© Linear Technology
Application Notes |
[chapter-index=Index Head link-to=24934,24935,24936]Here you find links to the other parts of the application note.[/chapter-index]Simplify and Reduce Signal Chain Elements Use of a high-resolution ADC can lead to an intriguing benefit: simplification of the analog signal chain. A higher-resolution ADC can reduce or even eliminate the need for analog signal-conditioning blocks. Since analog functions often exhibit nonlinearity, drift, and other sources of error, the resulting system design is both simpler and more accurate. Wide-dynamic-range sensors are often paired with variable-gain amplifiers to achieve adequate measurement resolution over the entire input span of the sensor. For example, an optical power sensor may have a usable range that spans six decades of measurement, from nanowatts (nW) to milliwatts (mW). A traditional approach is to use a logarithmic amplifier to scale the high-dynamic-range signal into the input range of a lower-dynamic-range ADC. The gain is high for small input amplitudes, and it rolls off at higher input amplitudes. The disadvantage of this approach is that the analog log function exhibits drift, and that bandwidth varies with the input. Thermal flow meters are another example of nonlinear sensors that traditionally require variable gain. Low thermal flows have higher sensitivities, resulting in higher resolution for the indicated measurement, while high thermal flows have lower sensitivities and resolution. The LTC2378-20 has over 5 decades of dynamic range in terms of noise, and it provides 6 decades of DC accuracy (1ppm), which is generally sufficient to digitize such signals directly. Digital signal processing techniques may be used to increase the noise dynamic range by reducing bandwidth, or to implement a log function (e.g., a simple right or left shift of the digital code), or to compensate for sensor nonlinearities. Programmable gain amplifiers (PGAs) and stepped attenuators are other approaches to achieve a wide dynamic range in a system with a low-resolution ADC. An auto-ranging voltmeter is an example of this; the meter starts in its most sensitive range, and switches to a higher range (often 10x larger) once the input exceeds the limit of the lower range. However, there will be a discontinuity when the range is switched. Ideally 100% of one input range should exactly equal 10% of the next larger range, but there will always be some error. Once again, the exceptional linearity and dynamic range of the LTC2378-20 allows several ranges to be combined, eliminating discontinuities associated with switching ranges. Control Systems Latency is a key parameter for ADCs used in mixed-mode control systems, because too much latency may cause instability. Delta-sigma ADCs with ppm-level linearity are available, but they can be used only in very slow control systems with low regulation bandwidth. The no-cycle latency characteristics of LTC2378-20, combined with its exceptional linearity, enables cost-effective implementation of much faster, highly accurate mixed-mode control systems. A control system’s regulation bandwidth is related to its noise bandwidth, and it is only the in-band fraction of the ADC’s noise that contributes to the control system’s overall noise. The LTC2378-20 provides 104dB SNR, implying that its 22.5µVrms input-referred noise corresponds to a noise power spectral density (PSD) of only 31.5nV/rtHz at a sampling rate of 1Msps. Accordingly, when employed in a 1Msps control system with a 10kHz regulation bandwidth, the in-band noise is only 31.5nV/rtHz*√10kHz= 3.2µV, corresponding to a dynamic range of 121dB. In this example, the noise resolution of 3.2µV is roughly the same as the nonlinearity induced uncertainty of only ±0.5ppm*10V = ±5µV. The control system effectively averages the noise across 1Msps/(2*10kHz) = 50 samples to achieve ppm-level noise and linearity performance.
Performance is independent of whether the averaging is performed by a digital filter (controller), or by an analog system component limiting the bandwidth. Figure 3 shows a mixed-mode control system where the bandwidth is limited, in part, by the inertia of a flywheel.
Part 3: Simplify
Many industrial systems require measurement of critical parameters with extreme accuracy. Examples include seismic monitoring, energy exploration, airflow sensing, and silicon wafer fabrication.
