LTC1966 Precision Micropower RMS-to-DC Converter Features Simple to Use, Requires One Capacitor True RMS DC Conversion Using Technology High Accuracy: 0.1% Gain Accuracy from 50Hz to 1kHz 0.25% Total Error from 50Hz to 1kHz High Linearity: 0.02% Linearity Allows Simple System Calibration Low Supply Current: 155A Typ, 170A Max Ultralow Shutdown Current: 0.1A Constant Bandwidth: Independent of Input Voltage 800kHz -3dB, 6kHz 1% Flexible Supplies: 2.7V to 5.5V Single Supply Up to 5.5V Dual Supply Flexible Inputs: Differential or Single-Ended Rail-to-Rail Common Mode Voltage Range Up to 1VPEAK Differential Voltage Flexible Output: Rail-to-Rail Output Separate Output Reference Pin Allows Level Shifting Wide Temperature Range: -55C to 125C Small Size: Space Saving 8-Pin MSOP Package Typical Application 2.7V TO 5.5V IN2 0.1F OPT. AC COUPLING EN OUTPUT LTC1966 OUT RTN VSS GND The LTC1966 also has a rail-to-rail output with a separate output reference pin providing flexible level shifting. The LTC1966 operates on a single power supply from 2.7V to 5.5V or dual supplies up to 5.5V. A low power shutdown mode reduces supply current to 0.5A. The LTC1966 is insensitive to PC board soldering and stresses, as well as operating temperature. The LTC1966 is packaged in the space saving MSOP package which is ideal for portable applications. Applications True RMS Digital Multimeters and Panel Meters True RMS AC + DC Measurements n n L, LT, LTC, LTM, Linear Technology and the Linear logo are registered trademarks and No Latency DS is a trademark of Linear Technology Corporation. All other trademarks are the property of their respective owners. Protected by U.S. Patents including 6359576, 6362677, 6516291 and 6651036. 0.2 LTC1966, 0 -0.2 VDD IN1 The LTC1966 accepts single-ended or differential input signals (for EMI/RFI rejection) and supports crest factors up to 4. Common mode input range is rail-to-rail. Differential input range is 1VPEAK, and offers unprecedented linearity. Unlike previously available RMS-to-DC converters, the superior linearity of the LTC1966 allows hassle free system calibration at any input voltage. Quantum Leap in Linearity Performance Single Supply RMS-to-DC Converter DIFFERENTIAL INPUT The LTC(R)1966 is a true RMS-to-DC converter that utilizes an innovative patented computational technique. The internal delta sigma circuitry of the LTC1966 makes it simpler to use, more accurate, lower power and dramatically more flexible than conventional log antilog RMS-to-DC converters. LINEARITY ERROR (VOUT mV DC - VIN mV ACRMS) n n n n n n n n n n n n Description -0.4 CAVE 1F 1966 TA01 + VOUT - -0.6 CONVENTIONAL LOG/ANTILOG -0.8 -1.0 60Hz SINEWAVES 0 50 100 150 200 250 300 350 400 450 500 VIN (mV ACRMS) 1966 TA01b 1966fb 1 LTC1966 Absolute Maximum Ratings Pin Configuration (Note 1) Supply Voltage VDD to GND.............................................. - 0.3V to 7V VDD to VSS ............................................. -0.3V to 12V VSS to GND.............................................. -7V to 0.3V Input Currents (Note 2)....................................... 10mA Output Current (Note 3)...................................... 10mA ENABLE Voltage........................ VSS - 0.3V to VSS + 12V OUT RTN Voltage................................ VSS - 0.3V to VDD Operating Temperature Range (Note 4) LTC1966C/LTC1966I.............................-40C to 85C LTC1966H........................................... -40C to 125C LTC1966MP........................................ -55C to 125C Specified Temperature Range (Note 5) LTC1966C/LTC1966I.............................-40C to 85C LTC1966H........................................... -40C to 125C LTC1966MP........................................ -55C to 125C Maximum Junction Temperature.......................... 