Hi,
I cannot answer the speculative questions, but I can tell you what I did.
I made my wn board with tons of sziklai & ring of two current source followers with multipin sockets to hold RC Cards (and attenuation resistors) that changed the crossover frequencies (and yes, the filters where Sallen Key but not 4th order LR but 3rd order Butterworth for their interesting positive properties and the capacitors selected with an RLC bridge!).
The gains where pre-set according to the measured efficiency of the drivers, with 10-turn cermet trimmers having a few dB fine adjustment range, so I could tweak the system for production tolerances.
I did not include EQ as non was (seemed?) needed with the drivers and amps I used I used and the measurement gear we had available semi-portable 3rd octave RTA I had to rent).
The later Pro-Audio versions retained the discrete followers and plug in cards but changed to 4th order LR and added the various limiters and some EQ for the HF of the horns and the low bass (also plug-in). Again the whole thing was carefully set to match the actual speaker system and was completely un-adjustable, save for small amounts of channel level trim internal to the unit.
So even then I would not want to have an adjustable system, if only for the difficulty to get the adjustment made easily and repeatable under field conditions. I was probably influenced by EV's and JBL's active crossovers at the time, which used specific cards for their specific speaker systems, so there was also no adjustability. The western "Processor PA" systems that influenced me later of course precluded any ability to adjust at the outset...
After that I moved over to digital speaker management on the pro side and fullrange drivers and SET Amp's personally.
Ciao T
I cannot answer the speculative questions, but I can tell you what I did.
I made my wn board with tons of sziklai & ring of two current source followers with multipin sockets to hold RC Cards (and attenuation resistors) that changed the crossover frequencies (and yes, the filters where Sallen Key but not 4th order LR but 3rd order Butterworth for their interesting positive properties and the capacitors selected with an RLC bridge!).
The gains where pre-set according to the measured efficiency of the drivers, with 10-turn cermet trimmers having a few dB fine adjustment range, so I could tweak the system for production tolerances.
I did not include EQ as non was (seemed?) needed with the drivers and amps I used I used and the measurement gear we had available semi-portable 3rd octave RTA I had to rent).
The later Pro-Audio versions retained the discrete followers and plug in cards but changed to 4th order LR and added the various limiters and some EQ for the HF of the horns and the low bass (also plug-in). Again the whole thing was carefully set to match the actual speaker system and was completely un-adjustable, save for small amounts of channel level trim internal to the unit.
So even then I would not want to have an adjustable system, if only for the difficulty to get the adjustment made easily and repeatable under field conditions. I was probably influenced by EV's and JBL's active crossovers at the time, which used specific cards for their specific speaker systems, so there was also no adjustability. The western "Processor PA" systems that influenced me later of course precluded any ability to adjust at the outset...
After that I moved over to digital speaker management on the pro side and fullrange drivers and SET Amp's personally.
Ciao T
Mr. Klippel had mentioned some form of way to implement speaker motor linearity improvements as well as speaker protection. Really nice approach.
It just so happens that I have a great article for a diy active xover in Linear Audio Vol 1. See pic.
But it'll be some more weeks...
jan didden
Attachments
ThorstenL
----I would suggest to use only single op-amp's with the PCB pattern for both SMD and TH Op-Amp's, all else TH parts, this way people can use whatever kind of op-amp's they like (instead of being limited to NE5532's and TL082's).----
I look forward what Douglas will tell about the LME49990 in his book to come. If the same circuit, hoped not to be too expensive, is not available in dual packages, maybe someone will sell assemblies à la Burson.
----I would suggest to use only single op-amp's with the PCB pattern for both SMD and TH Op-Amp's, all else TH parts, this way people can use whatever kind of op-amp's they like (instead of being limited to NE5532's and TL082's).----
I look forward what Douglas will tell about the LME49990 in his book to come. If the same circuit, hoped not to be too expensive, is not available in dual packages, maybe someone will sell assemblies à la Burson.
Q
That Behringer product isn't what I am talking about, it is NOT fully flexible. It has fixed Q filters, what I'm talking about is the ability to alter both of the resistors in a standard sallen key topology.
That gives you the flexibility to have the filter do whatever you want it to (within sensible reason), such as LspCAD would optimise during the design phase. All of the available off the shelf active xovers appear to come with fixed Q filters, maybe they can switch between Butterworth and LW alignment, but they can't say cascade two 2nd order high pass filters, the first having a Q of 0.37 with a corner frequency of 160hz and the second with a Q of 0.55 with a corner frequency of 100hz.
We all know how useful textbook filters are, it's what we're trying to get away from.
