Mobile Communications & Infrastructure Supplement
mmWave Propagation Measurements
Understanding Envelope Tracking
Six LTE Receiver Measurements
A Simple VVA RFIC Design with a Focus on Repeatability and Stability
Description of an FETbased voltage variable attenuator, its characteristics, design tradeoffs and common topologies
Technical Feature
A Simple VVA RFIC Design with a Focus on Repeatability and Stability
A novel voltage variable attenuator (VVA) integrated circuit reduces the sensitivity of the transfer curves to temperature and process variations. The new solution is achieved without compromising the simplicity of a single positive voltage control design and a large attenuation range. This article describes FETbased VVA characteristics, design tradeoffs and common topologies. Following a sensitivity analysis discussion, measured results are presented to show the lower sensitivity of the suggested circuit over a traditional design. Measurements of a large quantity of VVAs, based on the new topology, confirm the robustness of the solution.
Yair Shemesh
Alpha Industries
Woburn, MA
Voltage variable attenuators are important control elements, widely used for automatic gain control both in receive and transmit chains.^{1,2} When used in handsets, their overall cost and size are of paramount importance. A single positive voltage control scheme and a simple circuit design help in achieving this goal. When operated as part of an open loop gain control, a linear transfer curve simplifies the design of the chain's gain. The unittounit transfer curve repeatability and its stability over temperature are key factors in the ability to design and predict the overall system gain. In receive chains, the VVA is used to control the signal that enters the mixer. It must have a high enough thirdorder intercept point (IP3) by itself in order not to degrade the overall power linearity. FET VVA topologies such as TEE, PI and bridged TEE are often used,^{2,3} each with its own merits and problems.
This article reviews the tradeoffs among VVA features, and focuses on the issue of stability and repeatability. The reasons for high parameter sensitivities are identified, and based on these identifications, a new VVA topology is suggested. Compared to traditional topologies, the new configuration significantly reduces the sensitivity to process and temperature variations, without sacrificing simplicity, single positive voltage control approach, attenuation range and transfer curve linearity. Several VVAs have been designed and measured based on the new topology. The results clearly show the predicted stability and repeatability improvements.
Table 1  
Emphasis on...  Causes 
High attenuation range  Higher sensitivity to process and temperature variations 
Transfer curve linearity  Increased circuitry complexity 
Good repeatability and temperature stability  Higher circuit complexity 
Lower current consumption  Longer settling time 
Higher IP3 along the full transfer curve  Higher insertion loss 
VVA Design Parameters
The VVA performance is a tradeoff among contradicting requirements, as shown in Table 1 . Several FETbased VVA topologies are commonly used. An emphasis on a specific requirement affects the topology selection. For example, the bridgedTEE has superior return losses and a very linear transfer curve, at the cost of complex circuitry and sensitivity to process variations and temperature changes.

Fig. 1 A single positive voltagecontrolled TEE VVA. 
The large volume handset production in recent years has emphasized the need for repeatability when selecting a VVA topology. In FETbased VVAs, the device is used as a variable resistor at zero drain source voltage (V_{ds} ). The device parameters vary over temperature, across the wafer and from wafertowafer. Since channel resistance changes sharply for gatesource voltage (V_{gs} ) close to the pinchoff voltage (V_{p} ), the stability and repeatability of the transfer curve is governed mostly by V_{p} . Other FET parameters, such as the ON resistance (R_{on} ), have a secondorder influence on the VVA transfer curve. These parameters are usually correlated to V_{p} changes in a predicted manner.
In the commonly used topology of a single positive voltagecontrolled TEE VVA, shown in Figure 1 , the control voltage V_{c} is translated into two gatesource voltages.
For the series devices it is given by
V_{gsser} = Div · V_{c}  V_{center} (1)
and for the shunt device, by

Fig. 2 Attenuation transfer curve defining top and bottom knees, attenuation range and control range. 
V_{gssh} = Div · V_{c} (2)
where
Div = resistor divider ratio R1/(R1+R2)
V_{center} = fixed DC voltage applied to the series path at the TEE junction
Proper selection of Div and V_{center} for a given V_{p} allows a roughly linear attenuation dependence on V_{c} .
At V_{c} = V_{top knee} , the shunt device start conducting (V_{gssh} = V_{p} ), while at Vc = V_{bottom knee} , the series devices start conducting (V_{gsser} = V_{p} ). A V_{p} variation will translate, according to Equations 1 and 2, into a shift of the top and bottom knees. A V_{p} change by an amount of DV_{p} , shifts the nominal point N by a voltage DV_{c} , to the point R, where
(3)
The result, as seen in the ideal curve shown in Figure 2 , is that at V_{c} = V_{top} knee, the attenuation will drop by an amount D[dB] where
D = S · DV_{c} (4)

Fig. 3 An (a) ideal TEE attenuator and (b) implementation with FETdevices. 
S is the transfer curve slope given by DA/DV_{eff} . Therefore, minimizing the attenuation variation D is possible by designing a VVA with small attenuation range DA and a wide control voltage range DV_{eff} . A 12 dB range VVA for a 5 V control voltage was demonstrated by Boglione and Pavio.^{4}
Unfortunately, an attenuation range of more than 30 dB is a more common requirement. Cascading several low attenuation range VVAs is possible for an overall wide attenuation range; however, the overall transfer curve sensitivity will increase as well.
Applying two separate positive controls for the series and shunt devices, and setting V_{center} with a pinchoff follower, can reduce the sensitivity to V_{p} variations. This can be done at the cost of additional circuitry, based on an operational amplifier, which generates the proper two control voltages.^{4} In many applications, the need for a simple, low cost, low size VVA solution rules out this option.
A TEE VVA Sensitivity Analysis
In an ideal TEE attenuator, as shown in Figure 3 , the resistors R_{sh} and R_{ser} should satisfy the matching and attenuation conditions given in Equations 5 and 6, respectively.