150C Storage Temperature Range.................. -65C to 150C Lead Temperature (Soldering, 10 sec)................... 300C TOP VIEW GND IN1 IN2 VSS 1 2 3 4 8 7 6 5 ENABLE VDD OUT RTN VOUT MS8 PACKAGE 8-LEAD PLASTIC MSOP TJMAX = 150C, JA = 220C/W Order Information LEAD FREE FINISH TAPE AND REEL PART MARKING* PACKAGE DESCRIPTION TEMPERATURE RANGE LTC1966CMS8#PBF LTC1966CMS8#TRPBF LTTG 8-Lead Plastic MSOP 0C to 70C LTC1966IMS8#PBF LTC1966IMS8#TRPBF LTTH 8-Lead Plastic MSOP -40C to 85C LTC1966HMS8#PBF LTC1966HMS8#TRPBF LTTG 8-Lead Plastic MSOP -40C to 125C LTC1966MPMS8#PBF LTC1966MPMS8#TRPBF LTTG 8-Lead Plastic MSOP -55C to 125C Consult LTC Marketing for parts specified with wider operating temperature ranges. *The temperature grade is identified by a label on the shipping container. For more information on lead free part marking, go to: http://www.linear.com/leadfree/ For more information on tape and reel specifications, go to: http://www.linear.com/tapeandreel/ Electrical Characteristics The l denotes the specifications which apply over the full operating temperature range, otherwise specifications are at TA = 25C. VDD = 5V, VSS = - 5V, VOUTRTN = 0V, CAVE = 10F, VIN = 200mVRMS, VENABLE = 0.5V unless otherwise noted. SYMBOL PARAMETER CONDITIONS MIN TYP MAX UNITS 0.1 0.3 0.4 0.7 % % % 0.1 0.2 0.4 0.6 mV mV mV 0.02 0.15 % Conversion Accuracy GERR VOOS LINERR Conversion Gain Error Output Offset Voltage Linearity Error 50Hz to 1kHz Input (Notes 6, 7) LTC1966C, LTC1966I LTC1966H, LTC1966MP l l (Notes 6, 7) LTC1966C, LTC1966I LTC1966H, LTC1966MP l l 50mV to 350mV (Notes 7, 8) l 1966fb 2 LTC1966 Electrical Characteristics The l denotes the specifications which apply over the full operating temperature range, otherwise specifications are at TA = 25C. VDD = 5V, VSS = - 5V, VOUTRTN = 0V, CAVE = 10F, VIN = 200mVRMS, VENABLE = 0.5V unless otherwise noted. SYMBOL PARAMETER CONDITIONS PSRR (Note 9) LTC1966C, LTC1966I LTC1966H, LTC1966MP VIOS Power Supply Rejection Input Offset Voltage MIN TYP MAX UNITS 0.02 0.15 0.20 0.3 %V %V %V 0.02 0.8 1.0 mV mV l l (Notes 6, 7, 10) l Accuracy vs Crest Factor (CF) CF = 4 60Hz Fundamental, 200mVRMS (Note 11) l -1 2 mV CF = 5 60Hz Fundamental, 200mVRMS (Note 11) l -20 30 mV l VSS VDD V Input Characteristics IVR Input Voltage Range (Note 14) ZIN Input Impedance Average, Differential (Note 12) Average, Common Mode (Note 12) CMRRI Input Common Mode Rejection (Note 13) l VIMAX Maximum Input Swing Accuracy = 1% (Note 14) l VIMIN Minimum RMS Input PSRRI Power Supply Rejection 8 100 7 1 V/V V 5 mV 600 300 V/V V/V VDD V 85 30 95 k k 16 200 V/V 250 120 l l 200 1.05 l VDD Supply (Note 9) VSS Supply (Note 9) M M Output Characteristics l VSS VENABLE = 0.5V (Note 12) VENABLE = 4.5V l 75 Output Common Mode Rejection (Note 13) l Maximum Differential Output Swing Accuracy = 2%, DC Input (Note 14) OVR Output Voltage Range ZOUT Output Impedance CMRRO VOMAX l PSRRO Power Supply Rejection VDD Supply (Note 9) VSS Supply (Note 9) 1.0 0.9 1.05 250 50 l l V V 1000 500 V/V V/V Frequency Response f1P 1% Additional Error (Note 15) CAVE = 10F 6 kHz f10P 10% Additional Error (Note 15) CAVE = 10F 20 kHz f- 3dB 3dB Frequency (Note 15) 800 kHz Power Supplies VDD Positive Supply Voltage l 2.7 5.5 V VSS Negative Supply Voltage (Note 16) l -5.5 0 V IDD Positive Supply Current IN1 = 20mV, IN2 = 0V IN1 = 200mV, IN2 = 0V l 155 158 170 A A ISS Negative Supply Current IN1 = 20mV, IN2 = 0V l 12 20 A 0.5 10 A Shutdown Characteristics IDDS Supply Currents VENABLE = 4.5V l ISSS Supply Currents VENABLE = 4.5V LTC1966H, LTC1966MP l l -1 -2 -0.1 A A IIH ENABLE Pin Current High VENABLE = 4.5V l -0.3 -0.05 A 1966fb 3 LTC1966 Electrical Characteristics The l denotes the specifications which apply over the full operating