hpasternack said:
Work's been too busy to spend much time here, to limit the answer to two words I'ld have to say: Patrick Turner. Why he stays in that sty is a mystery.
I did squeeze a little time to examine the behaviour of the low LCR supply in PSUD and Spice. Since part of the trick is getting the front LC section resonating it raised two possible enhancements. The first was 'unterminating' the section by following it with a low F, high L section to keep the impedance high and islolate with inductance. The second was to cascade two low LCR sections tuned to the same frequency (~240 Hz) to augment the resonance. Both work and work well. Can't post particulars, they're all at home.
jlsem said:
Absolutely. He stated clearly that his work was limited to a series of simulations. Further he stated that he didn't have a lot of time to continue the work and hoped that someone else would carry on with it. All of this appears in the single thread that I linked to, if not in a single post.
Generally, I give a lot more weight to someone's ideas if they have taken the time to implement them and listen. In some cases that's not necessary. In this case I think that Henry's findings are more than interesting enough without him having built the circuit. Really, I think it would have been a shame if he had not shared his findings, but that's just my opinion.
I don't think so. There are a few members of the forum that are put off by his demand for rigor. There are a smaller number of members with which there is a "personality clash." But, I think the majority of asylum members appreciate his posts or, at worst, simply don't bother reading them.
-- Dave
Huh. I've had a number of extended technical debates with Patrick and have found him to be very stubborn and resistant to persuasion, even by very elegant engineering arguments. He doesn't know his own limits.
On the plus side, Patrick seems to be a very experienced technician. I think r.a.t. suits him because he doesn't feel challenged there.
A few years back, Jute unleashed himself on Patrick with all his fury. It was interesting to watch the irresistible force go up against the immovable object. In the end, the two seem to have formed an alliance, unholy as it may be. I think in a way they need one another.
Oh well.
-Henry
The simplest way to test this idea in the simulator is to load the first section with a pure current source, which is an infinite AC load. You'll see the performance doesn't change. In my opinion, the operation of the circuit is not well described as a resonant tank. I don't think you'll get better performance by cascading 200Hz resonant sections. After all, the frequency of operation is 120Hz.
It might help to do some reading on SMPS boost converter circuits. You'll see some similarities, and probably get some insight into how this circuit works. The key is the energy storage and subsequent release from the input inductor.
-Henry
Take a look, tell me what you think. The AC voltage across the second capacitor is higher than the first, the end ripple ~8 mv p-p, the reg not bad at all. C3 = 200 uF and C4= 133 uF represents parts on hand and have no effect on DC voltages or regulation, just settling and ripple. The DC output didn't change with C3 as high as 2000 uF.
The attachment is a zipped png to get around the forum's resolution limit.
The attachment is a zipped png to get around the forum's resolution limit.
This is an interesting complication. Now you have a resonant filter. The frequency you want is 240Hz, the second harmonic of the 120Hz ripple waveform. But you've forgotten that L2 is resonant with C1 as well at C2. Try changing L1 to around 117mH and see if you can't get a more pronounced 240Hz resonance out of the circuit.
If you tweak L2 carefully, you'll see some pretty dramatic changes in the shape of the L1 current waveform. I agree that this configuration seems to give better regulation than the one without the resonant second section. I'm a bit concerned about deliberately provoking a resonance at a harmonic of the power line frequency, though. If the operation of the supply depends critically on the tuning, you may run into trouble as the inductor values change with load current.
Still, the variation you propose is worth investigating.
-Henry
If you tweak L2 carefully, you'll see some pretty dramatic changes in the shape of the L1 current waveform. I agree that this configuration seems to give better regulation than the one without the resonant second section. I'm a bit concerned about deliberately provoking a resonance at a harmonic of the power line frequency, though. If the operation of the supply depends critically on the tuning, you may run into trouble as the inductor values change with load current.
Still, the variation you propose is worth investigating.
-Henry
hpasternack said:
Accounting for C2 was the reasoning behind choosing stock values of L (Triad or Panasonic, don't recall) which doubled in the second stage. The lower first L suggests lower Q and less critical values required of C1. Sims confirm this, raising C1 50% has minor impact. The larger L2 with a reduced C2 acts to reduce the second section capacitance seen by the first and raises the Q of the second to drive more 'action' - of whatever type is happening - from it. L3 is a much larger off-the-shelf value (Hammond) with low DCR to decouple stage two from C3. Sims also suggest it's working, as above increasing C3 to 2000 uF has no impact on DC out or regulation.
Increasing L1 to 117 mH causes the regulation to collapse to approximately the value expected from a straight C input filter but with ~30 volts lower Vdc. One iteration of this circuit (possibly with 3 cascaded low-L section) had almost double the regulation of the posted example. Increasing the load current from 50ma + 50ma to 100ma + 100ma resulted in ~4 Vdc drop. It was however overly sensitive to tuning.