Fig. 4 TEE attenuator resistors values for an attenuation range of 40dB, with r_{ser} =110W. 
where
A = attenuation in dB
Since for V_{gs} < V_{p} , the FET resistance is much higher than 50 W, the series devices should be shunted by a resistor (r_{ser} ) that limits the overall series arm resistance to 50 W. For a maximum attenuation of 30 to 40 dB, r_{ser} is of the order of 60 to 110 W, depending on the device's periphery and characteristics.
Figure 4 shows the required R_{ser} and R_{sh} for an attenuation range of 0 to 40 dB, together with the series FET resistance, r_{net} , for the case of r_{ser} = 110 W in shunt with the device.

Fig. 5 Deviation from nominal attenuation due to a 10% R_{sh} change (a) and 10% R_{ser} change with r_{ser} =100W (b) and r_{ser} =60W (c). 
Figure 5 shows the deviation from the nominal attenuation of the transfer curve, caused by a change of 10 percent in the shunt device (curve a) and 10 percent in the series device, with the resistor in shunt r_{ser} = 100 W (curve b) and r_{ser} = 60 W (curve c).
The transfer curve shape is more sensitive to the shunt device resistance variation than to the series device variation, as the shunting resistor r_{ser} decreases.
Though the deviation in dB from the nominal transfer curve is smaller for higher V_{c} (lower attenuation levels), the relative deviation dA/A is larger at low attenuation levels, with a maximum at A = 0.
The New Circuit
One possible approach^{4} for minimizing the variations is to apply a V_{p} follower voltage (V_{p} ) to V_{center} .
In this case, the R_{ser} deviation from the nominal R_{ser} due to the V_{p} changes is ideally eliminated since
R_{ser} = f(V_{gsser}  V_{p} ) = f(Div · V_{c} ) (7)

Fig. 6 The new VVA topology. 
Hence, stabilization is achieved for the bottom knee position. The disadvantage of this approach is that the top knee, affected mostly by R_{sh} , is very sensitive to V_{p} variations, especially for wide attenuation range, steep slope VVAs. In the newly proposed circuit, shown in Figure 6 , a V_{p} follower applies a V_{p} voltage to the drain and source of the shunt device. The sensitivity of the shunt arm is eliminated since
R_{sh} = f(V_{gssh}  V_{p} ) = f{[Div · V_{c}  (V_{p} )]  V_{p} } = f(Div · V_{c} ) (8)
Hence, stabilization is achieved for the top knee position. As previously explained, the series device resistivity mainly affects the bottom knee, which is less critical to the transfer curve shape. Selection of shunting r_{ser} resistors low and close to 50 W reduces attenuator sensitivity at the high attenuation levels. Such selection requires high OFF impedance, that is, small gate periphery. On the other hand, a small gate periphery for the series devices increases the loss at the minimum attenuation state. This shunting r_{ser} resistor is finally set as a compromise between these two requirements.

Fig. 7 Measured transfer curves of design A (old topology) and design B (new topology). 
The conflict of these requirements is easy to resolve for low attenuation range VVAs and for devices with low normalized draintosource capacitance (C_{ds} ).
Results
Several VVAs have been designed, fabricated and measured based on both old and new topologies. Measured transfer curves of the two types of 900 MHz VVAs built to similar specifications are shown in Figure 7 .
While the old topology (design A) has a maximum attenuation at low control voltage, the new one (design B) has maximum attenuation at high control voltage (inverted slope). Other measured parameters are similar for both designs.
V_{p} variations are caused by process variations and temperature changes.V_{p} is also the major cause for transfer curve changes. Therefore, transfer curve stability over temperature is an indicator for process variations.

Fig. 8 Change in VVA attenuation when the temperature is increased from 20° to 85° C. 
Figure 8 shows the measured change in the transfer curve when the temperature is raised from +20 to +85°C for the two designs.
It is clearly seen that the new design B deviation is smaller by a factor of 4, compared to design A. Furthermore, since design A maximum deviation is at the low attenuation point of Vc = 3.5 V, the relative deviation dA/A is even worse compared to the dA/A of design B.
VVAs based on the new topology with 3 V control and steeper slope have also been designed and built in large quantities. Measurements confirm the robustness of the solution.
Conclusion
A new VVA topology has been suggested, offering an improvement in stability and repeatability over temperature and process variations.
Acknowledgment
The author wishes to acknowledge the support of D. Johnson and the lab assistance of J. Bedrosian.
References
1. R. Goyal, Monolithic Microwave Integrated Circuits Technology and Design , Artech House Inc., 1989.
2. I. Bahl and P. Bhartia, Microwave Solid State Circuits Design , John Wiley & Sons, New York, 1988.
3. B. Maoz, "A Novel, Linear Voltage Variable MMIC Attenuator," IEEE Transactions on Microwave Theory and Techniques , Vol. 38, No. 11, November 1990, pp. 16751683.
4. L. Boglione and R. Pavio, "Temperature and Process Insensitive Circuit Design of Voltage Variable Attenuator IC for Cellular Band Applications," IEEE Microwave and Guided Wave Letters , Vol. 10, No. 7, July 2000, pp. 279281.
Yair Shemesh received his BScEE and MScEE degrees from the Technion, Haifa, Israel, and has worked for Rafael in Israel and Alpha Industries in the US. His specialties are MW devices, modeling, PAs, RF to MMW MMICs, modules and RF frontends. He is currently a senior principal engineer at Alpha and can be reached via email at yshemesh@alphaind.com.