Maximum quality isn't congruent with anything other then a very specific design. No doubt Mr. Self goes into quite some detail illustrating the requirements for maximum performance based around component values. If however you're producing a board that's supposed to be capable of fitting into a vast array of different end designs, you have to let some of the quality slip.
That Behringer product isn't what I am talking about, it is NOT fully flexible. It has fixed Q filters, what I'm talking about is the ability to alter both of the resistors in a standard sallen key topology.
That gives you the flexibility to have the filter do whatever you want it to (within sensible reason), such as LspCAD would optimise during the design phase. All of the available off the shelf active xovers appear to come with fixed Q filters, maybe they can switch between Butterworth and LW alignment, but they can't say cascade two 2nd order high pass filters, the first having a Q of 0.37 with a corner frequency of 160hz and the second with a Q of 0.55 with a corner frequency of 100hz.
We all know how useful textbook filters are, it's what we're trying to get away from.
Maximum quality isn't congruent with anything other then a very specific design. No doubt Mr. Self goes into quite some detail illustrating the requirements for maximum performance based around component values. If however you're producing a board that's supposed to be capable of fitting into a vast array of different end designs, you have to let some of the quality slip.
Hi,
Adjust both resistors independently?
With Pots?
How on earth are you actually intending to set this with any reasonable accuracy.
High order sallen key filters are complete PITA's on component tolerances...
Sorry, but I am absolutely not prepared to attempt that.
If you want to optimise the filter transfer functions to get non-standard filters, I'm with you, but doing it adjustable?
I think I prefer the nightmares bad acid give you. It is precisely this sh!t that modern DSP filters are for (like the pure software one discussed here - ultimate EQ or such).
But by all means have fun.
Sure. They have good reasons to do that.
Sure, my original proposal to DS earlier would allow precisely that, especially if he goes with classic SK filters.
Precisely.
My proposal was aimed at a maximum quality design that could be configured by changing passive component values to suit the filter curves desired.
I', just not mad enough to suggest variable components with external adjustment for each and every one - in fact I suggest to have the adjustability for all components only by un- and re-soldering TH Parts.
The only justification I can see for such an anlogue crossover is in fact the ability to go for "maximum quality" (even that is questionable though), all else really is a job for some form of DSP.
I disagree. Hence actually my suggestions.
If you have a PCB with maximum flexibility applicable during assembly you can take this "generic PCB" and simply assemble the components to fit the design required, you will have maximum quality AND maximum flexibility.
And no nightmares.
Ciao T
Adjust both resistors independently?
With Pots?
How on earth are you actually intending to set this with any reasonable accuracy.
High order sallen key filters are complete PITA's on component tolerances...
Sorry, but I am absolutely not prepared to attempt that.
If you want to optimise the filter transfer functions to get non-standard filters, I'm with you, but doing it adjustable?
I think I prefer the nightmares bad acid give you. It is precisely this sh!t that modern DSP filters are for (like the pure software one discussed here - ultimate EQ or such).
But by all means have fun.
Sure. They have good reasons to do that.
Sure, my original proposal to DS earlier would allow precisely that, especially if he goes with classic SK filters.
Precisely.
My proposal was aimed at a maximum quality design that could be configured by changing passive component values to suit the filter curves desired.
I', just not mad enough to suggest variable components with external adjustment for each and every one - in fact I suggest to have the adjustability for all components only by un- and re-soldering TH Parts.
The only justification I can see for such an anlogue crossover is in fact the ability to go for "maximum quality" (even that is questionable though), all else really is a job for some form of DSP.
I disagree. Hence actually my suggestions.
If you have a PCB with maximum flexibility applicable during assembly you can take this "generic PCB" and simply assemble the components to fit the design required, you will have maximum quality AND maximum flexibility.
And no nightmares.
Ciao T
I think we're thinking about this in different ways to one another. This isn't going to be some pretty box with values printed on the front of the fascia allowing you to dial in set frequencies. The xover board would have caps fitted of a suitable value so that you end up with an acceptable range of frequencies for an acceptable size of trim-pot, given the application (tweeter network, midrange, bass etc)
The end user at the very basic level would have to calculate the value for the pots using formulas as are present on Linkwitz website. They would then have to set all the pots manually using a multimeter and a small screwdriver.
At a more advanced level they would design the xover in something like LspCAD, like this..
Then get out the multimeter again and set all the pot values. This gives you ultimate design freedom and is what I currently use and what I am suggesting. The caps are high tolerance types, the only thing left is how accurately I can set the pots. If done correctly, the filters are bang on LspCADs predicted response.
You can even wire the xover up to the computer and measure the thing with something like ARTA using the real time single channel frequency response option. You can then alter the pot values and actually watch the transfer function changing on the screen, I actually find doing that fun, maybe I'm strange, but it's doing things like that that educate me the best too.
