Filtering an LED light's power supply

[tim9000]

Hi,
Out of interest, today I was mucking around with a 160W LED, 240V, 50hz, outdoor light.
I looked at the Power Factor (it was 0.96) and the THD was 14.something %.
I then wondered what would happen if I put a capacitor in parallel with the light. It was a three phase 90uF cap, so since they were in Delta configuration I believe the effective capacitance was 135uF (90+45uF), which is something like -j23Ohms (if recollection serves).

Sooo, obviously this corrected the heck out of the PF, so it was almost zero (basically leading). And I think the THD went down to 11.5%.

'Good', I thought.
Not sure what the current did, but I expect it went up, a fair bit.

Then I thought, 'I wonder what would happen if I put an inductor in series with it'.

I think when I put the inductor alone, in addition to the light, it decreased the PF and decreased the THD slightly, possibly, I'll need to re-check.

So I put a 14.25mH in series with the active. Subsequently, the THD went up to 23+%!
I don't remember what the PF did.

'Interesting'; then I thought about it and my working hypothesis is that the LC resonant frequency for this is about 114 Hz. So perhaps 114 Hz is too close to the first and third harmonic?

But in conclusion, I'm not entirely sure why it decreased the performance so much when I had the inductor in series, and the capacitor in parallel, I initially thought it would be like an LC filter, but I have a suspicion that it was resonant on the third harmonic. Even then I'm not sure how that works. Thoughts?

Also, I was wondering, if you're filtering something like semiconductors, can your filter be too big? What would you do if you wanted to almost perfectly filter a non-linear load like an LED?

Thanks

 

[berkeman]

Sooo, obviously this corrected the heck out of the PF, so it was almost zero (basically leading). And I think the THD went down to 11.5%.

The capacitor shifted the phase of the current with respect to the voltage, so that's why the PF improved. The cap will not do much for THD.

What would you do if you wanted to almost perfectly filter a non-linear load like an LED?

The LED driver sounds like it already is a Power Factor Correcting design, or you would not be up at 0.96 to start with. The THD number is likely just from the finite input capacitance (usually low for PFC supplies). I'm not sure you can do much to filter that out.

BTW, what instrumentation are you using to measure the PF and THD? Does the THD measurement give you a list of the harmonics and current amplitudes?

 

[tim9000]

The capacitor shifted the phase of the current with respect to the voltage, so that's why the PF improved. The cap will not do much for THD.

The LED driver sounds like it already is a Power Factor Correcting design, or you would not be up at 0.96 to start with. The THD number is likely just from the finite input capacitance (usually low for PFC supplies). I'm not sure you can do much to filter that out.

BTW, what instrumentation are you using to measure the PF and THD? Does the THD measurement give you a list of the harmonics and current amplitudes?

I understand that the capacitor shifted the current with respect to the voltage, but I don't understand why the THD won't really be improved by it.

Yeah, no doubt the light has power factor corrective components.

What do you mean by the "finite input capacitance"? You mean you get a diminishing return for the bigger the capacitor? And you'd need an infinite capacitance for 0% THD?

I was using a Hioki power quality analyser, yes it will give me the individual harmonic current amplitudes.

How do you mitigate such THD?

Please find attached a comparison of the waveforms and DMM values.
When the load was just the LED, was it slightly leading or slightly lagging? (Also, 14% THD still seems quite high to me.)
You can see the approx -j23 Ohms in parallel with the load does draw more quadrature current. Do you think resonant frequency is the cause of increasing the THD when there was both an L and a C in/on the circuit?

Thank youP.S. I presume that the light works something like Example 2. here:
https://www.allaboutcircuits.com/technical-articles/understanding-thd-total-harmonic-distortion-in-power-systems/

Does S₁ switch at a high frequency? If it is this circuit, I still do not understand how the alone capacitor on the supply could distort the supply so much.

 

[rbelli1]

so that's why the PF improved

The PF went from almost 1 to almost 0. That is not really an improvement. It's actually the exact opposite of improvement.

BoB

 

[berkeman]

The PF went from almost 1 to almost 0. That is not really an improvement. It's actually the exact opposite of improvement.

