My thinking is cascading filters introduces phase distortion.

[3db]

The point I'm trying to make is this,, I would think that the filter on the receivers effect the signal first before reaching the filter of the sub itself. So if dialing a cutoff frequency at the receiver say around 80Hz, dialing the sub around 90 Hz will have very little affect as the 90 Hz is already being dealt with by the receiver's low pass stage. Secondly phase becomes an issue because every filter be it low, band, or high pass introduces phase delay (don't want to use the word distortion hear cause I don't know if its technically correct). Unless you have a continoulsy variable phase control on your sub ( which mine does not, just 0 and a 180 degrees) your gonna find it harder to dial the sub in properly.

The kind a fiilter that is being employed as the low pass will also determine how difficult it will be to dial in the sub as well. Some filters have an emphasis on keeping the phase shift to a minimum at the expense of an even amplitude response while others trade off phase for an even amplitude response (Butterworth) .

What are your thoughts?

 

[Halon451]

Try this link, it may help.

http://www.st-andrews.ac.uk/~www_pa/Scots_Guide/experiment/lowpass/lpf.html

It's a pretty cool interactive tool - you can play around with settings in a simple low pass filter circuit and see the input and output waveforms and how it affects phase shift.

 

[Joe Schmoe]

But why do you care?

 

[3db]

But why do you care?

If one has a sub with variable phase control, then you can get around this isue (if its even an issue and thats what I'm trying to find out) . But my sub doesn't have a variable phase control, just 0 and 180 degrees. If what I'm thinking is correct, I may not even use the filter on the receiver (I only have 200Hz, 150Hz, and a 100Hz steps) and use the sub's filter alone to integrate my sub with the rest of the system.

 

[3db]

Try this link, it may help.

http://www.st-andrews.ac.uk/~www_pa/Scots_Guide/experiment/lowpass/lpf.html

It's a pretty cool interactive tool - you can play around with settings in a simple low pass filter circuit and see the input and output waveforms and how it affects phase shift.

Thanks Halon but my I understand phase shift in analog filters. I just don't know if phase shifts occur with digital filters.

 

[WmAx]

Thanks Halon but my I understand phase shift in analog filters. I just don't know if phase shifts occur with digital filters.

Phase shift can or can not happen, depending on the digital filter topology. IIR(infinite impulse response) filters, which are most common in speaker application filters, are homologous to the behavior of it's analog counterpart, including phase rotation. A FIR(finite impulse response) filter can be used to prevent phase rotation. It should be noted that a FIR filter can only work in a simple application where the phase of the signal is always constant between two independent sources. By this, I mean, it can work for a subwoofer where FR/phase is nearly identical regardless of angle to the sources, due to the relatively long wavelengths vs. separation distance and driver surface radiation area. However, used in a conventional multi-way system where you have for example, a mid-bass and a tweeter, the two drivers are large in relation to the overlapping bands of each other, and thus, phase alters appreciably as you examine from different angles off axis. In this case, the FIR filter could only do it's job on a single design axis. Move from this design point off axis and the phase response will be substantially poor. In fact, due to the inherent operation of an FIR filter, you will have a pre-ringing effect off axis, as each driver must cancel the others pre-ring impulse, but this cancellation can only be achieved within a very limited axial window with conventional applications.

-Chris

 

[3db]

Phase shift can or can not happen, depending on the digital filter topology. IIR(infinite impulse response) filters, which are most common in speaker application filters, are homologous to the behavior of it's analog counterpart, including phase rotation. A FIR(finite impulse response) filter can be used to prevent phase rotation. It should be noted that a FIR filter can only work in a simple application where the phase of the signal is always constant between two independent sources. By this, I mean, it can work for a subwoofer where FR/phase is nearly identical regardless of angle to the sources, due to the relatively long wavelengths vs. separation distance and driver surface radiation area. However, used in a conventional multi-way system where you have for example, a mid-bass and a tweeter, the two drivers are large in relation to the overlapping bands of each other, and thus, phase alters appreciably as you examine from different angles off axis. In this case, the FIR filter could only do it's job on a single design axis. Move from this design point off axis and the phase response will be substantially poor. In fact, due to the inherent operation of an FIR filter, you will have a pre-ringing effect off axis, as each driver must cancel the others pre-ring impulse, but this cancellation can only be achieved within a very limited axial window with conventional applications.

-Chris

So I'm assuming that each receiver/pre-amp manufacturer chooses which kind of digital filter they employ or is there an industry norm applied here. Secondly, because the wavelengths are so long, phase appears to be constatnt? If the answer to the 2nd part is yes, then we can ignore the phase delays intorduced (if introduced) by the receiver's filter.

 

[annunaki]

I played around with this briefly last night. There appear or seems to be a decrease in output associated with using two filters. This is what Home Theater magazine had suggested happens when cascaded filters are used. At least near the crossover point. That is what I could tell by ear anyway. It did help to remove more of the sound above the crossover point though as well.