If your transfer function is out by a bit you can import the predicted transfer function, overlay them and tweak the real time until they match, I've done this a few times with notch filters. None if it is hard to do, it just takes time and as DIYer I want this kind of tweak-ability.
The end user at the very basic level would have to calculate the value for the pots using formulas as are present on Linkwitz website. They would then have to set all the pots manually using a multimeter and a small screwdriver.
At a more advanced level they would design the xover in something like LspCAD, like this..
Then get out the multimeter again and set all the pot values. This gives you ultimate design freedom and is what I currently use and what I am suggesting. The caps are high tolerance types, the only thing left is how accurately I can set the pots. If done correctly, the filters are bang on LspCADs predicted response.
You can even wire the xover up to the computer and measure the thing with something like ARTA using the real time single channel frequency response option. You can then alter the pot values and actually watch the transfer function changing on the screen, I actually find doing that fun, maybe I'm strange, but it's doing things like that that educate me the best too.
If your transfer function is out by a bit you can import the predicted transfer function, overlay them and tweak the real time until they match, I've done this a few times with notch filters. None if it is hard to do, it just takes time and as DIYer I want this kind of tweak-ability.
Last edited:
Hi,
We are talking about self assembly of a circuit board here.
I really cannot for the life of me why it be preferable to to turn dozens of pots attempting to get the values right in a circuit where they are not always DC isolated is preferably to just soldering in some resistors.
Or DSP.
But that's just me. To me what you describe is a nightmare scenario, not a nirvana.
Well, one guy's meatloaf...
I'm sure DS finds this debate incredibly amusing.
Ciao T
We are talking about self assembly of a circuit board here.
I really cannot for the life of me why it be preferable to to turn dozens of pots attempting to get the values right in a circuit where they are not always DC isolated is preferably to just soldering in some resistors.
Or DSP.
But that's just me. To me what you describe is a nightmare scenario, not a nirvana.
Well, one guy's meatloaf...
I'm sure DS finds this debate incredibly amusing.
Ciao T
active crossover niche...
I am addressing this niche. See the modular crossover board overview pdf file linked to in this post:
http://www.diyaudio.com/forums/anal...bilities-i-want-your-input-3.html#post2454043
It uses the LME49740.
-Charlie
I am addressing this niche. See the modular crossover board overview pdf file linked to in this post:
http://www.diyaudio.com/forums/anal...bilities-i-want-your-input-3.html#post2454043
It uses the LME49740.
-Charlie
Last edited:
I never said it was nirvana, that would be a well executed digital design. And there's never been anything digital, commercially available that does exactly what I want it to do.
At the time when I built the analogue xover, building my own digital one wasn't possible, now however it looks like it is, something I am most delighted about, because as you've mentioned twiddling about with a hundred pots isn't fun, but it was the only practical way of getting what I wanted.
It's just a different point of view. I think you find what I am suggesting amusing.
The idea of a modular design where you can plug and play any number of filters together and purchase the number of modules you specifically require is a nice one, this also allows for adding more at a later date should it be necessary.
In an ideal world you'd have individual PCBs for each available filter type that would plug together ESI Juli@ style allowing you to easily daisy chain any combination of filters that you'd want.
Hi,
Actually, I find it scary!
My "socketed plugin" version allowed (with a modicum of manual dexterity and minimal soldering experimental hardwired component combinations to be assembled probably in less time than you spend twisting pots and absolutely repeatable...
Ciao T
Actually, I find it scary!
My "socketed plugin" version allowed (with a modicum of manual dexterity and minimal soldering experimental hardwired component combinations to be assembled probably in less time than you spend twisting pots and absolutely repeatable...
Ciao T
Ah but your way required quite a lot of hardware changes, with this, nothing changes, except pot values and that suits me better then swapping stuff in and out. It doesn't take that long to change enough pots around, it probably takes me about 5-10 mins to set the ones im interested in. I don't make sweeping system changes, mostly I'll change one thing. Like the xover point/rate between two drivers, that's a doddle to change, takes about 5 mins.
The whole fiddling and trimming hassle is where the advantage of DSP lies. As already mentioned some design programs allow the emulation of crossovers on the PC.
This would allow to build a crossover that is using the correct and exact values from the beginning. No need for tons of trimpots - just the few that are unavoidable will be left.
And exactly these crossover emulation capabilities are a subject that shouldn't be left out in Douglas' book IMO
Regards
Charles
This would allow to build a crossover that is using the correct and exact values from the beginning. No need for tons of trimpots - just the few that are unavoidable will be left.
