Measurement Length Changes Impulse Response? And a Sub Integration Question

[jtalden]

I was busy today. I will explain my process in more detail tomorrow.

I understood you adjusted the mains delay. I normally analyze the delay needed using the SW as the variable. I do this as because the phase is more easily read when 0 ms is defined at the high frequency driver impulse. More delay on the SW is the same as reducing delay on the main. Which one to actually adjust is dependent on the system setup.
I normally measure only L, R, SW, then set 0 ms at the tweeter Impulse, then copy the SW measurement. That way I can move SW-Copy delay within REW until I find the best delay. The change in SW delay needed is then found by measuring ms distance change between the SW Impulse and the shifted SW-Copy impulse.

Your data was already provided in a different format; taking several different delays of the main to compare to the SW. I thus changed my method. I decided to see if any of the delays your provided were appropriate. Several were obviously wrong so I didn't consider them. I first check out your 2 x 7 choice and found that it was a good alignment. I then checked out 2 x 9 because the impulses were located such that the initial rise was aligned. This is the normal recommended starting point for the fine a tuning process. In this case no fine tuning was needed. The phase tracking of the direct sound was as good as it was going to get.

If you don't mind, tomorrow I will detail the process I would normally use to find delay timing rather that what I actually did here. That way I will be less likely to get confused in my explaination. I will use this same data.

 

[Ultrasonic]

The first applies the windows and reference time of the selected measurement to all the other measurements, the second leaves the reference times for the other measurements as they are. Don't change the reference times.

So with the second option it will apply a window centered about the peak of the impulse response for each separate file? I'm just trying to be certain about distinguishing between the Window Ref Time at the impulse response peak vs the time of the acoustic reference used when making the measurement.

 

[Ultrasonic]

I understood you adjusted the mains delay. I normally analyze the delay needed using the SW as the variable. I do this as because the phase is more easily read when 0 ms is defined at the high frequency driver impulse. More delay on the SW is the same as reducing delay on the main. Which one to actually adjust is dependent on the system setup.

Given that the main speakers are closer to me than the subwoofer (by 3 foot or so) it makes sense to me that I should delay the mains to try to get the first arrival from both the mains and sub to match? Delaying the sub would always increase the first arrival time discrepancy.

What am I missing?

 

[Ultrasonic]

Checking back through my data I think I probably made a mistake in achieving the above result, in applying the delay to the speaker generating the acoustic reference as well as that being measured (doh!) Otherwise the black and red curves here should be showing a 3.6 ms delay, and there should be some delay to the sub signal.

For the later set with multiple delay times I kept zero delay time for the acoustic reference throughout, and here a delay of closer to 8 ms looks to get the first arrival times close (going by the impulse responses) and this goes along with what you were saying above @jtalden .

 

[jtalden]

It's good you resolved the issue. I was not able to see why there was confusion, but did not understand the process you used.
Keeping the reference time consistent and keeping track of the delay used for the measurements is key. The analysis then identifies the delay change needed to align the phase.

 

[Ultrasonic]

It's good you resolved the issue. I was not able to see why there was confusion, but did not understand the process you used.

No way you could have predicted my muppetry, which is why I thought I'd post to explain .

The bit that does confuse me from the impulse responses though is that the 2+9 combination seems to have a positive going first half-cycle for the main speaker as compared to a negative going first half-cycle for the sub. Which would suggest a mismatch? I guess what is critical is the relative phase at the crossover frequency, which the phase analysis shows to match around 100 Hz, and correspondingly the summed SPL transitions well with this combination.

I do still think the delay should be applied to the mains and not the sub though. Although if with your method you still start around the point where the impulse responses seem to coincide I must be missing something, as they surely never would by delaying the subs signal. Unless you're talking about a negative 'delay' I guess?

 

[jtalden]

Yes, the impulse and more importantly the step response suggest the Sub is a negative polarity. It may not be however as any steep high/low pass filters on the drivers rotate the shape of the impulse and can mislead us. One way to be absolutely sure on a woofer is to use a 1.5 V battery (outward cone displacement is positive). I never tried this on a powered sub however that may have a DC filter. Another way that should work reliability is to sweep only the mid bandpass range of the Sub (50-70 Hz?). Then the Impulse and step responses are not as influenced by any steep rolloff/filters. Regardless, you are correct it is the relative phase tracking that we targeting. Positive or negative polarity is not the target.

