I want to talk about circuit topologies for Analog Active Crossovers. Of course this applies most commonly to subwoofers, and this is why I am putting it here.
But it also can apply more broadly to multi-amped systems.
Now, we know about the Sallen-Key type circuits.
I'm interested in the full range of approaches being used. I'm particularly curious to know if anyone is using a subtractive approach, so that one filter, with one set of frequency determining components, does the job. Then you should get flat summation.
I'm also interested in general issues for multi-amping, if you can make the power amp itself part of the active filter circuit.
Multiamping and powered speakers seem to be one of the audio thresholds not yet crossed, even in very pricey stuff.
But it also can apply more broadly to multi-amped systems.
Now, we know about the Sallen-Key type circuits.
I'm interested in the full range of approaches being used. I'm particularly curious to know if anyone is using a subtractive approach, so that one filter, with one set of frequency determining components, does the job. Then you should get flat summation.
I'm also interested in general issues for multi-amping, if you can make the power amp itself part of the active filter circuit.
Multiamping and powered speakers seem to be one of the audio thresholds not yet crossed, even in very pricey stuff.
You're new here aren't you?
Active crossovers are talked about often on here, I think just about everybody is aware of their benefits. There has been a lot of good discussion on subtractive crossovers as well. Use the search function and start contributing to the threads
As for making power active filters with amps, this is possible, but hard to do in practice because of the amp stability issues when you start messing around with their gain.
Personally I would have put this in loudspeakers not subwoofers, but it's no big deal and I'm sure a friendly mod can move it.
Active crossovers are talked about often on here, I think just about everybody is aware of their benefits. There has been a lot of good discussion on subtractive crossovers as well. Use the search function and start contributing to the threads
As for making power active filters with amps, this is possible, but hard to do in practice because of the amp stability issues when you start messing around with their gain.
Personally I would have put this in loudspeakers not subwoofers, but it's no big deal and I'm sure a friendly mod can move it.
I'm interested in pursuing the active crossovers issue as it applies to subwoofers, and to multiamping in general.
Of course changing amp gain is a problem. I'm looking at it more as a full custom design.
I've never seen anything published about subtractive crossovers. Have you? Do you know some threads where that comes up?
Of course changing amp gain is a problem. I'm looking at it more as a full custom design.
I've never seen anything published about subtractive crossovers. Have you? Do you know some threads where that comes up?
zenmasterbrian said:Interesting. I'd never seen "subtractive crossover" in print. I was just looking for words to describe it. Every printed or online source I've ever seen uses a high pass and a low pass filter. I've never really liked this. I'm glad people are exploring this other way, "subtractive crossover"
(JPK) Try looking for Constant Voltage or CV crossovers. Small published a paper on this back in the 70's or 80's. They are subtractive. and the summed response is minimum phase. However, you might also want to look at state-variable filters. These are also referred to a subtractive but do not sum to a minimum phase repsponse.
Forgot to mention: There is also some guy called John Kreskovsky who shows some alternative methods (to subtractive crossovers) for transient-improved crossover solutions on his homepage.
Unfortunately a search on the web mostly finds allpass crossovers (like LR) if one is searching for constant voltage crossovers.
Regards
Charles
Unfortunately a search on the web mostly finds allpass crossovers (like LR) if one is searching for constant voltage crossovers.
Regards
Charles
Hi,
Using the power amplifier as part of an active high pass filter up
to 3rd order is relatively straightforward, the stability issues come
with lowpass filters, these need unity gain stable stages.
(You drive an unity gain op-amp filter circuit from the junction
of the gain setting resistors of the power amplifier, here the
power amplifier gain appears to be 1, like the op-amp output.)
Subtractive filters are not very power efficient because allthough
they sum to all pass they exhibit poor phase matching in the c/o
region. Whatever the order of the filter one roll-off is always 6dB
per octave, it also peaks, the higher the order the bigger the peak.
One advantage of the L/R is its power efficiency, i.e. phase matching.
