TEMPERATURE COMPENSATION METHOD
Technical field
This invention relates to a method for temperature compensating a radio frequency front-end CMOS circuit.
Technical background
With the scaling of CMOS technology, it is now possible to achieve good performance at RF frequency with reasonable circuit structure and biasing strategy. Lately, several RF CMOS implementations in 1-2 GHz range have been reported. The performance of the front-end circuit is critical to the successful implementation of a receiver. A low noise amplifier (LNA) is the first active stage in a receiver system, the total Noise Figure is defined as:
™"»™ . E P gg 1
NF ttonttaall = NF LINNAA + ■*- + — -**■ ( 1 )
^LNA ^LNA^3 "mixer
where GLNA is the gain of the LNA, GmiXer is the gain of the down converting mixer and NFBB is the noise figure at baseband frequency.
It can be seen that any variations in the gain or noise performance of the RF front-end will affect the whole system noise performance greatly. This is espe- cially undesirable for those systems which have a tight noise budget. One important source of the performance variations is the temperature dependency of transistor parameters. The performance of LNA can vary in a large range without temperature compensation. Until now, the temperature-compensation for RF front-end circuits has been addressed by few researchers. For a bipolar case the transconductance of the transistor is gm = Ic/Vτ, and NT
is Proportional To the Absolute Temperature (PTAT) . The method to compensate the performance variation is to bias the transistor with a PTAT current source. This is the method described in 'A 1-GHz BiCMOS RF Front-End IC" , by R. Meyer and W. Mack, published in IEEE J. Of Solid State circuit, vol. 29, No. 3, pp. 350-355, March 1994. As claimed by the authors of that publication, the gain variation of the LNA is about -0.008dB/°C. No noise figure variation result is available. For CMOS RF front-end circuits, the temperature issue is seldom addressed.
There appears to be a single publication dealing with the variation of front-end circuits, ""The Design and Implementation of Low-Power CMOS Radio Receivers", by D. Schaeffer and T. Lee, as published in Kluwer Academic Publishers, 1999. In that work, the method is to track the variation of power supply and Vτ by some approximation. However, the authors of that work only claim that their circuit will decrease variation, no data is available. In order to demonstrate the strategy of the temperature compensation, a conventional cascode CMOS Low Noise Amplifier is designed with AMSO .35$\mu$m, as shown in Fig. 1. Under a 3V power supply at 27°C, the simulation result shows that it can achieve 12.8dB power gain with 2.5mA biasing. The input port and output port return losses, Sll and S22 respectively, are -25dB and -14dB respectively with NF of about 1.7dB (Fig.2) . The linearity of the amplifier as measured by IIP3 is checked to be - 7dBm. In CMOS devices, the temperature dependent effects are mainly due to two factors: threshold voltage and mobility variations. The threshold voltage has a similar temperature tendency as bipolar devices, which have a TC about -2mV/°C. The mobility will decrease as temperature increases because of its exponential nature, and it will be the dominating variation factor. The fractional Temperature Coefficient (TPf) of mobility is simply given by
TCf = — - —μ(T) = -l,5T~l (2)
For the common source stage the transconductance can be expressed as
W_
Xm(T) = μ(T)Coχ -r (Vg5 - Vth(T)) (3) L where μ is mobility, Cox is the gate oxide capacitance, W and L are width and length of channel, Ngs is the gate- source voltage, and Nth is the threshold voltage. If we assume constant gate-source voltage biasing, then the TCf of the gm is:
τc 1 δgm(T) l d(T) 1 dVth (T) _ x 2mVI' C f gm dT ΘT Veff ΘT ' veff
where Veff =Vgs-Nth is the overdrive voltage of the common source stage. For large overdrive, the first item will dominate and thus gm will decrease with temperature. In LΝA design, if we assume that the input network quality factor (Qin) is fixed and temperature independent, then from Gm=gmQin (5) where Gm is the transconductance of the input stage, comprising Ml, Ls, Lg and Cin in Fig. 1, the gain of the LΝA will drop when the temperature increases, and from
ΝF = 1+ Y ■ (6)
R SgmQin where Rs is the source impedance and γ is a constant, the noise performance will be degraded accordingly. In the above derivation, we assume that the gate-source biasing voltage Ngs is constant, but for a conventional biasing circuit like that in Fig.l, Vgs does vary with tempera- ture. In other words, the current in the biasing branch varies with temperature. It is easy to show that the TCf
of biasing current will also have a negative value. So this will further degrade the performance of the low noise amplifier. Note that temperature variation will also affect the input and output matching, but simulation results show that this variation is not very significant and the return losses are always below lOdB. Fig.3 shows that the biasing current will decrease as temperature increases. The power gain drops more than 3dB in the temperature range as shown in Fig.4. While in Fig.5, we can see that the NF increases about 0.7dB. These simulation results agree with the derivation above .
From the above analysis, it is clear that in order to keep a constant performance of the CMOS Low Noise Amplifier over wide temperature range, a biasing current which has a positive temperature coefficient is required. At a given power consumption, in order to find the biasing current that can compensate the performance variation, theoretical analysis can be done to find the cur- rent vs temperature curve that can totally compensate performance degradation. But since in integrated circuits the passive components also have temperature variation, if we take them all into account, the problem in question will turn out to be prohibitive.
Summary of the invention
One object of this invention is to provide an improved method for temperature compensating a radio frequency front-end CMOS circuit. The object is achieved by a method according to the appended claims .
In a most general view the invention could be seen as a method for temperature compensating a radio frequency front-end CMOS circuit by finding the temperature characteristic of a specific front-end circuit, then finding the desired current source which compensates for
temperature variations, and then realizing said current source.