temperature range, otherwise specifications are at TA = 25C. VDD = 5V, VSS = - 5V, VOUTRTN = 0V, CAVE = 10F, VIN = 200mVRMS, VENABLE = 0.5V unless otherwise noted. SYMBOL PARAMETER CONDITIONS IIL ENABLE Pin Current Low VENABLE = 0.5V LTC1966H, LTC1966MP VTH ENABLE Threshold Voltage VDD = 5V, VSS = -5V VDD = 5V, VSS = GND VDD = 2.7V, VSS = GND VHYS ENABLE Threshold Hysteresis Note 1: Stresses beyond those listed under Absolute Maximum Ratings may cause permanent damage to the device. Exposure to any Absolute Maximum Rating condition for extended periods may affect device reliability and lifetime. Note 2: The inputs (IN1, IN2) are protected by shunt diodes to VSS and VDD. If the inputs are driven beyond the rails, the current should be limited to less than 10mA. Note 3: The LTC1966 output (VOUT) is high impedance and can be overdriven, either sinking or sourcing current, to the limits stated. Note 4: The LTC1966C/LTC1966I are guaranteed functional over the operating temperature range of - 40C to 85C. The LTC1966H/ LTC1966MP are guaranteed functional over the operating temperature range of -55C to 125C. Note 5: The LTC1966C is guaranteed to meet specified performance from 0C to 70C. The LTC1966C is designed, characterized and expected to meet specified performance from -40C to 85C but is not tested nor QA sampled at these temperatures. The LTC1966I is guaranteed to meet specified performance from -40C to 85C. The LTC1966H is guaranteed to meet specified performance from -40C to 125C. The LTC1966MP is guaranteed to meet specified performance from -55C to 125C. Note 6: High speed automatic testing cannot be performed with CAVE = 10F. The LTC1966 is 100% tested with CAVE = 22nF. Correlation tests have shown that the performance limits above can be guaranteed with the additional testing being performed to guarantee proper operation of all the internal circuitry. Note 7: High speed automatic testing cannot be performed with 60Hz inputs. The LTC1966 is 100% tested with DC and 10kHz input signals. Measurements with DC inputs from 50mV to 350mV are used to calculate the four parameters: GERR, VOOS, VIOS and linearity error. Correlation tests have shown that the performance limits above can be guaranteed with the additional testing being performed to guarantee proper operation of all internal circuitry. Note 8: The LTC1966 is inherently very linear. Unlike older log/antilog circuits, its behavior is the same with DC and AC inputs, and DC inputs are used for high speed testing. Note 9: The power supply rejections of the LTC1966 are measured with DC inputs from 50mV to 350mV. The change in accuracy from VDD = 2.7V to VDD = 5.5V with VSS = 0V is divided by 2.8V. The change in accuracy from VSS = 0V to VSS = -5.5V with VDD = 5.5V is divided by 5.5V. Note 10: Previous generation RMS-to-DC converters required nonlinear input stages as well as a nonlinear core. Some parts specify a DC reversal error, combining the effects of input nonlinearity and input offset voltage. The LTC1966 behavior is simpler to characterize and the input offset voltage is the only significant source of DC reversal error. l l MIN TYP MAX UNITS -2 -10 -1 -0.1 A A 2.4 2.1 1.3 V V V 0.1 V Note 11: High speed automatic testing cannot be performed with 60Hz inputs. The LTC1966 is 100% tested with DC stimulus. Correlation tests have shown that the performance limits above can be guaranteed with the additional testing being performed to verify proper operation of all internal circuitry. Note 12: The LTC1966 is a switched capacitor device and the input/ output impedance is an average impedance over many clock cycles. The input impedance will not necessarily lead to an attenuation of the input signal measured. Refer to the Applications Information section titled Input