The circuit I experimented with briefly in the simulator is shown in the following images. The first is more or less your original configuration:
The second is with a larger L2 sized to resonate with C1 and C2 in series at 240Hz, more or less:
I really haven't spent much time thinking about this. It seems in the second image that the C2 waveform is getting closer to a 240Hz sinusoid. The effect on the L1 current is interesting.
I agree the simulated output voltage is higher and the regulation is better in the configuration you've developed. I haven't spent enough time looking at this to get a handle on the essential differences between the original version and your two-stage resonant version. It would be nice to have a simple, comprehensive theory that takes the resonance into account.
I should add that it isn't a given that a choke will look like a high impedance when it's sitting in a resonant circuit. You might want to calculate that magnitude and phase as a function of frequency of the impedance looking into the filter at the inputs of L1 and L2.
-Henry
An externally hosted image should be here but it was not working when we last tested it.
The second is with a larger L2 sized to resonate with C1 and C2 in series at 240Hz, more or less:
An externally hosted image should be here but it was not working when we last tested it.
I really haven't spent much time thinking about this. It seems in the second image that the C2 waveform is getting closer to a 240Hz sinusoid. The effect on the L1 current is interesting.
I agree the simulated output voltage is higher and the regulation is better in the configuration you've developed. I haven't spent enough time looking at this to get a handle on the essential differences between the original version and your two-stage resonant version. It would be nice to have a simple, comprehensive theory that takes the resonance into account.
I should add that it isn't a given that a choke will look like a high impedance when it's sitting in a resonant circuit. You might want to calculate that magnitude and phase as a function of frequency of the impedance looking into the filter at the inputs of L1 and L2.
-Henry
What's interesting, on further consideration, is that in your circuit, C1 is actually resonant at 360Hz. A current pulse from L1 "kicks" the tank, which is a pi network consisting of C1-L2-C2, and then the whole thing rings at six times the power line frequency.
Increasing the choke size drops the resonant frequency to 240Hz, but introduces the peculiar bifurcated L1 current due to the phase relationship between the T1 secondary voltage and the C1 ripple voltage. (I apologize for the inconsistency between the schematics and the trace colors of the first and the second two images.)
Slowing down the tank even more restores a more normal looking charging current. The exact resonant frequency isn't important because the circuit synchronizes itself each time the diodes switch on.
I have a feeling that all of this would be very familiar to vacuum tube television set designers.
-Henry
An externally hosted image should be here but it was not working when we last tested it.
Increasing the choke size drops the resonant frequency to 240Hz, but introduces the peculiar bifurcated L1 current due to the phase relationship between the T1 secondary voltage and the C1 ripple voltage. (I apologize for the inconsistency between the schematics and the trace colors of the first and the second two images.)
An externally hosted image should be here but it was not working when we last tested it.
Slowing down the tank even more restores a more normal looking charging current. The exact resonant frequency isn't important because the circuit synchronizes itself each time the diodes switch on.
An externally hosted image should be here but it was not working when we last tested it.
I have a feeling that all of this would be very familiar to vacuum tube television set designers.
-Henry
It was anticipated going in. This was intended as a rough, proof-of-concept sanity check.
hpasternack said:
The proverbial 'ah-ha!'. This explains the low ripple with low to moderate final stage filtering. I didn't have time, or honestly much inclination, to confirm the ripple frequency. Whether lower level at higher frequency is an improvement is an open question. When time becomes available I'll drop these circuits into Spice and examine the impedances presented to the rectifiers.
Diggin' up an old thread...Jeff is at it again....-> "Also, do NOT ever build a 2A3 amp with any R's as the filter to the finals. Never!!
The 2A3 is a PEAK current hungry tube, and it demands you use as little series resistance in the power supply to the finals as possible. IF YOU DON'T HAVE ROOM, GET A NEW BIGGER CHASSIS."
Of course, I argue with him, but mostly just to create some opposition to his zealous design rules. If no one steps up and says 'that ain't right,' then people who read the forum may believe him.
Boris_The_Blade said:
-> "Also, do NOT ever build a 2A3 amp with any R's as the filter to the finals. Never!!
The 2A3 is a PEAK current hungry tube, and it demands you use as little series resistance in the power supply to the finals as possible. IF YOU DON'T HAVE ROOM, GET A NEW BIGGER CHASSIS."
Of course, I argue with him, but mostly just to create some opposition to his zealous design rules. If no one steps up and says 'that ain't right,' then people who read the forum may believe him.
I never question other people's belief systems, even when they are engineers with MS degrees. It is their right to believe in whatever they want to believe. I can explain my point of view, I can explain details I know, but it is not my business to change their beliefs, so as soon as I see beliefs VS facts or opinions I stop the discussion, though sometimes it is too late, so they already got mad. It is easy to get people mad questioning their beliefs.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.