/sreten.
Using the power amplifier as part of an active high pass filter up
to 3rd order is relatively straightforward, the stability issues come
with lowpass filters, these need unity gain stable stages.
(You drive an unity gain op-amp filter circuit from the junction
of the gain setting resistors of the power amplifier, here the
power amplifier gain appears to be 1, like the op-amp output.)
Subtractive filters are not very power efficient because allthough
they sum to all pass they exhibit poor phase matching in the c/o
region. Whatever the order of the filter one roll-off is always 6dB
per octave, it also peaks, the higher the order the bigger the peak.
One advantage of the L/R is its power efficiency, i.e. phase matching.
/sreten.The efficiency statement is indeed very true. But I wouldn't dare to say that the outputs are in phase. They are out of phase by 360 degrees actually. And how this influences efficiency on transients (the norm with music rather than sinusoids) has to be further investigated.
This has to be specified a little more: It is definitely possible to make symmetrical and asymmetrical subtractive crossovers of any desired order. In the "vicinity" of the crossover frequency they will always show a rolloff close to 1st order. And they definitely do have humps which are mathematically unavoidable for anything over 1st/1st order. These humps are higher for higer filter orders. They get smaller but broader for lower filter Qs involved.
The trick is to use the lowest order that is possible/necessary for a given driver combination and/or situation. So they are best used with sturdy drivers that are well behaved past the intended crossover frequency.
Regards
Charles
Hi,
I had a vague memory from many years back that subtractive filtering had some major failings, but my understanding then was much poorer.
However, Sreten & Phase accurate seem to have some (substantial) experience of this problem inherent in subtractive.
Sreten said
that 4pole filters are also impossible?
and how about 2pole L-R filters?
Some more extended clarification would be appreciated. It appears that if one already knows the answers then the message buried in the replies is clear, but for us poor amateurs, that level of understanding is lacking and I'm afraid I could be reading too much between the lines and drawing incorrect conclusions.
I had a vague memory from many years back that subtractive filtering had some major failings, but my understanding then was much poorer.
However, Sreten & Phase accurate seem to have some (substantial) experience of this problem inherent in subtractive.
Sreten said
and Phase accurate replied
Does the final statement
mean that subtractive cannot be used for symmetrical multiple pole filters?
that 4pole filters are also impossible?
and how about 2pole L-R filters?
Some more extended clarification would be appreciated. It appears that if one already knows the answers then the message buried in the replies is clear, but for us poor amateurs, that level of understanding is lacking and I'm afraid I could be reading too much between the lines and drawing incorrect conclusions.
I want the subractive approach because it seems elegant. It should be flat summing. Where as the HP and LP method really is not, and then there is component tollerance on top of that.
But this is why I posted here. Does it work? It would seem that you use HP to get the upper range driver signal. Then you subract that from the original to get the lower range driver signal.
If there is some component value error in the subraction, since it is the lower range driver, it does no harm.
But does it work. Is home or commercial equipment made this way?
All the books and online info I've ever seen and HP and LP.
But this is why I posted here. Does it work? It would seem that you use HP to get the upper range driver signal. Then you subract that from the original to get the lower range driver signal.
If there is some component value error in the subraction, since it is the lower range driver, it does no harm.
But does it work. Is home or commercial equipment made this way?
All the books and online info I've ever seen and HP and LP.
Andrew,
I don't know how you make symetrical slopes with subtractive crossovers, but Charles says it can be done. 4th-order for one side is possible, but the higher the order and/or the higher the Q the more pronounced the peak is on the opposite slope.
You cannot make a Linkwitz-Riley crossover subtractive because by definition a Linkwitz-Riley crossover is two 4th-order matched filters with a specific phase response. The 2nd-order variant hold the same as well.
I don't know how you make symetrical slopes with subtractive crossovers, but Charles says it can be done. 4th-order for one side is possible, but the higher the order and/or the higher the Q the more pronounced the peak is on the opposite slope.
