AN-1141 APPLICATION NOTE One Technology Way • P.O. Box 9106 • Norwood, MA 02062-9106, U.S.A. • Tel: 781.329.4700 • Fax: 781.461.3113 • www.analog.com Powering a Dual Supply Precision ADC with Switching Regulators by Rui Du INTRODUCTION Compared with an LDO device, a switching regulator always dissipates much less heat and provides higher efficiency. Therefore, a switching regulator is suitable for powering different kinds of portable devices or the nodes in a wireless sensor network to lengthen battery time. Unfortunately, a switching regulator intrinsically generates ripple and noise at both the output rail and the ground. At the same time, a switching regulator brings electromagnetic radiation. These interferences are almost inevitable due to the continuously on-off operation of the power switch. With the parasitic parameters involved, the noise will be present at unexpected frequency points besides the integer multiple of the switching frequency. 10481-001 NOISE RIPPLE Figure 1. Typical Ripple and Noise at the Output of a Switching Regulator Although damaged from the noise issues, if the interference of the switching regulator can be effectively controlled while the circuit is powered by a switching regulator, which has a strong anti-interference capacity, then the degrading of the system performance will be slight. For a typical application, when a data converter is powered by a switching mode power supply (SMPS), keep the total noise over the band of interest lower than the noise floor to prevent it from being seen by the converter. Although the in-band noise from the switching regulator is normally greater than the noise floor, because of the power rejection ratio of the ADC, the noise will be sharply attenuated before entering the signal path; as a result, the switching noise will not degrade the performance of the ADC. There are two options to power the ADC with switching regulators: A. Select a low noise switching regulator and then use carefully designed filtering and shielding methods to remove as much of the noise as possible. B. Estimate the noise suppressing capability of the data converter, and then select the product, which has good noise-immunity performance. In practice, both options can be used at the same time so that the power solution of using a switching regulator can be acceptable in most cases. The switching regulator solution includes the benefits of high efficiency and low temperature. In CN-0137, a dual-output synchronous buck switching regulator, ADP2114, is used to power the 16-bit, 125 MSPS analog-to-digital converter, AD9268. The outputs of ADP2114 are filters using an extra stage of an LC filter (ferrite bead). Compared with the linear supplies solution, the testing result shows that using a dc-to-dc supply has nearly no influences on the performance of the ADC (see Table 1). Table 1. Experiment Results Reported in CN-0137 Analog Input Frequency (MHz) 10.3 70.0 100.3 140.3 170.3 200.3 Linear Supplies SNR SFDR (dBFS) (dBc) 79.2 92.2 78.5 91.0 77.8 85.8 76.9 85.0 76.2 84.3 75.0 76.9 DC-to-DC Supply SNR SFDR (dBFS) (dBc) 79.2 92.3 78.4 90.8 77.7 85.6 76.9 84.8 75.9 84.6 75.0 77.0 The noise performance of the ADP2114 is guaranteed by multiple technologies implemented in the design. The typical voltage ripple is less than 1 mV. Using additional filtering, its noise performance can even align with linear supplies. Rev. 0 | Page 1 of 8 AN-1141 Application Note TABLE OF CONTENTS Introduction ...................................................................................... 1 Filtering Considerations ...................................................................7 Experiment Results........................................................................... 3 Conclusion..........................................................................................8 Experiment 1 ................................................................................. 5 References ...........................................................................................8 Experiment 2 ................................................................................. 6 REVISION HISTORY 2/12—Revision 0: Initial Version Rev. 0 | Page 2 of 8 Application Note AN-1141 EXPERIMENT RESULTS 100 In this application note, a switching regulator is used to power the ADC without any additional filtering or shielding measures, only the necessary external components kept for normal operation of the dc-to-dc power supply. According to the second option mentioned previously, two power-insensitive ADCs, the AD7610 and AD7612 are used for testing. This application note aims at finding out how much harm is brought by the SMPS to the ADC and whether it is acceptable. 