Bias Techniques for GaN and pHEMT Depletion Mode

advertisement
 Bias Techniques for GaN and
pHEMT Depletion Mode Devices
John Bellantoni
Technical Director Wireless Communications
TriQuint, Inc.
San Jose CA 95434
Introduction
Advances in process technology have led to GaN depletion mode devices finding their way out of research
labs and into communications markets. In many microwave wave applications for example, the LDMOS
devices that have dominated power amplifier designs are giving way to GaN solutions, particularly the final
output stage where efficiency and linearity are key requirements. The high power density, high efficiency
capability of GaN makes it valuable for microwave applications, and the trend of GaN devices finding their
way into future communication systems is predicted to continue. As a result, there are incentives to develop
novel bias techniques that provide proper bias control and temperature compensation for depletion mode
devices. These techniques must be compatible with communication applications where stable operation into
high capacitance loads is necessary for the wide video bandwidths found in Doherty configured amplifiers
that carry 4G and LTE signals. As pHEMT devices are well established in numerous applications, the
approach is to develop universal techniques that work for any type of depletion mode device.
Concepts, practical circuit recommendations, measurement data and discussion are presented in the
following sections:
1.0 Design Requirements for Depletion Mode Device Biasing
2.0 Temperature Characteristics of TriQuint GaN Devices
3.0 Temperature Compensation Methodology
3.1 Temp Comp Made Easy
3.2 Temperature Compensation Techniques
4.0 Stability Considerations
5.0 Low-Cost Analog Temperature Compensation Technique
6.0 DAC Control
7.0 Practical Reference Designs
8.0 Lab Verification
9.0 Conclusions
10.0 Acknowledgements
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 1 of 19 -
www.triquint.com
Bias Tech
hniques forr GaN and
pH
HEMT Depletion Mode
e Devices
1.0
0 Depletion Mode Bias Design Req
quirements
Miccrowave
mplifier
(P
PA)
devic
ce
power
am
perrformance is
s typically ev
valuated for a given drain
quiiescent current. Rela
ated to the
e conduction
ang
gle, the gate voltage of
o the device
e is adjusted
unttil the des
sired quies
scent drain current is
i
ach
hieved. The
e majority of
o applicatio
ons require a
bia
as point somewhere between
b
line
ear Class A
ope
eration, wh
here full 360
3
degree conduction
occcurs, and th
he most efficient Class C operation
where only 180
0 degrees off the signal is
i conducted
d.
ost common in the PA world
w
is Class
s AB biasing
g,
Mo
where the trade-off betwee
en linearity and
a efficienc
cy
sable range. Class AB biasing
b
could
falls within a us
s “light Class AB” say fo
or example in
be classified as
e carrier am
mplifier of a Doherty cirrcuit. On the
the
oth
her end, “de
eep Class AB”
A
is employed for the
pea
aking ampliffier in a Doh
herty configu
uration or the
fina
al PA stage of a transmitter. Here the
t efficienc
cy
is p
paramount to
o linearity, so
s biasing the device at a
low
w quiescent current
c
as possible is ad
dvantageous
s.
Cla
ass C biasin
ng is employ
yed in a wide variety of
o
con
nstant enve
elope modu
ulation applications tha
at
use
e a single de
evice.
The
e first step towards
t
ada
aptive device
e bias is tha
at
the
e gate voltag
ge must be adjustable
a
to
o the desired
class of operattion. Given the populariity of moderrn
mbedded con
ntrollers in microwave
m
systems,
s
the
em
con
ntrol method
d of choice is the ubiq
quitous DAC
C.










Eng. Rev. 3 07/24/1
14
© 20
014 TriQuint
The DAC value for the dessired quiescent current
be determined using au
utomated rou
utines on a
can b
unit tto unit basiss at final tesst. This is a necessary
align ment step ffor power amplifiers. A
Alternatives
DACs inclu
ude mecha
anical potentiometers,
to D
EEPO
OTs, and sw
witched resisstor networkss.
Beyo
ond the fun
ndamental sstep of sele
ecting and
settin
ng the bias point, there
e are otherr factors to
cons ider. While
e the gate of a deple
etion mode
negligible ccurrent at low power
devicce draws n
levelss, the Schotttky diode in
n the gate sttructure will
start to rectify as the input power increases,
send ing significa
ant current in
nto the gate. Additional
C circuitry o
on the gate, such as le
evel shifters
MMIC
or biias resistorss, can requ
uire either positive or
nega
ative gate current fflows. Ofte
en bypass
capa citors are placed at the gate in
n order to
ove video b
bandwidth, and these capacitors
impro
can b
be quite larg
ge, up to hundreds of m
microfarads.
Op-a
amps, a com
mmon bias ccontrol device, have a
tende
ency to o
oscillate into capacitiive loads,
requiiring gate b
bias circuits that are sttable under
ng. Tempera
ature compe
ensation is
capa citive loadin
n needed forr critical line
ear applicatiions, and it
often
advantageou
us if the temperatu
ure slope
is a
adjusstment is in
ndependentt of the bias voltage
settin
ng. A summ
mary of the re
equirementss for biasing
deple
etion mode devices is provided be
elow. The
bias circuits desccribed in this paper add
dress these
requiirements.
Applicable to any deplettion mode technology: GaN
N, pHEMT, M
MESFET, etc.
