Subwofer's LPF frequency response

[alexk0il]

Your entire setup used for measuring and for listening to music needs more detailed explanations. Same as in that other forum where people don't like to measure things as much as here (according to your perception).

Theory of low-pass (and high-pass) filters has been explained in some detail.

And the theory matches my expectations.

What is still up to debate is how you did perform your measurements

Ignore the measurements. I did the measurements only after my ears detected something suspicious. Read below.

and why you even thinks there's something wrong with your BK electronics XXLS400.

You would expect 200Hz to be get quiter by a noticable amount when you move the cutoff filter from 80Hz to 40 Hz.

My ears tell me it's not the case with my sub, the spl/loudness stays the same when I change the cut off frequency from 80 to 40 Hz.

I didn't trust my ears, so I did the measurements which confirmed what I hear.

Anyway, everyone here is saying that's not a normal, there should be at least 6dB, sine say 12 or 24dB which should be easy to hear.

If you think that I should not hear the difference between 80hz and 40 hz cutoff frequencies, please explain why.

Also, is there any chance you can do a hearing test with your sub and confirm it's behaviour. You don't have to measure, I will trust your ears. Easy to test.

Go to https://onlinetonegenerator.com/ on your phone/pc/mac and stream 200Hz to your sub.

Then rotate the cutoff filter knob and report if the sound coming out of your sub stays the same for at least one octave at lowest frequencies. If not, my sub is broken.

 

[CJPopovich]

Years ago I did a somewhat similar test with a Paradigm X30 (I think) crossover for a sub. I plotted out response with the low pass as high as it would go, and same with it as low as it would go.

It was not the slope specified by the mfg, and it was not stable at different volume (not crossover) adjustments. I took it back, got a replacement, and that tested per spec.

 

[harkpabst]

If you think that I should not hear the difference between 80hz and 40 hz cutoff frequencies, please explain why.

I don't say so. But there are other possible reasons why you might not be able to tell a difference, depending on how you did your listening test and your measurements. One of them is the influence of the listening room, in particular standing waves since 200 Hz is below the Schrödinger frequency of a typical living room. If a standing wave happens to be exited it could have a big influence on perception.

I do not disregard your listening tests completely, though. Sound was different from what you expected and searching for the reason is absolutely valid. But for testing by ear a much better test would e.g. be to just play back white noise and then slowly turn the frequency knob counter-clockwise. If there is still no difference between the 80 Hz and the 40 Hz markings then you can be very sure something is wrong with your sub.

I will happily ignore your measurements as it remains unclear if they have any value at all. When time permits I will make some proper nearfield measurements on one of my XLS200 subs (one is still in use in my second system, the other one is currently collecting dust). Will try a couple of filter settings, maybe also combined with different gain settings. But don't rush me, I have some other things to do in real life.

 

[alexk0il]

I don't say so. But there are other possible reasons why you might not be able to tell a difference, depending on how you did your listening test and your measurements. One of them is the influence of the listening room, in particular standing waves since 200 Hz is below the Schrödinger frequency of a typical living room. If a standing wave happens to be exited it could have a big influence on perception.

Irrelevant for at least two reasons:

• I do the hearing/measurement tests at nearfield distance
• Also, you can add the room gain to the attenuation of the cut-off filter for the same frequency. Meaning, once the measurements are normalized to 0dB at the initial reference point (180Hz in my case) , the chart effectively cancels the room gain for that particular frequency and only shows you the attenuation of the filter.

I do not disregard your listening tests completely, though. Sound was different from what you expected and searching for the reason is absolutely valid. But for testing by ear a much better test would e.g. be to just play back white noise and then slowly turn the frequency knob counter-clockwise. If there is still no difference between the 80 Hz and the 40 Hz markings then you can be very sure something is wrong with your sub.

We already know different frequencies are affected in a different way by the cut off filter, see all three lines in my chart. There is no doubt there will be an audible difference; the result will be inconclusive.

I will happily ignore your measurements as it remains unclear if they have any value at all. When time permits I will make some proper nearfield measurements on one of my XLS200 subs (one is still in use in my second system, the other one is currently collecting dust). Will try a couple of filter settings, maybe also combined with different gain settings. But don't rush me, I have some other things to do in real life.

No problems. I appreciate this.