More particularly the method according to the present invention comprises the step of generating a tempera- ture compensated bias current, which step in turn comprises the steps of:
- determining the temperature characteristic of said temperature compensated bias current by simulating it with an ideal simulation current source, while tuning the absolute current value as well as a temperature coefficient thereof, said ideal simulation current source thereby generating an ideal simulation current corresponding to said temperature compensated bias current; and - synthesizing said temperature compensated bias current by generating a first current, which has a fractional temperature coefficient similar to the fractional temperature coefficient of the ideal simulation current, and tuning said first current by means of a current source having substantially zero temperature coefficient.
As mentioned from the beginning, it is important to keep constant performance of the RF front-end circuit. The method according to this invention is specifically designed for an LNA as the front-end circuit, but can also be applied to any other front-end circuit where similar conditions apply. By employing the method according to this invention substantially constant gain and noise performance can be achieved over a large temperature range. Another advantage is that this method can get best performance at 27°C, which is the normal case.
Through the analysis of the required temperature characteristics in a simulation environment by means of an ideal simulation current source, while tuning the absolute current value as well as a temperature coeffi- cient of the current source, a template for a realization is obtained. Then a temperature compensted current is
generated by means of two current sources having complementary characteristics.
Preferably, the first and second current sources are a PTAT and a Bandgap current source respectively, which are used for realizing the specific ideal current source resulting from the simulation step and generating the temperature compensated current. As is well known within this technical field a bandgap current source includes both negative and positive TC sources in order to get zero TC.
Further objects and advantages of the present invention will be discussed below by means of exemplary embodiments .
Brief description of the drawing
Exemplifying embodiments of the invention will be described below with reference to the accompanying drawings, in which:
Fig. 1 is a schematic circuit diagram of a prior art low noise amplifier;
Fig. 2 is a diagram showing the performance of the LNA of Fig. 1;
Fig. 3 is a diagram showing the variation of a non- compensated biasing current; Fig. 4 is a diagram showing power gain variation for a non-compensated case;
Fig. 5 is a diagram showing noise figure variation for a non-compensated case;
Fig. 6 is a schematic circuit diagram of a tempera- ture compensated biasing circuit for performing an embodiment of the method according to the present invention;
Fig. 7 is a diagram showing a temperature compensated bias current generated by the method according to this invention;
Fig. 8 is a diagram showing a temperature compensated bias voltage generated by the method according to this invention.
Fig. 9 illustrates the power gain of a front-end circuit with and without compensation;
Fig. 10 illustrates the noise figure of a front-end circuit with and without compensation; and
Fig. 11 illustrates the connection between the circuit shown Fig. 6 and the basic LNA portion of the cir- cuit shown in Fig. 1.
Description of embodiments
In Fig. 6 a temperature compensated biasing circuit for performing an embodiment of the method according to the present invention is shown.
Initially a simulation is performed in order to determine the temperature characteristics required. Hence, from the derivation described above, it can be assumed that the biasing current will be generated by a PTAT current source comprising transistors Mbl-Mb5 and Ql, Q2 and resistor Rbl as shown in Fig. 6. Through tuning, we find that the biasing current curve should have a TCf of about 5800ppm/°C and 216μA at 27°C. Also in order to make the biasing voltage at the gate of M2 constant, there is a need for generating a temperature independent biasing voltage. In a simple realization a PTAT current source could be used to realize the desired current source directly, but for the reasons explained below, and in accordance with the invention, there is provided a PTAT current source and a Bandgap current source to synthesize the desired current source. The Bandgap current source is also employed to generate the constant biasing voltage at the gate of M2.
The pnp transistor of the circuit in Fig. 6 is real- ized by the p+ region in n-well and p-substrate in the n- well CMOS process. Transistor Mbl-5, Ql-2 and Rbl implement a PTAT current source. The area of Q2 is A times
that of Ql (A=10 in this design) . Thus the current in each branch will be
_ AVbe _ VTIn(A) l O t D D ' } Kb\ Kb
Since both Vτ and (dVτ/δT) are constant at certain temperature, once we decide the TCf of the PTAT current, the absolute value is also determined, or vice versa. It is found that we can not meet both requirements at the same time by using this circuit, so we make this PTAT current source to have the same TCf as the desired one, and use a Bandgap current source to tune the absolute value. Since we already have the PTAT source, all we need for the Bandgap source is a current source with negative TC. This is realized by Mb9-13, Q3 and Rb2, using the negative TCf of Nbe - The bandgap current source is also used to generate a 2N biasing voltage by Mbl4, 15 and Rb3.
Fig.7 shows the biasing current generated by the biasing circuit, and Fig.8 shows the biasing voltage of M2. This voltage has a TCf less than 30ppm/°C. Fig.9 and 10 show the variation of gain and ΝF respectively. The variation of gain is less than 0.4 dB (0.0031dB/°C) , and it achieves maximum value of 12.76 dB at 27°C. The ΝF variation now is less than 0.1 dB, at 27°C it has a minimum value of 1.66 dB. In Fig. 11 the compensation circuit described in conjunction with Fig. 6 is schematically indicated by a dashed line. The output terminals of the compenstion circuit are Vbl and Vb2 correspond to the input terminals Vbl and Vb2 of the very LΝA of Fig. 11. Above a preferred embodiment of the method according to the present invention has been described under reference to a circuit implementation. This should be seen as merely a non-limiting example. Many modifications will be possible within the scope of the invention as defined by the claims.