90 80 CMRR (dB) 70 40 20 0 1 10 100 FREQUENCY (kHz) 1k 10k 10481-004 10 Figure 4. Analog Input CMRR vs. Frequency of AD7610 or AD7612 More importantly, the AD7610 and AD7612 provide excellent power rejection ratio. They are very insensitive to power supply variations on AVDD over a wide frequency range (see Figure 2). 80 EXTERNAL REF 75 70 PSRR (dB) 50 30 The AD7610 and AD7612 are 16-bit charge redistribution successive approximation register (SAR) analog-to-digital converters. They feature true bipolar analog input range. The analog input signal should never exceed the supply rails by more than 0.3 V. For ±10 V input, a typical power supply is ±12 V. See the AD7610 and AD7612 data sheets available from www.analog.com. For powering the AD7610 and AD7612, a 5 V to ±12 V power module using the ADP1613 (the boost dc-to-dc converts 5 V to 12 V) and ADP2301 (the inverting dc-to-dc converts +5 V to −12 V) was designed. For more information about the inverting converter application of ADP2301, refer to the AN-1083 Application Note. The power solution is made to help those customers, who only have 5 V on-board to generate ±12 V with high-current ability and high-efficiency (see Figure 5). INTERNAL REF 65 60 60 L 55 50 D 45 VIN 40 EN ADP1613 CIN 35 100 FREQUENCY (kHz) 1k 10k Rf2 SS Figure 2. AVDD PSRR vs. Frequency of AD7610 or AD7612 Figure 3 shows an equivalent circuit for the input structure of the AD7610 and AD7612. The analog input is first handled by the high voltage branches (powered by VCC and VEE) and scaled down to 0 V to 5 V. VIN D2 D4 RIN BST L CIN SW IN+ OR IN– D Rf1 GND EN VEE AGND 10481-003 CPIN CBOOT ADP2301 CP D3 COUT RCOMP CCOMP CIN AVDD D1 COMP GND 0V TO 5V RANGE ONLY VCC GND CSS FB GND Rf2 VOUT– Figure 3. Simplified Analog Input Structure of AD7610 or AD7612 This analog input structure allows the sampling of the differential signal between IN+ and IN−. By using this differential input, small signals common to both inputs are rejected as shown in Figure 4, which represents the typical CMRR over frequency. COUT Figure 5. Schematic of the 5 V to ±12 V Power Module Using ADP1613 or ADP2301 Rev. 0 | Page 3 of 8 10481-005 10 10481-002 1 Rf1 FB FREQ 30 VOUT+ SW VIN AN-1141 Application Note For a SMPS, the output-voltage ripple can be suppressed by using a big inductor and output capacitor in the topology. To deal with the switching noise, an extra filter at the output can be used. By doing this, the PCB area will be sacrificed. T The basic configuration of the 5 V to ±12 V power modules is shown in Table 2. Configuration Switching Frequency Output Inductor Input Capacitor Output Capacitor Maximum Load +12 V 1.3 MHz 10 µH 10 µF 10 µF 400 mA Output −12 V 1.4 MHz 8.2 µH 10 µF 44 µF 200 mA CH1 20.0mV BW 0.00V Figure 8. AC-Coupled Output Voltage of the −12 V Rail T Note the following about the ripple and noise performance under the typical loading (50% of full load): • M 1.00µs A CH1 T 172.000ns 10481-008 Table 2. Basic Configuration for the ±12 V Power Modules For the +12 V rail of the power module: the ripple ≈ 20 mV p-p the noise ≈ 140 mV p-p (oscilloscope in 1 MΩ mode) For the −12 V rail of the power module: • the ripple ≈ 10 mV p-p • the noise ≈ 50 mV p-p (oscilloscope in 1 MΩ mode) • • CH1 20.0mV BW M 100µs A CH1 T 200.000ns 0.00V 10481-009 T Figure 9. Filtered Output Voltage of the −12 V Rail (AC-Coupled) CH1 50.0mV BW M 400ns T 0.00000s A CH1 0.00V 10481-006 When 2-stage filters are added to the output (the first stage is the LC filter, and the second stage is the bead + decoupling capacitor), most of the ripple and noise can be removed. However, in all the experiments mentioned in this application note, the original output, no extra filters, was used. Figure 6. AC-Coupled Output Voltage of the +12 V Rail T CH1 20.0mV BW M 40.0µs A CH1 T 160.000ns 0.00V 10481-007 • Figure 7. Filtered Output Voltage of the +12 V Rail (AC-Coupled) Rev. 0 | Page 4 of 8 Application Note AN-1141 NUMBER OF COUNTS The first experiment was performed based on AD7612-EVAL and ADuC7026-EVAL. The ADuC7026-EVAL was used to read the conversion results. The input range of AD7612 is configured as ±5 V. The analog inputs of AD7612 (IN+ and IN−) are both directly grounded, the input buffers on the evaluation board are bypassed (see Figure 10). 