Looking up something else I stumbled across something on the TI site suggesting with an active second order filter the second stage filter should have wildly different component values, say 10x the capacitance and 1/10 the resistance, or the other way round (it would help if I could remember which).
I've never heard this before or seen it on examples. Anyone know anything about it? As I said, I was looking for something else and at the time what I reda didn't sink in (I was looking for advice on feedback resistor values).
I've never heard this before or seen it on examples. Anyone know anything about it? As I said, I was looking for something else and at the time what I reda didn't sink in (I was looking for advice on feedback resistor values).
No, it wasn't, but that covers it very well, thank you. At first glance a second order filter counts as one stage and it is successive stages that you take Q down. So a second order filter can use the same values for both parts.
Mystery solved, thanks again! (Phew, don't have to change all the component values!)
Mystery solved, thanks again! (Phew, don't have to change all the component values!)
The "second stage" part of the second order filter had me confused since the widely used unity gain Sallen-Key filter does second order slopes with what is generally considered as a single stage. But I know what you mean. Using the same value for the parts is generally seen as an advantage I believe. Glad to have helped.
Quick Q if anyone knows the answer.
I'm trying to build a modular fully configurable 3 unit crossover that can be passive, active, second order or fourth order (I don't know what's best so make it flexible and see)!
If I run it as fourth order, if I put a unity gain buffer op amp between the stages will that negate the need for descending Q?
The way I'm designing it each module so it can be set to be:
Bypassed.
Passive.
Buffer.
Low pass.
High pass.
Low frequency (exact level variable by dip switch)
High frequency (exact level variable by dip switch)
If gain is on, level variable by dip switch.
Although conceptually complex and a pain to construct, it's only a matter of getting one module right with veroboard, then getting a load of boards printed up to fit a really very simple main board. I can then progressively turn things off and see at what point the musical advantage of not too much stuff in the way falls over in the face of complex impedances and signal loss.
However, if I have to make them variable Q, that would make my head hurt.
I'm trying to build a modular fully configurable 3 unit crossover that can be passive, active, second order or fourth order (I don't know what's best so make it flexible and see)!
If I run it as fourth order, if I put a unity gain buffer op amp between the stages will that negate the need for descending Q?
The way I'm designing it each module so it can be set to be:
Bypassed.
Passive.
Buffer.
Low pass.
High pass.
Low frequency (exact level variable by dip switch)
High frequency (exact level variable by dip switch)
If gain is on, level variable by dip switch.
Although conceptually complex and a pain to construct, it's only a matter of getting one module right with veroboard, then getting a load of boards printed up to fit a really very simple main board. I can then progressively turn things off and see at what point the musical advantage of not too much stuff in the way falls over in the face of complex impedances and signal loss.
However, if I have to make them variable Q, that would make my head hurt.
Rather that descending Q, it sounds as though you are talking about impedances where cascaded passive stages are loaded by other stages and their filter effect changed?
Ordinarily, variable cascading of stages to change the slope requires adjustment of the knee.
If you are interested in testing having clutter in your signal path, why not use a string of buffers?
Ordinarily, variable cascading of stages to change the slope requires adjustment of the knee.
If you are interested in testing having clutter in your signal path, why not use a string of buffers?
Just checking, are you familiar with this diyaudiostore kit?
LXmini Crossover / Analog Crossover Network – diyAudio Store
LXmini Crossover / Analog Crossover Network – diyAudio Store
I must remind, that if the crossover is supposed to work with a multi-way loudspeaker, acoustic response is what really counts.
All loudspeaker drivers are nonlinear and the baffle/box will have it's effect too. This means you must have ability to equalize each "way" too! And after that one must take care of time coherence which means to adjust delay/phase for each unit too! And don't forget gain/volume per channel.
All loudspeaker drivers are nonlinear and the baffle/box will have it's effect too. This means you must have ability to equalize each "way" too! And after that one must take care of time coherence which means to adjust delay/phase for each unit too! And don't forget gain/volume per channel.
Last edited:
No, I wasn't aware of that! Not exactly what I'm doing, I'm trying to build a speaker that requires the minimum correction, but still fascinating.
You should definitely have some kind of low shelf/BSC circuit. (one stage)
I tried the same thing before but it`s tough to come up with a generic crossover that fits "all" drivers.
That ends up being quite complex with lots of stages (slopes up to 24dB at least, several notch filter, low shelf, high shelf etc.)
See this nice summary of active filters from the great Linkwitz for inspiration:
Active Filters
Here is an example of a relatively simple XO with just BSC, tweeter delay and LR4-filter. (incl. gerber files):
TIPD134 Analog, Active Crossover Circuit for Two-Way Loudspeakers | TI.com
I built something similar (~generic&discrete) here:
Discrete crossover, sanity check
Also made a FET version which isn´t documented. Works well but doesn´t have notches and tweeter delay.
I tried the same thing before but it`s tough to come up with a generic crossover that fits "all" drivers.
That ends up being quite complex with lots of stages (slopes up to 24dB at least, several notch filter, low shelf, high shelf etc.)
See this nice summary of active filters from the great Linkwitz for inspiration:
Active Filters
Here is an example of a relatively simple XO with just BSC, tweeter delay and LR4-filter. (incl. gerber files):
TIPD134 Analog, Active Crossover Circuit for Two-Way Loudspeakers | TI.com
I built something similar (~generic&discrete) here:
Discrete crossover, sanity check
Also made a FET version which isn´t documented. Works well but doesn´t have notches and tweeter delay.