BoB

Interesting, I misread that. I thought he said it went to almost 1.0 or unity. Good catch.

@tim9000 -- can you clarify what the capacitor did to the PF? (not the THD)

 

[rbelli1]

What was the input voltage of the lamp after putting the extra components in front of it?

BoB

 

[tim9000]

Interesting, I misread that. I thought he said it went to almost 1.0 or unity. Good catch.

@tim9000 -- can you clarify what the capacitor did to the PF? (not the THD)

No, when I said 'good' in the initial post, I was referring to the THD decreasing. The PF went from 0.96 to about 0.04. (I think this means it's leading by 91.53 degrees?)

 

[tim9000]

What was the input voltage of the lamp after putting the extra components in front of it?

BoB

It was only an inductor before the load. As you can see from the #3 'waveforms.png' DMM table, it didn't really do anything to noticeably decrease the voltage on the LED. The supply oscillates from 240V-245V anyway, and the inductor was only 14.25 mH.

 

[berkeman]

The PF went from 0.96 to about 0.04

That's pretty crazy (not you, your data). Can you post your test setup?

 

[jim hardy]

Your plots show voltage and current to the LED alone out of phase.
That seems counterintuitive , unless it's a 'darkness absorbing diode'?

So I inverted all your current traces and added vertical black lines at a couple voltage zero crossings.
Phase shift due to capacitor is evident, so i guess your current measurement was upstream of that.

Presence of both inductor AND capacitor seems to offend your LED lamp. I added vertical purple lines at the upsets. Interesting, they occur just after voltage peak.

If it is as you suggest, active PFC, then
My guess is they interact with the closed loop controller inside your lamp base . Did it sound strange ?
https://www.allaboutcircuits.com/technical-articles/how-the-boost-pfc-converter-circuit-improves-power-quality/
the one referred to in your link

Here's the current waveform it's trying to draw. (Ignore amps numbers , you didnt measure yours)

You didn't say which you placed adjacent the lamp, inductor or capacitor.

 

[tim9000]

That's pretty crazy (not you, your data). Can you post your test setup?

In keeping with the waveforms.png, here is a list of the circuit diagrams and a photo of the setup:

 

[tim9000]

...You didn't say which you placed adjacent the lamp, inductor or capacitor.

Hopefully my last diagram and setup pictures will be of use to you Jim. I'd like to get closer to the bottom of this phenomenon, because filtering non-linear loads is interesting to me. As does THD.

Your plots show voltage and current to the LED alone out of phase.
That seems counterintuitive , unless it's a 'darkness absorbing diode'?

Maybe the active and the neutral were connected wrong? So the current clamp was around the wrong way?

So I inverted all your current traces and added vertical black lines at a couple voltage zero crossings.
Phase shift due to capacitor is evident, so i guess your current measurement was upstream of that.

Presence of both inductor AND capacitor seems to offend your LED lamp. I added vertical purple lines at the upsets. Interesting, they occur just after voltage peak.

H'mm that is an interesting observation.

If it is as you suggest, active PFC, then
My guess is they interact with the closed loop controller inside your lamp base . Did it sound strange ?
https://www.allaboutcircuits.com/technical-articles/how-the-boost-pfc-converter-circuit-improves-power-quality/
the one referred to in your link

My power electronics knowledge is a little bit rusty, but I remember a little bit (duty cycles etc.) No, the LED didn't sound strange, but the inductor certainly did on start-up, and it was humming quite a bit, more than I'd have expected given that I wound it around powder core toroids.

Here's the current waveform it's trying to draw. (Ignore amps numbers , you didnt measure yours)

Remind me, so if that's the current waveform it's trying to draw, is that sharp triangular characteristic just a result of it not being able to draw a perfect sinusoid, so that's the best it can do?

That flat blip around the zero crossing, when it's just the LED load, what's causing that?
I imagine that the LED is forward biased 100% of the time through the aid of inbuilt capacitors, so what is not able to keep up just before/after the zero crossing to cause it not to draw any current then?

If people like, I can try and dig out the Harmonic current recordings from the Hioki, the % of each harmonic component; I'm sure they're still on the SD card.

