Multi-stage LC Filter

[Unknown1980]

Hi all

I'm new to these forums.

I am planning to build a multistage LC low pass filter for the output of an AC to DC converter.

The output of the supply will be adjustable between 6V - 12V at 2A of steady current. The required cut off frequency required for the multi-stage filter is 2Hz.

The problem I am having is, that despite the calculations required to design such a filter, actual capacitors and inductors are not ideal, therefore the inductance and capacitance values that I actually need to use, may be very different than the values specified by the mathematical equations.

For example, It is almost a certainty that I will need to use electrolytic capacitors, which means that there will be no high frequency attenuation above a few hundred Kilohertz when the capacitors become inductive.

This won't matter, as I have other methods of dealing with the higher frequencies, but my main goal right now, is to achieve a filtered band from 2Hz up to 200KHz.

If anyone has an suggestions, they would be much appreciated.

 

[D.A.(Tony)Stewart]

2Hz under load implies a lot of stored energy better done by a rechargeable battery. or ultra-cap.

Low ESR is important for load regulaion, but for HF, parallel Caps with higher SRF is required. ( or lowESL )

This is why multiple caps are often used in different ranges and materials of ceramic, electrolytic and plastic.

polymer types then plastic filmtypes are most effective with lowest conductor ESR but also higher cost and size.

plastic includes PU, PE polyurethane, polyester and polystyrene

start with specs for V, vs I on DC before LC filter specs. this should include; load regulation and, losses with input dropouts and step loadresponse.

got it??8-O if no then ask.:thinker:

 

[FvM]

Low-pass filters with 2 Hz cut-off are not really practical and don't serve a reasonable purpose, I assume.

Filter are primarly interesting for switch-mode converters. But the lowest frequency you want to supress is the switch frequency fundamental. A reasonable cut-off frequency may be between 0.05 to 0.5 fswitch. Low-pass filters also cause problems in load step response. If they are part of the feedback loop, they can also cause stability issues. Finally inductor size and weight will definitely set a lower limit to feasible cut-off frequencies.

 

[Unknown1980]

I should have mentioned in my initial post.

The AC to DC supply will be linear, The multi-stage filter will be mainly to deal with DC ripple from the rectification stage, and any noise/unwanted signals within the 2Hz - 200KHz bandwidth.

I will use additional ceramic capacitors and ferrite beads for the upper end of the frequency spectrum, but my main concern is achieving what is possible with electrolytic capacitors first, this will give me an idea of how much board space I am going to need.

I am not sure exactly what you mean SunnySkyguy, could you elaborate please?

 

[FvM]

If it's an adjustable linear regulator, you won't need an output filter.

 

[Unknown1980]

If it's an adjustable linear regulator, you won't need an output filter.

What about the DC ripple, noise and RFI? Won't they still be present on the output?

 

[mtwieg]

If you have a DC regulator with some low frequency ripple on its output, then the proper solution is to improve its feedback loop performance, not just add an enormous filter on the output.

 

[Unknown1980]

If you have a DC regulator with some low frequency ripple on its output, then the proper solution is to improve its feedback loop performance, not just add an enormous filter on the output.

I have already consulted the data sheet for the DC regulator and optimized the regulator for the best performance that is possible. DC ripple is just one of the reasons for having such a wide band filter. Noise and any other unwanted signals are also a priority. The goal is to make the DC output as 'quiet' as possible. Seeing as it is impossible to know exactly which frequencies may see noise or some kind of harmonics, the plan is to filter as many of them out as possible.

 

[FvM]

I wonder what "DC ripple" is? Ripple is always an AC quantity, I think.

As already explained, a linear voltage regulator can give a nearly perfect DC output voltage if designed respectively. It has usually output bypass capacitors that reduce higher frequency noise and provide a low output impedance for frequencies near and above the feedback loop cutoff frequency. Even a high performance voltage regulator shows a certain amount of electronic noise, slow voltage fluctuations and long-term drift. But it's effectively impossible to remove it with LC filters.

 

[Unknown1980]

I wonder what "DC ripple" is? Ripple is always an AC quantity, I think.

