passive structure like monolithic spiral inductor (MSI) does not need more than one or two hidden layers.The neural network has successfully modeled the MSI, exhibiting good and acceptable average error performances. This proposed method can be conveniently applied for passive monolithic circuits.

No abstract

A chip set for compact transceivers meeting IS54 dual-mode cellular telephone standards consists ofa direct up-conversion modulator and a double conversion receiver in a 12GHz superself-aligned bipolar process. Operated from a 5V power supply, the total power dissipation is 310mW. Both circuits can be put into a power-down mode with a maximum off-current of 1OpA. The modulator chip is 4.1mm2, and the receiver, 4. 9"' . The modulator block diagram is shown in Figure 1. The advantage of direct up-conversion over indirect up-conversion is that there is only one local oscillator (LO) and no intermediate frequency (IF) filtering is required, but care must be taken to avoid producing in-band spurious signals. In this implementation the inputs are a single-ended local oscillator (LO) and balanced I (in-phase) and Q (quadrature) baseband signals. Nominal maximum output power is 0dBm.The local LO power requirement is -1OdBm typical. The baseband inputs are high impedance and require 2.5Vpp differential for maximum RF output.The LO input is buffered and sent to an R-C phase splitter to produce two signals approximately 90" out of phase. The two signals are passed through the Havens amplitude and phase corrector circuit to insure that the inputs to the mixer are equal in amplitude and 90" out of phase, with small deviation [ 11. The LO signals are then mixed with the I and Q baseband signals to produce the single-sideband suppressed-carrier RF output signal. The output stage consists of a low-distortion voltage amplifier followed by a power gain stage with 50R output impedance. The corrector circuit is shown in Figure 1. Operation is based on the fact that if two equal-amplitude signals are not exactly SO" out of phase, their sum and difference will be. The pair of gain stages following the phase splitter, overdriven emitter-coupled pairs, insure that the signals into the first pair of summers are of equal amplitude. Since the correction of phase introduces amplitude imbalance, the second pair of gain stages recorrects the amplitude. The summers are linearized emitter-coupled pairs with shared collector loads. The measured phase error is 0.120.084" and the mixer amplitude imbalance is 0.10+0.05dB.Part of the modulator design is concerned with avoiding the creation of spurious signals. Figure 2 shows the output spectrum generated with a single-tone baseband input (at 80kHz) as a function of baseband drive level. The desired signal is at the local oscillator frequency plus 80kHz (USB). The largest undesired signals are at the LO frequency, 2 basebands above LO (LO+BBB), and at 1 and 3 basebands below LO (LSB, LO-3BB). The behavior of LO feedthrough shows relatively flat response reflective of passive coupling. The component at LO-1BB is due to mixer imbalance and the LO phase error. The component at LO+2BB is due to feedback of the modulator output to the LO input and increases at a rate of 2dB per dB increase of output level. The output-signal component at the LO input is remixed with the baseband signal. This component decr...

No abstract

A circuit including a 100 MHr IF amplifier with a digitally controlled gain of 0-to-45 dB, a quadrature phase-shifter, balanced mixers, and activelsleep mode capability is described. The measured gain and quadrature phase variation was less than 1 dB and lorespectively. The 92 mW circuit was fabricated using a 4.0 GHz complementary bipolar process.