 

[jim hardy]

That flat blip around the zero crossing, when it's just the LED load, what's causing that?
I imagine that the LED is forward biased 100% of the time through the aid of inbuilt capacitors, so what is not able to keep up just before/after the zero crossing to cause it not to draw any current then?

Well, we're speculating that it's this kind of device

in which C is likely charged to > peak value of Vac by the boost converter.

When Vac is near its zero crossing, ie very small, there's nothing to push current through L and S1 so no current can flow during a brief interval surrounding each voltage zero crossing. Hence the flat spots in current waveform ?

if that's the current waveform it's trying to draw, is that sharp triangular characteristic just a result of it not being able to draw a perfect sinusoid,

That's explained in the link

These currents are the general shape that they should be (sinusoid/rectified sinusoid), but one thing that stands out is that the lines of the signals look thick. This thickness occurs because during one cycle the current ramps up and then ramps down as the average current is controlled to track the reference sinusoidal voltage.


Zooming in on the inductor current reveals the repeatedly increasing and decreasing currents of the inductor as the system switches between the two states.

.

 

[tim9000]

Well, we're speculating that it's this kind of device
View attachment 206610

in which C is likely charged to > peak value of Vac by the boost converter.

When Vac is near its zero crossing, ie very small, there's nothing to push current through L and S1 so no current can flow during a brief interval surrounding each voltage zero crossing. Hence the flat spots in current waveform ?

That's explained in the link.

So the unavoidable issue is that the rectifier is acting like an initial little speed bump which the small post-zero-crossing voltage can't push any current over.

What it needs for better harmonics are some sort of load that turns on at this small voltage, and for the frequency of the reference controller to be infinitely fast.

I'm still intrigued about the impact of having a capacitor before the rectifier. Similarly, I was expecting that by putting an inductor in series that as the diode stopped conducting, it might have forced the inductor to create an MMF in the direction of the bridge and re-forward bias it, so the inductor could dissipate it's energy. Instead, it just seemed to make things worse.
I was also wondering why having the inductor and capacitor both in the circuit increased the THD so much (i.e. my previous hypothesis RE resonant freq).

Thanks

 

[jim hardy]

I'm still intrigued about the impact of having a capacitor before the rectifier.

Move your current probe to after the capacitor and see whether capacitor makes the LED's behavior change.

Similarly, I was expecting that by putting an inductor in series that as the diode stopped conducting, it might have forced the inductor to create an MMF ? in the direction of the bridge and re-forward bias it, so the inductor could dissipate it's energy. Instead, it just seemed to make things worse.

MMF ? EMF?

Might have, traces are too small to see.
Amplitude of current looks smaller and something changed at current's zero crossing

I was also wondering why having the inductor and capacitor both in the circuit increased the THD so much (i.e. my previous hypothesis RE resonant freq).

What's resonant frequency of this thing?

If you have two near resonant things in proximity might they interact? Ever live with a piano?

Are you sure about location of your voltage probe?

 

[tim9000]

Move your current probe to after the capacitor and see whether capacitor makes the LED's behavior change.

MMF ? EMF?

Might have, traces are too small to see.
Amplitude of current looks smaller and something changed at current's zero crossing

View attachment 206650

Sorry, Emf.

Unfortunately I no longer have access to the light, at first I thought that the current would have flowed out when the supply was going over zero crossing. But now, I don't think it really did anything significant.

What's resonant frequency of this thing?

As I said in my OP, from what I calculate from my L and C, they're resonant frequency for is about 114 Hz. I have no idea if the light had a resonant frequency, or if it did, how I would have calculated it. If it did have a resonant frequency LC in it, I would imagine that it would be much higher than the supply 50 Hz.

If you have two near resonant things in proximity might they interact? Ever live with a piano?

Are you sure about location of your voltage probe?

Yes and no, I don't get the piano analogy, do you think my 114 Hz LC resonance did have an effect? (I do, but how would you describe it?)
How would it have interacted with the lights internal resonance?

I am sure about the location of the voltage probes location.