DAC adjusttable and ON/OFF controlllable
Bilateral current source capable
c
of su
upplying posittive or negativve current flow
w
Low IR drop from driver to gate
Stable driving large capa
acitive loads
Stable driving impedanc
ces from open
n circuits to he
eavy current lloads
Tolerates supply
s
voltage
e variations
Independen
nt temperaturre slope and bias
b
voltage a
adjustments
One wire te
emp sensor suitable for SM
MT or hybrid a
assemblies
Low cost co
ommodity com
mponents ava
ailable from m
multiple vendo
ors - 2 of 19 -
www.triiquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
What happens when a fixed voltage is placed at
the gate of a depletion mode GaN or pHEMT
device and the quiescent drain current measured
over a wide temperature range?
Figure 1
illustrates
the
expected
result,
arbitrarily
normalized to 1A range. As carrier concentrations,
mobility and Schottky barrier voltages change over
temperature, so does the drain current. The
absolute value of the current does vary from unit to
unit based on pinch-off voltage variations across
the wafer, as illustrated by three units 1 – 3 in Fig.
2(a). On the other hand the temperature slope is a
derivative that has much less variation. In addition
to a strong linear dependence, there are small
percentages of second and third order terms. This
semiconductor physics is remarkably consistent for
a given process.
Maintaining a fixed quiescent bias point over
temperature is one of the essential techniques to
realizing high performance transmitter chains
capable of meeting specifications over a wide
range of operating conditions. By adjusting the
gate voltage over temperature, the quiescent
current can be held constant. Figure 2 presents
data from the lab, the actual gate voltages required
to hold the quiescent current constant for several
different TriQuint GaN devices at various different
current densities. Similar to the fixed voltage plot,
the gate voltage required to maintain constant
current over temperature shows a strong linear
dependence with small second and third order
contributions. For saturated operation the above changes in
quiescent current have little effect on overall output
power or efficiency performance. Data presented
in Section 4 will verify that a fixed gate voltage is
sufficient for saturated power applications.
In the world of modern digital communications
linearity and efficiency are the key parameters.
Devices in receiver chains and in transmitter
blocks outside of DPD loops are required to be
highly linear.
Final stage Doherty amplifiers,
drivers and pre-drivers inside the DPD loop are
required to be highly efficient and capable of being
linearized with realizable polynomial functions.
Eng. Rev. 3 07/24/14
© 2014 TriQuint
1.5
Quiescent Drain Current (A)
2.0 Temperature Characteristics of TriQuint
GaN Devices
1.3
1.1
0.9
0.7
0.5
-40
-25
-10
5
20
35
50
65
80
95
110 125
Temperature (°C)
Fig. 1. Illustration of depletion mode device quiescent
drain current dependence over temperature for fixed
gate bias voltage.
- 3 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
-2.5
-4.1
12.5 mA/mm
-2.6
10 mA/mm
-4.2
20 mA/mm
Vgq (V)
Vgq (V)
-2.7
-2.8
Unit 1
-2.9
-4.3
-4.4
Unit 2
Unit 3
-3.0
-4.5
-3.1
-40
-15
10
35
60
-4.6
85
-40
Temperature (°C)
-15
10
35
60
85
Temperature (°C)
(a)
(b)
-3.3
10 mA/mm
Vgq (V)
-3.4
-3.5
-3.6
-3.7
-3.8
-40
-20
0
20
40
60
80
Temperature (°C)
(c)
Fig. 2. Gate voltage over temperature for GaN devices at constant current density. (a) 120W 32V
T1G4012032-FS, three devices at 12.5 mA/mm, (b) 30W 28V T1G6003028-FL, one device at 10 mA/mm
and one device at 20 mA/mm, and (c) the 55W 28V T1G4005528-FS, one device at 10 mA/mm.
3.0 Temperature Compensation Methodology
3.1 Temp Comp Made Easy
The simplest and most expedient way to
temperature compensate a depletion mode device
is to ignore the entire subject. Simply apply Vgate,
a fixed voltage to the gate and then accept any
subsequent
performance
variations
over
temperature. This bare-bone technique is effective
in a number of applications. The further a device
is biased into Class AB and Class C operation, the
more likely that performance using fixed gate
voltage bias will be satisfactory.
Saturated
Eng. Rev. 3 07/24/14
© 2014 TriQuint
modulation schemes and applications with relaxed
temperature requirements are all ideal candidates
for fixed gate voltage bias schemes. There is a lot
to be said for the simplicity shown in Fig. 3.
VDD
Vgate
C
C
L
RF_IN
L
PA
RF_OUT
Fig. 3. Standard bias network with fixed voltage Vgate
applied to device gate.
- 4 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
3.2 Temperature Compensation Techniques
Advancement of modern digital communications
systems often pushes the technology envelope. In
these systems the temperature performance of
power devices is an important factor. Cellular
base stations, microwave point-to-point radios and
many ISM applications employ bias techniques
that minimize linearity degradations as the ambient
temperature varies over the course of the day and
seasons.
There
are
two
primary
approaches
to
compensating the bias point of power amplifiers as
the temperature changes:


Closed Loop
Open Loop
Closed loop techniques monitor the quantity in
question, for example device drain current, and
feed that information back to a controller. When
the loop is closed by the controller, the quantity in
question is maintained at the set-point value, within
the accuracy of the monitor.
Open loop techniques apply the set point directly
to the device and change that set point based on a
temperature sensor without regard to the actual
drain current. The consistencies of semiconductor
devices make open loop control possible, as the
temperature characteristic does not vary
appreciably from unit to unit. Both closed and
open loop temperature compensation can be
further divided into analog or digital techniques.
