Hey guys,
I just wondered if there are discrete topologies to make active filters like Linkwitz-Riley lowpass or -highpass filters without op-amps?
As far as I see, all you need is a non-inverting buffer. I am not at all familiar with discrete designs.
If there is nothing like this: Are their discrete op-amp designs which can be used with single supply?
Best wishes,
Dennis
I just wondered if there are discrete topologies to make active filters like Linkwitz-Riley lowpass or -highpass filters without op-amps?
As far as I see, all you need is a non-inverting buffer. I am not at all familiar with discrete designs.
If there is nothing like this: Are their discrete op-amp designs which can be used with single supply?
Best wishes,
Dennis
Yes you can make active filters using discrete devices rather than monolithic ICs or hybrid multi-die modules.
Probably the least complicated implementation is to use a filter type called Sallen-Key which google can tell you all about. Then you just need one vacuum tube triode or pentode per section. Or if you prefer FETs, one MOSFET per section. (alternatively, one JFET per section).
And finally you can build Sallen-Key filters using one bipolar junction transistor per section.
However, I would recommend that you spend the extra money and the extra printed circuit board area, to include a constant current source in each section. It's not mandatory but many designers feel this little extra increases the linearity of the circuit. These CCS's could be a constant current diode (Mouser sells over 1000 different models), or a transistor (BJT, MOSFET, JFET). Now you've got two silicon devices per Sallen-Key section.
In your research you'll discover that the Sallen-Key filter topology is much older than the integrated circuit. So it was created with discrete devices {vacuum tubes} in mind.
Probably the least complicated implementation is to use a filter type called Sallen-Key which google can tell you all about. Then you just need one vacuum tube triode or pentode per section. Or if you prefer FETs, one MOSFET per section. (alternatively, one JFET per section).
And finally you can build Sallen-Key filters using one bipolar junction transistor per section.
However, I would recommend that you spend the extra money and the extra printed circuit board area, to include a constant current source in each section. It's not mandatory but many designers feel this little extra increases the linearity of the circuit. These CCS's could be a constant current diode (Mouser sells over 1000 different models), or a transistor (BJT, MOSFET, JFET). Now you've got two silicon devices per Sallen-Key section.
In your research you'll discover that the Sallen-Key filter topology is much older than the integrated circuit. So it was created with discrete devices {vacuum tubes} in mind.
Pretty much any opamp based circuit can be configured for single rail operation. That can be done either with an active 'virtual ground' or in some cases just by simple biasing of the circuit itself. It depends on the circuit complexity which is best.
Thank you for your replies. Simply using one transistor (BJT, FET or Tube) in place of the op-amp doesn't seem very Hi-Fi to me.
I studied Rod Elliot's Website (as always a great resource) and found this article quite informative. Figure 9 looks promising, but doing a quick spice simulation, it is not really up to my requirements when it comes to headroom, drive capability, input impedance etc.
I probably have to look for real discrete op-amps and in this context probably also use dual supplies.
If anybody can recommend literature, resource on discrete buffer or single supply op-amp topologies, I am happy to hear 🙂
I studied Rod Elliot's Website (as always a great resource) and found this article quite informative. Figure 9 looks promising, but doing a quick spice simulation, it is not really up to my requirements when it comes to headroom, drive capability, input impedance etc.
I probably have to look for real discrete op-amps and in this context probably also use dual supplies.
If anybody can recommend literature, resource on discrete buffer or single supply op-amp topologies, I am happy to hear 🙂
The Klein & Hummel OY active studio monitor loudspeakers use an active Sallen and Key crossover between the woofer and the rest, with a simple emitter follower per stage. Remarkably modern design for 1967.
If you don't mind dual supplies, the M2x project here on the Forums provides nine different unity gain buffers on daughter cards, 35mm x 40mm. Unity gain buffers are exactly what the Sallen-Key active filter uses. I've attached the schematics for two of those nine buffers below. You'd almost certainly NOT use their PCBs because you probably wouldn't want daughter cards using M3 bolts for electrical connections, but maybe their circuit design approaches might be illuminating.
You might find Norwood particularly amusing since it's a composite amplifier with a single global negative feedback loop wrapped around two individual amplifiers in series. The second one has a bandwidth of 110 Megahertz and a slew rate of 1300 volts per microsecond. Listeners report that it sounds great too, in the M2x project.
