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

[levimax]

I have 4 subs, 2 SVS SB3000 which are 13" small sealed heavily DSP'd subs and 2 very large 18" DIY seal subs with no DSP. I have tried to use these in a 4 sub set up but was never happy with the integration. I tried DIRAC DLBC, 2 different MSO programs, REW Time Alignment tool, REW wavelet tool, and a sweep with a delay on one driver to check time between impulse responses. All of these techniques gave different results for the delays and DIRAC went crazy with 125 ms delays which is clearly not right.

Through the process I have learned about the limitations of in room LF measurements and many other things but am struggling to understand exactly how the filters on the SVS sub (strong bass boost with strong high pass protection filter and it looks like a low pass as well) affect how the cone of the driver actually moves. I have some near field measurements of the 2 subs taken with REW, a Earthworks M23 MIC, and an interface with loopback. See below. All settings were the same between the 2 measurements. When I look at the Impulse and the Step response of the 2 subs they are very different as expected due to the DSP. The DIY sub impulse and step make sense to me but the SVS sub I can't understand at all what I am looking at. If someone could explain what I am looking at and how to think about setting the delays (for the DIY subs to match the SVS subs) it would be greatly appreciated. My questions are:

1. There is obviously some DSP delay with the SVS subs but when trying to time align subs what am I trying to align? The start of the impulse, the peak (and which peak the SVS had a lot of peaks), or something else. I know MSO programs look for "peak energy" or something like that.

2. Why does the SVS have an attenuated peak in the same direction as the DIY and then a larger peak in the opposite direction? What do the small and large peak represent?

3. Both the impulse response and step response of the SVS continue on much longer than on the DIY sub. I assume the SVS is continuing to "move" for a longer time period after an impulse than the DIY but Is that caused by the diver difference or the filters or both? How does that affect time alignment?

4. When trying to understand the difference between how the driver cone moves with and without filters how do I think about it? Does a filter delay or attenuate the initial cycle or all cycles or something else.

Maybe my questions show that I am not thinking about this correctly at all in which case I would be very interested in learning how to think about these issues. Thanks for any help.

DIY Sub

SVS SUB

 

[Keith_W]

Book 2 of the REW eBook, Chapter 4.4 (page 50) on Impulse and Step response should tell you what you need to know. Not everything you need to know, but enough to get you started.

1. There is obviously some DSP delay with the SVS subs but when trying to time align subs what am I trying to align? The start of the impulse, the peak (and which peak the SVS had a lot of peaks), or something else. I know MSO programs look for "peak energy" or something like that.

For subwoofer time alignment, it is a two step process. For minimum-phase DSP, align the start of the subwoofer(s) impulse response or step response with the tweeter. The purpose of this step is to get rid of egregious delays which may exist. The second step, which is more difficult, is to align the phase of all the subs to get rid of dips in the frequency response. You can do this manually in REW or use a tool like MSO.

2. Why does the SVS have an attenuated peak in the same direction as the DIY and then a larger peak in the opposite direction? What do the small and large peak represent?

The amplitude / height of the peak = volume of the sub. The width of the peak = the wavelength. If you measure a sub which is low passed at 100Hz, it will have a narrower peak than a sub that is low passed at 50Hz, for example.

3. Both the impulse response and step response of the SVS continue on much longer than on the DIY sub. I assume the SVS is continuing to "move" for a longer time period after an impulse than the DIY but Is that caused by the diver difference or the filters or both? How does that affect time alignment?

This is post-ringing. Assuming that the DIY and the SVS have been normalized to the same volume (because a softer sub will have less apparent ringing!) then the length of the ringing may be due to the sub or the room. Sometimes post-ringing can be due to bad DSP. To tell if it's due to DSP, examine the filter (load the filter into REW and look at the step response). To find out if it's the sub or the room ... take the sub outside and measure.

Post-ringing does not affect time alignment. It is the "head" of the impulse which is aligned, and not the "tail".

4. When trying to understand the difference between how the driver cone moves with and without filters how do I think about it? Does a filter delay or attenuate the initial cycle or all cycles or something else.

I am sorry, I don't understand this question. Would you be able to rephrase it? Or maybe some ASR member who is more coffee'd up than me might know the answer

 

[3ll3d00d]

The frequency response of the two subs must be pretty different so comparing IRs is nigh on impossible. For example, does the svs really ring more or is it just more visible because of the lack of higher frequency content?