10481-010 AD7612 Figure 10. Simplified Schematic for Experiment 1. The 3 dB Bandwidth of the Input Anti-Aliasing Filter is about 4 MHz. 0 Two power supply configurations were used in the experiment: Configuration A: VCC and VEE of the AD7612 are powered by the ±12 V power module based on ADP1613 and ADP2301. The +5 V AVDD and DVDD are powered by high quality linear dc power. Configuration B: For comparison, VCC, VEE, AVDD, and DVDD are all provided by the high quality linear dc power. • • During the test, 16,384 conversions were performed. 1 2 LSB LSB 0x0000 0x0001 0x0002 0x0003 0x0004 µ δ 3 NUMBER OF COUNTS 0 717 13769 1898 0 2.0721 0.393 4 Figure 12. Testing Results—Histogram for Power Supply Configuration B Following are the calculations for mean and variance: For Configuration A, mu (μ) = 2.0729 LSB sigma (δ) = 0.3857 LSB The peak-to-peak noise ≈ 2.5456 LSB 14500 14000 13500 13000 12500 12000 11500 11000 10500 10000 9500 9000 8500 8000 7500 7000 6500 6000 5500 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0 peak-to-peak resolution ≈ 216 ≈ log 2 10 × 4.411 ≈ 14.65 bits log 2 2.5456 For an interval estimation, The 95% confidence interval of μ is [2.0670, 2.0788] The 95% confidence interval of δ is [0.3816, 0.3900] For Configuration B, mu (μ) = 2.0721 LSB sigma (δ) = 0.3930 LSB The peak-to-peak noise ≈ 2.5938 LSB 0 1 2 LSB LSB 0x0000 0x0001 0x0002 0x0003 0x0004 µ δ 3 NUMBER OF COUNTS 0 664 13862 1857 1 2.0729 0.3857 peak-to-peak resolution ≈ 4 216 ≈ log 2 10 × 4.403 ≈ 14.63 bits log 2 2.5938 10481-011 NUMBER OF COUNTS The testing results for Configuration A and Configuration B are shown in Figure 11 and Figure 12. 14500 14000 13500 13000 12500 12000 11500 11000 10500 10000 9500 9000 8500 8000 7500 7000 6500 6000 5500 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0 10481-012 EXPERIMENT 1 Figure 11. Testing Results—Histogram for Power Supply Configuration A For an interval estimation, The 95% confidence interval of μ is [2.0661, 2.0781] The 95% confidence interval of δ is [0.3888, 0.3973] The change of peak-to-peak resolution is within 0.03 bit, using SNR = 6.02 N + 1.76. The change of the SNR is within 0.2 dB. Rev. 0 | Page 5 of 8 AN-1141 Application Note EXPERIMENT 2 –0.0003 The second experiment was performed based on the AD7610-EVAL. The FIFO board (EVAL-Control BRDXZ) and the evaluation software were used to analyze the conversion results. am –0.0004 –0.0005 The input range of AD7610 is configured as ±5 V. –0.0006 –0.0007 –0.0008 0 AD8021 AD7610 0.02 0.04 0.06 0.08 0.10 0.12 Figure 15. Time Domain Waveform for Power Supply, Configuration B 10481-013 4500 AD8021 0.14 10481-015 The input buffers are enabled (AD8021). The inputs of AD8021s are both grounded. The AD8021 is dual-supply operation, using the same ±12 V rails as AD7610 (see Figure 13). m 4000 Figure 13. Simplified Schematic for Experiment 2. The 3 dB Bandwidth of the Input Anti-Aliasing Filter is about 4 MHz. 3500 3000 Two power supply configurations were used in the experiment: 2000 1500 1000 500 0 7FF9 The testing results for Configuration A and Configuration B are shown in Figure 14 to Figure 19. –0.0003 2500 7FFA 7FFB 7FFC 7FFD 7FFE 7FFF 8000 10481-016 Configuration A: Using high quality linear dc power to provide +12 V VCC, −12 V VEE, +5 V AVDD, and +5 V DVDD (and power the input buffer). Configuration B: Using high quality linear dc power to provide +5 V AVDD, +5 V DVDD; using ±12 V SMPS to provide +12 V VCC, −12 V VEE (and to power the input buffer). Figure 16. Histogram for Power Supply, Configuration A 5500 m 5000 am 4500 –0.0004 4000 3500 3000 –0.0005 2500 2000 –0.0006 1500 1000 –0.0007 –0.0008 0 0.005 0.010 0.015 0.020 0.025 0.030 0.035 10481-014 500 Figure 14. Time Domain Waveform for Power Supply, Configuration A Rev. 0 | Page 6 of 8 0 7FF9 7FFA 7FFB 7FFC 7FFD 7FFE 7FFF Figure 17. Histogram for Power Supply, Configuration B 8000 10481-017 Application Note –70 AN-1141 FILTERING CONSIDERATIONS d –80 As a second-order filter, the LC filter provides a sharp roll-off above its resonant frequency and is widely used at the output of the dc-to-dc power supply. However, normally the performance of the switching regulator is only specified for the resistive load. If a LC filter is inserted between the dc-to-dc power supply and the resistive load, the dc-to-dc power supply sees a new complex load present at the output: –90 –100 –110 –120 –130 –140 –150 –160 –170 ZL –180 –190 –210 0 20 40 60 80 100 120 140 10481-018 –200 R sL s 2 LCR 1 sCR where s is the complex variable in Laplace transform. ZIN ZL Figure 18. Spectrum for Power Supply, Configuration