Figure 4 and Fig. 5 provide a summary of bias
compensation techniques in block diagram form.
achieved.
Open loop controllers cannot
compensate for long term drift of devices, a
phenomenon that has been observed in early
generations of semiconductor processes.
To
account for long term drift with an open loop driver,
designers can either ignore the drift or take
advantage of the fact that the majority of long term
drift occurs within the first hours of operation.
Devices are burned-in before final alignment, a
process suitable for limited volume production.
Digital closed loop techniques (Fig. 4) do not
require manual alignment or burn-in and can
compensate for long term drift. However material
costs are higher, the loops require development of
PID controller algorithms and the sensors need to
be stable over temperature. Complications arise
with closed loop techniques when the amplifier
must operate continuously, such as in cellular base
stations.
Information from RF power output
monitoring could be added into the equations for
the device drain current settings, or the RF
removed during periodic maintenance cycles. This
additional complexity is addressable only in the
digital domain. Large analog suppliers integrate
digital monitor and control functions into products
designed specifically for closed loop control of
multiple PA stages [1].
Each approach has its own unique set of
advantages and issues. Open loop techniques
have lower material cost than closed loop but
require alignment at final test, and can be less
accurate over temperature. Alignment can be a
manual process or automated with digitally
controlled devices, such as an EEPOT, adjusted
by the test stand until the desired operating point is
Eng. Rev. 3 07/24/14
© 2014 TriQuint
VDD
DIGITAL PID
EMBEDDED
CONTROLLER
CURRENT
MONITOR
ADC
DAC
C
GATE
DRIVER
C
L
RF_IN
Rsense
L
PA
RF_OUT
Fig. 4. Closed loop bias compensation. Bias stability
over temperature is determined by the current monitor
function. The embedded controller uses a Digital PID
controller loop to adjust the gate voltage to the required
drain current when RF is periodically turned off.
- 5 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
Analog closed loop techniques constantly respond
to changes in drain current, making them useful
only for small signal Class A constant current
designs. There are several well-known analog
Class A closed loop bias circuits that use a single
op amp, or even just a matched transistor pair [2,
3]. Block diagrams for open loop techniques are
provided in Fig. 5. The three techniques are
similar in that no monitoring of the drain current is
performed. Shown in Fig. 5(a), a mechanical
potentiometer is used to set the gate voltage, while
a DAC replaces the potentiometer in Fig. 5(b).
The temperature coefficient of the voltage output is
set by the resistor value from the temperature
sensor feeding into the summing function. This
design feature is a key contribution to the
discussion of Class AB device temperature
compensation. In the past tight coupling between
the temperature sensor and RF device was
required to track gate voltage. For the open loop
designs presented here, the temperature sensor
coefficient can be arbitrary because it is scaled
when entering the summing function. Therefore
the temp sensor does not need to be close to the
RF device and its output does not have to match
the RF device. Figure 5(c) differs in that the
embedded controller reads the temp sensor, and
then using a look-up table or interpolation trims the
gate voltage to the needed value. Since the lookup table scales the temperature sensor coefficient,
the response of the sensor can be arbitrary relative
to the RF device, allowing for a wide flexibility in
the selection of the temperature sensor.
EMBEDDED
CONTROLLER
TEMP
SENSOR
+V
DAC
VDD
C
GATE
DRIVER
Set_Point
VDD
C
L
L
RF_IN
C
C
GATE
DRIVER
TEMP
SENSOR
L
L
PA
RF_OUT
RF_IN
(a)
PA
RF_OUT
(b)
EMBEDDED
CONTROLLER
ADC
DAC
TEMP
SENSOR
VDD
C
C
GATE
DRIVER
L
L
RF_IN
PA
RF_OUT
(c)
Fig. 5. Open Loop Bias Control Techniques. Gate driver controlled by (a) mechanical set-point
and temperature sensor, (b) DAC set-point and temperature sensor and (c) all digital control.
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 6 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
4.0 Stability Considerations
Operational amplifier circuits are prone to
oscillations when driving large load capacitances.
When a capacitor is added to the output, a pole
appears in the response as shown in Fig. 6. While
the gain starts to drop at 40 dB/octave instead of
20 dB/octave, the potential for instability arises
when the phase exceeds 180 degrees before the
gain is less than 0 dB [4, 5], a sufficient condition
for oscillation.
There are a number of ways to avoid the oscillation
condition. These design “tricks” are summarized
in Fig. 7. Each has its own advantages. In Fig.
7(a) a series resistor isolates the capacitive load
from the op amp feedback loop. Simple and
guaranteed to always work, the series resistance
will need to be large enough to avoid oscillations.
However if there is appreciable gate current flow,
the series resistance is an impediment to getting
the maximum performance out of the RF device.
Current flow into the gate, typical of pHEMT
devices as they approach compression, will pull
the gate voltage more negative, shutting the device
off. For designs that have negligible gate current
flow the series resistor is a preferred approach.
Voltage regulators are designed to drive
essentially unlimited large capacitances as shown
in Fig.7(b).
This can be useful for modern
Fig. 6. Bode plot for op amp with and without capacitive loading CL [6]. The capacitor
loading adds a pole to the response. The pole quickly sends the phase to more than 180
degrees while the gain is still greater than 1, a sufficient condition for oscillation.
