Isolated Power Supply world-wide AC input

Part 6

spectrum 30MHz
Figure 94
Span 10kHz to 30MHz, we see switching frequencies starting 100kHz (wide spreading) Blue Capacitor voltage curve shows a significant pole, capacitor impedance rise up at 2MHz. 

Comparison Table Output Frequencies

85VAC rms 260VAC rms
AC-coupled output
& Capacitor Current
  • Switching starts at 50kHz
  • Switching stops at 3MHz
  • discrete frequencies
  • higher amplitude
  • Switching start at 100kHz
  • Switching stops at 3MHz
  • spreading frequencies
  • lower amplitude
Figure 95
EMC test should be done with low 85VAC rms, normal 230VAC rms and at maximum voltage.

AC-line Current Spectrum @85VAC rms

85VAC rms line current
Figure 96
line current @85VAC rms. Waveform from the supplying safety transformer looks similar, sharp needles near zero created by the DUT.

31MHz spectrum, AC-line current 85V
 Figure 97
Span 100Hz to 31.25MHz
Switching frequencies at 50kHz.

Figure 98
Span 10Hz to 100kHz
The odd line harmonics are dominant.
A perfect PFC would show only a single 50Hz frequency - impossible. 

AC-line Current Spectrum @260VAC rms

Figure 99
@260VACrms. Sharp needles still significant.

260VAC 31Mhz
Figure 100
@85VAC rms, AC-line current (line side), Switching frequencies at 100kHz

260V 100k
Figure 101
@260VAC rms, AC-line current (line side), increased distortion.

Line Current after LC Filter (PFC side)

85V input line current
Figure 102
Line current after input LC filter. Looks similar to rectified primary coil current. Core in saturation under max. amplitude.

85VAC line spectrum 30M
Figure 103
Input line current after LC filter, compare to Figure 97.

Effect of input filter
Figure 104
@85VAC rms, 16.3Wout. Both diagram under same scaling and conditions. Filter removes unwanted switching signals.
  • Top: Line Current with filter (line-side)
  • Bottom: Line Current after filter (PFC-side)

Changing Input Bridge Diode Rectifier to a smaller package

small bridge rectifier 85VAC rms
Figure 105
Using a smaller package for saving board space.

@80VAC rms -True RMS for a half sine wave = 290mA rms.
@85VAC rms -True RMS for a half sine wave = 250mA rms.
@105VAC rms -True RMS for a half sine wave = 210mA rms.
@160VAC rms -True RMS for a half sine wave = 140mA rms.
@230VAC rms -True RMS for a half sine wave = 100mA rms.
@260VAC rms -True RMS for a half sine wave = 92mA rms.

Rectifier is within absolute maximum ratings under all conditions. 85VAC rms cause highest input currents. Part require PCB copper for cooling, recommended. Consider lowest AC-line voltage and highest operating temperature.

Absolute Maximum Ratings:
max. 500mA rms @ Tc=25°C
max. 300mA rms @ TC=80°C

1µF/50V Ceramic Capacitor necessary at output? - YES

1uf removed
Figure 106
1µF/50V ceramic removed from output

1uf soldered
Figure 107
1µF/50V soldered to output

1uf both diagram
Figure 108
Figure 106+107, 1µF soldered direct on output, improves output spectrum. Both measurements done with the same settings. Capacitor useful, because a post added linear voltage regulator is in general not able to rejects higher frequencies. Capacitor prevents post electronic from higher frequencies.

Current in 1µF/50V Ceramic Capacitor

1uf current
Figure 109
Current in 1µF, 5A peak Under higher frequencies the electrolytic capacitor already behaviours as inductor and becoming useless for higher frequencies.

Capacitor can be enlarged to 1µF-4.7µF, check datasheet for max. allowed peak and rms currents.

Capacitor operates on a max. 16.3Wout power source, in case of cracking and shorted ceramic, danger - overheat or in worst case fire. Use a capacitor series with high resistance after cracking ceramic, available under different names. 

1uf spectrum
Figure 110
spectrum in the current of the 1µF ceramic.

Current in the Resonant Capacitor parallel to MOSFET Drain-Source

Figure 111
220pF/1kV COG capacitor peak current of 1.3Apk @85VAC rms

Figure 112
220pF/1kV COG capacitor, peak current 2.6Apk @260VAC rms. Maximum peak under actice clamping TVS diode at high Drain voltage conditions. Consider max. allowed capacitor voltage and be carefull when doing layout for enough distance between these high voltages.