You cannot make a Linkwitz-Riley crossover subtractive because by definition a Linkwitz-Riley crossover is two 4th-order matched filters with a specific phase response. The 2nd-order variant hold the same as well.
Hi Zen,
I bought and tried a BKelectronics subtractive crossover.
It sounded terrible.
I then tried to measure what it was doing.
The slopes were unmatched but seemed to vary as well.
There was a lot of feedthrough on some of the channels, i.e. the passband was not flat and was much wider than it should be, but of variable response.
I then tried to reverse engineer the cicuit board and eventually convinced myself that the subtraction had been sent to the wrong place on the PCB. It appears the designer did not check his work or could not hear the errors. I dumped the PCB and kept the case & transformer.
That was about 15 years ago, my first silly attempt at going active.
BK, the last time I looked, still sold that subtractive but maybe by now they have sorted the errors.
I bought and tried a BKelectronics subtractive crossover.
It sounded terrible.
I then tried to measure what it was doing.
The slopes were unmatched but seemed to vary as well.
There was a lot of feedthrough on some of the channels, i.e. the passband was not flat and was much wider than it should be, but of variable response.
I then tried to reverse engineer the cicuit board and eventually convinced myself that the subtraction had been sent to the wrong place on the PCB. It appears the designer did not check his work or could not hear the errors. I dumped the PCB and kept the case & transformer.
That was about 15 years ago, my first silly attempt at going active.
BK, the last time I looked, still sold that subtractive but maybe by now they have sorted the errors.
A subtractive crossover is just one method to derive a pair of lowpass and highpass that sum to unity. You can achieve the same thing also with seperate filters for high- and lowpass. But this would be much more complex than the subtractive variant since it would result in the summing of different filter outputs and it would be less tolerant to component variations than the subtractive method. This latter part is elegantly solved by the subtractive solution. With separate filters it would be easier to take driver response into consideration however. The third method for buliding these crossovers - the familiy name of which is constant voltage crossovers as John already mentioned - is the use of state-variable filters whose outputs are summed accordingly.
Basically you use a funtion that is always 1. I.e. its numerator and nominator polynoms are the same. The order of the polynom is the sum of all output orders minus 1. You then split the function into two halves having both the same nominator. And that is where the humps come from whoch grow larger with increasing filter order.
Though this is not very exotic from the mathematical point-of-view there are not that many sources showing this. I know only two personally. One is a recent paper by N. Thiele. The other one is this one, showing how it is done for a 2nd/2nd order symmetrical version implemented as state-variable filter (page 5):
http://www.national.com/an/AN/AN-346.pdf#page=6
How the same is achieved with the subtractive method is shown here:
http://www.diyaudio.com/forums/showthread.php?postid=292789#post292789
Regards
Charles
Basically you use a funtion that is always 1. I.e. its numerator and nominator polynoms are the same. The order of the polynom is the sum of all output orders minus 1. You then split the function into two halves having both the same nominator. And that is where the humps come from whoch grow larger with increasing filter order.
Though this is not very exotic from the mathematical point-of-view there are not that many sources showing this. I know only two personally. One is a recent paper by N. Thiele. The other one is this one, showing how it is done for a 2nd/2nd order symmetrical version implemented as state-variable filter (page 5):
http://www.national.com/an/AN/AN-346.pdf#page=6
How the same is achieved with the subtractive method is shown here:
http://www.diyaudio.com/forums/showthread.php?postid=292789#post292789
Regards
Charles
AndrewT said:
Hi,
Without consulting my (couple of) references I'm going to stick
my neck out and say the following, from what I remember :
Whatever the order and shape of the main filter, the subtracted
output roll-off is 6dB/octave away from the c/o region. The
subtracted output tends to peak, the peak is bigger the higher
the order and Q of the "normal" filter.
So for 2nd order or > symmetrical filters are impossible, so no L/R.
4 pole main filters are certainly possible but the subtracted output
is nothing like you'd imagine, looks like a peaking 1st order roll-off.
