Both Bob Cordell and Jim Brown write on this subject. Power cords, interconnects and speaker cables act as antennas for RFI. Once inside the box a well designed audio circuit may or may not be well behaved at RF frequencies. One new problem is that digital devices pulse the RF interference and that can affect bias at an audio rte.
This effect (frequency dependend PSSR) comes by discrete topologies through the fact, that the resistors for determine the reference current through the zener or normally diodes are connected to the GND line instead to the opposite voltage rail. But what is the reason by integrated monolitic op amps therefore?
Similar observations I have make, too. For me the royal way concerning the upper frequency range (i. e. for tweeter) is compliance with the rule to make the signal path as short as possible - have a look to
http://www.diyaudio.com/forums/pass...itional-op-amp-ultimate-sounding-phl1230.html
Last edited:
I haven't read this paper (URL therefore wanted) but I asume, that it refers to the Sallen Key topology (multiple PFP/positive feed back) I think this is very hard clearly to measure. But from the audible view I claim, that additional OP-Amps for the tweeter high pass path must be avoided in all cases. The combination of the solution from the URL of my previous post and additional few passive devices in the signal path to the tweeter must be the royal way to get ultimate sound quality.
The very high measured distortion values of the "ZEN" (compared to an OPA 627 distortion e.g.) are obviously not so critical from the audible perception.
An other way is the use of multiple NFB crossover network (MFB) - with this topology I haven't experience, but it isn't expect a distortion enhancement because there are no amplifyings of certainly frequency aeras (in opposite to Sallen-Key) - only attenuations are present.
Additional there is the possibility to use a passive crossover network as power amp front end (detailled project about this solution was released in the German "Speaker Builder" version "Hobby HiFi" or "Klang & Ton" (Mr. Bernd Timmermanns).
Last edited:
Hello Douglas,
maybe my topic has been already mentioned so forgive me for not reading all the 23 pages of this post. However, I read the TOC with increasing interest. What I am missing so far in all such publications about active crossover design is a guide or discussion about physical pcb-layout. How to route all the gnd-, signal- and power paths cleverly with regards to system noise. I guess with a smart trace layout (e.g. how to run the gnd, where to connect the gnds of several stages, where to connect to the gnd plane, etc.) a lot can be done in order to improve the overall noise performance of an active crossover.
maybe my topic has been already mentioned so forgive me for not reading all the 23 pages of this post. However, I read the TOC with increasing interest. What I am missing so far in all such publications about active crossover design is a guide or discussion about physical pcb-layout. How to route all the gnd-, signal- and power paths cleverly with regards to system noise. I guess with a smart trace layout (e.g. how to run the gnd, where to connect the gnds of several stages, where to connect to the gnd plane, etc.) a lot can be done in order to improve the overall noise performance of an active crossover.
Doug,
Baffle step response is one of the things it took me longest to fully get and integrate into my speaker designs. It should not be overlooked in your active crossover book (I'm sure you're on top of that - the trick is explaining it well - not many do). There's really no way to talk about designing an active crossover without talking about the speaker system as a whole. I've learned much about that and more from the linkwitzlab.com website. That's a great reference. Scaling the values of existing crossover circuits is my favorite way (often the easiest way) to get an active crossover circuit together and ready for modeling on SPICE just before building and verifying with calibrated mic and such. I would put in some examples on that concept. Why reinvent the wheel. Some of the math can go way over the heads of many. You might even have a chapter on how to do a tube type "active" crossover - one and two pole, with passive filter sections, (6SN7's making up for gain loss and buffering between sections?). Tube people love a circuit that has no negative feedback anywhere except probably the follower output buffer (if you call that neg feedback). Put a current source or sink on every tube, and the tube crowd might love it. Myself I designed and built a 3 way 4 pole Liinkwitz-Riley with EQ sections for open baffle speakers and closed box woofers, with the time delay on the tweeter all with op amps. Fairly similar to the Linkwitz Orion circuit, but tailored to my particular speakers/enclosure. My system measures flat from 20HZ to 20kHZ within a few dB with a calibrated mic and pink noise, at the listing position on my couch, despite listening room acoustics. It's fun stuff.
By the way, I love your poweramp design book. I'm better at doing grounding because of that book. Separating the high current grounds from low current, and then joining them at the star center, instead of just piling them all together helped significantly. And all the stuff on output stage crossover distortion, and so much more. An excellent book.