Waveform 800V, 1A Ultra Fast Rectifier Diode on Clamping network-primary coil

Figure 113
Diode current and voltage, fast diode turn-on peak. Not EMC friendly, good for efficiency. Peak has a 40ns pulsewidth only.

Figure 114
Turn-on peak, current increased to 1.6Apk. Pulsewidth increased.

Figure 115
Ultra fast diode current, 39mA rms for one sine cycle. rms and peak below absolute max. value.

Measure the Input EMC Filter

input filter
Figure 116
Photo shows the main parts of the line filter.
  • Evaluation Board use 2*47mH (coupled) + 2*100nF X2
  • shown filter here 2*47mH (coupled) + 2*220nF X2
Main task of this filter: reduction of the fundamental and the harmonics of the switching current. This filter is important to reach a good powerfactor with the target of an current waveform following exactly the waveform of the input voltage. Filter reduce high-frequency components in the line current, necessary for a EMC conform design.

network analyzer
Figure 117
Network Analyzer 3577A together with 50 ohm S-Parameter Measurement Set
  • 5Hz to 200MHz measurement range
  • Inputs selectable between 1 Megohm and 50 Ohm
  • high dynamic range and precision
  • easy to operate
  • fine instrument for low frequency up to 200MHz
  • its a perfect instrument for a wide range of applications
This network analyzer can response the transfer function amplitude vs. frequency, if necessary also the phase vs. frequency function (not shown here).  All S-parameter directions S11, S21, S12, S22 can be measured comfortable.

Forward and reverse direction - pressing only a few buttons. Measurement can be normalized for highest precision.

ampliitude response
Figure 118
Filter amplitude vs. frequency
  • 3kHz: Filter has a peak at 3kHz - this higher order low-pass has a good quality factor, there is not much damping resistance in the network. (Resistance cause heat and loss of efficiency).
    • fres= 1/ 2*Pi * sqrt ( 47mH*220nF) = 1565Hz , may be I have to half L and C in this calculation to hit the 3kHz peak, I don't want to think too much about it now, its too late in the evening.
    • It is impossible to design peakless LC filters without damping resistors. Any frequency in this peak area will be amplified by some dB. The peak add an additional fast changing phase-shift. Both are not good for reaching a better powerfactor, causing sine waveform distortion.
  • 15kHz: approx. at 15kHz the filter starts damping up to 900kHz, very fine response with -90dBc, excellent for a coil with many windings and high copper densitiy.
  • 900kHz: the filter change it´s face from low-pass to high-pass characteristic. At this frequency, complex capacitive and complex inductive impedance having the same amount and cancel out each other. The remaining ohmic resistance-@900kHz, determine the -90dBc.
  • 95MHz: filter impedance converts again to a lowpass filter.

Simulation of the Network Analyzer Measurement

Figure 119
Trying to find a schematic for the network analyzer measurement. The analyzer has a 50 ohm source impedance and 50 ohm load on the receiving channel.

The inductor has two 47mH windings coupled on the same core. I guess as coupling factor 95%. On both sides of the double-coil are 100nF foil capacitors.

The network analyzer short one winding. Coil has 47mH for currents flowing both in the same direction, these are not the operating currents, 47mH effective for common mode signals, today I found no way to measure the network for common mode with the analyzer. Input and output are on the same ground.
  • Coupling factor determines the effective leakage capacitance for this simulation. Coupling factor determines the 900kHz frequency together with C4, R8, L5. Coupling factor has the most influence.
  • Slope between 900kHz and 20MHz determined by R5, L4 and R5, L3.
  • 95MHz determined by C3, R9 and C6, L10.
Resistors are constant vs. frequency in the simulation programm, the parasitic resistors were estimated for a value under higher frequencies with respect to the skin effect, that´s why for example ESR R4 and R5 are higher than under DC. The 47mH coils having parasitics. The 47mH is the most non-linear part in the filter, but constant here in the simulation.

100nF plot
Figure 120
Simulation result of the circuit in figure 119.

Figure 121
Measurement and Simulation in comparison. There was no way to force the simualtion to create the overshooting near 3kHz, needs further investigation. Coils are difficult under simulation, non-linear parts over frequency and current.

Experiments with the LC filter

in eval fre
Figure 122
AC-line current shows a waveform distortion of 2.6kHz - why?

