I read from "Designing High-Fidelity Tube Preamps" and learned that LC filter may need to be "damped" by adding resistor in series to the choke, and multiple LC filter (e.g. LCLC), although can achieve a much higher filtering, may lead to multiple resonance frequency, which even under audio frequency, may still cause stability problem like motorboating. Nevertheless CLC is frequently used, at least, for the output stage as RC simply consume too much power to provide adequate filtering. I want to explore how to properly implement CLCLC filtering but seems there is little mention to the multiple resonance frequency and how should it be damped. I would be appreciated if you can share what you know on this topic and any good materials discussing about this.
You can use PSUD2 simulation to apply a step load change to the output of any form of CLCLC.... filter, and then determine the damped resonant frequency of the response waveform. The aim is not to have a resonance within the audible frequency range (ie. force it to be low), and have it well damped.
You can spend hours enjoying if mains frequency (50, 60, 400Hz), choke ESR, capacitor ESR, LC configuration, step load magnitude, and C and L values make a difference.
If that is not sufficient to satiate your simulation needs, then you can set up LTSpice to sweep periodic disturbances across the frequency spectrum and look for resonance peaks.
If you don't like simulation, but have a scope and loads and signal generator and power mosfet type switches, then you can set up a hardware rig to do the same thing.
You can spend hours enjoying if mains frequency (50, 60, 400Hz), choke ESR, capacitor ESR, LC configuration, step load magnitude, and C and L values make a difference.
If that is not sufficient to satiate your simulation needs, then you can set up LTSpice to sweep periodic disturbances across the frequency spectrum and look for resonance peaks.
If you don't like simulation, but have a scope and loads and signal generator and power mosfet type switches, then you can set up a hardware rig to do the same thing.
Last edited:
Or, just accept that a handful of cheap sand components can turn any reasonably filtered input into smooth clean output without an expensive pile of magnetics cluttering your build...
Well damped by high capacitance?
At least 60Vpp ripple is a realistic value for a power amp reservoir capacitor. I once replaced the MOSFET with an expensive Lundahl CM choke for the better. I observed less noise on voices and on loud program.
One approach is to have multiple pi-filter resonances at least a factor 10 apart. Would be impossible to obtain with three filters though. Just fill up the capacitance I would argue and stay with two.
I'd suggest there is no 'proper' way to implement filtering, and more likely design should aim to avoid improper filtering.
Capacitance is just one of 4 main parameters that can be varied, given a set mains frequency. It is really for the OP to explore what can be implemented given practical part and loading constraints. All that is stated so far is that the middle filtering section could be LCLC or CLCLC or CLC, with no identification of any part models/types or power source info or loading info.
Capacitance is just one of 4 main parameters that can be varied, given a set mains frequency. It is really for the OP to explore what can be implemented given practical part and loading constraints. All that is stated so far is that the middle filtering section could be LCLC or CLCLC or CLC, with no identification of any part models/types or power source info or loading info.
The maths is very complicated since you are dealing with a fourth-order filter (or more). The easiest thing is to treat each LC section separately, and use the equation to find a suitable damping resistance for each section:
R = 2*sqrt(L/C)
This will give you something close enough to critical damping. It might not produce a truly 'optimum' result (use a circuit simulator if you want to look at the frequency reponse) but it will be good enough.
Attachments
Last edited:
True, but a Lundahl choke sound so much better than any sand device in a HV PSU.
Practical example?
R = 2*sqrt (L/C)
R = 2*sqrt (10H/0.00022F)
R = 2*sqrt (45454,545)
R = 2*213,2 = 426,4Ω
If L have for example 10Ω only needs 416,4Ω?
Last edited:
Another practical example.
R = 2*sqrt (10H/470uF) = 290 ohm
That is a surprising high value for a practical build. Too much drooop.
R = 2*sqrt (10H/470uF) = 290 ohm
That is a surprising high value for a practical build. Too much drooop.
Some additions
It is good to have two chokes parallel: One with very small resistance, capable to take big current, and the other choke with higher resistance and much smaller current, perhaps also series connected resistor, to make optimized damping.
Same principle with the capacitors: One capacitor with very low ESR and another capacitor with higher ESR and also series connected resistor.
It is good to have two chokes parallel: One with very small resistance, capable to take big current, and the other choke with higher resistance and much smaller current, perhaps also series connected resistor, to make optimized damping.
Same principle with the capacitors: One capacitor with very low ESR and another capacitor with higher ESR and also series connected resistor.
Attachments
You may need to point to an example with detailed technical performance measurements, as I'd suggest that is risky due to unforeseen resonances in the audio range, and I can't really see any benefit from the additional complexity accruing for normal ripple reduction.
A typical laminated steel core power supply choke is likely to have its first max-impedance resonance in the low to mid kHz region, and then impedance will reduce and go through further resonances. And the choke inductance and resonances will move around for any pseudo DC level changes such as with class AB operation, or where filter output capacitance is insufficient.
Last edited:
I recently revisit the LCLC input filter resonance topic and I noted from the internet that, rather than adding damping resistor in series with the inductor to damp the resonance, the resistor is added in series with the capacitor. I hope some of you would share your view on this.
A real power choke will be constructed using hi b cut double c core strip wound.laminated core chokes bobbin wound are poor perfomers. They might get rid of 100hz ripple but let all the hash through due to high distributed capacitance which adds to low freq reasonances.you get what you pay for.my recommendation is hashimoto LC 15 200. Looks beatifull ans sounds marvelous.
Depends what you're building. For a PP pentode output amp, I've been successful with just a capacitor before B+ going to CT of the output transformer. Not my idea, this was common on commercial amps and works because the ripple is common mode.
For SE pentode amp I've had great success with a simple CLC, and with a relatively small inductance of about 1-2H. A lot of people over-do it. First cap is always small relative to the second. PSUD2 is your friend.
I don't do SET, so others have to chime in.
For SE pentode amp I've had great success with a simple CLC, and with a relatively small inductance of about 1-2H. A lot of people over-do it. First cap is always small relative to the second. PSUD2 is your friend.
I don't do SET, so others have to chime in.
Others have already give good info. It's worth a reminder that there are all sorts of tricks you can do in the audio circuit to make the circuit resistant to power supply noise.
Read the filter section of this article then get back to me.
Power Supply Design for Vacuum Tube Amplifiers
And stop worrying about "resonance", it's a red herring.Hello Januaryabc:
I agree with Matt.
KISS is a good principle to stick with.
or as another rather smart fellow once said. "As simple as possible to achieve the desired goal"
not sure whom he was but I agree.
I guess in other words beware of the "Rube Goldberg syndrome" it is a deep dark hole with no bottom or way out.
Just solder something up and be happy.
have fun
I agree with Matt.
KISS is a good principle to stick with.
or as another rather smart fellow once said. "As simple as possible to achieve the desired goal"
not sure whom he was but I agree.
I guess in other words beware of the "Rube Goldberg syndrome" it is a deep dark hole with no bottom or way out.
Just solder something up and be happy.
have fun