Yes, my analysis method may result in a suggested negative delay change to the Sub. I often have no visibility to the setup or current delay settings applied to the Sub and Main as the data is often provided without that info. If the analysis indicates a reduction is needed to the Sub delay and there is not enough current delay in that channel then all Mains delays need to be increased. In this case I thought that indicating the 2 + 7 and 1 + 9 timings worked and thus would make it clear that you could just set those same settings - whatever they were. Any other delay settings that provides the same relative offset between the 2 drivers will work as well - users choice.

It may be more clear when you see my analysis process.

 

[Ultrasonic]

Yes, my analysis method may result in a suggested negative delay change to the Sub.

My point was more that, provided the mains speakers are closer to the listener, I can't think of any case where this wouldn't be the case. In that whilst you could get things in phase by increasing the delay to the sub, it would always increase the first arrival time discrepancy, which for equivalent phase matching you would surely want to avoid.

What I suspect I may be missing is some feature of how AV amps are typically configured, which is something I know absolutely nothing about as I've never used one. From my perspective above you would always start from a situation where there is zero delay to the mains, with the first arrival for the sub being at some later time due to it's more distant position.

This isn't important. I'm just explaining why to me the approach of choosing to work on the basis of finding an optimal sub delay seems an odd way to go about things.

 

[Ultrasonic]

On reflection possibly it's just a question that I'm thinking too much in terms of prioritising first arrival times, and that for situations where priority is given to the steady-state phase then ending up with the sub signal arriving.

Edit: on further reflection, no it isn't. LOL! I need to get to bed .

 

[jtalden]

The analysis finds the recommended change in relative delay - not the absolute sub delay value. Because the process dictates the lower frequency driver is the one that is moved in the analysis it is normally how I express the change - more or less delay on the Sub. It is sometimes confusing to the OP. When I am provided actual settings of the delays as Main = 10 ms and Sub = 0 ms. I would normally indicate the actual change in values needed as maybe Main = 10 ms and Sub = 0 ms. In this case I was a little confused as to exactly how the test was done given the large numbers for the mains. I wasn't sure if I should used those as actual delay settings. I instead just referred to the actual measurements by number (1 + 9 and 2 + 7) assuming you knew the delay settings you used for them. The sub delay comment that must have confused you was only an observation regarding the difference in relative timing between the 2 options. It was not direction on how to make an adjustment as I don't know your current settings.

[If you follow my procedure and end out making several Main impulse vs Sub-copy impulse adjustments it will quickly become clear why moving upper frequency driver in the analysis complicates the process greatly. The t=0 needs to be shifted for any significant change to the higher frequency driver. I'm all for streamlining the process.]

 

[jtalden]

Here's the process detail:

 

[Ultrasonic]

The sub delay comment that must have confused you was only an observation regarding the difference in relative timing between the 2 options. It was not direction on how to make an adjustment as I don't know your current settings.

Apologies, I obviously didn't make myself clear. I wasn't actually referring to being confused by your comments on my specific data. I was just commenting on the overall logic of thinking in terms of a delay to the sub vs the main as a standard starting point.

Thank you very much for the detailed PDF you have uploaded. I really appreciate you taking the time to go through my specific data like that. I'm too busy to properly go through this right now but I will later, although possibly not till the weekend, when I can also make some more measurements.

 

[Ultrasonic]

I was sufficiently curious I just had a read through of the process pdf in a coffee break at work...

That makes sense. As I'm sure was obvious the main difference in what you did to what I did is that you are (intelligently!) talking about trying different delays by changing this in the software based on a single measurement, whilst I was doing the same by actually making measurements with different delays and then comparing these.

You have also confirmed that I was right in my hunch that we should use a consistent timing reference for all of the applied FDWs. I'll have another look at my data now I know this is the case.

The question about whether the sub output should be inverted or not is the one I'd like to resolve. I will try the limited bandwidth measurement you suggested but I note you thought the low frequency phase makes it look like my original interpretation based on the impulse response was correct i.e. it should be inverted.


