Resonant Choke Filter Revisited

Question

The Resonant Choke filter in its traditional form looks like an orthodox choke input filter but it has a cap across the choke tuning it to twice the power frequency. This means 100 or 120 Hz depending on what part of the world. Intuitively you can get lower output ripple for a given output cap value compared to the normal choke input filter. In fact old audiophiles would sometimes "tune the choke" to reduce hum. Quite simply the choke and resonant cap form a high impedence parallel tuned circuit that blocks the 2F ripple. The higher order terms of the full wave rectified sine wave are of much lower amplitude and are dealt with by the output cap. An interesting aspect of this postwar circuit is that it has the potential for a good power factor but the old choke input filter can have a power factor that beats the cap input filter anyway. Could the choke in a resonant choke be smaller than the choke in a traditional choke input filter because less inductance is needed ? What would be a sensible L/C ratio? Would the power factor compete with the valley fill? When this was originally done loads tended to have less variation. How would the resonant choke perform under changing loading?Is this circuit easy on the diode recovery making it a contender in a sine wave SMPS? How would it perform if it was slightly off resonance ? Would the component tolerances be reasonably broad for easy production ? Would the low THD input current waveform be useful for a quiet SMPS ?

Answer 1

I just simulated a simple parallel inductor capacitor circuit. A 4H choke (common size for a guitar amp) requires a 440n cap to get a notch at exactly 120Hz. But with a 470n cap for example, the notch would be at 116Hz and so you would only get about 12dB attenuation (as opposed to more like 40dB if it really is tuned). Load has no effect on the resonance of the circuit in the simulation. I don't think it would work with SMPS because those chokes are usually pretty small. For example, a 100uH choke and 101.3n cap would resonate at 50kHz. But because the choke is so small, the notch is very narrow. I had to use 10ths of nano-farads to really get the 50kHz.

Answer 2

I have been told that a frequency sweep is important because the filter has a deep notch, then good attenuation above the resonant frequency, but it has very poor rejection below the resonant frequency. I have not verified this but I believe the person who mentioned it and will have to check that myself.

In radio transmitters and RF amplifiers, the resonance is usually set a bit high, such as 12 5Hz for a 120 Hz ripple. The reason for this is so that as the choke inductance decreases with increasing load current, the resonant frequency does not dip below the ripple frequency and the power supply begin to behave as a capacitor input circuit.

It is prudent therefore to know the minimum inductance of the choke under all foreseen operating conditions. As power levels increase, especially voltages, these details become more important for safety reasons, i.e., non-exploding reasons. I have done numerous simulations in LTspice with various inductance and capacitance combinations. I am working on a power supply making 4800 VDC at 2 A, as part of a 3CX3000 RF linear power amplifier build. RF amps of this kind are tricky because they are almost always close to class B and so the very signal acts to modulate the output voltage of a carelessly or cheaply built power supply, in turn distorting the linearity of the signal.

One issue that shows up with non-resonant choke input power supplies is the generation of large voltage transients when abrupt changes in load current are present. Such transients are less common in audio equipment, but can appear in class B circuits used for single sideband, and become troublesome for 'pulsed' operation such as CW/Morse code, especially in high voltage circuits.

As an example, a simulation showed a 4 kV transient on top of the 4800V line when decreasing current from 2 A to 350 MA in 10 ms. The power supply was a full wave bridge rectifier followed by a 30 H choke and a 31 uF capacitor (not shown).

These transients can be greatly reduced with a resonant choke followed by a good size filter capacitor. Typical values seen were about 1 kV +/-300 V.

Those using LTspice may be interested in the way the current step with a resonant choke was simulated. There is a discussion about this supply on amfone.net http://amfone.net/Amforum/index.php?topic=44750.msg321900#msg321900

I am still researching this subject and that is what led me here. The only real remaining concern on my part is how the capacitor will handle the current. So, just wanted to share what I have learned about this so far. It's a very interesting topic. There is no reason not to post to an older question & answers, if more information can help. The amplifier build is slowly continuing on bunkerofdoom.com, and that is where any solutions to our own power supply (exchange choke for resonant filter) will end up: http://bunkerofdoom.com/3cx3000/index.html