The funny thing, the sub sounds amazing even with the miss-behaving cut-off filter. It blends with my speakers in an incredibly smooth way, the only way to tell it's there is to turn it off. Plus I get featherlight palpations when the music hits low notes. Though it's a bit more challenging to control it's temper with the movies, I guess this is where I really need to set the correct frequency crossover with that LPF thing.

 

[grogi.giant]

A general comment (despite not understanding the graph), a typical LPF is either 12dB/octave or 24dB octave.

So if the crossover is at 60hz, you would expect 120hz to be either 12dB or 24dB down, and you'd expect 240hz to be either 24dB or 48dB down.

1. Crossover is usually already at -3dB... So 120Hz would be at -15dB or -27dB...
2. THX requires LPF (for the Subwoofer) to be 24dB/octave and HPF (smaller speakers) to be 12dB/octave.

 

[TurtlePaul]

1. you graph is confusing.

2. If you only tested at 70 dB or so, I think room noise is the reason for the leveling off at -15 to -20 dB.

 

[TurtlePaul]

Also, confirm your SPL measurement device is unweighted and not a-weighted, c-weighted or any other non-linear setting.

 

[TurtlePaul]

To put things in context, my A/C is on and my room measures 55 dB background noise. A/C off and close the door and I can get down to 40 dB. You cant reliably measure within 6 dB of the noise so upper 40s is the quietest valid measurement in my room.

To measure a 50 dB drop from the reference level I would need to test at 96 dB like Amir and Erin do. They can hear their Klippel test tones throughout their home. 96 dB is concert level (not right in front of speaker). It is nightclub shouting to hear the person next to you level. It is loud as a lawnmower level. It is OSHA hearing damage after a few hours level. It is apartment neighbors call the cops level.

If you want to measure -40 to -60 dB, you need these levels. If you cant do this then you should try close mic techniques - every halving of distance should get ~+6 dB of signal to noise.

 

[grogi.giant]

Got a new BK XXLS400 sub, still learning how to set it up in my room. Anyway, found a strange behavior of the frequency filter of the sub while setting the crossover with my speakers.


I am generating 200, 100 and 60Hz audio signals on my phone and stream them to my AMP. I measure then the sub output attenuation vs sub's LPF settings.
Here is my setup:
Android frequency generator app -> BT LDAC receiver -> DAC -> Analogue Amp.

Here are the results:
View attachment 294703
My concerns are as following:

• The total attenuation range is about 20dB for 200Hz, and only 11dB for 60Hz. I'd expect it to be in the range above 40-60dB at least
• The 60Hz signal is attenuated by -3db at ~110-120Hz. I'd expect to see the LPF to kick in around 60dB
• 200Hz signal is still leaking at 40Hz. In fact, there is no difference after the 80Hz filtering, the attenuation doesn't change in the 80-40Hz area

Does it look normal and I need to readjust my expectations, or do my concerns have some merit?

Thanks

While I understand it, please pivot that chart... It will follow the very common convention and be much easier to what you're trying to show:

• X axis - frequency of the test signal
• Y axis - attenuation
• data series - various LPF settings, exp. one line for 40Hz, 60Hz, 80Hz, 100Hz, 150Hz and 180Hz.

 

[alexk0il]

While I understand it, please pivot that chart... It will follow the very common convention and be much easier to what you're trying to show:

• X axis - frequency of the test signal

Can't do. Don't have a calibrated mic to measure a meaningful frequency response chart. However I can meaningfully measure how uncalibrated value at a certain frequency is attenuated by a particular setting of the LPF.

• Y axis - attenuation
• data series - various LPF settings, exp. one line for 40Hz, 60Hz, 80Hz, 100Hz, 150Hz and 180Hz.

 

[alexk0il]

These are my theoretically calculated curves with second order Linkwitz-Riley (LR2) and fourth order (LR4) using your format. You can see the 6 dB attenuation when the signal frequency matched the cutoff. Also, for the 200 Hz and 100 Hz curve, at 1 octave below cutoff (i.e. 100 Hz and 50 Hz, respectively), the attenuation is -14 dB for the LR2 and -25 dB for the LR4. (These numbers are calculated exactly using their transfer functions, and at 1 octave below cutoff are a little higher than the nominal approximations of -12 dB/octave and -24 dB/octave.)