For clarity I'll restate that the crossovers used were LR 48 dB/octave filters centered around 100 Hz.

For the benefit of others who may read this to learn for themselves I'll also just correct a single point from step 1. of your document. The main and sub measurements were not full-range, but rather were measured with the crossovers applied.

Thanks again .

 

[John Mulcahy]

"Full range" in that context likely refers to the measurement span. Ideally measurements should extend well beyond the working range of the device being measured, otherwise the impulse response is affected by the truncated measurement range and has artefacts that are not part of the response of the device.

 

[Ultrasonic]

"Full range" in that context likely refers to the measurement span. Ideally measurements should extend well beyond the working range of the device being measured, otherwise the impulse response is affected by the truncated measurement range and has artefacts that are not part of the response of the device.

That makes sense. Yes, the frequency range of the sweeps was the same for both (probably 20 Hz to 20 kHz).

 

[Ultrasonic]

Thinking about this, might not the phase difference at low frequencies relate to a phase shift introduced by the HP filter on the main speaker, and so not actually tell us that the frequency components below the crossover (from the sub) are of opposite polarity to those above the crossover (from the main speaker)? With a quick Google I've been unable to find what the phase change through the LR crossover I used would be, but I'll check this out when I next do some measurements (comparing a measurement with and without the crossover applied).

 

[jtalden]

To confirm the absolute polarity of the sub with another measurement, I suggest:
> Disable all filters and XOs set either in the Sub itself or in the MiniDSP.
> Place the mic very close to the sub dustcap (maybe 1 - 2 cm).
> Set the sweep level down accordingly so the SPL is reasonable (maybe 65 - 80 dB)
> Sweep the Sub 30 - 500 Hz

That will provide a very clean Impulse response and wider bandpass without influence of the filters. The step response and phase response will then clearly indicate the absolute polarity.

 

[jtalden]

There is lots of info available on XO filter characteristics as: http://www.frazierspeakers.com/download/cross.pdf
To see almost any combination of high pass and low pass filters we can use rePhase.exe to create the 2 filters and export there resulting impulse responses. Those 2 files can be imported into REW and summed to see the XO characteristics. The 2 filters can be different types, different slopes, and different frequency if we like. In REW we can also change the relative timing to investigate the impact of different delays. This example file is just a 1k high pass and a low pass LR-96 that I summed using correct timing.

 

[Ultrasonic]

To confirm the absolute polarity of the sub with another measurement, I suggest:
> Disable all filters and XOs set either in the Sub itself or in the MiniDSP.
> Place the mic very close to the sub dustcap (maybe 1 - 2 cm).
> Set the sweep level down accordingly so the SPL is reasonable (maybe 65 - 80 dB)
> Sweep the Sub 30 - 500 Hz

That will provide a very clean Impulse response and wider bandpass without influence of the filters. The step response and phase response will then clearly indicate the absolute polarity.

Thanks. I'll give that a try at the weekend.

Would there be an benefit from placing the sub in the middle of the room for this test? It's a downfiring sub by the way, but I'll pop it on its side for the test.

 

[jtalden]

1. Mid Room would work. It shouldn't be necessary though with the mic real close.
2. It will likely work to just place the mic on the floor under the upright sub if you can get it under there.
3. If the Sub is close to a wall and you can lay the mic at the floor/wall intersection (baseboard) so it's closest to the nearby Sub. That should work out well as there can be no floor/wall reflections that way.
4. Placing the mic at a corner of the room that is very near the sub may work as well.

Try the easiest way and look a SPL/phase response. All is good if they are smoothly shifting through most all that range. Some rolloff or breakup at the higher frequency end (>300) is probably not a problem and the sub may have a lowpass filter up there that can't be disabled. We just want enough of the bandpass range to be as clean as possible. We can then adjust the impulse location as needed to make the phase of the bandpass frequencies horizontal on the chart at 0° or 180°. The one that is the flattest is the polarity of the driver. The step response should also be a reliable indicator.

 

[Ultrasonic]

1. Mid Room would work. It shouldn't be necessary though with the mic real close.
2. It will likely work to just place the mic on the floor under the upright sub if you can get it under there.
3. If the Sub is close to a wall and you can lay the mic at the floor/wall intersection (baseboard) so it's closest to the nearby Sub. That should work out well as there can be no floor/wall reflections that way.
4. Placing the mic at a corner of the room that is very near the sub may work as well.