Why do you need to know how the driver physically moves?

 

[levimax]

The frequency response of the two subs must be pretty different so comparing IRs is nigh on impossible. For example, does the svs really ring more or is it just more visible because of the lack of higher frequency content?

Why do you need to know how the driver physically moves?

Yes the FR of the DIY is much more extended, F3 ~800 Hz vs F3 ~225 Hz for the SVS. I ran a sweep from 5 Hz to 2000 Hz. I have read sweeps should always be "full range" but I don't know if that is "full range" of the driver or 20Hz to 20 Khz or even if that is correct. Does that mean I should run a sweep based on the range of the "lowest performing sub" on both subs?

I am trying to understand what I am looking at when I look at the SVS impulse and step response. The DIY looks like what I would expect but the SVS I have no idea. For instance is the polarity reversed on the SVS (The "big" peak is in the opposite direction of the DIY) or is the "first peak" of the SVS just attenuated? If the later than why is the second peak so much bigger than the initial peak. And in either case how would I try to "line them up" for timing alignment. It is easy when they look the same but these seem to me to have nothing to do with each other.

 

[gnarly]

Ime, it's not really possible to use impulse responses to align subs when one is unfiltered and the other is processed.
Knowing the filtering being applied to the SVS subs is key I think, to help match your DIY.
For that I would not look at acoustic impulse. Best would be the electrical filter impulse if there is way to tap into amp module's internal line-level feeding the amp section. Or run a transfer on the amp output at driver terminals...(being careful not to blow your soundcard up.)

I figure doing either of those is probable difficult, so most likely looking at the acoustic transfer function mag and phase is best option.
Ime, mag and phase are far easier to discern what are the likely hpf, low pass filter, EQ's etc,

At any rate, getting the transfer function of the DIY, or impulse, to resemble the SVS will help bring them together.
They don't have to be identical, but the closer to the same filter topology your can get will help (this is assuming the are both sealed)

Here is the impulse and transfer function of an electrical BW 36dB/oct hpf at 20Hz, and a LR 24dB/oct low pass at 100Hz.
I'm taking a guess it might me close to what SVS is using. The steep hpf is because I figure the sub has a loot of bottom end boost that has to be cut off abruptly.
Anyway, it appears to be somewhat close to the acoustic impulse shape you posted.
(Both it and your acoustic impulse, show polarity inversion.)

When we are working with driver sections having the same passband, as is the case with subs, we can time align either impulse starts or peaks.*
It does not matter whether we use start or peak once we have achieved reasonably similar looking acoustic impulses, but again I would suggest using mag and phase trace matching as an easier to make sense of method. Key is getting mag and phase traces to match under same measurement conditions (preferably ground plane outdoors). Once that's done simply time align the more easily identifiable impulse peaks.

* It's when we are time aligning dissimilar passbands, like to sub to main, that impulse starts need to be aligned. This if for minimum phase /IIR time alignment.
For either similar or dissimilar passbands, when using linear phase, it's impulse peaks that time align.

 

[levimax]

The second step, which is more difficult, is to align the phase of all the subs to get rid of dips in the frequency response. You can do this manually in REW or use a tool like MSO.

I have tried all of the tools and none of them give reliable or repeatable or sensible results and I am starting to get the idea that there is no way these 2 subs are going to play nice together. The group delay is massively different so you will only have the phase aligned at one frequency and it will then quickly get so far off that you will get cancelation. For example eyeballing the GD graphs if I align the 2 subs at 80 Hz by the time you get to 35 Hz the timing will be off by ~14 ms which would put the 2 subs 180 degrees out phase causing cancellation across the exact range subs do most of their work. See below GD graphs. Does that make sense to you that these 2 subs will never work well together or am I not thinking about this correctly.

DIY Sub GD

SVS Sub GD

 

[levimax]

Ime, it's not really possible to use impulse responses to align subs when one is unfiltered and the other is processed.
Knowing the filtering being applied to the SVS subs is key I think, to help match your DIY.
For that I would not look at acoustic impulse. Best would be the electrical filter impulse if there is way to tap into amp module's internal line-level feeding the amp section. Or run a transfer on the amp output at driver terminals...(being careful not to blow your soundcard up.)

