1
\$\begingroup\$

I've got quite a few issues with harmonic currents on a VFD so I designed a (fat) harmonic filter. It's single phase so it's designed to block the third and the fifth harmonic (150Hz and 250Hz, here).

schematic

simulate this circuit – Schematic created using CircuitLab

The left side models the mains supply, the load is here represented by a 15 ohm resistor. For a transient analysis it needs to be replaced with something more realistic (like a diode bridge and a capacitor) but for AC this is OK.

Usually these filters are done in a shunt topology with a blocking reactor but I opted for a series filter because I can't put a disconnect contactor on the shunt branches (otherwise they'll suck current even when it's not needed).

Now, it works, both on SPICE and in the lab (it's big and heavy as it should be) with a resistive load of 15-30 Ohm (which is the design value) but there's a thing worrying me. The seventh and ninth are under control, until I rise the source impedance.

AC Filter with low inductive source impedance Transient with low inductive impedance

It's a known fact that these filters can interact with the upstream distribution system since the inductive component of the distribution can be significant. With a nominal LISN-like 250µH the result is good (that would be a 0.4+j0.08 source impedance); given that the supply for harmonic testing has a really low impedence I should pass compliance with this. However real world supplies have really worse inductive component. I found in a paper (Evaluation of Fault Levels and Power Supply Network Impedances in 230/400 V 50 Hz Generic Distribution Systems, DOI 10.1109/TPWRD.2016.2609643) a worst residential case of 0.46+j0.45 which is about 1.5mH. With such a source, the series filters seems to resonate on the higher harmonics

Transient with high inductive impedance

Every paper I read warns of this issue but nobody gives useful indication ("just be careful"); I could lower the quality factor but the loss on the filter would be even more substantial (we already are at 25W, excluding the parasitics). A commercial shunt 16A single phase filter dissipates about 150W so maybe that's actually the right solution...

It seems that the issue is the second 'hill' in the AC response but that moves with the input impedance. Any suggestion on how to stabilise this thing?

\$\endgroup\$
8
  • \$\begingroup\$ Do put in your single phase bridge and filter cap ,Then look at your current wave in the time domain, \$\endgroup\$
    – Autistic
    Commented Apr 15 at 9:41
  • \$\begingroup\$ How can you get the 2nd waveform with the test circuit you have shown? What does the AC response look like when the source impedance is low and, where did you put the generator to produce that AC response? \$\endgroup\$
    – Andy aka
    Commented Apr 15 at 9:50
  • \$\begingroup\$ The transient analysis simply has a diode bridge on the right side; the current in the time domain is the one shown \$\endgroup\$ Commented Apr 15 at 10:05
  • \$\begingroup\$ > Usually these filters are done in a shunt topology with a blocking reactor but I opted for a series filter because I can't put a disconnect contactor on the shunt branches (otherwise they'll suck current even when it's not needed).< What do you mean with this? \$\endgroup\$
    – Antonio51
    Commented Apr 15 at 12:40
  • \$\begingroup\$ @Antonio51 shunt harmonic filters are only put online when the load is active otherwise they use a considerable amount of power for nothing. Series filter do not have this issue but the behaviour is different \$\endgroup\$ Commented Apr 15 at 12:56

2 Answers 2

1
\$\begingroup\$

With such a source [i.e. one with 1.5 mH inductance], the series filters seems to resonate on the higher harmonics

Not quite.

Here is your circuit with some slight changes. I have changed the value of L3 to 1.5 mH, to correspond with the "worst case" source impedance, and I have changed L1 and L2 to 11 mH and 4 mH, respectively, to give resonant frequencies that more closely match 150 Hz and 250 Hz. I have also omitted the load, so that we can see the admittance (the inverse of the impedance) of the notch filters together with the source admittance. This will allow us to observe a resonant peak at which the circuit may ring that is otherwise obscured.

schematic

simulate this circuit – Schematic created using CircuitLab

Here is the magnitude of the admittance:

enter image description here

From this plot, one can easily see a peak in the magnitude of the admittance at around 617 Hz. The filter network together with the input inductance is likely to ring at this frequency if the circuit receives an appropriate stimulus and the load has a sufficiently low impedance at that frequency. Note the difference between this plot and the plot shown in the original question (which includes a load resistance), where the 617 Hz peak is not noticeable. Note also, that peak frequency is determined by the filter network, together with the source impedance, and is completely unrelated to harmonics of the input frequency.

From the comments:

The transient analysis simply has a diode bridge on the right side; the current in the time domain is the one shown

Actually, I believe the your simulation was performed with both a diode bridge and a smoothing capacitor, plus a load.

The full wave rectifier plus the smoothing capacitor, provide the stimulus, and the necessary low impedance of the load at that frequency.

Here is a circuit with the 1.5 mH source inductance, bridge rectifier and smoothing capacitor.

schematic

simulate this circuit

Here is a time domain simulation of the input current, the current through the smoothing capacitor, and the current through the load resistor.

enter image description here

Now, here is your circuit without the smoothing capacitor

schematic

simulate this circuit

The absence of ringing when the smoothing capacitor is omitted, is due to the low admittance of the load at the ringing frequency. If the smoothing capacitor is present, but is of a smallish value, it will tend combine with the net capacitative reactance of the filter network at higher frequencies, and raise the ringing frequency.

enter image description here

Any suggestion on how to stabilise this thing?

The use of passive filters (rather than active/switching) results in high values for inductances and capacitances at low (i.e. mains) frequencies. However, if passive filters are what you want, and you can live with the high values, a solution may be to add a band-pass and/or low-pass filter to the filter cascade. Because the natural frequency of the ringing will vary with the input line impedance, a notch filter would be problematic.

Here is a potential solution.

schematic

simulate this circuit

enter image description here

The 100 uF and 100 mH values for the 50 Hz band-pass filter can be changed, as long as the product of the two values remains relatively constant. You should have

$$LC \approx \frac{1}{4\pi^2 50^2} \approx 10.13 x 10^{-6} HF$$

\$\endgroup\$
0
\$\begingroup\$

Series filter will cause a voltage drop that probably will cause another problems to the drive. Most of VDF have some user free outputs, so you can adjust one of them to be active when the drive is running, and close a line contactor for the filters.

To lower losses, see C-type filter, tuned at 3rd harmonic and check for 5th and higher. The advantage is that L and C are parallel with the damping resistor, reducing the current through R. Also you save some elements. See drawings and plot, it is an example, not calculated components.

You can take a look at this web: https://www.pscad.com/webhelp/Master_Library_Models/Passive/Filters/Passive_Filter_Design.htm

enter image description here

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.