearfield) and then misaligned as the total summation 265Hz and bellow develops towards the listening position:

View attachment 498136

This is because at this measurement position, pressure is lower at 0 and summed at 6dB higher at 10-12ms sound has to travel to this measurement position. In reality, it's quite the opposite, nearfield pressure sums a lot higher but is lost with distance.

Considering goals I was trying to achieve, (besides stereo bass and AE), that is floor bounce control where mid bass frequencies sum at the floor, whereas at the listening position at ear level they don't (highlighted area on the pink graph also shows physical spacing between the mains). And wise versa and more importantly, low frequencies are very much attenuated on the floor, but sum at ear ear level with a pressure loss as compared to the nearfield. Otherwise, room resonances are much worse and quite destructive to the room structure.

As for "perfect" step response, this is the same measurements as the first picture, but with Multi Time Window (MTW):

View attachment 498141

Where it is obvious that measurement distance and low frequency extension would greatly affect measured step response.

I hope this helps.

Thank you, there is a lot to think about in your post.

After reading this thread and some of the links I went back to my 2 measurements and played around with the REW alignment tool. Online the "common wisdom" is that SVS has 6 ms DSP delay which "intuitively" looks right from the impulse response but looking at the very start of the impulse response it is more like 2.5 ms. I was surprised that by aligning the 2 subs with 2.5 ms delay on the DIY sub I got the "smoothest" summed response. At 6 ms (internet wisdom) it was much worse. For me the look of the aligned step responses is completely counter intuitive, based on intuition it looks like the SVS sub is "way behind" the DIY. See below what the time aligned impulse responses look like, red is DIY green is SVS. If someone asked me to explain what the 2 step responses represent in the physical world I don't think I could but I know not to trust my intuition.

 

[3ll3d00d]

Seems like "common wisdom" is "people on the internet are wrong again" with the source of the wrongness being something like "align the peaks". Meanwhile the above sr seems to be better basically just what was described earlier (fundamental bandwidth differences)

 

[audiofooled]

Thank you, there is a lot to think about in your post.

After reading this thread and some of the links I went back to my 2 measurements and played around with the REW alignment tool. Online the "common wisdom" is that SVS has 6 ms DSP delay which "intuitively" looks right from the impulse response but looking at the very start of the impulse response it is more like 2.5 ms. I was surprised that by aligning the 2 subs with 2.5 ms delay on the DIY sub I got the "smoothest" summed response. At 6 ms (internet wisdom) it was much worse. For me the look of the aligned step responses is completely counter intuitive, based on intuition it looks like the SVS sub is "way behind" the DIY. See below what the time aligned impulse responses look like, red is DIY green is SVS. If someone asked me to explain what the 2 step responses represent in the physical world I don't think I could but I know not to trust my intuition.

View attachment 498243

I hope you understand that I just thought there could be some value in what I could post, but bare in mind that my complete system is DIY, that I did my best to build a matching sub/bass module from the start that can be crossed as high as 200Hz for this application and that there are no "small boxes and DSP" in my system.

On top of that, it's that different rooms may require different strategies. If I were to integrate, say, a pair of commercial subs into my system, I'm quite positive I would open yet another can of worms as far as compromises. My system as such, is not even close to it's full potential in current room. Setup is very inefficient, but still the order of things as far as dynamic capabilities is that first I give up, then the room, then the amp, and not the loudspeakers. As far as testing, the only thing I didn't do to them is perhaps submerging them into water

No, I wouldn't trust "Internet wisdom", but I would follow advice from members over here who are more than smart and proficient in DSP. If you post some .mdat files I'm sure someone would be able to help. Cheers

 

[audiofooled]

Nice work. Thx for showing. I need to look into JJ's test signals.

For me, phase is no longer an issue. I simply tune my DIY's for flat phase. Reasoning whether audible or not, I know flatter smoother phase is technically correct.
And is therefore one less potential source of signal degradation coming out of the speaker. One less factor to be concerned with.

