Gain Compression and Breakdown

It is clear that premature gain compression in AlGaN/GaN HEMTs is a source of many ailments: reduced output power and efficiency, and most importantly, reduced linearity. Roll your mouse over thrust III, fig. 1 to see the illustration where a typical power and gain data for a passivated AlGaN/GaN HEMT is plotted, along with the desired constant gain behavior in red. The reason for the soft power saturation is unclear and is the object of intense research.

Thrust III, Figure 1- (Left): Soft-saturation or early gain compression observed in a passivated GaN HEMT (measured dark, without UV illumination). Red curve emphasizes desired (flat) gain. (Right): Gain and power in an unpassivated GaN HEMT with and without UV illumination.

Presently we can identify many possible sources of gain compression; these are summarized in thrust III, fig. 2. Bulk and surface traps in the HEMTs are one possibility, as well as several device and circuit issues, many suggested recently by R. Trew.  We believe that all may be important.  For example, thrust III, fig. 1b shows RF gain and power data for an unpassivated GaN HEMT under both dark and UV-illuminated conditions.  UV illumination reduces gain compression to some extent. Note also that the UV illuminated device and passivated device (thrust III, fig. 1a) show virtually identical characteristics. These observations suggest that surface traps play a role in early gain compression, however, although gain compression has been reduced, it has not been completely eliminated.  The remaining compression could be due to the transport mechanisms recently put forth by Trew, which we address here.

Thrust III, Figure 2- Current understanding of premature gain compression in GaN devices.

The first issue of device design raised by Trew and identified in thrust III, figs. 3-4 is the space-charge choke near the source.  This arises because the combination of high current densities (>1A/mm) and low electric field in the source region causes the electron density to increase beyond the equilibrium value to maintain current continuity. This in turn causes the source resistance to increase in the region of the I-V plane near the knee-voltage at high current, causing the gain to drop. This would be reflected in gain compression as P_in was increased, since the problem region of the curve is increasingly accessed at high input drive. (thrust III, fig. 3). These problems are alleviated when the conductivity of the access region between the gate and the source is increased so that it never limits the current. In the next section we address two methods of reducing access resistance – those of polar dielectric deposition and recess etching – as a means to increase device g_m in concert with the Hajimiri model requirements. These efforts (also listed on the left of  Thrust III, fig. 4) alleviate the source choke issue put forth by Trew and hence address phase noise on all fronts.

The second issue is that of breakdown in HEMTs. Hajimiri predicts that the contribution of white noise to phase noise in some embodiments is as important as the up-conversion of 1/f noise. The white noise spectrum is a combination of

·     Johnson noise in the channel and electrodes
·      Shot noise due to gate leakage and drain current
·      Multiplication noise in the high field region of the channel.

In high charge density devices such as AlGaN/GaN HEMTs the electric fields at pinch-off are close to the ionization fields and hence it is reasonable to expect that the device experiences impact ionization over some portion of the load line. This raises the white noise floor and is very detrimental to phase noise. Also the ionization introduces an inductive delay in series with the FET analogous to that existing in an IMPATT diode. This is indicated in the high-field portion of thrust III, fig. 4.  The magnitude of this delay is proportional to the magnitude of ionization. This delay will detract from the gain and hence contribute to gain compression as P_in increases. It is therefore imperative to enhance the breakdown voltage of the HEMTs to reduce white noise and gain compression contributions to phase noise.

Low Phase Noise Combiner Systems (Hajimiri & York)

Combiner structures are especially important for achieving desirable power levels in GaN electronics.  The use of a large number of devices in a combiner can have a positive influence on the noise properties of the system.  In transmitter applications, particularly for certain types of radar, the amplifier must not seriously degrade the phase noise of the signal to be amplified, which is typically generated from a source that is phase-locked to a highly stable reference oscillator.  The phase noise reduction can be derived as follows [1], which roughly parallels earlier work aimed at noise reduction in oscillator systems [2]. With reference to the notation in fig. 5, we assume the input signal to be amplified is a noisy signal of the form:

where _describes the time-dependent phase fluctuations of the input signal.  The phase noise at the output of each amplifier is degraded, primarily from upconverted 1/f noise, due to the nonlinear devices in the amplifiers.  Amplitude noise in the bias supplies can also be upconverted to near-carrier phase noise.    For our purpose the origin of the noise is unimportant, and we simply describe the total excess noise contribution of each amplifier by a time-domain fluctuation , so that the total output signal is then given by

assuming small phase fluctuations.  We have also assumed an ideal symmetric, broadband, linear power splitter for simplicity. We now assume that the input and amplifier noise sources are uncorrelated random (ergodic) processes with zero time average, and apply the Wiener-Khintchine theorem to compute the power spectrum of the noise fluctuations.  If the amplifiers have roughly the same noise power spectral density, the power spectral density of the output signal phase fluctuations (i.e. the phase noise of the output signal) will be given by

where represents the noise spectrum associated with the  input signal,  represents the excess phase noise contributed by a single amplifier, and the tilde (~) denotes a Fourier transform (defined in the usual way for a random process). This result shows that the phase noise contributed by the amplifier ensemble is reduced by 1/N.  This is an intuitive result since the input signal being amplified adds coherently at the output (power increasing as N²), whereas the uncorrelated noise fluctuations add incoherently, and hence the peak amplitude of the carrier increases N times more rapidly than the noise skirts.        

Figure 6 – X-band combiner module developed at UCSB [4].  Residual phase noise of this amplifier was measured at Redstone Arsenal, Alabama (Courtesy Dr. Will Caraway).

Note that this result is sometimes confused with the  dependence predicted for noise due to gain fluctuations, [3] ie. amplitude noise.  Such noise contributions can sometimes be eliminated by operating the amplifier in or near saturation.  Amplitude noise due to bias supply fluctuations can similarly be managed through improved bias circuit design and filtering.  By improving the linearity of the devices as discussed in Thrust II, we can also minimize the conversion of AM-to-PM noise

Return to top of page