These curves are very different from your measurements. When we comparing attenuations, room response shouldn't matter if the sub and the mic weren't moved.

View attachment 294738

View attachment 294740

OP here, here is an update.

Got a response from the manufacturer. Looks like my measurements are kind of aligned with their implementation, except for #2 below.

1. They actually have two filters:

• a variable filter (controlled by the dial) with 12dB / Oct
• a fixed filter with 12dB octave

Depending on the dial position the effective attenuation is either 24db or 12db. Never zero, I think, still clarifying.

2. The steps on the filter dial are not linear. Close to log but not. So the values on my X- axis are incorrect which explains why almost no audible difference for the last few steps on the dial, the cut of frequency there is almost the same.

Just curious, what tool did you use to generate your charts? I'd like to model the new two filters combo, or maybe you have time and curiosity to do so?

Thank you for your help anyway, I really appreciate it.

 

[NTK]

OP here, here is an update.

Got a response from the manufacturer. Looks like my measurements are kind of aligned with their implementation, except for #2 below.

1. They actually have two filters:

• a variable filter (controlled by the dial) with 12dB / Oct
• a fixed filter with 12dB octave

Depending on the dial position the effective attenuation is either 24db or 12db. Never zero, I think, still clarifying.

2. The steps on the filter dial are not linear. Close to log but not. So the values on my X- axis are incorrect which explains why almost no audible difference for the last few steps on the dial, the cut of frequency there is almost the same.

Just curious, what tool did you use to generate your charts? I'd like to model the new two filters combo, or maybe you have time and curiosity to do so?

Thank you for your help anyway, I really appreciate it.

I use Mathematica to generate the plots. Attached is a PDF of the Mathematica notebook. Mathematica is free (for non-commercial use) for Raspberry Pi's.

Let me know what changes to my plots you'd like to see. I'd say most "simple" changes to how the data are presented should be relatively painless and quick to do.

 

[grogi.giant]

Can't do. Don't have a calibrated mic to measure a meaningful frequency response chart. However I can meaningfully measure how uncalibrated value at a certain frequency is attenuated by a particular setting of the LPF.

Well, I did it for you then... It shows what you wanted to show - at around 100Hz the filter set to ~40Hz, ~55Hz or ~65Hz flattens out at around -18dB and don't do any more attenuation.

 

[alexk0il]

I use Mathematica to generate the plots. Attached is a PDF of the Mathematica notebook. Mathematica is free (for non-commercial use) for Raspberry Pi's.

Let me know what changes to my plots you'd like to see. I'd say most "simple" changes to how the data are presented should be relatively painless and quick to do.

Oh... The Wolfram research Mathematica! Hasn't touched it since late 90s. Good to see it still being around.

I just gave it another thought and came to conclusion your charts are as good as they are, no need to change anything.

My first thinking was to ask for your chart with the two cascaded filters, one with a fixed fc, one with variable fc. Then I realized that the fixed filter will only move the lines up and down on the chart and won't change anything fundamentally, why would it if the fc is fixed. I'm slow but I eventually get it.

Anyway, if I need to model it I will use your code as a reference/inspiration for Matlab/Octave.

Thanks a lot.

 

[NTK]

Oh... The Wolfram research Mathematica! Hasn't touched it since late 90s. Good to see it still being around.

I just gave it another thought and came to conclusion your charts are as good as they are, no need to change anything.

My first thinking was to ask for your chart with the two cascaded filters, one with a fixed fc, one with variable fc. Then I realized that the fixed filter will only move the lines up and down on the chart and won't change anything fundamentally, why would it if the fc is fixed. I'm slow but I eventually get it.

Anyway, if I need to model it I will use your code as a reference/inspiration for Matlab/Octave.

Thanks a lot.

Simulating 2 serially cascaded filters shouldn't be too difficult. For two 2nd order (12 dB/oct slope) low pass filters that mean the FR curve will start flat, change to sloping down by 12 dB/oct when it reaches the first (lower) cutoff, and then "accelerates" to 24 dB/oct as it reaches the second (higher) cutoff.