 

[jim hardy]

I have no idea if the light had a resonant frequency, or if it did, how I would have calculated it. If it did have a resonant frequency LC in it, I would imagine that it would be much higher than the supply 50 Hz.

If that thing works as you suggested, with a full time internal boost regulator for PF correction,
that internal boost regulator is an active device with its own transfer function.
It has feedback so its transfer function has a denominator
and that implies gain and damping and a natural frequency.
All three of which we know zero about.
Changing conditions at its input, like adding series inductance, changes its response and we can only guess how.

As I said in my OP, from what I calculate from my L and C, they're resonant frequency for is about 114 Hz. I have no idea if the light had a resonant frequency, or if it did, how I would have calculated it. If it did have a resonant frequency LC in it, I would imagine that it would be much higher than the supply 50 Hz.

we can only speculate from your observations, the graphs.

Yes and no, I don't get the piano analogy, do you think my 114 Hz LC resonance did have an effect? (I do, but how would you describe it?)
How would it have interacted with the lights internal resonance?

Piano - in the middle the night when you're trying to sleep you'll hear the strings go into sympathetic vibrations with noises , like a conversation in next room or a radio nearby or something outside.
How interact ? Inductance in series (without parallel capacitance) adds to that of boost regulator's internal L , affecting regulator response.

I'll speculate too now.

All your voltage traces look the same. Text block on right is too small and blurry to make out.

LED alone

voltage = ~2 div P-P , reasonable approximation of a sinewave
current = ~ 2.3 div p-p, in phase with voltage, flattened at top and at zero crossing. Looks symmetrical so distortion is odd harmonics.

LED and capacitor

Voltage looks same so current drawn by LED shouldn't change.
Current now up to about 5.9 div p-p and shifted nearly 90 deg ahead of voltage. That suggests increased current is all drawn by capacitor.
Exaggerated distortion in current wave looks to me like high harmonics , 7th or 9th is evident which suggests this thing is powreed by a Ferroresonant Voltage Regulating transformer (Sola or something).
Capacitor bank is much lower impedance at harmonic frequencies so the harmonics are visible there in the current signal long before you'll notice them in the voltage waveform. A crude high-pass filter if you will.
Current through capacitor alone probably won't look much different.All 3 loads ---- same as last one but with series inductance 14.25 mh = 4.48 ohms @ 50 hz, higher at each harmonic by the order of that harmonic

Voltage looks the same to me still ~ 2 div p-p.

Current wave still about 5.9 div p-p but with those new spikes shortly after voltage zero crossing... what's going on there ?
I'll venture a USWAG(Unscientific Wild A** Guess)
That thing's boost converter does not run for complete line cycle, just near voltage zero crossings.
It starts after peak to maintain current into LED's internal filter capacitor and stops before next peak
the spike is overshoot on booster startup , inductor is higher z at switching frequency so the first few gulps of current deplete the capacitor.

That's my best guess from your observations.
What do you conclude ?

 

[tim9000]

Sorry, I didn't realize the picture would not upload so compressed and blurry. Not that I expect anyone is interested enough, but if they are I can upload just the RHS with the DMM measurement tables, so they're legible.

active device with its own transfer function.
It has feedback so its transfer function has a denominator
and that implies gain and damping and a natural frequency.

My Automatic Control theory is very rusty. So the TF is the Output/Input from memory, which IS the gain? So Why does the gain having a denominator imply the feedback having a damping and natural frequency? (Hopefully that question was reasonable)

Could you please elaborate on what a 'Ferroresonant Voltage Regulating transformer' is and what it is comprised of?

Interesting theory. One thing I still don't quite get is when you said "Capacitor bank is much lower impedance at harmonic frequencies so the harmonics are visible there in the current signal long before you'll notice them in the voltage waveform. A crude high-pass filter if you will."
wouldn't it be a low pass filter? Because the Capacitor is in parallel with the load, so at each higher harmonic, won't it be closer to a short across the load?

Another similar thing that is still playing on my mind is, if you have an LC circuit, (L in series, and C in parallel) circuit such as this, which feeds the load, if your supply is operating at 60 hz less than the resonant frequency of this LC circuit, does being that close to the resonant frequency have any effect? (Assuming so, how would I calculate it? Say I'm operating at 50 hz, and rf is 114 hz)

Well, if nothing else, I've learned the danger of trying to play with active components.