Baffle step response is one of the things it took me longest to fully get and integrate into my speaker designs. It should not be overlooked in your active crossover book (I'm sure you're on top of that - the trick is explaining it well - not many do). There's really no way to talk about designing an active crossover without talking about the speaker system as a whole. I've learned much about that and more from the linkwitzlab.com website. That's a great reference. Scaling the values of existing crossover circuits is my favorite way (often the easiest way) to get an active crossover circuit together and ready for modeling on SPICE just before building and verifying with calibrated mic and such. I would put in some examples on that concept. Why reinvent the wheel. Some of the math can go way over the heads of many. You might even have a chapter on how to do a tube type "active" crossover - one and two pole, with passive filter sections, (6SN7's making up for gain loss and buffering between sections?). Tube people love a circuit that has no negative feedback anywhere except probably the follower output buffer (if you call that neg feedback). Put a current source or sink on every tube, and the tube crowd might love it. Myself I designed and built a 3 way 4 pole Liinkwitz-Riley with EQ sections for open baffle speakers and closed box woofers, with the time delay on the tweeter all with op amps. Fairly similar to the Linkwitz Orion circuit, but tailored to my particular speakers/enclosure. My system measures flat from 20HZ to 20kHZ within a few dB with a calibrated mic and pink noise, at the listing position on my couch, despite listening room acoustics. It's fun stuff.
By the way, I love your poweramp design book. I'm better at doing grounding because of that book. Separating the high current grounds from low current, and then joining them at the star center, instead of just piling them all together helped significantly. And all the stuff on output stage crossover distortion, and so much more. An excellent book.
I'd put emphasis on baffle step response, scaling values of existing circuits for different frequencies, so you don't have to be super adept with the math, I'd do an example of a tube crossover with passive filter sections separated by 6SN7 gain stages, no neg feedback anywhere, current sources or sinks on each tube, and a follower buffer on the outputs. I'd talk much about speaker design and how it determines what's the best way to go. By the way, I loved your power amp book. Very excellent.
Here it is :
The Design of Active Crossovers by Douglas Self
According Amazon, mine should be sent the 11th of july.
The Design of Active Crossovers by Douglas Self
According Amazon, mine should be sent the 11th of july.
now available!
This is a must have book for anyone interested in using active crossovers in their loudspeaker projects.
You can download an electronic version for an e-reader, and for PC, for only $38.50 from Amazon USA:
The Design of Active Crossovers (Self): Amazon USA
The actual book seems to be delayed by a month at this time.
-Charlie
This is a must have book for anyone interested in using active crossovers in their loudspeaker projects.
You can download an electronic version for an e-reader, and for PC, for only $38.50 from Amazon USA:
The Design of Active Crossovers (Self): Amazon USA
The actual book seems to be delayed by a month at this time.
-Charlie
Rf energy and high feedback amplifiers
From what I've seen, very few preamps, amps, etc. have very good Rf filtering at the input, output or power supply. This is an area that may explain some of why many people like the sound of low feedback tube amplifiers. They are likely to handle what little Rf gets in more elegantly due to the low or no feedback. Put just a tiny bit of Rf into a high feedback amplifier, and you may get spurious grit, or ? Since phase shift at high frequencies (above 10kHZ anyway) is hard to detect with ears (so I don't really care about that), I always put a passive Rf filter at an input that rolls off 3dB by 100kHZ (usually more like 50kHZ). An Rf filter can also be put in a poweramp feedback output takeoff point (in case the speaker wires could act as an antenna and bring some in) by splitting the series feedback resistor, and adding a cap to ground. If the feedback series resistor is 20k ohms, just add 1K on the end that comes from the output, hang a cap to Gnd. there between the 20K and the 1K (calc. cap. based on what it sees for impedance - roughly 2n for 80kHZ in this example), and you're there. Nobody does this. I don't know why.
Because this is a "cool" way of converting a stable amplifier into perfect oscillatorFrom what I've seen, very few preamps, amps, etc. have very good Rf filtering at the input, output or power supply. This is an area that may explain some of why many people like the sound of low feedback tube amplifiers. They are likely to handle what little Rf gets in more elegantly due to the low or no feedback. Put just a tiny bit of Rf into a high feedback amplifier, and you may get spurious grit, or ? Since phase shift at high frequencies (above 10kHZ anyway) is hard to detect with ears (so I don't really care about that), I always put a passive Rf filter at an input that rolls off 3dB by 100kHZ (usually more like 50kHZ). An Rf filter can also be put in a poweramp feedback output takeoff point (in case the speaker wires could act as an antenna and bring some in) by splitting the series feedback resistor, and adding a cap to ground. If the feedback series resistor is 20k ohms, just add 1K on the end that comes from the output, hang a cap to Gnd. there between the 20K and the 1K (calc. cap. based on what it sees for impedance - roughly 2n for 80kHZ in this example), and you're there. Nobody does this. I don't know why.
