2nd-order filter phase confusion

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

stickjam

Well-known member
Joined
Jun 17, 2004
Messages
325
Location
Grand Rapids MI
Please help me understand what this article is saying...

Above the first figure, it states that "The low pass section is inverted, since 12dB crossovers always invert the phase of one signal. If desired, the high pass section may be inverted instead." I guess I'm not terribly hip to filter theory, because I'm confused. Why is the high-pass "in phase" and the low-pass "inverted?"

Thanks

--Bob
 
it's just part of the world of physics

active or passive
intended or un-intended ...
digital or analog

a 2nd order fiter will end with a polarity flip
yes maths can explain it

a web site was posted here within the last couple of days ... about AC circuits
It had some FLASH demonstrations that might help point the finger
not specifically directed at a Bessel or butterworth alignment but does show how caps can lag in phase response

:shock:

err

how deep do you want this to go ??

A PRR sort of thing or a superfitial Kev style, how it effects the design of passive or active speaker systems ...
:roll:
 
As little math as possible please, let's start "Kev Style." :wink:

I thought I sorta understood that 2nd order filters invert phase, but what threw me into confusion was the statement that one 2nd order filter (the lowpass) flips it and the 2nd-order highpass doesn't.

Thanks :thumb:
 
A second order filter will have phase 180 degrees out at frequencies well above the break point of the filter. A high pass will generate phase lead, and a low pass will generate phase lag on frequencies above the break point.

There is no 'inversion' of phase per se, as it is variable with frequency. It's not the same as flipping polarity. Figure 3 on the site puts the high pass output through an inverting amp....not sure why though as this will just put the highpass output at -180 deg phase instead of +180 deg.

Cheers,

Kris
 
[quote author="stickjam"] ... , but what threw me into confusion was the statement that one 2nd order filter (the lowpass) flips it and the 2nd-order highpass doesn't.[/quote]
:roll:
ahh
mmm

show me the spot ... paragraph

I need to read it in context
you may already understand and Rob's wording may have thrown you
 
Kris is correct

this is one case where I should not have use the term polarity flip ...
I was thinking the speaker terminals

the RC combination is an area that phase does change ... 90 and 180 are specific cases

Solutions wise ... the speaker flip is all you have
BUT
through the crossover point things are so very complicated
AND the phase response of the speaker driver helps to stuff it all up
text book LR filters just don't work

I'm not saying that it's not a fast easy solution to PA sound ... just aying it's not the end of the story

the combined active filter - passive filter - speaker - Accustic output ... should be LR

and then that's not the only solution

I foergot to mention the relative speaker mounting point and propergation pattern.
:roll:

it's an art
 
Is there a better solution to bring them back into phase?

Looping back to my application: a 5.1 bass management unit, if I put 2nd-order 80Hz highpasses on the five channels and buffer the highpass inputs into a summing amp into a 4th-order 80Hz lowpass for the sub (for non LFE-channel monitoring of course) what, if anything will I need to do to fix the phase for the sub? (considering that the summing amp will invert the signal too.

I'm sure I'll need to deal with this considering that I want to include the ability to switch the sub off the lowpass to discrete LFE-channel monitoring. I'd rather not have to adjust the phase control on the sub every time I switch.

--Bob
 
you could be over thinking this

start again
what you do want to do

many of us run multiple monitoring systems for exactly this reason
 
Basically this...



I'm thinking of a switch instead of the last summing. I'm still looking for a suitably steep filter for the 110Hz lowpass on the LFE. There don't seem to be many 7th-order elliptical lowpass schematics laying aroung online. :roll:
 
can't see the picture

why so step on the low pass ?

you can't expect Laboratory perfomance on the left/rights to mono OR stereo sub in a small room and NON time aligned Main to Sub systems

I have produced very usable results from just turning the sub towards the wall or floor.

I'd need to know more about what standard you are aiming at ...
a workable solution
OR
Skywalker Sound .. THX ... Dolby licence etc
 
The pictures on the bottom of this page: http://www.5dot1.com/articles/bass_management.html

The lowpass is a way to audition what the brick wall filter on a DTS or Dolby AC-3 encoder will do to the LFE mix without having to go through the hassle of encoding it first.