Thanks again

 

[jim hardy]

So Why does the gain having a denominator imply the feedback having a damping and natural frequency?

Trabnsfer function includes time response , forward transfer function is just say G , no denominator
Wrap feedback H around it and the transfer function becomes G / (1 + GH) , aha, it sprouted a denominator

Both numerator and denominator are complex.
Any frequency that makes the denominator small compared to numerator increases its gain.

Could you please elaborate on what a 'Ferroresonant Voltage Regulating transformer' is and what it is comprised of?

It's a commonly used voltage stabilizing transformer, usually just called "Sola" transformer.
http://www.solahevidutysales.com/cvs_hardwired_series_power_conditioner.htm
http://www.emerson.com/documents/automation/163818.pdf
I guessed from the harmonic evident in current wave that your lab bench might have one.
http://www.electroncoil.com/ferroresonant_transformers.php

wouldn't it be a low pass filter? Because the Capacitor is in parallel with the load, so at each higher harmonic, won't it be closer to a short across the load?

Indeed the higher the harmonic the closer to a short is your capacitor bank. So the harmonic content of CURRENT will be higher than it would for a resistive load.
You plotted current and it's mostly through your capacitor bank, so its harmonic content compared to a voltage plot will be exaggerated . The capacitor high passes current and that's what you plotted.
Sola transformers have third harmonic filters and claim 3% total . Much of that is higher harmonics. I measured after an R-L high pass in my rod position system 7th 9th and 11th of several per can't i don't recall exactly how much..

Another similar thing that is still playing on my mind is, if you have an LC circuit, (L in series, and C in parallel) circuit such as this, which feeds the load, if your supply is operating at 60 hz less than the resonant frequency of this LC circuit, does being that close to the resonant frequency have any effect?

Use your basics.
About an octave away from resonance? That depends on Q, doesn't it ?

 

[tim9000]

I think the "forward transfer function" is also referred to as the 'open loop transfer function', which rings a bell.

You plotted current and it's mostly through your capacitor bank, so its harmonic content compared to a voltage plot will be exaggerated . The capacitor high passes current and that's what you plotted.

With respect to the supply, does this mean that the detrimental impact of the THD on the supply, did not actually get worse when the capacitor was inserted into the circuit? I thought it still would, because the supply doesn't care if the current goes through the capacitor or if it goes through the LED, (but it means the performance of the LED itself didn't deteriorate)

About an octave away from resonance? That depends on Q, doesn't it ?

I'm not sure what defines the Quality factor...? I think maybe resistance defines the width...
I have not plotted a logarithmic graph in a great while. I am envisioning a graph for the LC-acting as a filter, I suppose like a band-stop at resonance; where on the Y-axis is the amount of current passed, and on the X-axis is the frequency content of the supply. And what this would look like is a trough, where the lowest part is when the supply is 1/(2*pi*√(LC)) which I think is 114 hz.
But I'm not sure how to translate from this conceptual grasp/understanding, into real world application, where I can say 'ah yep, at 50 hz the attenuation from the LC to the load will be ...%'. Because at this frequency you treat the capacitor as a short. But near it, you can calculate the LC impedance at a specific frequency, but I'm not sure how you'd separate out just the impedance of the capacitor, because it's the capacitor that would make the most difference to the load. I'm not sure if this proximity to resonance will be insignificant, or significant, given that resonance is double the supply frequency.

If it were 3x the supply frequency than I imagine it would be good at removing the 3rd harmonic?

I'll look into ferro-resonant transformers, because voltage regulation, saggs and swells...filtering harmonics is interesting to me

 

[jim hardy]

does this mean that the detrimental impact of the THD on the supply, did not actually get worse when the capacitor was inserted into the circuit?

That's what i surmised from the voltage wave.
I assumed you're plugged into a wall outlet that's pretty stiff.
The change in current wave when capacitor added suggested to me the voltage wave has higher order harmonics in it .