I guess I was too general. I'd have to see where the poles and zeros are on the bench before picking the 3dB down frequency. Phase margin isn't always simple. This technique worked great in a low feedback (12dB) tube poweramp I made a few years ago. I calculated all the phase margin aspects, and then verified it on the bench. No problem. Personally, I'd be surprised if the speaker wire acting as an antenna, could inject significant Rf into such a low Z as the output stage of a poweramp is likely to be. What's the source impedance of such an antenna? The output Z of most poweramps is below 0.1 ohms.
This is true only at relatively low frequencies. It is not likely for audio power amplifier to have bandwidth more than a couple of MHZ, usually much smaller. This means that out of band RF components may not be suppressed by amp output impedance, until special precautions are taken.
You can calculate influence of non-linear output resistances of opamps in active filters.
Hi,
First, if we had RF injected into the feedback loop, we could indeed have all sorts of interesting problems.
Secondly, speaker cables certainly have the length to act as aerials at a range of frequencies. This will be the worse, the more the cables are stretched out linear.
The output impedance of most amplifiers and pre amplifiers is low only at low frequencies.
Now, if we want to minimise RF ingress I think some ferrite beads on the output wires and a well optimised LC filter could work.
Basically, we need some "build-out" network anyway, ideally we add something like a 100pF cap (low ESL) directly across the outputs and ideally also to earth (so X/Y filtering) and use the build out inductor/resistor with the zobels as a second line of filtering.
So, many amplifiers and some preamplifiers will already have suitable filtering, others may miss it...
Ciao T
First, if we had RF injected into the feedback loop, we could indeed have all sorts of interesting problems.
Secondly, speaker cables certainly have the length to act as aerials at a range of frequencies. This will be the worse, the more the cables are stretched out linear.
The output impedance of most amplifiers and pre amplifiers is low only at low frequencies.
Now, if we want to minimise RF ingress I think some ferrite beads on the output wires and a well optimised LC filter could work.
Basically, we need some "build-out" network anyway, ideally we add something like a 100pF cap (low ESL) directly across the outputs and ideally also to earth (so X/Y filtering) and use the build out inductor/resistor with the zobels as a second line of filtering.
So, many amplifiers and some preamplifiers will already have suitable filtering, others may miss it...
Ciao T
Hi,
The common "Thiele network" may or may not suffice. It would take extensive testing in an RFI test chamber (which very few audio makers have - Marshall in Milton Keynes does BTW) to be sure.
An alternative would be to apply standard methods of RF ingress suppression, as I suggested, that is a full set of X/Y capacitors on the amplifier binding posts and a correctly terminated PI filter on the amplifier output, to avoid negative interactions between the feedback amplifier and the capacitances.
BTW, when using output inductors, it is tempting to wind then on the damping resistor and it is tempting to use the resistor as former... If the resistor has steel endcap's, this is not the best idea.
Plus, it is probably a sub-ideal idea to use resistors that have significant inductance in parallel with this inductor.
Ciao T
The common "Thiele network" may or may not suffice. It would take extensive testing in an RFI test chamber (which very few audio makers have - Marshall in Milton Keynes does BTW) to be sure.
An alternative would be to apply standard methods of RF ingress suppression, as I suggested, that is a full set of X/Y capacitors on the amplifier binding posts and a correctly terminated PI filter on the amplifier output, to avoid negative interactions between the feedback amplifier and the capacitances.
BTW, when using output inductors, it is tempting to wind then on the damping resistor and it is tempting to use the resistor as former... If the resistor has steel endcap's, this is not the best idea.
Plus, it is probably a sub-ideal idea to use resistors that have significant inductance in parallel with this inductor.
Ciao T
Hi,
But that last part is an RC combo, not just a capacitor, last time I looked.
Ciao T
But that last part is an RC combo, not just a capacitor, last time I looked.
Ciao T
Yes, the three Rs and the two Cs and the L can all be chosen to suit the amplifier and the characteristic the designer wants to achieve.
That last R+C across the speaker terminals can use an R value of 0r0. It is still a Pi filter.
In this particular case an array of caps to present near zero impedance to a very wide range of interfering signals could be implemented just as your opening comment pointed out.
That last R+C across the speaker terminals can use an R value of 0r0. It is still a Pi filter.
In this particular case an array of caps to present near zero impedance to a very wide range of interfering signals could be implemented just as your opening comment pointed out.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.