[chapter-index=Index Head link-to=24934,24935,24936]Here you find links to the other parts of the application note.[/chapter-index]Simplify and Reduce Signal Chain Elements Use of a high-resolution ADC can lead to an intriguing benefit: simplification of the analog signal chain. A higher-resolution ADC can reduce or even eliminate the need for analog signal-conditioning blocks. Since analog functions often exhibit nonlinearity, drift, and other sources of error, the resulting system design is both simpler and more accurate. Wide-dynamic-range sensors are often paired with variable-gain amplifiers to achieve adequate measurement resolution over the entire input span of the sensor. For example, an optical power sensor may have a usable range that spans six decades of measurement, from nanowatts (nW) to milliwatts (mW). A traditional approach is to use a logarithmic amplifier to scale the high-dynamic-range signal into the input range of a lower-dynamic-range ADC. The gain is high for small input amplitudes, and it rolls off at higher input amplitudes. The disadvantage of this approach is that the analog log function exhibits drift, and that bandwidth varies with the input. Thermal flow meters are another example of nonlinear sensors that traditionally require variable gain. Low thermal flows have higher sensitivities, resulting in higher resolution for the indicated measurement, while high thermal flows have lower sensitivities and resolution. The LTC2378-20 has over 5 decades of dynamic range in terms of noise, and it provides 6 decades of DC accuracy (1ppm), which is generally sufficient to digitize such signals directly. Digital signal processing techniques may be used to increase the noise dynamic range by reducing bandwidth, or to implement a log function (e.g., a simple right or left shift of the digital code), or to compensate for sensor nonlinearities. Programmable gain amplifiers (PGAs) and stepped attenuators are other approaches to achieve a wide dynamic range in a system with a low-resolution ADC. An auto-ranging voltmeter is an example of this; the meter starts in its most sensitive range, and switches to a higher range (often 10x larger) once the input exceeds the limit of the lower range. However, there will be a discontinuity when the range is switched. Ideally 100% of one input range should exactly equal 10% of the next larger range, but there will always be some error. Once again, the exceptional linearity and dynamic range of the LTC2378-20 allows several ranges to be combined, eliminating discontinuities associated with switching ranges. Control Systems Latency is a key parameter for ADCs used in mixed-mode control systems, because too much latency may cause instability. Delta-sigma ADCs with ppm-level linearity are available, but they can be used only in very slow control systems with low regulation bandwidth. The no-cycle latency characteristics of LTC2378-20, combined with its exceptional linearity, enables cost-effective implementation of much faster, highly accurate mixed-mode control systems. A control system’s regulation bandwidth is related to its noise bandwidth, and it is only the in-band fraction of the ADC’s noise that contributes to the control system’s overall noise. The LTC2378-20 provides 104dB SNR, implying that its 22.5µVrms input-referred noise corresponds to a noise power spectral density (PSD) of only 31.5nV/rtHz at a sampling rate of 1Msps. Accordingly, when employed in a 1Msps control system with a 10kHz regulation bandwidth, the in-band noise is only 31.5nV/rtHz*√10kHz= 3.2µV, corresponding to a dynamic range of 121dB. In this example, the noise resolution of 3.2µV is roughly the same as the nonlinearity induced uncertainty of only ±0.5ppm*10V = ±5µV. The control system effectively averages the noise across 1Msps/(2*10kHz) = 50 samples to achieve ppm-level noise and linearity performance.
Performance is independent of whether the averaging is performed by a digital filter (controller), or by an analog system component limiting the bandwidth. Figure 3 shows a mixed-mode control system where the bandwidth is limited, in part, by the inertia of a flywheel.
EDITOR'S NOTE_ Don't forget the links to part 1 and 2 of the application note. You find them in the small box at the top right corner of the article.Conclusion Precision industrial system designs have a new choice to improve signal chain performance. The 20-bit SAR ADC, LTC2378-20 provides an unprecedented level of accuracy (INL guaranteed at 2ppm) and low noise (104dB SNR) at a high conversion rate (1Msps) and low power consumption (21mW). The combination of high accuracy, low noise, and no-cycle latency makes LTC2378-20 highly versatile for use in precision measurements and control systems, enabling a new generation of highly accurate, flexible and cost-effective precision industrial systems. ----- Authors: Atsushi Kawamoto (Design Manager), Jesper Steensgaard (Staff Scientist), Mark Thoren (Staff Scientist) and Heemin Yang (Design Section Leader) at © Linear Technology Corporation