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 7 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
Doherty Amplifiers in multi-carrier environments,
where video bandwidths greater than 100 MHz
require significant bypassing. Commonly found in
these communication amplifiers are large
capacitors placed in shunt as close to the gate as
possible.
The disadvantage to the voltage
regulator approach is additional component count
and associated cost.
In addition, a power
consuming shunt resistor is required to maintain
minimum current through the regulator and sink
any negative current flow from the gate. Despite
the limited negative current handling capability,
many familiar with using bench power supplies and
shunt resistors will find appealing similarity to this
approach.
Common practice is to employ various
combinations of isolation, gain and dual feedback
for universal gate driver designs.
Gain may be added as shown in Fig. 7(c).
Capacitive
loading
capability
increases
proportionally to the gain [6].
As the gain
increases over 5X the loop dynamics are such that
the added capacitive pole falls outside the loop
bandwidth. Often the use of large gains is not
ideal, as the dynamic range of the DAC is
compressed by the gain in the driver.
Another approach employed in this work is leadlag compensation shown in Fig. 7(d), a dual loop
topology. A second loop added to the feedback
path places a zero in proximity to the pole thus
canceling them out [6, 7]. The references give the
impression that this cancellation approach is
difficult to implement. While true for high speed op
amps and stringent overshoot specifications, the
application here is at DC where distortion of high
frequency signals is not a consideration. With the
right component value selection, including the low
GBW op amps employed here, a wider range of
load capacitances can be tolerated with dual
feedback. The design can handle gate current
flow without appreciable voltage error as the series
resistor is within the feedback loop. The dual loop
feedback configuration has its advantages,
however there is still the potential for instability
under excessive capacitive loading conditions.
Eng. Rev. 3 07/24/14
© 2014 TriQuint
R_Iso
Vgate
C_LOAD
(a)
Voltage
Vi Regulator Vo
Vin
Vgate
Gnd
Rs
C_LOAD
-V
(b)
4R
R
Vgate
C_LOAD
(c)
Rfb
Cfb
Rs
Vgate
C_LOAD
(d)
Fig. 7. Stability techniques for driving capacitive loads.
(a) Isolation Resistor, (b) voltage regulator, (c) gain, and
(d) dual feedback loop.
- 8 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
PN junction is adjusted to achieve 0 V output at
whatever temperature equilibrium is achieved at
the ambient temperature setting.
5.0 Low-Cost Analog Temperature
Compensation Techniques
Conventional PN junctions are often employed as
temperature sensors. Empirical work has shown
that the PN junction of a BJT device is linear over
a wide temperature range. In this work a generic
PNP device is employed as the PN junction
temperature sensor.
A low-cost, single wire
sensor can be formed with the collector of the PNP
transistor connected to ground. Using a PNP
transistor allows die attachment to ground in hybrid
assemblies and use of packaged parts with
surface mount technology. The single wire sensor
simplifies layout & RF bypassing compared to
thermistors employed in feedback loops [8].
The temperature sensor 0 V output at +25 °C
ambient temperature provides a convenient input
to a ground referenced summing junction. The
advantage of the summing junction is that at room
temperature the current through R_Slope is zero,
allowing the DAC to set the gate voltage
independent of the temperature sensor. As the
temperature increases, V_TS will also increase,
sending current into the summing junction through
R_Slope. When temperature decreases, V_TS will
decrease below 0 V, pulling current out of the
summing junction. By adjusting the value of
R_Slope the change in output voltage Vgate over
temperature can be set.
Figure 8 illustrates the principle of operation for the
low-cost
gate
driver
with
temperature
compensation. Buffered by an inverting amplifier
that uses half the diode voltage as reference
VREF, the output of the complete temperature
sensor is 0 V at room temperature, often taken to
be +25 °C. In reality power amplifier heat sinks
experience a temperature rise in the vicinity of the
power device when the ambient temperature is
held to +25 °C. So the bias resistor R_Bias to the
Since the temperature response is set by the value
of R_Slope, the temperature sensor does not have
to be tightly coupled to the RF device. Nor does
the temperature coefficient of the sensor need to
be known. Whatever the coefficient and thermal
coupling may be, it is scaled into the appropriate
current at the summing junction by R_Slope.
R_Offset
DAC
+5V
R2
R_Bias
PNP
R3
R1
R_Slope
A1
A2
V_REF
V_TS
Vgate
0V
Fig. 8. Low cost temperature PNP sensor and ground referenced summing junction
forms temperature compensated gate driver. R1=R2, R3=R_Offset. R_Bias is adjusted to
set V_TS=0 at ambient temperature conditions. V_REF is set to ½ the PNP Vbe voltage.
The PNP temperature sensor often requires a bypass capacitor to avoid RF interactions.
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 9 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
That allows arbitrary temperature slopes to be
achieved with a fixed temperature sensor
coefficient, as shown in Fig. 9. In contrast many
previous Class AB open loop drivers require tight
thermal coupling between a matched temperature
sensor and the transistor fingers of the RF device.
The use of a summing junction eliminates the need
for tight thermal coupling of the temperature
sensor and interactions between the offset and
slope adjustments. What the approach described
here does not eliminate is the need to align
quiescent current at final test, a process that is
easily automated using DAC or EEPOT control.
The final result is provided in Fig. 10. The voltage
from the gate driver is superimposed on the actual
gate voltage required to hold a constant quiescent
current over temperature. This first order linear
correction holds quiescent current constant over
temperature, with only the small 2nd order term
contributing to error.
