Design Ideas: April 27, 1995

Unlike lowpass, bandpass, and other magnitude-altering filters, allpass filters can shift the phase of a signal without affecting its amplitude. For a first-order allpass circuit, the transfer function is

As you sweep the variables from zero (dc) to infinity, the sign of H(s) changes from plus to minus, indicating a change in phase from 0 to 180°. You can realize this transfer function in two wideband transconductance amplifiers (WTAs). The circuitry inside the dashed lines in Fig 1 is one allpass network.

A WTA's transfer function is I_OUT8V_IN/Z, where 8 is simply an internal constant and Z is an external gain-setting component connecting the WTA's Z+ and Z- pins. The transfer function for voltage amplification is V_OUT/V_INI_OUTxZ_OUT. Most applications require a resistive Z. But the WTA also accepts an inductor, a capacitor, or any other impedance network for Z.

The allpass circuit combines a resistive-Z WTA (IC₁) with a capacitive-Z WTA (IC₂). At low frequencies, IC₁ dominates the circuit's output because the capacitor's high impedance allows only a low I_OUT from IC₂. Rising frequencies lowers this impedance, causing the current from IC₂ to dominate at high frequencies. Moreover, IC₂ inverts, and IC₁ does not, producing the desired noninverting unity gain at dc and inverting unity gain at high frequencies.

Communications and signal-processing applications use allpass networks widely. An example is a 90°-phase-shift network, which, with appropriate mixers, produces a single-sideband signal. In Fig 1, the two allpass circuits have corner frequencies that differ by a factor of 7.5. The output RC networks determine these corner frequencies. The result is an output-phase difference that remains close to 90° over a wide frequency range. Measurements show 0.2-dB amplitude variations and a phase difference of 90°±7° from 180 to 740 kHz-a 4:1 range. (DI #1696)