The advent of multicarrier receivers and SDR applications are forcing board-level and integrated circuit (IC) designers to create systems with ever-increasing dynamic range. This means that receivers must accommodate a large range of input signals while maintaining good distortion and noise characteristics. As the resolution of IF-sampling analog-to-digital converters (ADC) increases from 12 bits to 14 bits, the amplifiers that drive them must keep pace by delivering lower distortion and noise. This presents a challenge to IC designers and test engineers, as the required performance levels often push the capabilities of the test equipment beyond what they were originally designed to measure. This article explores the challenges associated with measuring high-performance ADC driver amplifiers. Anti-aliasing filters can improve most driver interfaces and overall IF performance, but since much has already been written regarding this, it will not be discussed here.

Almost all IF-sampling ADCs feature differential inputs, but most RF and IF stages are single-ended, thus making the conversion from single-ended to differential essential. While transformers can provide single-ended-to-differential transformation with low distortion, some applications require additional gain or isolation from the signal chain. The gain and isolation can be provided by an active ADC driver, thus requiring an amplifier that can perform single-ended to differential conversion, with the flexibility of having adjustable gain while maintaining a good transmission coefficient (S21) performance for reasonable isolation. Also, note that baluns often cause problems at higher frequencies. When the IF frequency exceeds the first Nyquist zone, the even-order harmonic distortion products can be elevated due to the imbalance created by the internal parasitic capacitive coupling of the balun (Figure 1).

Measuring the third-order intercept for high-performance low-distortion amplifiers represents a challenge for available test equipment, with dynamic range being the primary challenge for spectrum analysis. For RF signal generation, the spectral purity of the two input tones required exceeds the performance of typical generators at medium and high drive levels. Also, generators can cause intermodulation (IM) in each other if there is not sufficient isolation between them, thus adding unwanted harmonic products. This intermodulation can be effectively eliminated by using circulators to reduce the unwanted signal from the opposing generator that would otherwise couple into the detector, interfere with the automatic level control (ALC) loop, and create unwanted sidebands. From a practical perspective, it is difficult to obtain IF band circulators (10 MHz to 300 MHz), so another alternative must be chosen. One could simply attenuate the outputs if sufficient power and gain were available from the signal generators. This is not practical, however, as the maximum output levels available from most generators are in the 15 dBm range. To achieve the required gain, power and isolation, a clean, low-noise linear power amplifier (LPA) with high IP3 and good reverse isolation can be used (Figure 2). The LPA allows the generator to be used at lower output levels, thus minimizing unwanted distortion. Further isolation and broadband matching can be achieved by adding attenuation before and after the amplifier. This output padding can be optimized to maintain good harmonic performance. Although this sounds like overkill, it is a good practice. Further low-pass filtering is required to ensure that a clean signal with reduced harmonics is presented to the amplifier under test. In addition, a broadband resistive combiner provides a way to combine the two signals that will give a good match at all power levels and frequencies.

Finally, the spectrum analyzer front-end must have the optimal input signal level in order to minimize additional unwanted IM products. Optimizing the spectrum analyzer is the key to getting the best dynamic range of the system, and can be achieved by keeping the nominal input power at around -35 dBm.

With acceptable levels to the analyzer, the settings should now be optimized for accurate measurements. Reducing the span and resolution bandwidths, and zeroing in on the discrete tones to measure absolute (dBm) rather than relative (dBc), will lower the effective noise floor, providing increased resolution for more accurate measurements. The reference level and attenuation settings must also be optimized, as higher output levels require more dynamic range and higher fundamental power levels. Unfortunately, while higher fundamental levels can overdrive the spectrum analyzer, too much attenuation can limit the dynamic range to the point where you cannot see the low-level harmonic distortion product of interest. As with many RF test setups, a thorough calibration must be completed to determine the overall capabilities of the test system. This measurement becomes the benchmark for the test setup.

1 2 Next

RSS    Save to Del.icio.us  Digg This