Yeah, I don't see why not

Two sine waves with different frequencies: Beats​

Two waves of equal amplitude are travelling in the same direction. The two waves have different frequencies and wavelengths, but they both travel with the same wave speed. Using the principle of superposition, the resulting particle displacement may be written as:

This resulting particle motion is the product of two travelling waves. One part is a sine wave which oscillates with the average frequency . This is the frequency which is perceived by a listener. The other part is a cosine wave which oscillates with the difference frequency . This term controls the amplitude "envelope" of the wave and causes the perception of "beats". The beat frequency is actually twice the difference frequency, .

In the animation at left, two waves with slightly different frequencies are travelling to the right. Since the two waves are travelling in the same medium, they travel with the same speed. The resulting superposition sum wave travels in the same direction and with the same speed as the two component waves, but its local amplitude depends on whether the two individual waves have the same or opposite phase. The "beat" wave oscillates with the average frequency, and its amplitude envelope varies according to the difference frequency.

Source: https://www.acs.psu.edu/drussell/Demos/superposition/superposition.html

 

[audiofooled]

there is a lot to think about in your post.

Then I hope you don't mind thinking a bit more (sorry for this being out of scope of this thread):

As for tradeoffs with my setup, the idea of floor bounce control, even though resulting in messy step response for that measurement location, has shown to be quite effective for controlling the most prominent room mode at about 43Hz. In return, at double the frequency it's much worse, however, considering the energy of the two frequencies as found in real music signals, and listening tests, I decided the tradeoff was a keeper.

This is another view of the wavelet from post #64 (linear, %peak)

The reasoning behind it, or the physics at hand, can be somewhat explained that, even with standing waves in rooms, things are not always black and white. The attenuation is almost coming from this kind of situation for the main listening position, where the three seat sofa is located:

Reflection from an Impedance Discontinuity: the Standing Wave Ratio​

Now let's see what happens when an incident wave train encounters a boundary with an impedance that is somewhere between hard and soft. Now, only part of the incident wave is reflected; the rest is either absorbed by or transmitted into the second medium with the different impedance. The orientation of the reflected wave (whether upright or inverted) depends on whether the new material has a larger or smaller impedance. For this particular animation, the reflected wave has half of the amplitude of the incident wave, and is inverted.


The resulting superposition looks kind of like a standing wave, but it doesn't stay still. It appears to lurch forward toward the right. This should make sense since only a part of the energy carried by the original wave is reflected from the boundary; the rest of the incident energy is either absorbed by the impedance termination or transmitted into the second medium. As a result there is more energy moving to the right than there is being reflected to the left. There is a net energy flow to the right.

If you watch the above animation very carefully you will notice that the locations where the wave pattern reaches a maximum value are fixed in space. Likewise, the locations where the wave pattern has a minimum value are also fixed locations. So, while there is a net transfer of energy toward the right, there is an aspect of this wave pattern that does not change with time or space.

Source, including math on the bottom of the page: https://www.acs.psu.edu/drussell/Demos/SWR/SWR.html

The energy transmission measured (moderately accurate) with accelerometer at the listening position, for a real time music signal:

Bass direction is audible

I would love to hear ASR's opinion about this. https://www.audiosciencereview.com/forum/index.php?threads/visualizing-how-different-loudspeaker-lf-directivity-patterns-couple-to-room-modes.45518/ I don't know why NTK didn't provide this link))

www.audiosciencereview.com

Note the difference in energy coming from the floor (less energy, so less to be absorbed for this frequency range - BLUE line), and rear wall bounce (RED). The highest absolute amplitude of the red line is upon constructive interference near the modal frequency. This is also in line with most prominent body sensation. However, the resulting pressure (what is heard) is not at all boomy, nor it results in much higher perceived loudness.

Subjectively, this can be easily mistaken for "fast bass", while there is actually no such thing