Took a look at the link, and it looks like he doesn't know what he's doing. Ignore his BS.
Here's a better reference (attached) There's a TMI risk here since there are discussions about polyphase rectifiers you're not likely to encounter with reasonable, for-home-use, designs. Everything you'll need right here.
Always interesting to hear that after decades of designing, testing, and deploying electronics that I don't know anything. So by this I can only assume that you believe that O. H. Schade and Herbert Reich also were clueless. Well please, by all means, enlighten us with your wisdom. We all wait with bated breath.
It's likely that if you construct a CLCLC filter with low values of L, C and series R, that filter will have pretty large resonances.
You can add series R to damp those resonances, but now you'll be dropping more voltage, which negates the most sought-after advantage of using power supply chokes.
One should design the CLCLC filter for both good filtering and reasonably good damping of resonances through the power supply. Fortunately, we can easily buy 330uF 400V capacitors these days, so it should be possible to make a CLCLC filter with good damping and still not requiring added series resistance(s).
What is the advantage of a CLCLC filter over using a CRC filter into a capacitance multiplier if the audio circuit (load) operates in class A?
You can add series R to damp those resonances, but now you'll be dropping more voltage, which negates the most sought-after advantage of using power supply chokes.
One should design the CLCLC filter for both good filtering and reasonably good damping of resonances through the power supply. Fortunately, we can easily buy 330uF 400V capacitors these days, so it should be possible to make a CLCLC filter with good damping and still not requiring added series resistance(s).
What is the advantage of a CLCLC filter over using a CRC filter into a capacitance multiplier if the audio circuit (load) operates in class A?
It's good that Matt Renaud (in the link provided by Suncalc in post # 19) takes the reader through Schade and then Reich. Imho, anyone trying to design such a power supply needs a good awareness of such aspects of design, and to understand what is being proposed. However, I would have preferred to see additional comment such as:
- hot switching can occur just from accidentally turning the mains switch off then on.
- how simple modern software like PSUD2 can avoid the need for lengthy and often error prone mathematical and chart interpretation efforts, and achieve a better insight into waveform characteristics and practical design.
- how the filter output capacitor is the one that handles amplifier signal current, and how ripple level on it may be significantly discounted for some types of amplifier.
- how some parts like chokes have significant non-ideal behaviour, and how modern e-caps don't need small value bypass caps which can be themselves a resonant problem.
- how poor layout can easily raise the ripple and noise level above what is being estimated by such equations.
This is all very dazzling, but I can't see how it relates to either power supply dynamic regulation or supply impedance to the signal path. Perhaps you could explain it for us slower learners.
All good fortune,
Chris
I deleted my post to this because I was mistaken about the details of the PSU I was writing about. My reply was not relevant. My apologies for the interruption.
Congratulations on mastering Ohm's law.
But I see noting about "resonance" or transient analysis. How about showing the impedance (or scattering) parameters of the filters so that you can assess the response. Z21 and Z22 should handle it but use any network parameters you wish. Your writeup assumes dual pole rolloff but never defines the transfer function poles or their damping factors. Your design has multiple roots in the OLH plane as well as at least 3 zeros at ∞. You haven't described the filter response at all but have just assumed performance. Not looking for theoretical synthesis, just enough basic filter analysis to verify your assumptions.
Before you openly dismiss 130 years of electrical theory and practice, at least try to make a convincing argument. 🙄
Filter resonance can cause problems.
I built a 12AX7-based RIAA preamp with a CRC pi filter common to both channels followed by independent RCRC filters for each channel.
The problem was that the final RC before the first stage 12AX7 had much too high impedance. I used a 4.7k resistor with a 33uF capacitor. That was adequate for ripple attenuation, but with the first stage's Rp of 200k ohms and rp of approx. 80k ohms the power supply was on the verge of motorboating. The preamp would burst into a brief infrasonic oscillation every once in a while, resulting in audible, short, 'boom' sounds during playback. I re-jiggered the decoupling parts values to a 43k series R for the RC filter while retaining the 33uF cap, with a 150k Rp on the input stage 12AX7. It was way overkill, but the first stage is now adequately decoupled, which cured the infrasonic oscillation issue. Why do I bring this up?
Because if you look at CLCLC filters, you will see that they introduce large resonant peaks at infrasonic or very low audio frequencies. This would not be a good thing, even if you don't hear them. A good PSU design should make sure these peaks aren't too large. One way to do that, as stephe suggested, is to switch to a CLCRC filter so there's only one of those large infrasonic resonances to deal with. With monster parts like 470uF 450V capacitors now readily available, this is achievable. Back in the day when affordable caps larger than 100uF were practically unavailable, sufficiently smooth power supply impedance was harder to implement with LC filters.
Does that make sense?
I built a 12AX7-based RIAA preamp with a CRC pi filter common to both channels followed by independent RCRC filters for each channel.
The problem was that the final RC before the first stage 12AX7 had much too high impedance. I used a 4.7k resistor with a 33uF capacitor. That was adequate for ripple attenuation, but with the first stage's Rp of 200k ohms and rp of approx. 80k ohms the power supply was on the verge of motorboating. The preamp would burst into a brief infrasonic oscillation every once in a while, resulting in audible, short, 'boom' sounds during playback. I re-jiggered the decoupling parts values to a 43k series R for the RC filter while retaining the 33uF cap, with a 150k Rp on the input stage 12AX7. It was way overkill, but the first stage is now adequately decoupled, which cured the infrasonic oscillation issue. Why do I bring this up?
Because if you look at CLCLC filters, you will see that they introduce large resonant peaks at infrasonic or very low audio frequencies. This would not be a good thing, even if you don't hear them. A good PSU design should make sure these peaks aren't too large. One way to do that, as stephe suggested, is to switch to a CLCRC filter so there's only one of those large infrasonic resonances to deal with. With monster parts like 470uF 450V capacitors now readily available, this is achievable. Back in the day when affordable caps larger than 100uF were practically unavailable, sufficiently smooth power supply impedance was harder to implement with LC filters.
Does that make sense?
Last edited:
As for voltage regulation, leave that to the engineers who designed your PTX. They'll get it right, and if they didn't, you need a different supplier. The only other situations where it would count would be if you were doing your own PTX or repurposing something like a "control" xfmr for a DC supply. For that, you'd be most concerned with the DCR of the windings, and form and utilization factors which were addressed in the Rectifier Applications Handbook (attached above). Otherwise, trust Hammond, Edcor or Stancor.
Most applications don't require extreme voltage regulation. One exception would be CRT colour TV sets since you need to maintain precise electron beam alignment if you're not looking to screw up the colour. Here's where you found
As for AC impedance, the Vixen PS operates at 365VDC and 320mA (design nominal) for an equivalent resistance of: 1K14. The output filter capacitor is 220uF, and at 30Hz, XC=24R11 -- well below the 0.1XC traditionally used for sizing bypass capacitors.