-3.0
-0.5
Rs=25 kΩ
-0.6
Rs=50 kΩ
-3.2
Rs=100 kΩ
Gate Voltage (V)
Gate Voltage (V)
Rs=50 kΩ
Rs=75 kΩ
-0.7
-0.8
-0.9
Rs=200 kΩ
-3.4
-3.6
-3.8
-1.0
-4.0
-55
-35
-15
5
25
45
65
85
105
125
-55
Temperature (°C)
-35
-15
5
25
45
65
85
Temperature (°C)
(a)
105
125
(b)
Fig. 9. Output voltage V_Gate over temperature for three values of R_Slope. (a) Voltage output at −0.75 V for pHEMT
device, and (b) voltage output at −3.4 V for a GaN HEMT device.
-3.40
-3.45
Constant IDQ Gate Voltage
Vgg (V)
-3.50
Gate Driver Voltage
-3.55
-3.60
-3.65
-3.70
-40
-15
10
35
Temperature (°C)
60
85
Fig. 10. Gate driver voltage provides 1st Order (linear) approximation to the
required gate voltage for constant IDQ. T1G4005528-FS, IDQ=10 mA/mm
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 10 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
state.
With the DAC at zero volts during
initialization the output of A3 is at the full scale
DAC voltage. Inverting through A2, the pinch-off
gate voltage is applied to the depletion mode
device.
6.0 DAC Control
Initialization of digital to analog converters takes
time while the processor boots up, loads the image
and executes commands. During this startup
interval most DACS will provide 0 V output, or tristate. With zero volts at the DAC the output of A2
in Fig. 8 will also be at 0 V, an undesirable
saturated drain current (IDSS) condition for depletion
mode devices. Under saturated channel operation
inadequate heat sinking in linear applications will
result in junction overheating and decreased
device reliability. What’s needed is a simple circuit
technique that provides pinch-off gate voltage
rather than zero volts while the DAC is being
initialized. The addition of a simple inverting buffer
as shown in Fig. 11 accomplishes this initialization
DACs often drive the gate directly in closed loop
systems and automated test stands. During the
initialization period when the DAC is at zero volts,
a non-inverting circuit with offset provides pinch-off
voltage levels. Figure 12 shows the non-inverting
buffer and output voltage over the DAC range.
The resistor divider R1-R2 at the DAC output is not
only needed to make the voltage ratios work; for
DACs that tri-state during initialization the shunt
resistor keeps the DAC output at 0V and thus the
gate at pinch-off.
R5
R4
DAC
R_Offset
A3
V_HS
+5V
R2
R_Bias
R3
R1
R_Slope
A1
PNP
Vgate
A2
V_REF
V_TS
0V
Fig. 11. Inverting the DAC around its half-scale voltage V_HS sets Vgate to −4 V while DAC
initialization is completed. R4=R5, R_Offset=R3, V_TS=0 V at ambient temperature conditions.
0
VGATE (V)
-1
R1
5V
DAC
Rs
R2
Vgate
-3
Cfb
V_REF
-2
-4
Ros
Rfb
0
1
2
DAC Voltage (V)
3
4
Fig.12. DAC controlled gate driver (left) and gate voltage as a function of DAC voltage for V_REF=+4 V,
R1=R2=10 kΩ, Ros=Rfb=2 kΩ, Rs=25 Ω, Cfb=330 pF.
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 11 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
7.0 Practical Reference Designs
The principles outlined here for low-cost universal
depletion mode devices bias circuits were followed
to develop the two reference designs in Fig. 13 –14
and the corresponding BOMs provided in Tables
1 – 2. Each design has a safeguard to insure that
the gate voltage does not exceed 0V, which can
damage the device.
In the bipolar output version in Fig. 13, the
collector of Q2A is tied to ground. If U1B rails
positive during the power up sequence, then the
forward biased base collector junction holds the
emitter to approximately 0 V. When just the op
DAC
Q1
R2
22.0K
1%
R7
3.30K
1%
R9
3.30K
1%
+5V C1
0.1uF
8
U1C
4
2
+5V
3
R4
22.0K
1%
In both designs, R1 will need to be determined
such that the output voltage of U1A is identical to
the summation reference voltage at ambient
temperature conditions and quiescent current. The
potentiometer R8 is set to provide the optimal
temperature performance. Once the setting is
determined, the potentiometer can be left as is or
replaced by a fixed value resistor. For applications
with
demanding
temperature
performance
requirements, an EEPOT may be substituted and
automated alignment performed on each unit at
final test.
R3
22.0K
1%
+5V
R1
24.9K
1%
amp is used as shown in Fig. 14, a Schottky diode
clamp prevents excessive voltages at the gate.
U1A
1
C3
330pF
R6
1.00K
1%
R8
10K
Q2A
U1B
R10
2.00
1%
C2
-V 0.1uF
Vgate
Q2B
R5
1.40K
1%
-V
Fig. 13. Universal depletion mode bias circuit using complementary bipolar output stage. Stable driving
capacitances in the uF range and capable of providing more than +/- 200 mA bi-polar current drive.
Table 1. BOM, Universal Bias Circuit, Depletion Mode; Bipolar Output Version
Ref. Des.