As already explained, a linear voltage regulator can give a nearly perfect DC output voltage if designed respectively. It has usually output bypass capacitors that reduce higher frequency noise and provide a low output impedance for frequencies near and above the feedback loop cutoff frequency. Even a high performance voltage regulator shows a certain amount of electronic noise, slow voltage fluctuations and long-term drift. But it's effectively impossible to remove it with LC filters.

I see... but what if someone wanted to filter out all these frequencies just for the sake of filtering them out, how would they go about doing it.

 

[godfreyl]

Linear regulators do an excellent job of removing low frequency stuff like mains ripple (100Hz or 120Hz), but don't cope well with high frequency garbage at the input, so a good plan is to put a passive filter before the linear regulator to prevent any high frequency garbage reaching it.

 

[mtwieg]

So basically you're talking about two things, PSRR and output noise. A decent regulator should have extremely high PSRR at low frequencies (like below 1kHz), enough to practically eliminate any low frequency ripple. The regulator itself is a filter for this sort of ripple. If you're seeing poor PSRR, then the problem is with the regulator itself. Noise on the other hand is a trickier issue. The low frequency output noise can't be effectively filtered, but high frequency noise can, using an external LC or RC filter. What sort of noise levels are you looking for?

Ultimately the quietest DC source is a plain old battery, it's pretty much impossible to beat it.

 

[Unknown1980]

Linear regulators do an excellent job of removing low frequency stuff like mains ripple (100Hz or 120Hz), but don't cope well with high frequency garbage at the input, so a good plan is to put a passive filter before the linear regulator to prevent any high frequency garbage reaching it.

Thanks for letting me know about that godfreyl, I can put in some ceramic capacitors, and possibly some ferrite inductors at the input of the regulator to prevent high frequencies from reaching the regulator.

- - - Updated - - -

So basically you're talking about two things, PSRR and output noise. A decent regulator should have extremely high PSRR at low frequencies (like below 1kHz), enough to practically eliminate any low frequency ripple. The regulator itself is a filter for this sort of ripple. If you're seeing poor PSRR, then the problem is with the regulator itself. Noise on the other hand is a trickier issue. The low frequency output noise can't be effectively filtered, but high frequency noise can, using an external LC or RC filter. What sort of noise levels are you looking for?

Ultimately the quietest DC source is a plain old battery, it's pretty much impossible to beat it.

The noise I am looking for, is any level of noise on any frequency, that's why I plan on using a comprehensive 2Hz - 200KHz filter to begin with, then to focus on the higher frequencies at the output of the PSU.

I understand that a battery will be impossible to beat, but there is no harm in doing the best I can.

 

[samEEEf]

I think, we all should be clear about designing LC filter for a converter to get almost pure dc. It would be great if any expert designer waste some time here and describe it with some simple equations.

 

[Audioguru]

For many years, I have used switched-capacitor lowpass filter ICs for multi-stage filters. They are accurate and can have an adjustable cutoff frequency.
With 8 stages in a Butterworth lowpass filter you can have a squarewave input and get a pure very low distortion sinewave output.

 

[D.A.(Tony)Stewart]

I am planning to build a multistage LC low pass filter for the output of an AC to DC converter.
The output of the supply will be adjustable between 6V - 12V at 2A of steady current. The required cut off frequency required for the multi-stage filter is 2Hz.

What is your budget for this 12V, 2A (24W) power supply? A 2 Hz LPF seems reasonable to get low ripple until you find out how many thousand uF and uH you need to get 2Hz and how much that costs to supply 2Adc then you will give up and choose Buck regulator instead which will cost <1% of the Linear filter since it operates at 1000x the frequency.


How much peak current does a 2A DC inductor use ? Answer 350%x average. How much does this cost
Now go find an inductor rated for 7A (>7A saturation) It will be an steel laminated core as large as the transformer. Choosing a &A Ferrite Choke only gets you 10uH at reasonable cost.

If you rely on an RC filter instead how much ripple current does a Capacitor see with 15% ripple voltage. Answer AC current increases as AC ripple decreases, thus with 15% ripple voltage then the AC ripple current shown below is +10A,-2A =12Ap-p or about 6x the Adc.

This is not an optimal solution to design an LC filter around a 50Hz AC to DC conversion. This is why Buck converters are preferred.