Cascading crossovers cause a number of problems, including a 6-dB loss in bass level. Your preamp should also eliminate bass from the main speakers' amplifier, which will prevent an overabundance of bass where the speakers and sub overlap and, as I mentioned earlier, increase power and decrease distortion. The sub's crossover controls will be useful, however, if your preamp or receiver either doesn't offer a sub output or doesn't filter both the main and sub outputs. In the latter case, make sure your preamp's outputs (whether they're line- or speaker-level) match the sub's inputs.

Article: http://www.hometheatermag.com/subwoofers/3/index2.html

 

[WmAx]

So I'm assuming that each receiver/pre-amp manufacturer chooses which kind of digital filter they employ or is there an industry norm applied here.

As I stated, IIR is usually used, which is homologous to analog filters. FIR is only used in special applications, typically.

Secondly, because the wavelengths are so long, phase appears to be constatnt?

Yes. As for example, a subwoofer could be crossed at 80hz. This is about a 13 or 14 foot wavelength in air. The small (1-2foot) cabinet is not going to cause any substantial attenuation of this wavelength, as it is small in relation to the wave length. The driver is also tiny in relation to the wavelength, so you don't have to be concerned with off axis response differences as a factor here. In other words, the FR vs. phase response would be virtually identical if you measured the subwoofer at 0 degrees, vs. 90 degrees, on a pole 50 feet in the air, away from acoustic interactions of a room/boundary. As long as the subwoofer is within a half wavelength of the main speaker(s), a FIR filter would probably work with no issues.

If the answer to the 2nd part is yes, then we can ignore the phase delays introduced (if introduced) by the receiver's filter.

You can never ignore phase when summing two different filters(the filter on the mains vs. the filter on the sub). But ideally, you would use only the receiver filters and bypass the sub's filter -- if possible, as the receiver has filters specifically intended to sum properly. In some cases, you may still be better off using the sub filter instead -- but you can only determine this if you find this to be true as a result of measured behaviour.

 

[WmAx]

I played around with this briefly last night. There appear or seems to be a decrease in output associated with using two filters.

As any filter is specified at a particular frequency, is actually reduced in level at that specification, of course, two set to the same frequency in cascade would reduce the level at that point further. We will consider an even order L-R(Linkwitz-Riley) topology crossover for examples from this point on. At Fc(crossover point), level is always -6dB. When you combine a low pass function with a high pass function, both are in phase at Fc, thus sum to +6dB, which results in 0dB at Fc. Now, if instead, you cascade two equivalent LP or HP together, then at Fc, it will be -12dB instead of -6dB. It would not be possible for the LP and HP to combine to 0 at Fc in this circumstance, resulting in a dip in the response. If a further roll off is needed beyond Fc, and you want to avoid a response error, you can target the cascaded filter to take effect beyond Fc and it will leave the range near FC unaffected if executed correctly.

-Chris

 

[3db]

You can never ignore phase when summing two different filters(the filter on the mains vs. the filter on the sub). But ideally, you would use only the receiver filters and bypass the sub's filter -- if possible, as the receiver has filters specifically intended to sum properly. In some cases, you may still be better off using the sub filter instead -- but you can only determine this if you find this to be true as a result of measured behaviour.

My receiver has only 3 low pass settings, 200Hz, 150Hz, and a 100Hz. In this case, it would be more advantageousI think to NOT use the low pass setings on my receiver and simply use the low pass filter on the sub. I'm assuming the receiver will send the bass out to the sub whether its filtered or not by virtue that all speakers have been set to small.

 

[WmAx]

My receiver has only 3 low pass settings, 200Hz, 150Hz, and a 100Hz. In this case, it would be more advantageousI think to NOT use the low pass setings on my receiver and simply use the low pass filter on the sub. I'm assuming the receiver will send the bass out to the sub whether its filtered or not by virtue that all speakers have been set to small.

That is unfortunate. Personally, I would sell the receiver and replace it with a better equipped unit -- because you are not going to be able to integrate with the mains properly here unless you have two subs to use in stereo, one near each main, due to the relatively high minimum crossover point. However, if you have pre-outs for all channels, and that receiver has a clean pre-amp section, I would use it to feed outboard crossovers/amplifiers. A far better option for optimizing the system. Another option is to bypass the internal circuits from the internal pre-amp to amplifier sections and install input/output jacks externally so that you can add an outboard crossover to the existing architecture.

-Chris

 

[mtrycrafts]

.. It did help to remove more of the sound above the crossover point though as well.

This can be a compelling reason to use both filters offset as was indicated, to quickly reduce the out of band higher frequencies that may distract by localization of those higher band frequencies to the sub.

