Ole Kiel Jensen

Ground-shielded measuring technique for accurate on-wafer characterization of RF CMOS devices

This paper presents a new test fixture with associated de-embedding procedure for efficient and accurate on-wafer device measurements at microwave frequencies. The fixture is based on a substrate shield and (i) provides an accurate common ground for N-port measurements, (ii) effectively reduces substrate carried coupling, (iii) gives well-defined parasitics for simplified de-embedding, and (iv) fits arbitrarily large devices. Due to these characteristics, the accompanying de-embedding technique requires only few in-fixture standards that can be fabricated with very high accuracy; even in standard CMOS processes. The technique can advantageously be applied to a wide range of commonly used processes, but highest performance improvement is achieved with low-resistivity substrates. The performance of the technique is demonstrated to 12 GHz in a 0.25 /spl mu/m CMOS technology and conclusions are drawn.

Assessing and Modeling the Effect of RF Impairments on UTRA LTE Uplink Performance

In this paper we investigate the effect of modelled RF transmitter impairments on UTRA Long Term Evolution (LTE) uplink link performance. Nonlinear power amplifier, gain imbalance, phase imbalance, and phase noise are considered. The main contributions of this work is assessment of the relation between Error Vector Magnitude (EVM) and link performance loss due to each impairment and the investigation of a white Gaussian noise source in the transmitter as an equivalent model of the impairment. The results obtained are based on the UTRA LTE uplink parameter settings. Results show that the EVM is a good measure to estimate the link level performance loss for numerous impairments. The performance degradation in AWGN and a frequency selective fading when using a white noise source as an equivalent transmitter RF impairment model, fits well with the other impairments in the anticipated range of EVM. An analytical formulation describing the effect of the white noise source in the SNR loss fits similarly with the simulation results. With 16 QAM rate 3/4, an EVM value of 8% corresponds to an SNR loss of approximately 0.5 dB.

Crosstalk coupling effects of CMOS co-planar spiral inductors

The coupling effects between two adjacent co-planar spiral inductors are characterized in two cases, one where no guard structure is used and one where simple guard-rings are used. In addition, the effect of guard-rings is evaluated at different distances (190 /spl mu/m to 1090 /spl mu/m) between inductors. The model traditionally used to predict this crosstalk is found to be insufficient and an extended model including mutual inductive coupling and direct capacitive coupling is shown to provide accurate fit. Measuring low levels of crosstalk is difficult and in this context the effect of the test fixture itself is evaluated. Return current paths are here found to have significant influence on low frequency results. Also, based on laser cutting of test fixtures, a surrounding ground-ring is found to increase the crosstalk level. The use of simple guard-rings is shown to improve isolation by approximately 10-15 dB for closely spaced adjacent inductors. At larger distances the gain from having a guard-ring reduces and eventually reduces to zero at a distance of 1000 /spl mu/m. For closely spaced devices a doubling of distance is found to provide an additional 20 dB attenuation of crosstalk.

Measurement and modeling of coupling effects of CMOS on-chip coplanar inductors

The coupling effects between two adjacent coplanar spiral inductors are characterized for two cases, one where no guard structure is used and one where simple guard-rings are used. In addition, the effect of guard-rings is evaluated at different distances (190 /spl mu/m to 1090 /spl mu/m) between inductors. Measuring low levels of crosstalk is difficult, and in this context the effect of the test fixture itself is evaluated. Return current paths are found to have significant influence on low frequency results. In addition, based on laser cutting of test fixtures, a surrounding ground-ring is found to increase the crosstalk level. With the effect of ground-ring coupling removed, the use of simple guard-rings is shown to improve isolation by approximately 10-15 dB for closely spaced adjacent inductors. At larger distances, the gain from having a guard-ring reduces and eventually reduces to zero at a distance of 1000 /spl mu/m. For closely spaced devices, a doubling of distance is found to provide an additional 20 dB attenuation of crosstalk. An extended model, including mutual inductive coupling and direct capacitive coupling, is shown to provide an accurate fit.

Modeling and Design Guidelines for

This paper presents a compact model for P+ guard rings in lightly doped CMOS substrates featuring a P-well layer. Simple expressions for the impedances in the model are derived based on a conformal mapping approach. The model can be used to predict the noise suppression performance of P+ guard rings in terms of S-parameters, which is useful for substrate noise mitigation in mixed-signal system-on-chips. Validation of the model has been done by both electromagnetic simulation and experimental results from guard rings implemented using a standard 0.18- μm CMOS process. In addition, design guidelines have been drawn for minimizing the guard ring size while maintaining the noise suppression performance.

{\rm P}^{+}

This brief presents an ultra-wideband pulse generator topology featuring adaptive control of power spectral density for a broad range of applications with different data rate requirements. The adaptivity is accomplished by employing a limited monocycle precharge approach to control the energy used for pulse generation at different desired data rates. By doing so, the need for tuning circuits is eliminated and the radiated power is maintained at the highest level allowed by the Federal Communications Commission. A prototype pulse generator has been implemented using the UMC 180-nm CMOS process for validation. The measured results show that the pulse generator can be used for a wide pulse repetition rate range from 100 Mpps to 1 Gpps. In addition, the pulse generator consumes 0.76 pJ/pulse at 1 Gpps, equivalent to 760 μW and has a compact size of 0.09 mm2.

Guard Rings in Lightly Doped CMOS Substrates

A low power RF multiplier with ultra-wide signal response band is presented for ultra-wide band system applications such as an UWB demodulator (FM-UWB) or RF-correlator (impulse-radio). The principle of operation and the bandwidth theory is presented and discussed. The practical circuit is implemented using a 0.25μm CMOS process from UMC. The test results show an average gain of 22.5dBV-1at 1.2 GHz and 20.8dBV-1at 3 GHz. Across a full bandwidth of more than 700 MHz the design provides high in-band gain flatness. The circuit consumes a total of 1.3 mA from a 2.5V supply. The total circuit area is 200μmx300μm.

A 0.76-pJ/Pulse 0.1–1 Gpps Microwatt IR-UWB CMOS Pulse Generator With Adaptive PSD Control Using a Limited Monocycle Precharge Technique

Abstract New advanced Envelope Tracking (ET) techniques can provide RF (Radio Frequency) transmitters with high-efficiency Power Amplifiers (PAs). On the other hand, system complexity substantially increases, requiring more advanced PA models for the representation and compensation of ET PA distortion effects. In this context, this paper proposes some solutions for behavioral modeling and digital predistortion of ET PAs. The adopted modeling strategy consists in including the modulated supply voltage as an additional independent model variable to define more accurate behavioral models capable of an increased accuracy when applied to model and compensate ET PAs. The new model variable is included in a polynomial model with memory whose nonlinear structure is derived from a binomial power series, whence the name of Memory Binomial Model (MBM). Another modeling approach is subsequently proposed, where the Cann model for static PA AM/AM nonlinearities is extended to model both AM/AM and AM/PM dynamic distortion occurring in ET PAs. The Extended Cann model includes an MBM structure for modeling dynamic AM/PM distortion effects. Both modeling approaches are tested on measured data-sets acquired using an ET measurement set-up including a commercial PA from RFMD and an envelope modulator designed using a commercial IC from Texas Instruments. The measured results showed that the proposed models could obtain a better modeling and predistortion performance when applied to ET PAs, with respect to the Memory Polynomial Model, here considered as a reference to represent the state-of-the-art of PA modeling and digital predistortion.

A 0.25μm CMOS Low Power RF Multiplier for Ultra-wide Band System Applications

X-parameters have been introduced as the natural extension of S-parameters capable of characterizing a nonlinear device excited by a large-signal input. This paper describes validation of the X-parameter model of a switching cascode power amplifier (PA), which has strong nonlinearity. The X-parameter model of the PA was measured and extracted by an Agilent N5245A PNA-X. Measurements were done on wafer and deem-bedded to the input and output pads of the device. An Enhanced Data rates for GSM Evolution (EDGE) signal was applied to the model for simulations. The simulated relative levels of output spectrum and RMS value of error vector magnitude (EVM) were compared with the measured data in order to validate the X-parameter model. A good match was achieved between the simulation and measurement. The maximum difference between the simulated and measured relative levels of output spectrum is 4 dB. The maximum error between the simulated and measured EVM is less than 3 %-point.

Memory models for behavioral modeling and digital predistortion of envelope tracking power amplifiers

For the UTRA long term evolution (LTE), the transmit signal spectrum must comply with stringent spectrum emission mask requirements because of very low guard band size for high spectral efficiency. Compared to e.g. WCDMA, the requirements are tightened, which calls for efficient user equipment (UE) transmitter design, to balance the tradeoff among performance and hardware complexity. In this paper a spectrum shaping technique suitable for LTE user equipment is proposed and evaluated. The key component of this technique is an inverse Chebyshev digital IIR filter which has maximally flat in the passband and sharp transition from passband to stopband. We compare this technique with the well-known time windowing method which offers low implementation complexity. We study the two techniques impact on both error vector magnitude (EVM) and spectrum emission for two different LTE system bandwidths and for various bandwidth allocations for the user. Results show that for wideband setting (10 MHz) both techniques are able to produce acceptable EVM values less than 0.2%. However, for narrowband setting (1.4 MHz), only digital filtering performs sufficiently well since the time windowing method fails when the window overlap size is greater than the cyclic prefix length.