Try the easiest way and look a SPL/phase response. All is good if they are smoothly shifting through most all that range. Some rolloff or breakup at the higher frequency end (>300) is probably not a problem and the sub may have a lowpass filter up there that can't be disabled. We just want enough of the bandpass range to be as clean as possible. We can then adjust the impulse location as needed to make the phase of the bandpass frequencies horizontal on the chart at 0° or 180°. The one that is the flattest is the polarity of the driver. The step response should also be a reliable indicator.

The sub is in a corner.

 

[Ultrasonic]

I've realised there is a simple way to avoid any iterative adjustment of time delays to find the best phase agreement at the crossover frequency. (For a particular time-window anyway: a balance will still need to be considered between early arrival and steady-state conditions.)

I'll show this for a quick example, aligning the right main (R) and Sub in the attached mdat file data.

1) Use the overlays view to compare the phases for the two measurements you want to phase align.
2) Measure the phase difference at the crossover frequency. For this example this is 100 Hz, and as shown below the phase difference is 16.5°, with the sub being ahead of the right main. (It says 16.74° on the image but I zoomed in to be more accurate.)

3) The time shift (dt) associated with any phase difference (dPhs) at a frequency f is then given by dt = dPhs/(360*f). In this example a delay to the sub of 16.5/(360*100) s = 0.46 ms will ensure alignment at 100 Hz.

4) Applying a 0.46 ms delay to the sub signal will provide phase agreement at 100 Hz. This 0.46 ms shift is just the smallest time shift that will achieve matching though, with this time plus or minus any integer number of periods of the crossover frequency achieving the same phase match (at the crossover frequency). This period is 1/f, which as 100 Hz is 10 ms. So the same agreement would occur by delaying the sub 10.46 ms, 20.46 ms etc. More relevantly advancing the sub by 9.54 ms (0.46 - 10) will also ensure a match, which is the same as delaying the main speaker by this amount. Inspection of the impulse responses shows that such a delay will give the best agreement in terms of first arrival times and so is going to be the most appropriate shift. Below are impulse response and phase plots showing this:

I'll note I did actually get to this 9.5 ms result by my trial and error approach but it took rather longer!

6) If we then add in the consideration of using a FDW to distinguish early arrivals from steady-state conditions we can quickly see how much the phase difference changes with the window length and use this to come up with a time delay range to consider.

I hope this makes sense?


(I'll update more re. my specific situation later. I've not made the nearfield measurements on the sub yet.)

 

[jtalden]

Yes, perfect sense.

 

[Ultrasonic]

Quick update.

Below is a comparison of nearfield impulse response measurements of my main speaker mid-bass driver (blue) and subwoofer (red), for sweeps from 30-500 Hz. So a pretty clear confirmation that the sub output is of opposite polarity to the main speakers I think. I also took a series of measurements of different low-pass filters on the sub just to check that these didn't invert the signal, but they didn't, rather just increasing the delay. (This is probably obvious but I wanted to be sure.)

My gut reaction is that I should therefore run the sub with an inverted signal so that they are of the same polarity as the mains, but perhaps this is too simplistic?

In terms of working on the wider integration I do have rather different phase for the left and right channels for the 100 Hz crossover frequency I was initially trying, meaning whatever I go with is a compromise between the two. I'm going to investigate different speaker positions a bit more with this in mind to see if I can help the situation, as well as different crossovers.

In terms of compromising between steady-state and early arrival (short EDW window) phase agreement I'm inclined to favour the former, in the same way that at low frequencies it is the steady-state amplitude response that is usually 'corrected' to achieve better sound. Does this sound sensible?

 

[Ultrasonic]

Just thought I'd add that below is how the phase of the left and right mains compares in their current positions (without EDW). Essentially worst over the 80-100 Hz range I hand in mind for the crossover . There is a region of much better agreement from 105-130 Hz but I'm assuming this will be too high in terms of having issues with the sub being localisable. I will try it out of curiosity though, probably in combination with moving the sub to a more central position.

Definitely a work in progress.