 

[tim9000]

That's what i surmised from the voltage wave.
I assumed you're plugged into a wall outlet that's pretty stiff.
The change in current wave when capacitor added suggested to me the voltage wave has higher order harmonics in it .

Wait, so you mean the THD on the supply didn't get worse only because I'm plugged into the infinite bus on the wall? Or because the total load actually didn't draw/create more THD?

Are you saying the THD from when the capacitor was coming in was higher order harmonics already in the supply? Because, it only occurred when the inductor was also present (hence why I thought it was related to resonant frequency).

With a ferroresonant transformer, what is the significance of the capacitance value? I assume it's related to the inductance, but is this the inductance of the secondary, or both primary and secondary...or just primary inductance? And how far is the secondary part of the core into saturation, by design for normal operation?

Cheers

 

[jim hardy]

Wait, so you mean the THD on the supply didn't get worse only because I'm plugged into the infinite bus on the wall? Or because the total load actually didn't draw/create more THD?

that's what i would guess.
Wall outlet capacity is large compared to the light .
135 uf at 50 hz is 23.5 ohms? So you had almost ten amps through the capacitor and one through the LED? So that current is 90% capacitive and 10% LED ?
How much of each harmonic is available to drive current through the capacitor? What is Z of capacitor at each harmonic ? If you had Fourier coefficient of each harmonic in the voltage and the current waveforms you could check to see whether they're in proportion.

Are you saying the THD from when the capacitor was coming in was higher order harmonics already in the supply?

That's exactly what i hypothesized to explain your observed current wave.
What experiment could you propose to test that hypothesis?
What did you conclude from the waveform ?

what is the significance of the capacitance value?

It determines how many amps per volt you get at every frequency present in the voltage wave.

That your 90% capacitive current changed shape when you added inductance in series with the source suggests that it blocked some higher harmonic currents, which should change the current wave shape.
That it looked different on opposite sides of the current peak suggests something else was going on too. That's why i hypothesized LED has part cycle power factor correction.

Have you never done the exercise of constructing on graph paper an arbitrary waveform by adding together a fundamental and harmonics? If not , try it.

 

[rbelli1]

Try putting just the L and C components on the powerline with no light.

BoB

 

[tim9000]

I expect that Zc would decrease with every odd multiple of the fundemental, given that 1/(jwC) where w is 1,3,5,7 etc.

That your 90% capacitive current changed shape when you added inductance in series with the source suggests that it blocked some higher harmonic currents, which should change the current wave shape.

Well that is what I would have expected, however the THD went UP to 23%, so as you say, some other problem, which we may not be aware of, is popping up.

That's exactly what i hypothesized to explain your observed current wave.
What experiment could you propose to test that hypothesis?
What did you conclude from the waveform ?

Makes perfect sense, but doesn’t help clarity.
If I still had the LED light, I could maybe set up an alternate supply, from a generator, and see if the same effects were observed. Either that or modify the set-up, I could still set-up but with a regular incandescent light.

I conclude from my waveforms that if you want to add reactance, that it has to be in in proportion with the load, otherwise you make a bloody mess

It determines how many amps per volt you get at every frequency present in the voltage wave.

That your 90% capacitive current changed shape when you added inductance in series with the source suggests that it blocked some higher harmonic currents, which should change the current wave shape.
That it looked different on opposite sides of the current peak suggests something else was going on too. That's why i hypothesized LED has part cycle power factor correction.

Have you never done the exercise of constructing on graph paper an arbitrary waveform by adding together a fundamental and harmonics? If not , try it.

I'm going to check which harmonics went up when the inductor was in series, because I'd have thought that the higher order ones would decrease (yet the total harmonic dist did actually go up).

'At every frequency'...hmm, Is there a good layman explanation of the operation of a ferroresonant TX? I’m wondering if you were designing one for 50 hz or 60 hz, supply frequency for something small, like 1kW at 230V or 120V, how would you know how big to have your capacitor?

So is the secondary part meant to only just be in saturation region under normal operation? Or anywhere in particular? If they're quite inefficient than I'd expect that under normal operation the secondary region is quite saturated.

I remember doing superposition back in high school and I have a modicum of appreciation for Fourier series, but I should probably brush-up.