Value
Description
U1
Q1
Q2
C1, C2
C3
R1
R2, R3, R4
R5
R6
R7, R9
R8
R10
n/a
n/a
n/a
0.1 uF
330 pF
24.9 kΩ
22.0 kΩ
1.40 kΩ
1.00 kΩ
3.30 kΩ
10 kΩ
2.00 Ω
LM2904, Op Amp, Dual, General Purpose, MSOP-8
MMBT2907A, Transistor, PNP, 60V, SOT-523
FMB3946, Transistor, NPN/PNP, 40V, 700mW, SOT23-6
Capacitor, Ceramic, 16V, 10%, X7R, 0402
Capacitor, Ceramic, 50V, 10%, X7R, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Trim Potentiometer, 1/2W, Top Adj, +/-10%
Resistor, Thick Film, 1/16W, 1%, 0402
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 12 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
R7
3.30K
1%
DAC
R1
24.9K
1%
R2
22.0K
1%
3
R4
22.0K
1%
U1A
U1C
C3
330pF
2
+5V
Q1
+5V C1
0.1uF
8
R3
22.0K
1%
+5V
R9
3.30K
1%
1
R6
1.00K
1%
R8
10K
R5
1.40K
1%
U1B
4
R10
2.00
1%
C2
-V 0.1uF
Vgate
D1
BAT60A
Fig. 14. Universal low-cost depletion mode bias circuit. The dual loop topology around U1B maintains stability
when driving capacitive loads in the uF range. This approach is suitable for GaN biasing, and other applications
under +/-20 mA gate current flow. For production R8 may be replaced with a chip resistor once the optimal value
is determined.
Table 2. BOM, Universal Bias Circuit, Depletion Mode; Op Amp Output Version
Ref. Des.
Value
Description
U1
Q1
D1
C1, C2
C3
R1
R2, R3, R4
R5
R6
R7, R9
R8
R10
n/a
n/a
n/a
0.1 uF
330 pF
24.9 kΩ
22.0 kΩ
1.40 kΩ
1.00 kΩ
3.30 kΩ
10 kΩ
2.00 Ω
LM2904, Op Amp, Dual, General Purpose, MSOP-8
MMBT2907A, Transistor, PNP, 60V, SOT-523
BAT60A, Diode, Schottky, 10V, 3A, SOD-323-2
Capacitor, Ceramic, 16V, 10%, X7R, 0402
Capacitor, Ceramic, 50V, 10%, X7R, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Resistor, Thick Film, 1/16W, 1%, 0402
Trim Potentiometer, 1/2W, Top Adj, +/-10%
Resistor, Thick Film, 1/16W, 1%, 0402
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 13 of 19 -
www.triquint.com
Bias Tech
hniques forr GaN and
pH
HEMT Depletion Mode
e Devices
The
e circuits we
ere fabricated on a single PCB with a
com
mmon temperature sens
sor and a quad op amp
p.
Fig
gure 15(a) shows the bllock diagram
m of the bia
as
boa
ard and Fig. 15(b) is
s a photog
graph of the
com
mpleted as
ssembly.
Table 3 provides a
com
mparison off the two circuits.
c
Th
he capacitiv
ve
loa
ading provid
ded is a guideline de
etermined by
b
sim
mulating the
e overshoot and settlin
ng times fo
or
instantaneous step changes in load curren
nt
e maximum rating.
r
bettween zero load and the
+5V
DAC
DAC
Temp
Sensor
TS
-5V
1
Bipollar
Drive
er
2
3
4
DAC
TS
Op Am
mp
Drive
er
5
6
7
8
Fig. 1 5(a). Block d
diagram of the
e experimenta
al bias board.
ve space the temperature s
sensor and DA
AC control is
To sav
comm
mon to both c
circuits. A ju
umper selects
s between a
PNP temperature sensor on th
he PCB or e
external PNP
or input at pin
n 4 of the con
nnector. A potentiometer
senso
was s
substituted forr the DAC.
F
Fig. 15(b). Com
mpact bias bo
oard with both versions of
lo
ow-cost bias circuit.
c
Actual size is 1.25” x 1.325”.
Ta
able 3. Bias
s Circuit Co
omparison
Circuit
Positive Current
C
(mA
A)
Bipolar
Op Amp
200
0
20
Eng. Rev. 3 07/24/1
14
© 20
014 TriQuint
Negative
N
Current Max. Voltage
(mA)
(V)
200
20
0
M
Min. Voltage
e
(V)
Capacitive
Loa
ading
−V + 2.3
−V + 1.5
<1
10 uF
< 1 uF
- 14 of 19 -
www.triiquint.com
Bias Tech
hniques forr GaN and
pH
HEMT Depletion Mode
e Devices
8.0
0 Lab Verific
cation
Verification con
nsisted of demonstrating that the op
am
mp circuit in Fig. 14 did
d not degra
ade linear or
o
satturated perfformance compared
c
to
o gate bia
as
sup
pplied in the
e traditional manner fro
om laboratorry
sup
pplies. Two bias conditions were ev
valuated ove
er
tem
mperature, a fixed ga
ate voltage VG, and a
con
nstant drain quiescent current IDQ. A TriQuin
nt
T1G
G4012036-F
FS 120 W second
s
generation GaN
N
RF
F power transistor served as the depletion mode
devvice. At +3
36 V the 18 dB gain Ga
aN device is
i
cap
pable of ge
enerating 24
4 W of line
ear power &
120
0 W of peak power in pu
ulsed operatiion (Fig. 16).