 

[annunaki]

As any filter is specified at a particular frequency, is actually reduced in level at that specification, of course, two set to the same frequency in cascade would reduce the level at that point further. We will consider an even order L-R(Linkwitz-Riley) topology crossover for examples from this point on. At Fc(crossover point), level is always -6dB. When you combine a low pass function with a high pass function, both are in phase at Fc, thus sum to +6dB, which results in 0dB at Fc. Now, if instead, you cascade two equivalent LP or HP together, then at Fc, it will be -12dB instead of -6dB. It would not be possible for the LP and HP to combine to 0 at Fc in this circumstance, resulting in a dip in the response. If a further roll off is needed beyond Fc, and you want to avoid a response error, you can target the cascaded filter to take effect beyond Fc and it will leave the range near FC unaffected if executed correctly.

-Chris

Thanks a lot Chris. This explains a lot. So If crossing over in the receiver at 60hz you could activate the sub's crossover at 80hz with no ill effects, correct? This is what I thought. I will continue to use this option.

 

[TLS Guy]

Phase and time

You have asked a complex question, but a timely and important one. First, please understand that filters and drivers have roll off. A first order roll off is 6db per octave. Each order adds 6db of roll off. So a fourth order filter rolls off at 24db per octave. Say you have speakers that roll off at 12db per octave at 60 Hz (second order), then the speaker will be 12db down at 30Hz. Assuming that a filter is set in the region of the cut off of the driver the order of roll off of the driver and filter are additive. Unfortunately manufacturers specs are not adequate or helpful in this regard. However if your speaker has a port, then it will roll off at 24db per octave below the manufacturers stated F3 point frequency. If the speakers are sealed they will roll of at 12db per octave.

Now analog filters have 45 degrees of phase shift per order. Low pass filters (connecting to the woofer) the phase shift lags. In high pass filters the phase shift leads. So if you have a second order roll off on the high and low pass filters, the drivers are 180 degrees out of phase at crossover. So there is cancellation at crossover. Digital crossovers can be made with zero phase and time shifts. Analog filters have the same phase shifts whether active or passive.

Now to the sub woofer question. Ideally you want to splice the roll off of the subs to match the roll off of the main speakers, both in terms of frequency and order. In other words if you have ported speakers, you want the sub to roll off fourth order at the F3 point of the main speakers. This is important so as to not have irregular bass response that can interact in a negative way with room effects.

Now to phase. In the example above if the analog filters are used to the sub, and the roll off is optimal fourth order then the sub will be 180 degrees out of phase with the mains at crossover. There will be highly audible cancellation. I believe many systems to be improperly set up in tis regard. The connections to the other speakers need to be reversed if the phase of the sub can not be reversed.

A word about subs. Some subs use coupled cavity designs. If your sub outputs only via a port, and not from the driver cone directly, you have a variant of a coupled cavity design. Space does not permit me to go into this in detail, but in the crossover regions these designs will have second, third or fourth order acoustic roll off depending on the design. They will also have phase, and time shifts to match. The distortion of these designs is huge and increases hugely with order. In my view only second order coupled cavity designs can give just acceptable performance. I regard the other configurations a low Fi.

Now the point of all this is, that I have come across many installations were phase was not checked in real life. This is a vitally important and usually missed step.

Drive a signal from an oscillator, or CD with sine wave tones, at the crossover frequency of sub to main. Start with the sub turned right down. Then advance the volume of the sub, and see if the sound output increases or decrease as you turn the volume up. If it decreases as you turn up the volume at any point the sub and mains are out of phase, even if every thing is connected correctly, red to red and black to black. Reverse the phase to the sub or main speakers, not both and repeat. This time the sound will gradually increase as you turn up the sub. There will be NO NULL POINT.

I can't stress how important it is to do this. I have an idea, that may be as many as 50% of systems may contain this serious error. If you correct it you won't believe the improvement.

One other point, amps confuse the issue also, as many amps have the input in phase with the output. Many however have the output 180 degrees out of phase with the input. The manual does not usually state whether the output is in phase with the input or not.

When cascading filters, the orders of the filters and phase shifts are additive. I strongly suspect the comment about reduced output, with cascading filters was due to creating an out of phase situation.

As you can see there are opportunities galore to make a serious mistake. Fortunately checking for, and correcting this all too common error is straight forward.

 

[WmAx]

Thanks a lot Chris. This explains a lot. So If crossing over in the receiver at 60hz you could activate the sub's crossover at 80hz with no ill effects, correct? This is what I thought. I will continue to use this option.

Yes, your solution could be viable. But measurement should always be performed to verify.

-Chris

 

[TLS Guy]

Measure if you can

Dear Chris,
I agree measurement is best if you have the equipment. You and I obviously do. I can see the phase changes on my dual channel scope. However for those that can not do this, my method actually works very well in practice.
Those contributors with an array of test equipment will not likely need the advice I gave, it will be second nature. But to many who are not highly technical this issue is confusing and mysterious to them. I tried to explain a moderately complex issue as simply and in as much detail as I thought would cover most peoples situations.
Regards, Mark.