Simulate Power Supply Filter

lindula

Joined Sep 23, 2016
94
Hello, I need some help coming up with a LCR circuit to filter a power rail from a Switching Power Supply that will provide a 1.8V to an FPGA rail. The Switching Supply has a frequency of 800Khz and has a Vout of 1.8V. How do I design an LRC circuit to filter the switching noise?

I'm trying to use LTSpice to simulate the filter. Does this look what I should be doing? I've seen some designs where they only use 1-100 ohm resistor and various number of 100uF caps and then the inductor.

Can anyone give me some help to get started please?

malli_1729

Joined Nov 28, 2012
32
What is noise level you are getting in input side and what is noise level you are expecting on output side.

Papabravo

Joined Feb 24, 2006
20,378
The normal way to do this is to establish the requirements for attenuation in the passband, stopband, and transition band. This will give you the form of the filter and allow you to pick the order, the rest is plug and chug. With a switching frequency of 800 kHz you should be able to set the corner frequency at a reasonable value so the filter will be realizable with easily available components. You would do it like this:

Maxiimum attenuation in the passband -0.5 dB out to 8 kHz., maximally flat
Transition band, 1 decade to 80 kHz
Minimum attenuation in the stop band 80 kHz to ∞, -60 dB
You can also specify the input and output impedance

This specification would imply a passive Butterworth filter for example.

One more thing. The SPICE .noise analysis won't give you anything useful in that circuit, except the noise from the resistor as a function of temperature.
If I were you I would model the SMPS as a DC voltage of 1.8VDC in series with an AC source with 100 VAC P-P triangle wave output and a series impedance of 100 milliohms. I would drop the load resistor to some value representing the expected current draw. Lose the L & C that you have and put a multi section odd order LC filter in between the source and the load.

Last edited:

lindula

Joined Sep 23, 2016
94
The normal way to do this is to establish the requirements for attenuation in the passband, stopband, and transition band. This will give you the form of the filter and allow you to pick the order, the rest is plug and chug. With a switching frequency of 800 kHz you should be able to set the corner frequency at a reasonable value so the filter will be realizable with easily available components. You would do it like this:

Maxiimum attenuation in the passband -0.5 dB out to 8 kHz., maximally flat
Transition band, 1 decade to 80 kHz
Minimum attenuation in the stop band 80 kHz to ∞, -60 dB
You can also specify the input and output impedance

This specification would imply a passive Butterworth filter for example.

One more thing. The SPICE .noise analysis won't give you anything useful in that circuit, except the noise from the resistor as a function of temperature.
If I were you I would model the SMPS as a DC voltage of 1.8VDC in series with an AC source with 100 VAC P-P triangle wave output and a series impedance of 100 milliohms. I would drop the load resistor to some value representing the expected current draw. Lose the L & C that you have and put a multi section odd order LC filter in between the source and the load.
Hello, thank for responding to my message. I found some example of a filter that is used with a Switcher.

In the schematics I see they always use a 100-ohm resistor, various 100uF caps and a unique inductor. This Switcher has the switching frequency set to 350Khz. I need to find the equation to determine the value of the inductor or the transfer function.

Thank you very much for helping me.

Joe

Attachments

• 146.2 KB Views: 7

lindula

Joined Sep 23, 2016
94
Hello, thank for responding to my message. I found some example of a filter that is used with a Switcher.
View attachment 223939

In the schematics I see they always use a 100-ohm resistor, various 100uF caps and a unique inductor. This Switcher has the switching frequency set to 350Khz. I need to find the equation to determine the value of the inductor or the transfer function.

Thank you very much for helping me.

Joe
I've been looking at the datasheet for the MAX15301:
https://datasheets.maximintegrated.com/en/ds/MAX15303.pdf

On page 25 they cover the selection of the Inductor and Cap but I don't see anything for the 100 ohm resistor.

Does this sound like a good filter to use?

Joe

lindula

Joined Sep 23, 2016
94
I've been looking at the datasheet for the MAX15301:
https://datasheets.maximintegrated.com/en/ds/MAX15303.pdf

On page 25 they cover the selection of the Inductor and Cap but I don't see anything for the 100 ohm resistor.

Does this sound like a good filter to use?

Joe
I've been looking at the datasheet for the Switcher I am using, ISL8205MIRZ and I don't see any recommendation to use a output filter, so should I not put a filter in my design?

Thank you

lindula

Joined Sep 23, 2016
94
I've been looking at the datasheet for the Switcher I am using, ISL8205MIRZ and I don't see any recommendation to use a output filter, so should I not put a filter in my design?

Thank you
Here is my LTSpice:

I'm trying to come up with a filter for a Switcher that has a freq of 800Khz. I need to find the math to determine L1 and C2.

I'm using the ISL8205 switcher that does not call for an output filter so I might not really need this but it is something that I should learn.

Thank you

Attachments

• 1.1 KB Views: 1

Papabravo

Joined Feb 24, 2006
20,378
Please define your requirements. Pay no attention to other circuits. You are going about this all wrong. What you have is a second order low pass filter. I cannot tell what kind of filter it is without doing some work. You cannot apply any formula for the inductor unless you specify a corner frequency and the required attenuation in the stopband. From the existing filter you can easily calculate the corner frequency and the stopband attenuation, but it does not APPLY to your situation. Since you are using a higher switching frequency you have a great deal more flexibility.