(a)
20
80
2000 MHz,
M
100 us 20%
VDS=+3
36 V. IDQ=360 mA
The
e evaluation
n fixture wa
as designed to interfac
ce
dire
ectly with the
t
bias bo
oard. The temperaturre
sen
nsor is locatted just outs
side the mou
unting clamp
p,
app
proximately 1 cm from th
he GaN device as show
wn
in Fig. 17.
A photogrraph of the
e completed
aluation platform is show
wn in Fig. 18
8.
eva
70
18
60
17
50
16
40
15
30
14
20
ZS=1.64−j 2.89 Ω
ZL=2.80−j 0.19 Ω
13
Gain
Drain Eff.
E
PAE
10
12
0
36
Efficiency (%)
Gain (dB)
19
38
40
42
44
46
Pout (dBm
m)
48
50
0
52
(b)
Fig. 1 6. T1G401203
36-FS (a) devic
ce package an
nd (b) Gain,
wer in pulsed o
operation.
Drain Efficiency an d Output Pow
Fig. 17. PCB layo
out for T1G4012036-FS 120W
W GaN Device
e.
Eng. Rev. 3 07/24/1
14
© 20
014 TriQuint
Fig. 1
18. Assemble
ed high power GaN evaluatio
on board
with b
bias controlle
er card.
- 15 of 19 -
www.triiquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
Operating at 2140 MHz with a drain voltage of +32
V, both CW and a fully loaded WCDMA signals
were employed to collect data. Thermocouples
were placed where the device clamp screws into
the fixture and on the thermal forcing plate.
Saturated performance with constant VG and IDQ is
provided in Fig. 19 while linear performance under
the same conditions is shown in Fig. 20. For
saturated CW operation the 3 dB compression
point (P3dB) over temperature was consistent
between the two bias methods. Not shown is the
51% efficiency of the device that varied less than
0.5% for the two bias conditions. Fixed gate
voltages are sufficient for saturated operation;
there is little advantage to maintaining constant
quiescent current over temperature.
temperatures. Efficiency is expected to relatively
constant for the constant IDQ case given that the
current is not changing applicably over
temperature.
Figure 21 shows that to be the case. For the
constant VG bias, efficiency drops at hot
temperatures as the current increases for a fixed
40 dBm output power, while there is improvement
at cold temperatures due to the lower drain
current. Also plotted is the efficiency using the op
amp circuit, showing that it does not appreciably
change the efficiency compared to laboratory
supplies. 52.0
51.8
Linear operation places a premium on efficiency
and distortion performance. Figure 20 shows the
drain current over temperature for the two bias
schemes at (a) 40 dBm of WCDMA output power,
and (b) quiescent bias.
Differences in drain
currents are observable between the two bias
methods.
Constant IDQ = 0.574 A Avg.
VG = −2.7386 to −2.6322 V
P3dB (dB)
51.6
51.4
51.2
51.0
Constant VG = −2.6981 V
IDQ = 0.315 to 0.765 A)
50.8
50.6
50.4
Given that drain current at a fixed output power is
inversely proportional to efficiency, there is an
obvious expectation from this data. Constant VG
bias will result in lower efficiency at hot
temperatures and improved efficiency at cold
Eng. Rev. 3 07/24/14
© 2014 TriQuint
-30
-15
0
15
30
45
60
Thermocouple Temperature (°C)
75
90
Fig. 19. P3dB vs. Temperature for constant VG &
constant IDQ. The 0.1 dB offset between the two plots
attributable to bench-to-bench differences. VD =+32 V,
F=2150 MHz CW drive.
- 16 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
1.58
1.0
VD=+32 V
POUT=+40 dBm
0.9
Quiescent Drain Current (A)
1.56
Drain Current (A)
1.54
1.52
Power Supply
Constant VG
1.50
1.48
1.46
Power Supply
Constant IDQ
1.44
1.42
Power Supply vs. OpAmp Bias Circuit
VD=32 V
0.8
Op Amp Bias (Const. IDQ)
0.7
0.6
0.5
0.4
Power Supply (Const. VG)
0.3
0.2
0.1
1.40
0.0
-40
-20
0
20
40
60
80
100
-40
-20
0
Temperature (°C)
20
40
60
80
100
Temperature (°C)
(a)
(b)
Fig. 20. T1G4012036-FL drain current over temperature for (a) 40 dBm WCDMA output power and (b)
quiescent current state where the op amp circuit was aligned to hold the IDQ constant over temp.
VD=32V, F=2150 MHz.
Efficiency vs. Temperature
23
VD=+32 V
POUT=+40 dBm
Efficiency, EFF (%)
Power Supply vs. Op Amp Bias Circuit
22
Power Supply
Constant IDQ
21
OpAmp Bias
20
Power Supply
Constant VG
19
-40
-20
0
20
40
Temperature (°C)
60
80
100
Fig. 21. Drain efficiency over temperature for constant VG set by a power supply, and constant
IDQ set by both power supply and the op amp bias circuit. F=2150 MHz, Pout=40 dBm.
Additional efficiency data that compares laboratory
power supply bias to the op amp circuit is given in
Fig. 22. Efficiency over the drive-up level test is
comparable for the two bias methods. Linearity
data is presented in Fig. 23 for 40 dBm of fully
loaded WCDMA output power. This data indicates
the advantage of maintaining a constant quiescent
current over temperature when linearity is a critical
specification. There are no significant differences
in linearity between a power supply and the op
amp circuit supplying the gate voltage to the
Eng. Rev. 3 07/24/14
© 2014 TriQuint
device. Constant VG results in degraded distortion
performance while constant IDQ maintains linearity
over the temperature range. Gain drift over
temperature is expected from RF devices
regardless of the bias technique. Fig. 24 confirms
that constant VG or constant IDQ bias does not
influence gain drift to any great extent. The op
amp circuit bias provides the same result as the
lab power supply. Regardless of the bias method
gain drift occurs due to other factors such as
transconductance variation over temperature.