(i.e. a 1st order roll-off + parametric peak at the c/o frequency.)
Say you use a 2nd order subtractive filter for a sub, 2nd order L/R
or Butterworth for the main filter @ 100Hz. The sattelite speakers
would need to work well and with minimal phase shift down to
50Hz at least to get correct all pass crossover summing of outputs.
I'd say the above is probably its only sensible use, when you
want to add a sub to a pair of main speakers with substantial
bass capability, and trying to get an all pass response.
(I say trying because the mains rolloff and consequently phase
response will probably scupper any real all pass capability, though
for lowish Q 2nd order main filter, transient response will be good.)
/sreten.Sreten
The part about "the higher the order the higher the humps" is right.
But it IS definitely possible to build these filters with whatever order one desires. But the higher the order the better the behaviour away from the crossover frequency at the cost of worse behaviour around the crossover frequency.
See:
http://www.diyaudio.com/forums/showthread.php?postid=495629#post495629
Regards
Charles
The part about "the higher the order the higher the humps" is right.
But it IS definitely possible to build these filters with whatever order one desires. But the higher the order the better the behaviour away from the crossover frequency at the cost of worse behaviour around the crossover frequency.
See:
http://www.diyaudio.com/forums/showthread.php?postid=495629#post495629
Regards
Charles
Phase-accurate
---A subtractive crossover is just one method to derive a pair of lowpass and highpass that sum to unity.---
I think the Phase-accurate's phrase above applies to the voltage domain at the output of the active filters. However, I apologise for criticising that it is not explicitly mentionned. Even not-beginners, me included, were or are very easily confused with that matter.
Filters for loudspeakers have three different outputs, at least : in the voltage domain, in the forward acoustic pressure domain and in the total acoustic power domain. This is complicated by the amount of mutual radiation of the drivers, ie. the distance between them.
Once all this is mastered, there is an easily forgotten fact : in the high pass section, there is an invisible additionnal filter which is the driver itself (medium or tweeter). It is a hi-pass device which has its own level and phase responses. That's why "ideal" crossovers (ie substractive) crossovers yield to results far away for ideal.
I know a guy who proudly says he is using a square wave perfect crossover. He mixes the voltages of the signals at the output of his active circuit and gets the same perfect square waves on his scope at the output of his active circuit that they were at the input. He never replied to my questions about the acoustic response of his whole system...
Zenmasterbrian
---I'm also interested in general issues for multi-amping, if you can make the power amp itself part of the active filter circuit.---
To avoid stability problems, there's a need for a high level input and just a unity gain power buffer. That's a good challenge for designers ! Pavel Macura designed one.
---A subtractive crossover is just one method to derive a pair of lowpass and highpass that sum to unity.---
I think the Phase-accurate's phrase above applies to the voltage domain at the output of the active filters. However, I apologise for criticising that it is not explicitly mentionned. Even not-beginners, me included, were or are very easily confused with that matter.
Filters for loudspeakers have three different outputs, at least : in the voltage domain, in the forward acoustic pressure domain and in the total acoustic power domain. This is complicated by the amount of mutual radiation of the drivers, ie. the distance between them.
Once all this is mastered, there is an easily forgotten fact : in the high pass section, there is an invisible additionnal filter which is the driver itself (medium or tweeter). It is a hi-pass device which has its own level and phase responses. That's why "ideal" crossovers (ie substractive) crossovers yield to results far away for ideal.
I know a guy who proudly says he is using a square wave perfect crossover. He mixes the voltages of the signals at the output of his active circuit and gets the same perfect square waves on his scope at the output of his active circuit that they were at the input. He never replied to my questions about the acoustic response of his whole system...
Zenmasterbrian
---I'm also interested in general issues for multi-amping, if you can make the power amp itself part of the active filter circuit.---
To avoid stability problems, there's a need for a high level input and just a unity gain power buffer. That's a good challenge for designers ! Pavel Macura designed one.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.