Thank you

 

[tim9000]

Try putting just the L and C components on the powerline with no light.

BoB

Hi Bob,

See, this is what I'm not sure about: The resonant frequency for this will be 114.7 hz, the Cap will be -j23.6 Ohms @ 50 hz, the inductor j4.48 Ohms @ 50 hz.
So I expect this would mean that the current would be 240V/(j4.48 - j 23.6) = j 12.6 A (which is a fair bit, but less due to a small amount of resistance)

But what I was expecting was that there was a relationship between the resonance between LC and the parallel load, so you could calculate how much of a 'short' that cap was acting as, at whatever frequency. But maybe resonance is just a rough guide and maybe you have to calculate it as per basic circuit theory (i.e. parallel impedances, KVL, KCL).

Cheers

 

[rbelli1]

But what I was expecting was that there was a relationship between the resonance between LC and the parallel load,

Well if you put the LC in there with no light and get identical results as before you know it is not the light. If you get perfect sine waves then you know it is the light. In between is part both.

Do you still have access to any of the parts?

BoB

 

[jim hardy]

I conclude from my waveforms that if you want to add reactance, that it has to be in in proportion with the load, otherwise you make a bloody mess

We learn AC circuit theory with linear components and sine waves.

As soon as we encounter a nonlinear device and distorted sine waves we have to decide how much of a compromise we will settle for and analyze to that.

Ferroresonant transformers operate well saturated and run hot to the touch.
Most have a third harmonic filter to improve the sinewave output but higher ones sneak through.

http://www.sunpower-uk.com/glossary/what-is-a-ferroresonant-power-supply/
http://www.powerqualityworld.com/2011/04/constant-voltage-ferroresonant.html
http://www.oltronix.nl/en/ferroresonant-principle
https://www.ustpower.com/comparing-...esonant-transformer-advantages-disadvantages/

 

[jim hardy]

Wiki has a brief mention of ferroresonant transformer operation here

https://en.wikipedia.org/wiki/Voltage_regulator

Beware , one place they say output is a "perfect" sinewave. Well,your eye won't see a few % distortion on a 'scope. A few sentences further they admit :

output distortion, which is typically less than 4%,

Sola custom built some for us with 2%. Waveform was beautiful. High passing it t exaggerated the distortion.

 

[dlgoff]

Sola custom built some for us with 2%. Waveform was beautiful.

[offtopic]Just curious what they were used for.[/endofftopic]

 

[jim hardy]

[offtopic]Just curious what they were used for.[/endofftopic]

For a reactor control rod position indicating system. We called it RPI .

Control rods on a PWR come out the top of the core. They are lifted by a drive shaft which extends into a pressure housing atop the reactor vessel.
Here's my rough functional sketch of one, we had 45 of them.
Dome is reactor head, rectangle atop it is a nonmagnetic stainless steel pipe that's the pressure housing.
Rectangle inside the pressure housing is the drive shaft that lifts the control rod out of the core twelve feet below.

Orange winding is secondary which carries effectively zero current. So it's a flux detector.
Its induced voltage gets rectified and filtered to indicate position of the drive shaft (and its attached rod.)That 120 VAC supply and the 500ohm resistor set the current through the primary winding.
Observe that since flux Φ is in proportion to i which is set mostly by R(because X_L was only around 50 ohms),
and secondary voltage is dΦ/dt , we have effectively differentiated (high passed if you prefer) it exactly like Tim9000 did except with inductance not capacitance .
So harmonic content is exaggerated in the secondary voltage signal.

Our RPI system was specified to operate with maximum of 1% total harmonic distortion in the supply . Our custom built inverters met that.
The Sola transformers were the backup supply to this system and we settled for best they could do, 2%, and that worked fine.

Someplace i have 'scope traces . But that's another story for another thread.
Wish i knew enough about something to write a PF Insights Article - these old power plant lessons were painful to learn but are fun to look back on.
Excuse an old man's boring anecdotes - they're fond reminiscences You are very kind.