- 17 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
9.0 Conclusions
45
40
VD=+32 V
Op Amp Bias at +85°C
Op Amp Bias at +25°C
Power Supply (Const. IDQ) at +85°C
Power Supply (Const. IDQ) at +25°C
Power Supply (Const. IDQ) at −30°C
Efficiency, EFF (%)
35
30
25
20
15
10
5
0
28
30
32
34
36
38
40
42
44
46
48
Output Power, POUT (dBm)
Fig. 22. Efficiency vs. output power for gate bias supplied
by the op amp circuit and lab power supply. F=2150 MHz,
VD=+32 V.
Adjacent Channel Power (dB)
-40
POUT=+40 dBm, VD=+32 V
-41
PS (Const. VG) - Adj. H
PS (Const. VG) - Adj. L
-42
-43
-44
-45
-46
PS (Const. IDQ) - Adj. H
PS (Const. IDQ) - Adj. L
Op Amp Bias - Adj. H
Op Amp Bias - Adj. L
-47
-48
-40 -30 -20 -10 0
10 20 30 40 50 60 70 80 90 100
Temperature (°C)
Fig. 23. WCDMA Adjacent Channel Leakage Ratio (ACLR)
vs. temperature using power supply and op amp bias.
The requirements and fundamental operating
principles behind open-loop bias methods for
depletion mode devices have been presented.
The outlined approaches are suitable for any
depletion mode devices including GaN and
pHEMT RF power devices. Summation junctions
are employed to provide linear temperature
compensation of drain quiescent current. The
temperature slope unique to depletion mode
devices
is
easily
compensated
by
the
determination of a single resistor value, and the
bias level may be set using a DAC. Three different
low cost reference designs were presented, each
with its own advantages for a given application.
Experimental verification has been provided using
a 120 W GaN Device under both CW and WCDMA
excitation. Validation of the circuit designs was
accomplished by comparing op amp gate driver
data laboratory power supplies. Conclusions from
the data collection are that saturated operation
does not require temperature compensation, while
distortion and efficiency in linear operation benefit
from maintaining a constant IDQ over temperature.
The circuits & principles described here form the
basis of a complete suite of biasing options for
depletion mode RF devices.
18
POUT=+40 dBm, VD=+32 V
Gain (dB)
17
PS Only (Const. IDQ)
PS Only (Const. VG)
16
15
Op Amp Bias
14
-40 -30 -20 -10 0
10 20 30 40 50 60 70 80 90 100
Temperature (°C)
Fig. 24. Gain vs. temperature for constant VG and
constant IDQ bias conditions. F=2150 MHz, VD=+32 V. Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 18 of 19 -
www.triquint.com
Bias Techniques for GaN and
pHEMT Depletion Mode Devices
10.0 Acknowledgements
The author would like to thank Matt Ramberg, Peter Xia, Jeff Gengler & Jeff Taylor for their help building
and verifying the GaN test fixture. Thanks to Brent Crittenden for GaN VG data. Chris Blum edited and
formatted the application note. To these and others who contributed to this project the author extends his
thanks & appreciation.
References
[1] Texas Instruments “Precision Data Converters Guide” 2013, Slide 4.
http://www.ti.com/lit/sg/slyt475a/slyt475a.pdf
[2] ZNGB4000 FET Bias Controller Data Sheet, Diodes, Inc. June 1998
http://diodes.com/catalog/fixed_bias_generators_98/znbg4000.html
[3] BCR400W Active Bias Controller, Infineon, 05/29/2007
http://www.infineon.com/cms/en/product/rf-and-wireless-control/rf-transistor/active-bias-controllers-for-rftransistor/BCR410W/productType.html?productType=db3a3044243b532e0124d90ad281681b
[4] Stability Analysis of Voltage-Feedback Op Amps, Application Report, TI March 2001.
www.ti.com/lit/an/sloa020a/sloa020a.pdf
[5] Fortunato, Mark, “Simulation Shows How Real Op Amps Can Drive Capacitive Loads,” EDN Network &
Maxim Integrated Application Note 5597, April 28, 2013.
http://www.edn.com/design/analog/4412899/Simulation-shows-how-real-op-amps-can-drive-capacitiveloads
[6] King, Grayson, “Op Amps Driving Capacitive Loads,” Analog Dialog, Vol. 31, No. 2, 1997.
http://www.analog.com/library/analogdialogue/archives/31-2/appleng.html
[7] Buxton, Joe, “Careful Design Tames High Speed Op Amps,” Application Note AN-257, Analog Devices &
also published in Electronic Design, April 11, 1991
http://www.analog.com/static/imported-files/application_notes/AN-257.pdf
[8] “GaN HEMT Biasing Circuit with Temperature Compensation” CREE Application Note AN-011.
http://cree.com/~/media/Files/Cree/RF/Application%20Notes/Appnote%2011.pdf
Eng. Rev. 3 07/24/14
© 2014 TriQuint
- 19 of 19 -
www.triquint.com
Download