Go Down

Topic: Low pass design process. (Read 424 times) previous topic - next topic

Wintermuteldn

Apr 16, 2019, 09:15 pm Last Edit: Apr 16, 2019, 09:20 pm by Wintermuteldn
Welcome all. I am having trouble with designing low pass filter for my precision rectifier. So, in picture "1" I have my rectified waveform and envelope I am interested in. Frequency of rectified sine is about 40KHz. Currently my circuit has two low pass filters cascaded picture "2". Result of smoothing of LPF is in picture "3" after first RC branch and picture "4" after second. I am happy with that result, but trouble is I guessed them and wondering how I could improve or choose them properly. I have been advised that I could use FFT to see spectrum of rectified signal and use it to determine where to put cut-off frequency. Result of FFT is in picture "5". My trouble is that I potentially measured FFT incorrectly as in my LTSpice I get different result picture "6" and "7". So my question is what are the rules to find reasonable cut off point in that case?

herbschwarz

Hello Wintermuteldn,
The cut-off frequency is your operating frequency.
The capacitor should have reactance (Xc) equal to
the resistor value at that frequency.  This forms
a low-pass filter with a  3 dB loss. The output
voltage will be 0.707 times the input voltage.
Herb

Wintermuteldn

@herbschwarz. Yes, I agree. This is good method to consider when calculating response for sine wave (single harmonic). However, in my case I have complex wave form due to rectification and exponential rise and decay. FFT of this waveform contains many harmonics. Therefore, I need to make decision to where is my cut off point. Therefore, my approach is to try to find spectrum in which most of signals energy is contained to cut rest of spectrum in frequency domain. Therefore, I suspect convolve filter's and signals response in time domain.

Grumpy_Mike

Quote
This is good method to consider when calculating response for sine wave (single harmonic). However, in my case I have complex wave form due to rectification and exponential rise and decay
It makes no odds as to what the waveform is, the filter is still designed in the same way. The RC will only give you a first order filter. If you want it to roll of quicker you need a higher order filter.

Ether way the waveform you want to filter consists of sin waves of different frequencies.

herbschwarz

Hi.
This RC low pass filter is an integrator. It
has an attenuation of 6 dB per octave.
Higher frequencies are attenuated more
than low frequencies. Thus, it will tend
to remove any harmonics. (If you find that
you have a sub-harmonic, then a high pass
filter can be used to remove that since
it lies below the Fc of the low pass filter.)


jremington

Quote
wondering how I could improve or choose them properly
Please specify your criteria for proper operation.

Wintermuteldn

Thank you all. 1st has to be passive no spare op-amps in circuit. Preferably 2nd order. Ripple of about 5mV. Last and very important constrain is time constant of a filter. Since my goal is to have steep slope as this reduces my error.

jremington

That is not much help. Here is some background reading on passive filters: https://www.electronics-tutorials.ws/filter/filter_2.html

One of several line filter calculators: http://sim.okawa-denshi.jp/en/CRlowkeisan.htm

Wintermuteldn

Thank you for materials. Perhaps I didn't word my enquiry right. I have done some filter design both analogue active passive and digital bot FIR and IIR. My worry here is backing my design process. As I achieved my goal of filtering very well by observing oscilloscope and matching R and C. However, I am not able to draw the line mathematically for my low pass filter at which point would my output be distorted. Hence my question about FFT spectrum (picture below) of my input as this was hinted to me. However I so far faild to find bond between two.


jremington

#9
Apr 18, 2019, 02:37 am Last Edit: Apr 18, 2019, 02:42 am by jremington
Quote
at which point would my output be distorted.
A precision rectifier is a profound distortion of an AC signal. What do you mean by "distortion"?

Quote
So, in picture "1" I have my rectified waveform and envelope I am interested in
You forgot to post picture "1".

I don't think any of us understand what you are actually trying to do.

Grumpy_Mike

Quote
1st has to be passive no spare op-amps in circuit. Preferably 2nd order.
So that will be an LC filter then.

Quote
Ripple of about 5mV
You don't get ripple from a 2nd order passive filter.

Quote
Last and very important constrain is time constant of a filter. Since my goal is to have steep slope as this reduces my error.
I don't think this has any bearing on the slope.

Quote
at which point would my output be distorted
If the input waveform has any harmonics in the roll of region then the output will be distorted, that is the job of a filter.

Wintermuteldn

Sorry about confusion. So perhaps ill start from beginning I created circuit which I want to operate in similar fashion to envelope detector in AM radio. Picture 1 is my circuit. Picture 2 shows input and output of my precision rectifier. Now in order to process it further I need to have nice "envelope" of carrier signal. Here my trouble begins as I simulated such a circuit as RC ladder in picture 1 and it resulted in ok waveform on the output picture 3. However, it was pure guess and speculation rather than deduction. So, my question is what technique to analyse it should I use to get from picture 2 bottom waveform to picture 3 waveforms?
Reason for my question about ripple ect as it reminded me of calculation of ripple in rectifiers in respect to frequency and load.

jremington

#12
Apr 19, 2019, 07:49 pm Last Edit: Apr 19, 2019, 07:51 pm by jremington
Your low pass filter needs to have a cutoff near the modulation frequency, not the carrier frequency.

An AM demodulator also removes the DC component (a very low frequency cutoff, high pass filter).

Wintermuteldn

Thank you for your reply. Now it makes sense as clearly my cut off frequency (6.7kHz) is much lower than carrier frequency (40kHz). Now for DC part I would think of simple solution of high value capacitor in series with low-pass filter would that approach be suitable? Kipping in mind that someone already pointed that it is not best practice to connect op amp's output directly to capacitor. I would think of low-pass and capacitor arrangement Lastly how can I measure frequency of modulating signal?

jremington

#14
Apr 19, 2019, 09:33 pm Last Edit: Apr 19, 2019, 09:35 pm by jremington
Quote
Lastly how can I measure frequency of modulating signal?
How are you producing the signal?

The envelope shown in "3" does not have a well defined frequency.

Go Up