Can you get from your sub manufacturer the cutoff frequency of the fixed filter? I think some subs have fixed LPF at the high end (e.g. ~180 Hz or above) for protection, but with the cutoff frequencies being that high, they shouldn't have much effect on the typical frequency range the subs operate at (<80 or 100 Hz).

 

[alexk0il]

Simulating 2 serially cascaded filters shouldn't be too difficult. For two 2nd order (12 dB/oct slope) low pass filters that mean the FR curve will start flat, change to sloping down by 12 dB/oct when it reaches the first (lower) cutoff, and then "accelerates" to 24 dB/oct as it reaches the second (higher) cutoff.

Can you get from your sub manufacturer the cutoff frequency of the fixed filter? I think some subs have fixed LPF at the high end (e.g. ~180 Hz or above) for protection, but with the cutoff frequencies being that high, they shouldn't have much effect on the typical frequency range the subs operate at (<80 or 100 Hz).

They stopped responding to my emails after I asked for more info about the fixed filter , Don;t think a real value is important for the variable attenuation. You can take 200Hz, it should be just fine.

Anyway, they gave me three points on how the filter dial settings are roughly related to the actual cut off frequencies. Here is my original chart with the corrected X-axis (I used Piecewise cubic Hermite interpolation to calculate the values between these three ports, worked better than Lagrange for this set). The X- axis reference points I was given resulted in 15db/octave instead of 12db, so I did a bit of "reverse engineering" to find the values that give the 12db/octave as the manufacturer specified. It's still not super accurate in terms of the -3dB crossover (a good match for the 60Hz signal though), but it's good enough as a first approximation and for the inaccuracies of my measuring system.

Perhaps you want to try to plot these measurements on top of your charts (realigning the starting point to zero though) to see how they match. Please note that the max cut off frequency is 120Hz, not 180.

 

[NTK]

They stopped responding to my emails after I asked for more info about the fixed filter , Don;t think a real value is important for the variable attenuation. You can take 200Hz, it should be just fine.

Anyway, they gave me three points on how the filter dial settings are roughly related to the actual cut off frequencies. Here is my original chart with the corrected X-axis (I used Piecewise cubic Hermite interpolation to calculate the values between these three ports, worked better than Lagrange for this set). The X- axis reference points I was given resulted in 15db/octave instead of 12db, so I did a bit of "reverse engineering" to find the values that give the 12db/octave as the manufacturer specified. It's still not super accurate in terms of the -3dB crossover (a good match for the 60Hz signal though), but it's good enough as a first approximation and for the inaccuracies of my measuring system.

Perhaps you want to try to plot these measurements on top of your charts (realigning the starting point to zero though) to see how they match. Please note that the max cut off frequency is 120Hz, not 180.


View attachment 295641

View attachment 295649

Here are my calculations using serial cascading filters. To show their basic frequency responses, below are plots of a single LR2 LPF with Fc @ 200 Hz (blue), Fc @ 40 Hz (gold), and both in a serial cascade (green).

These are the FR of a series of these filters. The top one is a single LR2 with Fc = 200 Hz (the "fixed filter"). The other curvess are for the fixed filter in combination with another LR2 LPF with the indicated Fc.

This is the simulation of your measurements (I also overlaid your data) with the fixed filter (LR2 Fc = 200 Hz) and an another LR2 LPF with Fc indicated by the horizontal axis. I adjusted the vertical positions of the curves so that their "departing points" are the same as your measurements.

 

[alexk0il]

This is the simulation of your measurements (I also overlaid your data) with the fixed filter (LR2 Fc = 200 Hz) and an another LR2 LPF with Fc indicated by the horizontal axis. I adjusted the vertical positions of the curves so that their "departing points" are the same as your measurements.

View attachment 295734

Thanks a lot, looks very interesting. I would actually expect the theoretical lines to be more aligned with the 12dB/Octave of an ideal LR2 filter. You blue line shows about -8dB for blue and red lines, and -6dB for the yellow. 7.3db average, +/-1.15 std.

Unless I missing something very basic (which is of course plausible, I'm rusty at filter's theory), it's been a while. To my defense, my measured lines look exactly as I would expect in that respect; it's 14dB for blue and 12dB for the rest., i.e. 12.7 average, +/-1.15 dB std.

Not sure if I want to waste more of your time on this, it's now just a mental exercise for me, and I really appreciate all your help.