I'm going for the "might as well do it as close to right as I can since I'm doing it myself unless it's gonna cost an arm and a leg" standard. :wink:
 
can't see picture

You don't have permission to access /articles/bass_management.html on this server.

but I kind of get where you are comming from with the
" without having to go through the hassle of encoding it first. "
 
I've gotta tuck the offspring into bed for the night. I'll post a diagram of what I've come up with once I get it off paper and into the computer.

In the meantime, if anyone has any great really steep 110Hz lowpass filter schematics, elliptical or otherwise, please pass them on with great gratitude.

Thanks
--Bob
 
[quote author="stickjam"]The lowpass is a way to audition what the brick wall filter on a DTS or Dolby AC-3 encoder will do to the LFE mix without having to go through the hassle of encoding it first.[/quote]
My dts encoder doesn't have any low-pass. I assume all are the same, though we have a different encoder at work.

If I send 6 full-range channels into my encoder, I can recover 6 full-range outputs. I know, I've done it!

Just FYI.

And I'm inclined to agree with Kev, you might be over-thinking this a bit.

A couple of hints that might help: You can't flip phase. You can flip polarity. You can't shift polarity, you can shift phase.

Most filters shift phase, but people write about ohase inversions sometimes when they mean thtat the waveforem is inverted at a particular frequency only. Be very careful when reading about this, Linkwitz-Reilly crossovers are primarily defined by their phase response at the crossover frequency. You're building a crossover, you're primarily concerned with summing at the crossover froequency but not only that one.

It's tricky. Some common designe invert polarity, but you can always re-invert. Inversion is easy. Polarity is easy. So is listening. The tough part is development.

Keith
 
[quote author="stickjam"]I've gotta tuck the offspring into bed for the night. I'll post a diagram of what I've come up with once I get it off paper and into the computer.

In the meantime, if anyone has any great really steep 110Hz lowpass filter schematics, elliptical or otherwise, please pass them on with great gratitude.

Thanks
--Bob[/quote]

Check out figure 21 here:

http://www.analog.com/UploadedFiles/Data_Sheets/70859332AD713_c.pdf

You just need to scale it in frequency appropriately (caps and resistors larger to move the rolloff from the [roughly it looks like] 25kHz to 110 Hz. I would pick cap values to some convenient multiple of what is shown and then adjust the R's in the appropriate direction for the exact amount needed.
 
My dts encoder doesn't have any low-pass. I assume all are the same, though we have a different encoder at work.
I stand corrected--it's only Dolby AC-3 that uses the brickwall filter on the LFE.

Check out figure 21 here:

http://www.analog.com/UploadedFiles/Data_Sheets/70859332AD713_c.pdf
You just need to scale it in frequency appropriately (caps and resistors larger to move the rolloff from the [roughly it looks like] 25kHz to 110 Hz. I would pick cap values to some convenient multiple of what is shown and then adjust the R's in the appropriate direction for the exact amount needed.
Interesting, but they're pretty sparse on exactly which components need to be scaled to change the frequency.

I notice the M**kie HRS120 sub features the same stuff I'd like to put on the LFE channel of this box, including that 7-pole elliptical. Anyone have a schematic of one of those? :twisted:

-Bob
 
[quote author="stickjam"]Interesting, but they're pretty sparse on exactly which components need to be scaled to change the frequency.[/quote]

The circuit is obviously a buffer in & a buffer out (A1 & B4).

All the 4700pF and the associated resistors would look like a good starting point.

Keith
 
Keith is correct. Consider the stuff in the middle, from the output of A1 to the input pin of B4. You want to go down in freq by a factor of about 25kHz/110Hz = 227---again I am eyeballing the freq response graph. The parts that don't need to change are the 1k's in the middle of the vertical stacks of Rs and Cs---they are isolated from interaction with the other stuff by virtue of being connected to the op amp outputs.

