Low-Voltage, Low-Noise Direct-Upconverter Upconverters can be used in the transmitter section of a wireless phone to modulate a high frequency carrier with the baseband signal containing the information to be transmitted. To minimize cost and boost performance direct upconversion could be used, avoiding an intermediate frequency extra block in the signal path between baseband and the output power amplifier. The main challenges for a transmit upconverter used in a WCDMA or GSM phone are high linearity, low noise, high output power, low power consumption and accurate control over a wide range of output power. The high linearity is imposed by stringent requirements regarding leakage into adjacent channels. Out-of-band emissions in cellular bands require very low noise to be transmitted simultaneously with high output power. As an example, an UMTS time-division duplex mode phone should not put more than -121dBm/Hz power spectral density of noise into the DCS receive band, which is only 20MHz away from the transmit band. Low power consumption and technology scaling limit the bias voltage for these upconverters to 1.5V. For a WCDMA transmitter, to avoid interference from one user to another, more than 74dB of gain control is needed, with less than 1dB steps. To implement gain control, prior-art direct-upconverters use for example a mixer type differential structure biased with a tail current generator controlled by the baseband signal, while the differential structure is driven by the high frequency signal derived from a voltage controlled oscillator, similar to fig. 1 bellow:
Fig. 1
The switches in the gates of the current generators perform the gain control. These current generators might be scaled such that the output power is varying linearly in dB. One drawback of this configuration is the fact that the mosfets in these current generators need to be biased in the saturation region and this is stealing from the headroom of the switching core of the modulator. Another problem is the added noise of mosfet devices needed to implement the current generators. Also, matching problems associated with these devices create concerns about the minimum controllable output power, carrier rejection in the output spectrum and control linearity. These current generators require relatively large area on the chip layout. Mosfets in the current generators should be relatively large in order to prevent their 1/f noise to be upconverted in the output spectrum. This is a problem since their drain parasitic capacitance is loading the sources of the switching pair, directly impairing the linearity and noise performance of the upconverter. Moreover, these parasitic capacitances are nonlinear. This is another reason why it is desirable to remove them, if possible. According to the present invention, the mosfets in the current generators in fig. 1 above can be removed. Now the resistors only in the sources of the switching core of the upconverter determine the currents through the core, and consequently the output power. In order to precisely control these currents, a negative feedback control loop is built with the help of an opamp or OTA, as in the fig. 2 bellow:
Fig. 2 Baseband signals, through the opamp, reach the gates of the switching core, and consequently the sources of the mosfets in the ON state, determining the voltage drops on the source resistances (and currents through the switching mosfets). The capacitors in the gates
of the mosfets couple the high frequency signal from the VCO (LO signal). The output power is varied in steps by switching the source resistors. Matching between these resistors could be very good in integrated circuit technology, depending mostly on their area. Now the headroom on the mosfets is limited only by the drop on the source resistors. The lack of extra parasitic capacitances in the switching core considerably improves the linearity and noise performance of this upconverter. Another disadvantage of the prior-art schematic in fig. 1 is the fact that normally the baseband signal is sent to the current generators’ gates through a current-mode type digitalto-analog (DAC) converter and a replica bias mosfet. This noisy solution can easily be replaced using the upconverter disclosed above, with a R-2R-based DAC having less noise and much more linearity due to lack of extra active devices in its schematic. Not to mention the power saving of this type of DAC and area saving by not using mosfets in the DAC. Also, the R-2R ladder, having constant output resistance independently of the code value, can be the resistive part of the anti-aliasing RC-filter used between the DAC and the upconverter. The resistors in the gates of the core prevent the LO signal to reach the baseband opamp. -
-
Claim The upconversion method described in fig. 2 (using mosfets or bipolar transistors) where the switching core is closed in a negative feedback loop being controlled by the baseband signal, without using current generators for biasing the core, having accordingly only two devices consuming headroom between the power supply and ground (besides the load): the switching core itself and the switched-resistors for establishing the bias currents through the core (and consequently the output power). Similar configurations to this disclosure, where between the Vcc and ground there are only two devices cascaded on the signal path (besides the output load). One example could use PMOS transistors instead of NMOS ones in the switching core, driving the output transformer which is connected to the ground line, while the switched-resistors are connected between the Vcc line and the sources of the PMOS transistors.
Note I didn't apply for a patent in the one year time-frame from the provisional application for a patent, so this now is public domain. Raducu Lazarescu,
[email protected]