And exactly these crossover emulation capabilities are a subject that shouldn't be left out in Douglas' book IMO
Regards
Charles
Last edited:
Here's my view on component adjustability:
"... these resistor positions on a prototype PCB can be fitted with single-way turned-pin sockets like those that ICs are plugged into. It is then possible to very easily plug resistors in and out during development. This is much quicker than pulling the PCB out of the case, desoldering one resistor and then re-soldering another one in, and PCBs will only take so much of this before the pads are damaged. When the design is finalised the PCBs will have resistors soldered into the same positions in the usual way."
"... these resistor positions on a prototype PCB can be fitted with single-way turned-pin sockets like those that ICs are plugged into. It is then possible to very easily plug resistors in and out during development. This is much quicker than pulling the PCB out of the case, desoldering one resistor and then re-soldering another one in, and PCBs will only take so much of this before the pads are damaged. When the design is finalised the PCBs will have resistors soldered into the same positions in the usual way."
Last edited:
Hello Douglas,
After your "Shelf On Audio", i will wait for this new one !
I don't know a dedicated book on Crossover.. but i read an interesting article (4pages) about cleaver (for me) precise LoCost analog Phase-EQualized 2 & 3 ways with 2 & 4th order filters, in Elektor magazine (Paper, French translation from German or Dutch) of January1986, by Thomas Scherer. The only drawback is that the smooth Q=0.5 @ Fc.
It is old and i can only give you that recent link on internet (PowerAmp 2010... in french also)
Catalogue en ligne HENAM.
In case of interest, tell me !
Pierre.
After your "Shelf On Audio", i will wait for this new one !
I don't know a dedicated book on Crossover.. but i read an interesting article (4pages) about cleaver (for me) precise LoCost analog Phase-EQualized 2 & 3 ways with 2 & 4th order filters, in Elektor magazine (Paper, French translation from German or Dutch) of January1986, by Thomas Scherer. The only drawback is that the smooth Q=0.5 @ Fc.
It is old and i can only give you that recent link on internet (PowerAmp 2010... in french also)
Catalogue en ligne HENAM.
In case of interest, tell me !
Pierre.
Douglas,
I am eagerly awaiting the release of this book. I am especially curious about the contents of Chapter 5, "Notch Crossovers", e.g. filters with zeros. These seem to be more complicated to design than other types because of the additional design variable(s). I have a copy of Zverev's book from 1967 with all the design info, however this focuses on "filters", not necessarily intended for "crossovers". Since the object is to have a set of filters that are matched with respect to phase and amplitude at the crossover point, it's not always the case that one can take the standard designs for a set of HP and LP elliptical filters having the same corner frequency and pair them up to get a good crossover. Sometimes some sort of scaling of the corner frequency or other adjustment is required. This seems to be the case for Bessel and Chebychev crossovers as well.
Will you, in the Elliptical filter chapter, be approaching the subject from the point of view of crossover design or filter design?
Will you be describing one or more active realizations of Elliptical filters?
Will you also go in to crossover scaling for Bessel and Chebychev types?
Going deep in to these topics would definitely put out some new and useful info in this area...
-Charlie
I am eagerly awaiting the release of this book. I am especially curious about the contents of Chapter 5, "Notch Crossovers", e.g. filters with zeros. These seem to be more complicated to design than other types because of the additional design variable(s). I have a copy of Zverev's book from 1967 with all the design info, however this focuses on "filters", not necessarily intended for "crossovers". Since the object is to have a set of filters that are matched with respect to phase and amplitude at the crossover point, it's not always the case that one can take the standard designs for a set of HP and LP elliptical filters having the same corner frequency and pair them up to get a good crossover. Sometimes some sort of scaling of the corner frequency or other adjustment is required. This seems to be the case for Bessel and Chebychev crossovers as well.
Will you, in the Elliptical filter chapter, be approaching the subject from the point of view of crossover design or filter design?
Will you be describing one or more active realizations of Elliptical filters?
Will you also go in to crossover scaling for Bessel and Chebychev types?
Going deep in to these topics would definitely put out some new and useful info in this area...
-Charlie
Salut Pierre
If you are refering to the phase-subtractive elektor crossover then I can tell you that the same cn be achieved with Sallen-Key sections and (depending on Q values used also a subtractor) also other Q values than just 0.5.
Regards
Charles
If you are refering to the phase-subtractive elektor crossover then I can tell you that the same cn be achieved with Sallen-Key sections and (depending on Q values used also a subtractor) also other Q values than just 0.5.
Regards
Charles
Yes, I'd be interested to see that article if you have a copy in French. German would be too much work for me.
- Status
- Not open for further replies.