 

[Unknown1980]

What is your budget for this 12V, 2A (24W) power supply? A 2 Hz LPF seems reasonable to get low ripple until you find out how many thousand uF and uH you need to get 2Hz and how much that costs to supply 2Adc then you will give up and choose Buck regulator instead which will cost <1% of the Linear filter since it operates at 1000x the frequency.

View attachment 108407
How much peak current does a 2A DC inductor use ? Answer 350%x average. How much does this cost
Now go find an inductor rated for 7A (>7A saturation) It will be an steel laminated core as large as the transformer. Choosing a &A Ferrite Choke only gets you 10uH at reasonable cost.

If you rely on an RC filter instead how much ripple current does a Capacitor see with 15% ripple voltage. Answer AC current increases as AC ripple decreases, thus with 15% ripple voltage then the AC ripple current shown below is +10A,-2A =12Ap-p or about 6x the Adc.

This is not an optimal solution to design an LC filter around a 50Hz AC to DC conversion. This is why Buck converters are preferred.

Cost really isn't a real concern for me, but the size of the circuit will be. The enclosure space I have to work with for the whole device is 450x70x300mm WxHxD (approx)

To look at it from another perspective. The goal is to create a power supply that can deliver clean power between 6V and 12V at a steady 2A.

When I say clean power, I mean free of noise, RFI, oscillations or anything else that we wouldn't want on a DC supply.

I am definitely open to discussing alternative methods of achieving this. I will have a look into Buck converters and see if they can deliver what I am looking for.

I originally put the upper end of the filter limit at 200Hz, as I assumed I would be using only electrolytic capacitors for a multistage LC filter, and therefore, they would become inductive at higher frequencies.

 

[D.A.(Tony)Stewart]

Your question is about an LC filter yet all you want is a clean variable lab supply. If I were to suggest a solution, there are many assumptions not stated, so I will expect you to make a list of requirements on availability of all other parts such as a transformer solution or better, a full spec.

Generally one does not design an LC filter without knowing what is being filtered. 100Hz or 100kHz or ?

But for very clean lab supplies, there are many solutions which depend on safety, input voltage, regulation accuracy, noise in uV or mV.

I cannot give an explicit answer until ALL the requirements are completed. I suggest you look around to see HW lab supplies are created or specified and decided what you want to make/buy and expected compliance to safety, cooling, regulation etc etc.

- a linear regulator suppresses noise well, but at high current the voltage drop between input and output wastes heat that requires a large heatsink.

- SMPS using Buck mode or step-down is 80-95% efficient but can be noisy at light loads so a pre-load and good LC filter or an Linear regulator is needed.

-low ESR Caps are essential for any low noise filter supply, but also short circuit protection.

- here are some schematics on power supplies
- here is some terminology required in a spec.
- some general discussions

i suggest you consider this question (answered) and do some research and come up with a better question.

 

[Unknown1980]

Your question is about an LC filter yet all you want is a clean variable lab supply. If I were to suggest a solution, there are many assumptions not stated, so I will expect you to make a list of requirements on availability of all other parts such as a transformer solution or better, a full spec.

Generally one does not design an LC filter without knowing what is being filtered. 100Hz or 100kHz or ?

But for very clean lab supplies, there are many solutions which depend on safety, input voltage, regulation accuracy, noise in uV or mV.

I cannot give an explicit answer until ALL the requirements are completed. I suggest you look around to see HW lab supplies are created or specified and decided what you want to make/buy and expected compliance to safety, cooling, regulation etc etc.

- a linear regulator suppresses noise well, but at high current the voltage drop between input and output wastes heat that requires a large heatsink.

- SMPS using Buck mode or step-down is 80-95% efficient but can be noisy at light loads so a pre-load and good LC filter or an Linear regulator is needed.

-low ESR Caps are essential for any low noise filter supply, but also short circuit protection.

- here are some schematics on power supplies
- here is some terminology required in a spec.
- some general discussions

i suggest you consider this question (answered) and do some research and come up with a better question.

The question hasn't been answered, because I still don't know how to correctly build a filter to attenuate the 2Hz - 200KHz bandwidth.

I already have a schematic, but I didn't think it was necessary to post it, considering that all I needed was realistic L and C values for a multi stage filter that would be necessary to attenuate the frequencies specified, given the voltage range and current.



I had considered rethinking the approach only when you had suggested that it could be done an alternative way.