I figure doing either of those is probable difficult, so most likely looking at the acoustic transfer function mag and phase is best option.
Ime, mag and phase are far easier to discern what are the likely hpf, low pass filter, EQ's etc,

At any rate, getting the transfer function of the DIY, or impulse, to resemble the SVS will help bring them together.
They don't have to be identical, but the closer to the same filter topology your can get will help (this is assuming the are both sealed)

Here is the impulse and transfer function of an electrical BW 36dB/oct hpf at 20Hz, and a LR 24dB/oct low pass at 100Hz.
I'm taking a guess it might me close to what SVS is using. The steep hpf is because I figure the sub has a loot of bottom end boost that has to be cut off abruptly.
Anyway, it appears to be somewhat close to the acoustic impulse shape you posted.
(Both it and your acoustic impulse, show polarity inversion.)

View attachment 495973


When we are working with driver sections having the same passband, as is the case with subs, we can time align either impulse starts or peaks.*
It does matter once we have achieved reasonably similar looking acoustic impulses, but again I would suggest using mag and phase trace matching as an easier to make sense of method. Key is getting mag and phase traces to match under same measurement conditions (preferably ground plane outdoors). Once that's done simply time align the more easily identifiable impulse peaks.

* It's when we are time aligning dissimilar passbands, like to sub to main, that impulse starts need to be aligned. This if for minimum phase /IIR time alignment.
For either similar or dissimilar passbands, when using linear phase, it's impulse peaks that time align.

Thank you. So just to be clear since I don't really want to take the SVS subs apart you are suggesting to apply the estimated SVS filters to the DIY sub and then take acoustic measurements at which point they should be similar enough to "align". I am taking in room near field measurements which I figured would be OK because both are sealed. I don't have a way or place to drag the big DIY subs outside unfortunately.

Take a look at the GD graphs I just posted. To me it seems like maybe these 2 subs can never play nice together unless I cripple the DIY subs with unnecessary and performance reducing filters?

 

[gnarly]

So just to be clear since I don't really want to take the SVS subs apart you are suggesting to apply the estimated SVS filters to the DIY sub and then take acoustic measurements at which point they should be similar enough to "align".

Yes, guess the SVS filters, apply them, measure, .... repetitive guess tweaking /remeasuring for best match possible.

Take a look at the GD graphs I just posted. To me it seems like maybe these 2 subs can never play nice together unless I cripple the DIY subs with unnecessary and performance reducing filters

Highly likely the 18 DIY has to be made a bit sucky to match up.
Problem of course is with the SVS trying to reach too low for its britches...can't make chicken salad out of chicken shit.. Lol

edit...fwiw, i think impulse, step, and group delay are extremely difficult and unreliable for sub work....dual channel mag and phase traces, especially from a continuous pink noise transfer are much easier to align / make sense of. Plus you can dial them in real time as you apply EQs. dual channel OpenSoundMeter might be worth a try i think.

 

[DWPress]

I have the same problem @levimax. My SB-1000 measures the same way compared to my 15" DIY sub and I've been fighting it for several years now. I'm still living with it but plan to replace the SVS with another 12" passive sub in the coming months. I've observed quite a few threads that show the built in DSP on these plate amp subs can totally clash with advanced system wide DSP or room correction. I'm tired of the battle but will keep watching this thread in hopes @gnarly isn't completely correct!

 

[3ll3d00d]

Yes the FR of the DIY is much more extended, F3 ~800 Hz vs F3 ~225 Hz for the SVS. I ran a sweep from 5 Hz to 2000 Hz. I have read sweeps should always be "full range" but I don't know if that is "full range" of the driver or 20Hz to 20 Khz or even if that is correct. Does that mean I should run a sweep based on the range of the "lowest performing sub" on both subs?

this is quite a general statement but sweeps should cover the bandwidth they need to show whatever it is you want to look at. In this particular case, I would add the same low pass to each sub, their IR will now be more similar and hence should be a bit easier to compare. Are they also measured in the same position or not? As the oscillation in the svs one is clearly in the ~30Hz range so I would guess that's some modal resonance you're seeing and it's either not present in the DIY case (because the position is different) or obscured by the wider bandwidth.

For instance is the polarity reversed on the SVS (The "big" peak is in the opposite direction of the DIY) or is the "first peak" of the SVS just attenuated? If the later than why is the second peak so much bigger than the initial peak. And in either case how would I try to "line them up" for timing alignment. It is easy when they look the same but these seem to me to have nothing to do with each other.

It's not inverted, hard to say for sure but t suspect the larger subsequent peak is just driven by that resonance. If you were trying to time align the 2 subs, I'd just use the initial rise for that purpose but often you combine subs by allowing their exact timing to vary and then aligning the combination to the woofer so I would say that often you don't even need super precise alignment of 2 subs to each other.

 

[Keith_W]

I have tried all of the tools and none of them give reliable or repeatable or sensible results and I am starting to get the idea that there is no way these 2 subs are going to play nice together. The group delay is massively different so you will only have the phase aligned at one frequency and it will then quickly get so far off that you will get cancelation. For example eyeballing the GD graphs if I align the 2 subs at 80 Hz by the time you get to 35 Hz the timing will be off by ~14 ms which would put the 2 subs 180 degrees out phase causing cancellation across the exact range subs do most of their work. See below GD graphs. Does that make sense to you that these 2 subs will never work well together or am I not thinking about this correctly.

This situation is normal for multiple subs - align phase at one frequency, and it will be out-of-phase at another frequency. This is why we have brute force calculators like MSO to find the correct permutation of PEQ's, AP Filters, and delays to give you the flattest response possible.

In general, the principle for multiple subs looks something like this:

1. Decide if you want to high-pass your main speakers or run them full range. You might want to high-pass them if you have no choice (your speakers don't do bass) or if they have a lot of bass distortion. You might want to keep them full range to use them as additional bass sources in the room so you MIGHT be able to get away with fewer subwoofers. Once you have made your decision, sweep them together and take note of where the nulls in the FR are.

2. Place one subwoofer and sweep it. The goal is to use the sub to fill in the null noted previously. Use REW trace arithmetic A+B to sum the responses and look to see whether there are new nulls and what has happened to the original null.

3. Repeat with all your remaining subwoofers - place each one and sweep, adding the responses, and removing the nulls. Where do you place your subs? There are studies (for e.g. by Olive and Welti), but those won't help you if you do not have a non-rectangular room or you are constrained by room features like access and furniture. They should be used as a guide for sub placement. It is always better to place your subs and check with measurement that it is placed correctly. No disrespect to Olive and Welti.

4. Begin the DSP process. Your goal is to find the correct combination of delays, phase adjustment, and amplitude equalisation (in that order!) until you have the flattest curve possible. You can do this manually or use a tool like MSO.

 

[dualazmak]

Hello OP @levimax,
I assume you have been being aware of my rather primitive methods of not only using time-shifted multiple-Fq sine-tone-burst sequence (kick-off/trigger timing alignment) but also using sine wave (1-wave, 3-wave, 8-wave) shape matching (phase matching/optimization) all of which I summarized in my post here.

 

[audiofooled]

3. Both the impulse response and step response of the SVS continue on much longer than on the DIY sub. I assume the SVS is continuing to "move" for a longer time period after an impulse than the DIY but Is that caused by the diver difference or the filters or both? How does that affect time alignment?

In case you've missed this:

A Broad Discussion of Speakers with Major Audio Luminaries

Just curious. I attach my version, Good evening Dr. Floyd. Alton Everest acoustics manual. and you will see that even the first room reflections are around the threshold of detection for image shift. So, don't get concerned about "image shift and focus" caused by even later room Crosstalk is...

www.audiosciencereview.com

"Most subwoofers also create audible harmonics, you really can't do much with a smaller enclosure, even with positional feedback, as far as I've seen, to prevent that. Consider, for instance, a run of the mill 15" dayton "max" sub. Put it in a 15x15x15 enclosure, and you find that the interior volume changes enough during extension and compression (way over the 120 to 140dB SPL limit for linearity in air) to make an even order distortion system. This is why a lot of the "small, powerful" subwoofers can be quickly located, which should not be the case, because of audible harmonics above 90 hz or so. Even if most of the distortion is below 90 Hz, if the phase of the harmonics lines up on the cochlea, the CNS manages to help you locate this. (note, this is not like a 'steady tone' at one frequency, which isn't going to 'locate' anything, nor is it a variation in phase across the two ears, which again is audible, but which does not appear to provide directional sensation, only spatial sensation)."