If you left everything the same except for the caps you would change them to 1.068uF. This is probably a bit brutal and takes up a lot of board space for decent quality parts, so let's go to 100nF parts. This takes care of a factor of 47, leaving a factor of 227/47 = 4.83. So multiply the other resistors by this factor and pick the closest standard 1% values and you are there.
 
> "The low pass section is inverted, since 12dB crossovers always invert the phase of one signal. If desired, the high pass section may be inverted instead."

Not the "phase of one signal", that's a mis-statement. BOTH filters shift, and the 2nd-order case comes out OUT-of-phase between outputs, so you flip one output.

Consider the first-order crossover, and just one side, say the low-pass. At DC, output equals input. At infinite frequency, output is falling 6dB/octave, and is 90 deg out of phase with the input. There is a frequency we like to mention where response is 3dB down from DC. At that point the phase is shifted 45 degrees.

Use a 1st-order low-pass and high-pass together, both with the same -3dB frequency. At crossover, both are 45 degrees out of phase from the input. The drivers are fed 90 degrees out of phase with each other. This "could" produce a little cancellation at crossover, but in real life never enough to dominate the design.

Use a set of 2nd-order filters, and now each has a 90 deg phase shift at the -3dB frequency. (True for Butterworth; Bessel a little less, Chevyshev a little more, but still near 90 deg.) The two drivers are fed 180 degrees OUT of phase. If you actually had them working equally well at crossover frequency before you put in the filters, then the filters would give a perfect cancellation at crossover, and in real life a deep dip is possible.

Of course the drivers are also filters and are rarely at the same effective distance from the listener. Sometimes it all works out so the drivers crossover best when "in" phase, sometimes best when one is inverted. So on the 12dB crossovers you do want a phase-flop switch. It probably does not matter which side it is on: absolute phase is 99% inaudible, and for those few musical settings and recording chains where it is audible, it will be "wrong" about 50% of the time. If you care that much, put phase-floppers on all Xover outputs and fiddle till happy.

When you get to 4th order crossovers, ideally the outputs sum in-phase, and in practice any notch at crossover is so narrow that the null won't be noticed in concert. In precision monitoring, it is an issue to struggle with.

> 7th-order elliptical lowpass

And when mixing 2nd-order with 4th or 7th order, especially non-Butter shapes, phase WILL get bent and there may be no "perfect" answer.

Eliptical makes little sense to me to cut the squawk out of a sub. Your ear response is rising fast. Subwoofers are sometimes too-effective midrange speakers. Any filter that goes down and bounces up is going to leak audibly, unless the bump-up is 90dB down or more than 4 octaves up. You want a filter that falls and keeps falling. Get Jung's Filter Cookbook. A 3rd-order 110Hz Butterworth and a 3rd-order 0.5dB 150Hz Chevyshev in cascade will be 10dB down at 150, 45dB down at 300, 80dB down at 600, and keeps going down down down.

Don't overlook acoustic filtering. Running the sub output into a fuzz-lined chamber with a selected output gives a 12dB/oct cutoff, not only on the input signal but also on any squeaks and slappings that the woofer makes. Just facing the sub into a carpeted corner gives a sloppy ~6dB droop.

> 2nd-order 80Hz highpasses on the five channels and buffer the highpass inputs into a summing amp into a 4th-order 80Hz lowpass for the sub

Ah. Then you have additional phase-shifts. 80Hz is 14 feet. So a 7 foot path difference between the sub and one main is a 180 deg phase shift; just 3.5 ft is a 90 deg shift. You probably can't put 5 mains and a sub and your head all inside a <<7-foot sphere, so you are loaded with phase shifts, different for each speaker. In general, all your mains go where they have to go and "must" stay in phase with each other. Then throw your sub where-ever, put a phase-flip switch on it, and try both ways while listening to a well-known walking bass line. It may make no difference at all (the sub can't cancel the bottom-note of ALL the mains at varying distances from the sub and your head). If you do find 80Hz (the note ~an octave up from the bottom of a bass) go weak one way and strong the other way, you know what is best.
 

Latest posts

Back
Top