Changing the LC Filter Caps from 100nF to 220nF

LC filter with 220nF
Figure 123
AC-line current waveform distortion of 2.1kHz after changing from 100nF to 220nF
  • 2.6kHz with 100nF
  • 2.1kHz with 220nF

260V 220nF
Figure 124
AC-line current waveform distortion of 2.7kHz after changing from 100nF to 220nF @240VAC rms.

Peaking AC-line current distortion, where it comes from?

Figure 125
Simulation of the input filter with post bridge rectifier. If 100nF or 220nF build-in no matter much.

Figure 126
Generator current shows Peaking AC-line distortion, similar to the measurement.

Figure 127
Frequency approx. 3-4kHz

Figure 128
Distortion in all Diode Currents. Diode i vs. u curves are not ideal, 0.7 volt deadband. Zero area highly nonlinear with fast changing currents and voltages, resulting in stimulus under many frequencies, powers LC-circuits powered and oscillation starts if not damped.

Figure 129
Distortion cancelled almost out when output capacitor C3 (220nF) changed to zero capacitance.

Removing the Common mode 47mH Inductor

Figure 130
Inductor had common 2*47mH with a coupling factor of 95%. Let us remove the 95% common inductance and calculate the leakage capacitance:

5/100  ==  x/(47mH+47mH)   ==>  x=(5/100)*(47mH+47mH)  ==> x=4.7mH leakage inductivity.

The effective lowpass filter for the AC-line distortion created by L1=4.7mH and C3=220nF.

plot6  plot7
Figure 131 + 132
Distortion back with stray inductance 4.7mH & 220nF, same frequency compared to figure 127

Distortion caused by - capacitor on the rectifier output plus serial inductance. This is the way back to the traditional 50Hz Transformer ==> Bridge Rectifier ==> large Electrolytic, resulting in the common known terrible peaking AC-line currents. When increasing C3 and playing togehter with L1 all these effect can be seen.

Sometimes when obeserving an unknown effect you can`t explain, try to amplify the effect, sometimes a way for a better understanding.

Figure 133
Plot based on circuit figure 130

This is the result I want to see.

Here is the missing unexpected peak in the amplitude response, below 10kHz. Changing a litte the coupling factor/leakage inductance and peak frequency change a lot. Second order LC lowpass with high quality factor and 20dB peaking. Damping is low in this simulation, generator impedance low.

The nonlinear rectifier diode i vs. u curve behaves like a heavyside function with a wide range of stimulus frequencies. Filter response with AC-line current distortion. Final result a decreased powerfactor - the main purpose of an powerfactor controller IC.

As mentioned before it is difficult to design an input-filter. Peakfree LC-filters without damping resistors are not possible. This filter design has just started now.

Remember again my four rules for a better design (from Part 2):

"Best developments methods consider these important issues:
  1. build-up and handmade-evaluated under no missing test equipment and lab conditions.
  2. explanation for each single part: "why using exactly this part and why not another".
  3. explanation Absolute-Maximum-Ratings for each single parameter, how many % have been reached under all conditions.
  4. fully tested outside the maximum operating conditions, search for HALT test failures".
See point 1. , it was the network analyzer showing the filter curve and amplitude peak, the analyzer showed unexpected issues. Without having the necessary equipment, I had not the idea measuring the filter amplitude response. Easy with the analyzer, pressing a few buttons - 5 minutes. Now the understanding is clear, one step more for a better filter design.

Do you have all necessary test equipment in your lab, NO?
Hobbiest: if no - understandable, expensive, not important in your life, electronic just for fun, no need, ok.
Commercial: if no - yo waste a lot of money - for not understanding your products in detail, using much time seaching for errors in a faulty design. Many engineers and 50% of "deciding people" do not understand what´s important. Do not spend long with them about this topic - it´s a waste of time. Search for the other half and talk with them, a greater pleasure.

Thank you for reading part 1 - part 6. I´ve told you in the beginning of part 1, the story do not show a circuit. The story: how to design and learn which methods are important. This is your gain for reading all this long stuff: hoping you understand the four rules above.

Some Network Analyzer plots from another filter

Figure 134
LC Filter modificated to 100nF (yellow) and 220nF (blue, after rectifier), similar values as in Evaluation Board and in figure 130.



Figure 135-139
LC filter measured with 50 ohm source impedance, reduce amplitude peak compared to figure 130.
Phase shift cause additional power factor distortion.

Go to Part 1
Go to Part 2
Go to Part 3
Go to Part 4
Go to Part 5
This is Part 6
Go to Part 7

Technische Berichte

Impressum und Haftungsausschluss