Dear Sir, good day!
I made calculation for low-pass filter. Please, see attach.
The Filter Design Tool program suggested amplifier TLC2252. But it doesn't matter. The circuit has 10 nF and 20.5 nF capacitors. But these are inconvenient values. Does the program allow to change the capacitor values?
TIA
Sincerely,
Vladimir Naumenkov
Hi Vladimir,
the TLC2255 is a very low power OPAmp and doesn't like very much low impedant Sallen-Key filters. See the erosion of frequency response at high frequencies:
So one remedy is to choose higher impedant feedback components:
A filter using more convenient components could look like this:
Or this:
Furtherly increasing the impedances will furtherly improve the high frequency behaviour:
Kai
Hi Vladimir,
The gain fold-back that Kai reports for the Sallen Key (SK) low-pass is primarily the result of the op amp's decreasing open-loop gain (Aol), and increasing open-loop output impedance (Zo) as the frequency increases. The fold-back characteristic tends to be more pronounced with very low operating current (Iq) op amps such as the TLC2252, because they often exhibit low bandwidth and higher, sometimes complex, Zo across frequency. The fold-back often can be pushed out to a higher frequency by using a higher bandwidth op amp which typically has lower Zol, but at the expense of higher Iq. Additionally, scaling the resistors and capacitors as Kai has recommended improve's the SK's filter responses so that is an easy improvement.
If your circuit can tolerate somewhat higher Iq, TI has modern op amps that feature wide bandwidth, low noise, and low Zo across frequency - all of which will improve the Sallen-Key's low-pass response vs frequency. For example, the OPA320 and lower current OPA325 are two modern 5 V CMOS op amps I would recommend. I re-simulated your Sallen Key low-pass with the OPA325 in place of the TLC2252 and the response improvement is evident from the graph shown below. Additionally, I included the case where the resistors were scaled up by 10x, and the capacitors down by 10x, in the simulation.
The fold-back with the OPA325 doesn't occur until about 61.7 kHz, with an attenuation of -72 dB for the 1x components. The fold-back frequency increases to about 135 kHz, with an attenuation of -85.5 dB for the 10x components. I expect this would be even more improved with the higher current OPA320 which has wider bandwidth and even lower and flatter Zo across frequency.
Do note that the Multiple Feedback (MFB) exhibits better fold-back behavior than the Sallen Key. If you need to stay with a low current op amp such as the TLC2252 there is response improvement from the MFB. And if your need higher supply voltages there are many op amps that will fill this role.
Regards, Thomas
Precision Amplifiers Applications Engineering
Dear Tamara, dear Kai, dear Raymond, dear Thomas!
Thank you.
In fact, everything is somewhat more complicated.
I need a band-pass filter at 1873 Hz. I calculated manually an MFB filter at a -3dB band 200Hz. This is a “MFB filter 1” circuit. I need to receive a signal with phase manipulation (phase-shift keying signal - BPSK). I drew a circuit in TINA-TI, that generates BPSK signal, please, see attachs. I set the size of a Filter 1 quality Q= 10.
The signal after the filter 1 had the form: after phase change some time undesirable oscillatory process took place. It stirs to normal work of the receiver - we cannot correctly define the moment of change of a phase. This can be clearly seen if you look at the output signal with an oscilloscope.
Then, I have changed parameters of the filter, I have increased considerably a pass-band (approximately to 1500 Hz and even more) and have set filter quality Q = 1, please see “MFB filter 2” circuit. Then, process of change of a phase began to occur no unwanted generation. Such form of a signal suits us, but here the filter 2 already ceases to be the normal filter! So I tried to connect low-pass filter and high-pass filter in series. Please see "Low-pass 2200 plus high-pass 1300.TSC". With such a filter, there is also no unwanted generation at the time of the phase change, but the bandwidth of such a filter is also large.
QUESTION: Whether it is possible to avoid long oscillatory process at the moment of change of a phase and thus not to increase strongly pass-band of the filter?
TIA
Sincerely
Vladimir Naumenkov
Minsk, Belarus
Low-pass 2200 plus high-pass 1300.TSCLow-pass 2200 plus high-pass 1300_BPSK.TSCMFB filter 1_BPSK.TSCMFB filter 2_BPSK.TSCMFB_Filter 1.TSCMFB_Filter 2.TSC
Dear Thomas, thank you.
That's what I still can't understand. Take the circuit "MFB Filter 2", please see attach. On the oscilloscope in TINA, you can see that the output signal is AHEAD of the input signal. How can this be? I even simulated the circuit in LTspice and it's the same there! Please, see "Figure 1" - the blue trace is the out signal. In the range from 0 to 0.7 ms, this is clearly visible. Or does it just seem so?
Sincerely,
Vladimir Naumenkov
Minsk, Belarus
Hi Vladimir,
maybe I would do it this way:
A passive bandpass has a more linear phase response than an active one.
Kai
Hi Vladimir,
Here is another passive filter reference. This is how powerline communication uses in the application with 4th order passive BPF, which is typically done at much higher frequency. At lower filtering frequency, the filter circuit may not have significant advantages.
Enclosed is a simulation for the BPF.
Best,
Raymond
Hi Vladmir,
The 4th order bandpass filter is not going to be as precise as the active filter, since the tolerances of LC are fairly large from off the shelf components, unless you are willing to fine tune the values on your own for a given frequency.
Typically, I use the equations on p.29 to approx. BPF region, then I am altering the damping resistor value to remove Q and narrow the BW down to my design specification. For your filter design requirements, f1=1575Hz, f2=2025Hz, the filter characteristic impedance is approx. sqrt(L1/C2)=sqrt(1.5mH/1.5nf)=1kOhm (you can change to what your application is called out).
For your filter design, C1 = 101nf, C2 = 78.6nf, L1 = 77.6mH and L2 = 101.1mH, the capacitor C on the right side of Figure 32 is used to match the impedance of the Rx receiver. Here is what I got after I enter above values into the simulation.
Once the filter is close to what you'd like, you can start to play around simulation and refine LCR parameters to meet the design requirements.
Enclosed is the simulation of the above filter.
If you have any questions, please let us know.
Best,
Raymond
This is another filter that could be used:
What filter can be used also depends on the demodulation scheme :-)
Kai