If you fail to state, or worse, refuse to state your requirements you will not have any idea what "done" means.

lindula

Joined Sep 23, 2016
94
Please define your requirements. Pay no attention to other circuits. You are going about this all wrong. What you have is a second order low pass filter. I cannot tell what kind of filter it is without doing some work. You cannot apply any formula for the inductor unless you specify a corner frequency and the required attenuation in the stopband. From the existing filter you can easily calculate the corner frequency and the stopband attenuation, but it does not APPLY to your situation. Since you are using a higher switching frequency you have a great deal more flexibility.

If you fail to state, or worse, refuse to state your requirements you will not have any idea what "done" means.
Thanks for responding to my post. I have a 5A DC/Dc Step-Down Power Module ISL8205M that I use to generate a 1.8V rail for an FPGA. I've set the set the switching frequency to 750Khz. I look over the datasheet and I don't see any recommendation for an output filter. It does recommend using 2 22uF caps. I do not know if I really need to provide a output filter but installing one should not cause a problem. I've seen some people use Ferrite beads and some use Inductors. The FPGA 1.8V rail can be 1.79V - 1.89V. The Switcher datasheet shows Output Ripple Voltage of 11mVp-p at Iout=0A, Vout=1.2V, fsw = 1.6Mhz. On those calculations I would expect to see 1.7945V to 1.8055V with the switcher vout set to 1.8V. So maybe I really don't need to filter.

Papabravo

Joined Feb 24, 2006
20,378
Thanks for responding to my post. I have a 5A DC/Dc Step-Down Power Module ISL8205M that I use to generate a 1.8V rail for an FPGA. I've set the set the switching frequency to 750Khz. I look over the datasheet and I don't see any recommendation for an output filter. It does recommend using 2 22uF caps. I do not know if I really need to provide a output filter but installing one should not cause a problem. I've seen some people use Ferrite beads and some use Inductors. The FPGA 1.8V rail can be 1.79V - 1.89V. The Switcher datasheet shows Output Ripple Voltage of 11mVp-p at Iout=0A, Vout=1.2V, fsw = 1.6Mhz. On those calculations I would expect to see 1.7945V to 1.8055V with the switcher vout set to 1.8V. So maybe I really don't need to filter.
Well. I certainly think that if you can get 11 mVp-p at the output with no filter that seems like a good start, but I might try for just a bit higher output voltage to give yourself some headroom over the minimum 1,79. I think 1.82 , that's just under 3 time the ripple voltage. I am concerned about what the ripple will be when the FPGA is drawing it's heaviest load. Of course you will want to check these numbers on the actual parts, and in various conditions. If the ripple gets to where the voltage dips below 1.79V then we should revisit the question of what to do.

You still have not mentioned what corner frequency you would wan to use. Normally you might start with a corner frequency of fsw/10. this will give you small and economical components. A filter at this frequency might or might not do the job. At this pint you have two options:
2. Reduce the corner frequency, making it more difficult for high frequency ripple to get through.
Unfortunately these are not linear relationships.
For the circuit of post #7 you want L1 and C1 to have equal AC impedances at some low frequency between say 1/10 and 1/100 of your switching frequency
$X_L\;=\;2\pi fL$
$X_C\;=\;\frac{1}{2\pi fC}$

sghioto

Joined Dec 31, 2017
4,915
So maybe I really don't need to filter.
Probably not using the two 22uf output caps as recommended. If you want to add a filter I would suggest the one in post #4.
Eliminate R36 and change L66 to 100nH.

lindula

Joined Sep 23, 2016
94
Hello, I've working on the filter some more. Here is what I have:

Switcher is 800Khz and I choose the cutoff frequency of 8Khz.

Zoom in a little:

Looks very good.

How else should I verify the design?

I have using inductor part# 7443091100 that is 1µH Shielded Wirewound Inductor 34A 0.42mOhm Nonstandard and a little large: 14.50mm x 13.50mm.

lindula

Joined Sep 23, 2016
94
Here is a Bode Plot:

Should be worried about the 40dB Gain?

Thank you

lindula

Joined Sep 23, 2016
94
At 800Khz I have a -40dB attenuation.

lindula

Joined Sep 23, 2016
94
I was looking over the post and maybe should have the cutoff frequency of 80Khz not 8Khz. Is that correct?

Respectfully,
Joe

sparky 1

Joined Nov 3, 2018
734

The TI application report figure 5 is one noise at least for the board being used. The noise was captured on oscilloscope using recommended probe tip. The subject can be quite extensive. The idea of throwing different modules together can lead to logical conclusion of using simulation for redesign.
The ideal to real is better seen on test bench with a variety of good designs.
https://www.ti.com/lit/an/snoa543a/snoa543a.pdf

In figure 3 An E-field of a power supply is being evaluated.
https://www.we-online.com/catalog/media/o109027v410 ANP047b EN.pdf

Last edited: