0
\$\begingroup\$

I'm trying to design a filter to reduce the EMI from two H-bridges driving a Peltier element load (also called thermo electric module) with a PWM switching output at 20 kHz. The cut-off frequency of my filter should be approx. 30 kHz, depending on what suitable filter components I can find for the PCB design. I'm experimenting with different filter design tools such as this from Panasonic (https://util01.industrial.panasonic.com/ww/utilities/ds/lc-sim/) and this tool from Marki Microwave (https://rf-tools.com/lc-filter/).

Both of these tools include input and output resistances in the simulation, and I'm trying to figure out what I should include in these resistances.

Here's the schematic of the two SIC631 H-bridge drivers that I'm using to drive the Peltier elements with a PWM switched output at 20 kHz:enter image description here

I'm guessing that as output resistances I should include the resistance of the entire load, both the Peltier elements and the cabling to the Peltier elements. I also assume that I should take the full impedance of the load into account in some way, and not only the resistance? In my specific case the load is 3.7 Ohms, 81 uF and 3.1 uH, measured at a frequency of 10 kHz. The simulators seem to only model the load as a pure resistive load, how would I take the inductance and capacitance of the load and cabling into account?

When it comes to the input resistance I'm a bit more confused though. My guess is that I should take the full resistive path from the power supply that is driving the circuit into account, and not only the path from the H-bridge drivers, is this assumption correct?

In the schematic above I have included a T-filter as a placeholder for now, since this site from Panasonic recommended T-filters when you have both a low input and output resistance (which I believe I have in this case?) https://industrial.panasonic.com/ww/ss/technical/b4

\$\endgroup\$
1
  • \$\begingroup\$ You can design a filter with a zero input resistance, and finite output resistance - it will by design always have a series L on the input. If your chosen design tool does not support this, plenty of others do. \$\endgroup\$
    – Neil_UK
    Oct 27, 2022 at 13:04

1 Answer 1

0
\$\begingroup\$

The first tool (Panasonic) is for a π filter and that is definitely NOT what you want. You don't want your H bridge output to be directly connected to an input capacitor in the π filter at all. You need an LC filter that is chosen to have moderate Q-factor with the load impedance.

The second tool is for an RF impedance matching filter and, again, this is not what you need.

A T-filter is OK if you need a third order response.

As with any filter design, you decide what you require based on what spectral response is needed.

\$\endgroup\$
5
  • \$\begingroup\$ The Panasonic simulator supports both Pi-, T- and L-type filters: industrial.panasonic.com/ww/lc-simulator The second tool also supports Butterworth filters, which tought was just a variant of LC-filters, and could be suitable for my type of application? \$\endgroup\$ Oct 27, 2022 at 12:58
  • \$\begingroup\$ My last sentence says: As with any filter design, you decide what you require based on what spectral response is needed. <-- I can't decide that for you and, accordingly, I can't say what filter characteristic is suitable for your application. I am just explaining why a pi filter is not what you want and therefore any filter that needs a value for the source resistance is missing the point @LarsPetersson \$\endgroup\$
    – Andy aka
    Oct 27, 2022 at 15:12
  • \$\begingroup\$ Ok, thank you for explaining! Yes, after some more reading I'm leaning towards either just a simple, well-tuned LC-filter, or maaaaaaybe a T-filter (which I interpret as a third-order LC-filter?) I'm still confused about my original question on the input- and output impedances though, this article from InCompliance Mag talks about the source and load impedances when choosing filter topology, but are you saying that if I'm choosing an LC topology for example, the source and load impedances doesn't matter? incompliancemag.com/article/… \$\endgroup\$ Oct 28, 2022 at 8:58
  • \$\begingroup\$ More source impedance will reduce the Q factor of an LC low pass filter. This can be both a good thing and a bad thing so, you do need to factor it into your solution. See this LC tool I put on my website: stades.co.uk/RLC%20filters/RLC%20LPF.html <-- you can mess around with R, L and C to get various responses but, the "R" bit only affects the peakiness of the low pass filter. OK, that's talked about the source impedance but, the load impedance also has an effect on Q-factor too and, if you scroll down that page you can find a tool that models a LC low pass filter with a load resistor. \$\endgroup\$
    – Andy aka
    Oct 28, 2022 at 9:56
  • \$\begingroup\$ So yes, source and load impedances do affect the peakiness of the response but not the fundamental natural frequency of the filter @LarsPetersson and, using an LCL T-filter will be better than using an LC but, may not be needed. \$\endgroup\$
    – Andy aka
    Oct 28, 2022 at 9:57

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.