Not much helpful, but anyway this is how my room reacts to different alignments, or shall I say, deliberate misalignments:

Stereo Bass using subwoofers

Um, sound pressure is proportional to f^2 * volume velocity. There's 2 accelerations involved, cone displacement and microphone displacement, reacting to pressure, so from end to end, it's f^2 * displacement. That's why speakers can be flat, after all, the second-order lowpass above resonance...

www.audiosciencereview.com

 

[Heinrich]

I have 4 subs, 2 SVS SB3000 which are 13" small sealed heavily DSP'd subs and 2 very large 18" DIY seal subs with no DSP. I have tried to use these in a 4 sub set up but was never happy with the integration.

The SVS may delay the signal a bit due to the inbuilt DSP compared to the non DSP'ed DIY types. My first question was, what you are after with the 'impulse' graphs? A later comment gave a hint in that you state the extension of the subs towards the mids. It coincides with the graphs. The needles, the sharp rise in amplitude, the peaks of 'impulse' are midband related. Only midband content shape that, and most of all its position on the time line. Due to filtering the mids' timing is not the same as the bass timing.

 

[levimax]

Thank you everyone for your helpful comments and insight. Currently I am happily using just the 2 big DIY subs (co-located with the mains) for stereo music. The reason I was thinking about the SVS subs was because I have been reading and seeing some very interesting results from DIRAC ART users and since I have DIRAC DLBC (which I am not using) I was hoping to give DIRACT ART along with the 2 SVS subs a try when it comes out for PC. Since ART uses cancelation and reinforcement to smooth out low bass it seems like the speakers used have to be somewhat compatible with each other. After testing the SVS subs they seem like they behave very differently than the DIY subs so I am probably going to sell them if I can't get permission to move them into the TV room.

 

[3ll3d00d]

Since ART uses cancelation and reinforcement to smooth out low bass it seems like the speakers used have to be somewhat compatible with each other

I don't see why that's the case but it's a black box sobetter to try it and see what happens

 

[levimax]

I don't see why that's the case but it's a black box sobetter to try it and see what happens

I agree black box so don't know for sure but my thought was in order to accurately "cancel" out some frequencies you would need to have accurate timing. The group delay difference between the 2 subs between 20 Hz and 150 Hz (ART's operating range) is ~80 ms so unless ART calculates a different delay for each frequency for each driver (which it might) it seems like it would not be able to work accurately. I can tell you for sure DLBC had no chance trying to use the different subs together. It worked for either the 2 DIY subs or the 2 SVS subs but when mixed it just went crazy and set delays over 100 ms and every time I would run it gave radically different results. Not surprising really considering how different the time domain behavior is between the 2 subs. I guess I could wait and try it as you say as most reports are ART seems to be more reliable than DLBC but it is such a project to re-do a system and then run DIRAC measurements that I would like to maximise my chances for success. In addition, after listening to large well behaved subs I am also disappointed with the SVS subs in general.

 

[3ll3d00d]

Fwiw I use materially different subs and have no real problem using mso to support each other. I don't bother with exact timing between the two as that is one variable that helps mso do what it can. Basically if you have art, just try it and see what happens would be my advice.

 

[levimax]

As the oscillation in the svs one is clearly in the ~30Hz range so I would guess that's some modal resonance you're seeing and it's either not present in the DIY case (because the position is different) or obscured by the wider bandwidth.

DIY and SVS were very close to same position. I just noticed that Nuyes measured the step response of the SVS SB3000 sub outside and got similar results to what I did measuring nearfield, see below. https://www.audiosciencereview.com/...s/i-measured-10-subwoofers.49042/post-1756732

Maybe I am being paranoid and not understanding but it seems like a terrible result to have ringing that strong and long. I can only imagine what happens in room with it's ringing added to the sub ringing.... maybe that is why so many report not being happy with their sub integration.

 

[DWPress]

I get the same impulse from my SVS indoors or outside. Now I REALLY want to replace it!

NO, stay focused DWP, gotta wait 'til after May obligations but I'll report back!