Improve Test Performance with Low-Cost Signal Sources Using Inline Filters

Do you need a pure sine wave for a test but only have an arbitrary function generator with a high harmonic level available? Perhaps you’re mixing the outputs of two signal generators and must select the upper sideband component at the mixer output. How can you do this? The solution is to use inline RF filters, such as Crystek Corporation’s CLPFL-0200 (Figure 1, left) with an SMA connector and CLPFL-0021-BNC (Figure 1, right) with a BNC connector.

Figure 1: Inline coaxial filters, such as the CLPFL-0200 with an SMA connector (left) or the CLPFL-0021 with a BNC connector (right), can reduce signal harmonics and noise on signal sources. (Image source: Crystek Corporation)

RF filters clean up signals by selectively attenuating unwanted frequencies while passing desired frequency components. Inline filters, intended for use with coaxial lines, are designed with a 50 ohm (Ω) nominal impedance. These filters reduce noise by reducing the signal bandwidth. They also control the signal spectrum to reduce harmonics, images, and interfering signals.

Types of filters

There are several types of inline filter configurations, including low pass, high pass, and bandpass (Figure 2).

Figure 2: Shown are the frequency responses of low-pass, high-pass, and bandpass filters. (Image source: Art Pini)

Low-pass filters pass frequencies below a fixed cutoff and can eliminate the harmonics of a signal with the cutoff set just above the fundamental frequency. High-pass filters pass frequencies above a fixed cutoff and can eliminate an interfering signal with the cutoff set above the power line frequency.  Bandpass filters attenuate unwanted signals by passing frequencies within a desired band and can be employed as a preselector for an RF front-end. The region where the signal is transmitted with little loss is called the passband, and the region where the signal is highly attenuated is the stopband. The region(s) between the passband and the stop band is the transition region(s).

Selecting the right filter

Filters are designed for specific frequency response characteristics. These include the sharpness of the transition from the passband to the stopband, the flatness of the passband and the stopband, and the phase response as a function of frequency. There are several classic designs shown in Figure 3.

Figure 3: The frequency response of several types of classic filters shows the differences in roll-off and flatness characteristics. (Image source: Art Pini)

The Butterworth filter has a flat passband response and a moderate roll-off rate. The Bessel filter has the most linear phase response but the slowest roll off; it would typically be used when a band-limited pulse waveform must be transmitted with minimum distortion. The Chebyshev filter has a fast roll off but has ripple in the passband. The inverse Chebyshev filter has a flat passband response and a fast roll off but exhibits ripple in the stop band. The Butterworth and the Chebyshev are two of the most widely used inline filters.

The roll-off characteristics of any filter type are affected by its order. The order is derived from the filter's transfer function and indicates the number of poles in the design. In general, the higher the filter order, the faster the roll off (Figure 4).

Figure 4: Shown is a comparison of a Butterworth low-pass filter response for a filter with orders of 5 through 9. The higher the filter order, the faster the roll off in the transition region. (Image source: Art Pini)

Crystek’s CLPFL-0200 is a 7^th-order Butterworth low-pass filter with a passband of DC to 200 megahertz (MHz) and an insertion loss of 2.2 decibels (dB) at a frequency of 210 MHz. This filter could be used to clean up the output of a signal generator when making an effective number of bits (ENOB) measurement on an 8-bit analog-to-digital converter (ADC) (Figure 5).

Figure 5: Shown is the result of a 200 MHz low-pass filter being used to remove harmonics and noise from a signal generator. The filtered signal (lower trace) has significantly reduced noise and harmonic levels. (Image source: Art Pini)

The upper trace shows the signal generator output spectrum with a second harmonic only 22 dB below the fundamental. With the filter (lower trace), the second harmonic is down over 70 dB, and other harmonics are below the noise floor. Note also that the noise floor above the filter cutoff frequency is lowered by better than 40 dB.

High-pass filters eliminate interfering signals with a lower frequency than the desired signal (Figure 6).

Figure 6: Shown is a high-pass filter being used to eliminate a 13 MHz interfering signal from the desired 30 MHz signal (upper trace). The filtered signal appears in the lower trace. (Image source: Art Pini)

In Figure 6, a high-pass filter attenuates a 13 MHz interfering signal and passes the 30 MHz signal of interest. The interfering signal's effect can be seen in the time-domain view (upper left) as an amplitude variation of the signal peaks. The filtered signal (lower left) has flat peak amplitudes.

A filter such as Crystek’s CHPFL-0025-BNC, a 7^th-order 25 MHz Chebyshev high-pass filter with BNC connectors, could attenuate the interfering signal.

Crystek filters are offered in up to 9^th-order configurations. For example, the CLPFL-0021-BNC mentioned earlier is a 21 MHz Chebyshev response, 9^th-order, low-pass filter. It delivers a transition region that rolls off at about 55 dB per octave.

Bandpass filters typically require more components than low or high-pass filters, which take up space and add to the BOM. Crystek addresses this using surface acoustic wave (SAW) technology to allow its bandpass filters to fit in the same package as low-pass or high-pass filters. An example SAW bandpass filter is the Crystek CBPFS-0915 with SMA connectors and a 26 MHz bandwidth centered on 915 MHz.

Conclusion

Inline RF filters improve test performance by eliminating harmonics, noise, and interference from signal sources. Companies like Crystek offer a wide range of inline filters to match your signal-conditioning needs.