_
You might find Norwood particularly amusing since it's a composite amplifier with a single global negative feedback loop wrapped around two individual amplifiers in series. The second one has a bandwidth of 110 Megahertz and a slew rate of 1300 volts per microsecond. Listeners report that it sounds great too, in the M2x project.
_
Attachments
Some time ago I simulated 18 different buffer circuits for comparison, including White, diamond, CFP, complimentary EFs, etc.
That's an excellent book but entirely based on opamp circuits. Also Self's Small Signal Audio Design covers active filters as well (not as exhaustively though) and has a whole chapter dedicated to discrete circuitry. For a unity gain buffer, he starts with a basic, one-BJT emitter follower and successively adds complexity to it improving its performance until he ends up with a seven-BJT circuit that measures pretty much flat at 0.0004% THD at 5 VRMS on a 2k load.
Make the caps very large (e.g. input 1 and output 100) to get a meaningful FFT (see it with original and large cap values) and add .option plotwinsize=0, otherwise there's data compression going on that throws the THD measurement out of whack. Doing this I get 0.000047%.
Ya. I thought of that later. I was focused on the stability. A slower part at Q3 may also be stable, ie slower than Q2.
You can do simple active crossover very simply. And it sounds very transparent.
Hi,
attached is the data about my old JFET-buffer filter, which works with simple 2-JFET buffers of which the lower JFETs function as modulated current sources to allow for higher load currents, lower THD and lower output impedance.
The ESL-panels of my electrostats required such a rather complex filtering.
The topologies used are unity-gain structures and buffered passive filters.
Even though the filters amplitude response comprises of a deep notch, followed by a peaking highpass and a light notch (@2.5kHz, but here almost invisible) the number of buffers in the direct signal path is just 2!
The low-active-parts-in-the-signal-path goal could only be reached by deviating from textbook filters .... which btw. is a common way in designing passive loudspeaker filters, but rather uncommon with active filtering, even though such textbook filters are typically off optimal in speaker design.
For example creates cap C7 after the first buffer stage the peaking amplitude response in co-working with the double-T notch, thereby omitting with at least one textbook filter stage.
R15,16, C11 and the parts around buffer J5/J6 synthezise the inductance of the second, very light notch filter ... which could be replaced by a true physical inductance, reducing the number of active stages even further.
The output driver buffer does just that ... buffering ... it might also be omitted with if the circumstances allow for, or be replaced with a more current-potent buffer like the Calvin buffer.
In any case ... the sound of this ´reduced´ active parts number count discrete filter is really alot more authentic and real.
Compared to an early test-filter utilizing textbook structures and many OPAmp stages built with OPAmps like OPA134, OPA604, etc. that always had a technical note, replaying a conserve.
The JFET filter in contrast is clear winner hands down, performing like listening to real music, real voices.
jauu
calvin
attached is the data about my old JFET-buffer filter, which works with simple 2-JFET buffers of which the lower JFETs function as modulated current sources to allow for higher load currents, lower THD and lower output impedance.
The ESL-panels of my electrostats required such a rather complex filtering.
The topologies used are unity-gain structures and buffered passive filters.
Even though the filters amplitude response comprises of a deep notch, followed by a peaking highpass and a light notch (@2.5kHz, but here almost invisible) the number of buffers in the direct signal path is just 2!
The low-active-parts-in-the-signal-path goal could only be reached by deviating from textbook filters .... which btw. is a common way in designing passive loudspeaker filters, but rather uncommon with active filtering, even though such textbook filters are typically off optimal in speaker design.
For example creates cap C7 after the first buffer stage the peaking amplitude response in co-working with the double-T notch, thereby omitting with at least one textbook filter stage.
R15,16, C11 and the parts around buffer J5/J6 synthezise the inductance of the second, very light notch filter ... which could be replaced by a true physical inductance, reducing the number of active stages even further.
The output driver buffer does just that ... buffering ... it might also be omitted with if the circumstances allow for, or be replaced with a more current-potent buffer like the Calvin buffer.
In any case ... the sound of this ´reduced´ active parts number count discrete filter is really alot more authentic and real.
Compared to an early test-filter utilizing textbook structures and many OPAmp stages built with OPAmps like OPA134, OPA604, etc. that always had a technical note, replaying a conserve.
The JFET filter in contrast is clear winner hands down, performing like listening to real music, real voices.
jauu
calvin