Anyhow, in a nuke plant a portable FFT analyzer is really a lot of fun .old jim

 

[Asymptotic]

[offtopic]Just curious what they were used for.[/endofftopic]

Sola CVTs were heaven-sent when programmable logic controllers and other early industrial computers still used linear DC power supplies sensitive to AC line variation , and particularly when they shared an AC bus with high power SCR DC motor drives. Secondary waveform was practically unaffected even with fairly ugly commutation notches and spikes on the primary windings.

A CVT also makes a severe overload or short circuit on the secondary circuit painfully obvious. Rather than clearing the secondary fuse, what often happened was voltage collapsed, PLC outputs turned off, (with the short turned off) voltage was restored, PLC rebooted, and the sequence repeated. You could tell when a solenoid coil had shorted by watching cabinet pilot lamps flashing in unison every few seconds.

 

[jim hardy]

A CVT also makes a severe overload or short circuit on the secondary circuit painfully obvious. Rather than clearing the secondary fuse, what often happened was voltage collapsed, PLC outputs turned off, (with the short turned off) voltage was restored, PLC rebooted, and the sequence repeated. You could tell when a solenoid coil had shorted by watching cabinet pilot lamps flashing in unison every few seconds.

They indeed collapse so aren't much good for clearing faults.
One can however use that to advantage.
Consider a panel of branch circuits feeding instrumentation in a big factory or power plant.
Sometimes they're fed by battery powered inverters for ultimate reliability. Inverters are also weak when it comes to fault clearing.
A branch circuit that's fed by a CVT won't draw very much fault current from the upstream source. That means other branch circuits fed from same upstream source won't see a voltage perturbation from the fault on the CVT isolated branch .
That's useful when your upstream source also has limited fault clearing capability, such as inverters feeding instrument buses in a power plant. A fault on one branch circuit must not "glitch" the other branch circuits.
That's tripped many a nuke plant.
APRIL 1973 twice in two days...

https://cdnc.ucr.edu/cgi-bin/cdnc?a=d&d=DS19730403.2.171Inrush current to a SMPS equipped computer looks an awful lot like a fault. CVT is handy there too, to limit inrush to some value that doesn't glitch everything else on the bus..
CVT can be useful also for starting a motor that has high inrush. Makes the start a lot more gentle.

Like any other transformer CVT's are susceptible to first half cycle saturation should you have the bad luck to energize right at voltage zero crossing. So if you're using one to limit inrush, operate it at about 70% nominal input voltage else you'll get an occasional current spike on energize.

Personally I do like CVT power supplies where ultimate reliability is needed. They're just too simple to fail. No output regulator to latch up from a "spike" and trip the overvoltage crowbar.

my two cents, hope it helps somebody somewhere.

Wow really got off topic didnt i ? Sorry, guys.

old jim

 

[tim9000]

If CVTs weren't so inefficient they'd be perfect. So

I’m wondering if you were designing one for 50 hz or 60 hz, supply frequency for something small, like 1kW at 230V or 120V, how would you know how big to have your capacitor?

That 120 VAC supply and the 500ohm resistor set the current through the primary winding.
Observe that since flux Φ is in proportion to i which is set mostly by R(because XL was only around 50 ohms),
and secondary voltage is dΦ/dt , we have effectively differentiated (high passed if you prefer) it exactly like Tim9000 did except with inductance not capacitance .
So harmonic content is exaggerated in the secondary voltage signal.

Sorry, can you explain how the control rod is a high pass filter? So you're saying that the flux is in phase with the primary current, which is basically a 'real' impedance, but the secondary voltage is the differential of that, but I don't see the connection. Also, what did you mean, what I did expect with inductance, but not capacitance?

Thanks

 

[jim hardy]

but the secondary voltage is the differential of that, but I don't see the connection.

induced voltage e = L di/dt . Since primary and secondary link the same flux they see same induced voltage.

I = V/ Z = V / (R +jX) and since R >> jX,,
I ≈ V/R
so e = L d(V/R) dt = L/R dV/dt and that's differentiation. Each harmonic gets exaggerated by its order.

Also, what did you mean, what I did expect ?? with inductance, but not capacitance?

I don't remember asking that.

I drew a parallel - your example differentiated using capacitance, my example differentiated using inductance.