Internet Explorer is not a supported browser for TI.com. For the best experience, please use a different browser.
Video Player is loading.
Current Time 0:00
Duration 7:57
Loaded: 0%
Stream Type LIVE
Remaining Time 7:57
 
1x
  • Chapters
  • descriptions off, selected
  • en (Main), selected

Hello, and welcome to Active Filter Design in Minutes with TI's Filter Design Tool. In this video series, we will be exploring the use of the Texas Instruments Filter Design tool to design active filters to solve several example problems. We'll begin by considering a scenario in which a resolver is under test in a laboratory setting.

For those not familiar, a revolver is a sort of rotary transformer that can be utilized to determine the angle of a shaft. Essentially, the shaft rotates the primary winding or coil of a transformer, while two secondary coils are fixed and offset from each other by 90 degrees. Thus, the coupling between the primary and secondary is dependent on the shaft angle, such that the sine and cosine secondary outputs are the excitation signal modulated by the sine and cosine of the shaft angle, respectively.

This particular resolver is excited with a six-fold peak to peak 5-kilohertz signal. The resolver transformation ratio, which describes the ratio of the number of turns in the primary and secondary coils, is 0.25, so the outputs will be only 1.5 volt peak to peak. The resolver shaft is spinning at 600 RPM or 10 hertz. The sine and cosine secondary winding outputs should therefore be 5-kilohertz 1.5-volt peak to peak signals modulated by the sine and cosine of the shaft angle, such that they are 90 degrees out of phase with each other.

However, when this system is tested in the lab, it is discovered that 200 millivolts peal to peak of approximately 55-hertz power line noise is coupling onto the measured signals. Our objective is to implement an active band-stop or notch filter to eliminate this 50- to 60-hertz noise without significantly attenuating the output signals. The supply voltages available are plus and minus 5 volts.

Before we proceed to the Filter Design tool, we will briefly explore the scenario via simulation in TINA-TI. As shown, we have a 5-kilohertz excitation signal, sine and cosine waves that vary in accordance with the specified shaft RPM, and a 55-hertz wave representing the power line hum noise. The ideal sine and cosine secondary winding outputs, as well as the actual or non-ideal results exhibiting noise coupling, are also shown.

To eliminate the noise due to the power line hum, we will implement a notch filter centered at 55 hertz. Ideally, the passband will have no attenuation or gain at 5 kilohertz. We will employ a generous stopband of 50 hertz with an attenuation of minus 20 decibels, which will allow us to use a low-order filter.

With these requirements in mind, we proceed to the TI Filter Design tool. We choose a gain of 0 and center frequency of 55 hertz. We set the passband bandwidth at 500 hertz in order to minimize the attenuation out at 5 kilohertz. Our stopband bandwidth is 50 hertz. We can set our stopband attenuation to minus 20 dB, which helps reduce our filter order and thus reduce the circuit complexity.

After adjusting the specification table to match our desired specs, we can consider some of the options presented by the tool. We can compare the magnitude, phase, and group delay for several filter types, as well as the step responses. For this example, we will select a Chebyshev filter for the steepness of its roll-off.

The tool suggests a fourth-order filter with a Bainter topology. This topology is favored because of its clean and consistently deep notch and low sensitivity to component mismatches. For more information, see Band-stop Filters and the Bainter Topology, by Bonnie C. Baker.

The responses of each filter stage can be viewed on this page, if desired. But we'll proceed to the design page. Because we have plus and minus 5-volt supplies available, we will leave the supply voltage settings as they are. In order to explore the effects of tolerance on our design, we'll set the resistor series to e24, or 5% tolerance, and the capacitor series to e12, or 10% tolerance. We'll set the type of sensitivity analysis to Monte Carlo and Corner, then click Update Design.

You'll notice the actual values in the Target/Actual columns have updated. In this case, the tool notes the selected passive component tolerances could compromise the integrity of our first stage. By zooming in on the various plots, we can see how component tolerances will impact the filter by introducing uncertainty bands.

To restore our design integrity, we'll change the resistor series to e48, 2%, and the capacitor series to e24, and update the design again. By default, the tool has suggested the LMC7111 op amp. We can swap this part for one of the higher-gain bandwidth by selecting Choose Alternate Op Amp. Note that in order to export your design to TINA-TI, you need to select an amplifier that has a spice model. We'll be going with a TLC27L7 amplifier, which has a dual package to save space, higher gain bandwidth, and lower offset.

Now that our design is ready, we can click Export. This brings us to a summary page from which we can export the design to TINA-TI, view the response plots, view our bill of materials, and more. Click Export Design, and the tool will begin generating a TINA-TI design. This may take a while.

When it has finished, click Download at the bottom of the page to download the filter design. Open the exported design in TINA-TI. We'll test it by running an AC transfer characteristics sweep.

As the results show, we achieve a notch of negative 118 dB in the center of the stopband with an attenuation of at least minus 65 dB across the 50 to 60 hertz range. At 5 kilohertz, we have a slight gain of 0.066 dB and a phase shift of negative 719 degrees. For our resolver signals, the resulting phase and time delays will be negligible.

We can now test our filter by applying each of our resolver outputs to its input, as shown. The circuits are slightly condensed versions of the Filter Design tool output in order to reduce the schematic area, but no design changes have been made. We see the filtered cosine signal takes a little while to settle, but within about 2 milliseconds, our power line noise is successfully being filtered out.

Zooming in and comparing the filtered output to the ideal winding output, we see the phase delay is negligible, as expected. Even increasing the noise levels 10-fold to 2 volts peak to peak, we see the filters are still able to successfully attenuate the power line hum out. This concludes our example of a band-stop filter design using TI's Filter Design tool. In our next example, we'll explore the design of two band-pass filters in order to implement part of a DTMF receiver filter bank.