A DC-DC d 10481-020 –70 –80 –90 Figure 20. DC-to-DC with Input/Output LC Filters –100 From this point of view, as a closed-loop system, the load condition affects the dc-to-dc’s loop transfer function. The bandwidth and the phase margin of the closed-loop system are altered, which may even cause stability issues. The influence of the additional filter on the dc-to-dc is complicated. As an approximation, within an appropriate range, the transient behavior of the dcto-dc power supply with an extra LC filter is similar to the step response of a series RLC tank. –110 –120 –130 –140 –150 –160 –180 0 20 40 60 80 100 120 140 10481-019 –170 Figure 19. Spectrum for Power Supply, Configuration B For Configuration A, SNR = 93.40 dB SINAD = 93.39 dB Following are some semi-experiential guidelines for choosing the LC filter, which may help to improve the stability of the design. For Configuration B, SNR = 93.20 dB SINAD = 93.18 dB The impact on the noise performance of the AD7610 and AD7612 caused by the SMPS is very limited. The SNR has only about 0.1 dB to 0.2 dB variation and there is almost no change in ENOB. If an extra filter is added for the SMPS, the results should be even better. In the AD7610 and AD7612 data sheets, the frequency response of PSRR for AVDD is specified. From the experiment results, it seems that VCC and VEE also have impressive PSRR specifications. Rev. 0 | Page 7 of 8 Normally it’s safe to set the resonant frequency of the LC filter to be higher than the original loop bandwidth of the dc-to-dc. If the resonant frequency had to be made lower, try to use smaller inductance and bigger capacitance (lower Q). AN-1141 Application Note Figure 21 and Figure 22 shows the simulation results for the frequency response and the transient response of different LC filters. A group of inductance and capacitance is used, while the resonant frequency of the LC tank remains unchanged. Figure 21 and Figure 22 show the waveform of the output voltage during load-transient courses: the excessive waveform is measured before the additional LC filter; the lagging waveform is measured after the additional LC filter. With the increasing of inductance, ringing is present during the load-transient. A measured result for ADP1613 (12 V output) is shown in Figure 23. With an extra LC filter added to the output (L = 4.7 µH, C = 10 µF), the noise is greatly reduced, while the transient response doesn’t change a lot, and the system is stable. ORIGINAL OUTPUT 1T 2T FILTERED OUTPUT 50 0 –50 LOAD STEP 3 10481-023 IZI (dBΩ) –100 –150 –200 Figure 23. Transient Response of ADP1613 (12 V Output) with the Output LC Filter Inserted –250 CONCLUSION –350 –400 1 2 3 4 FREQUENCY (kHz) 5 6 7 8 9 10 10481-021 L = 30µH, C = 33µF L = 20µH, C = 50µF L = 10µH, C = 100µF –300 3.6 3.5 3.4 3.3 3.2 3.1 3.0 3.50 3.45 3.40 3.35 3.30 3.25 3.20 3.15 3.10 3.05 3.45 3.40 3.35 3.30 3.25 3.20 3.15 3.10 L = 30µH C = 33µF ORIGINAL OUTPUT FILTERED OUTPUT L = 20µH C = 50µF ORIGINAL OUTPUT FILTERED OUTPUT REFERENCES AD7610 Data Sheet. Analog Devices, Inc. 2006. 2.6 2.8 AD7612 Data Sheet. Analog Devices, Inc. 2006. ORIGINAL OUTPUT FILTERED OUTPUT L = 10µH C = 100µF 3.0 3.2 3.4 3.6 TIME (ms) 3.8 CN-0137. Powering the AD9268 Dual Channel, 16-bit, 125 MSPS 4.0 4.2 4.4 Analog-to-Digital Converter with the ADP2114 Synchronous Step-Down DC-to-DC Regulator for Increased Efficiency. Analog Devices, Inc. 2009. 10481-022 VOLTAGE (V) Figure 21. Frequency Response of Different LC Filters (with a Fixed Resistive Load) The AD7610 and AD7612 have excellent power rejection performance. Their differential inputs ensure the commonmode rejection capability within a certain frequency range. When the power supply is designed for these kinds of ADCs, a switching regulator can be considered. With the help of external filtering and shielding units, the noise characters of the SMPS will be improved further. For energy-constrained applications, if the system to be powered has good noise rejection ability, coupled with the filtering and the shielding measures, use of SMPS will improve the energy efficiency but not degrade the performance of the system. Figure 22. Transient Response of the DC-DC with Different Output LC Filters Inserted Kessler, Matthew C. AN-1083 Application Note. Designing an Inverting Buck Boost Using the ADP2300 and ADP2301 Switching Regulators. Analog Devices, Inc. ©2012 Analog Devices, Inc. All rights reserved. Trademarks and registered trademarks are the property of their respective owners. AN10481-0-2/12(0) Rev. 0 | Page 8 of 8