Impedance for more information. Note 13: The common mode rejection ratios of the LTC1966 are measured with DC inputs from 50mV to 350mV. The input CMRR is defined as the change in VIOS measured between input levels of VSS to VSS + 350mV and input levels of VDD - 350mV to VDD divided by VDD - VSS - 350mV. The output CMRR is defined as the change in VOOS measured with OUT RTN = VSS and OUT RTN = VDD - 350mV divided by VDD - VSS - 350mV. Note 14: Each input of the LTC1966 can withstand any voltage within the supply range. These inputs are protected with ESD diodes, so going beyond the supply voltages can damage the part if the absolute maximum current ratings are exceeded. Likewise for the output pins. The LTC1966 input and output voltage swings are limited by internal clipping. The maximum differential input of the LTC1966 (referred to as maximum input swing) is 1V. This applies to either input polarity, so it can be thought of as 1V. Because the differential input voltage gets processed by the LTC1966 with gain, it is subject to internal clipping. Exceeding the 1V maximum can, depending on the input crest factor, impact the accuracy of the output voltage, but does not damage the part. Fortunately, the LTC1966's topology is relatively tolerant of momentary internal clipping. The input clipping is tested with a crest factor of 2, while the output clipping is tested with a DC input. Note 15: The LTC1966 exploits oversampling and noise shaping to reduce the quantization noise of internal 1-bit analog-to-digital conversions. At higher input frequencies, increasingly large portions of this noise are aliased down to DC. Because the noise is shifted in frequency, it becomes a low frequency rumble and is only filtered at the expense of increasingly long settling times. The LTC1966 is inherently wideband, but the output accuracy is degraded by this aliased noise. These specifications apply with CAVE = 10F and constitute a 3-sigma variation of the output rumble. Note 16: The LTC1966 can operate down to 2.7V single supply but cannot operate at 2.7V. This additional constraint on VSS can be expressed mathematically as - 3 * (VDD - 2.7V) VSS Ground. 1966fb 4 LTC1966 Typical Performance Characteristics VDD = 5V VSS = -5V 0.5 0.5 0.4 0.4 0.2 0.1 0 -0.1 -0.2 GAIN ERROR 0.1 0 VOOS -0.1 VIOS -0.2 VIOS 0.3 0.2 0.4 0.3 0.2 VOOS 0.1 0.1 GAIN ERROR 0 0 -0.1 -0.1 -0.2 -0.2 1.0 VDD = 2.7V VSS = GND 0.8 VIOS 0.3 0.2 0.6 0.4 GAIN ERROR 0.1 0.2 0 0 -0.1 -0.2 VOOS -0.2 -0.4 -0.3 -0.3 -0.3 -0.3 -0.6 -0.4 -0.4 -0.4 -0.4 -0.4 -0.8 -0.5 -0.5 -0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 INPUT COMMON MODE (V) -0.5 -5 -4 -3 -2 -1 0 1 2 3 INPUT COMMON MODE (V) 4 5 1966 G03 0.3 0.2 GAIN ERROR VOOS 0.1 0 0 VIOS -0.1 -0.1 GAIN ERROR (%) 0.1 VIOS 0.2 0 0.5 0.4 0.4 0.3 0.3 0.2 VOOS 0.1 0.5 0.1 0 GAIN ERROR -0.1 -0.1 -0.2 -0.2 GAIN ERROR (%) 0.3 VDD = 5V VSS = GND 0.2 0.1 VDD = 2.7V VSS = GND 1.0 0.8 VIOS 0.6 0.4 GAIN ERROR 0.2 0 0 -0.1 -0.2 VOOS -0.2 -0.4 -0.2 -0.2 -0.3 -0.3 -0.3 -0.3 -0.3 -0.6 -0.4 -0.4 -0.4 -0.4 -0.4 -0.8 -0.5 -0.5 -0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 OUTPUT COMMON MODE (V) -0.5 -5 -4 -3 -2 -1 0 1 2 3 OUTPUT COMMON MODE (V) 4 5 1966 G06 Gain and Offsets vs Temperature 0.3 0.3 0.2 0.2 0.1 GAIN ERROR VOOS 0 0 -0.1 0.1 VIOS -0.1 GAIN ERROR (%) 0.3 VDD = 5V VSS = GND VIOS 0.2 VOOS 0.1 0.4 0.4 0.3 0.3 0.2 0.1 0 0 -0.1 0.5 GAIN ERROR -0.1 0.2 0.1 1.0 VDD = 2.7V VSS = GND 0.8 VIOS 0.6 0.4 GAIN ERROR 0.2 0 0 -0.1 -0.2 VOOS -0.2 -0.2 -0.2 -0.2 -0.3 -0.3 -0.3 -0.3 -0.3 -0.6 -0.4 -0.4 -0.4 -0.4 -0.4 -0.8 -0.5 -0.5 -60 -40 -20 0 20 40 60 80 100 120 140 TEMPERATURE (C) 1966 G09 -0.5 -0.5 -60 -40 -20 0 20 40 60 80 100 120 140 TEMPERATURE (C) 1966 G08 -0.2 -0.4 OFFSET VOLTAGE (mV) 0.4 -1.0 Gain and Offsets vs Temperature 0.5 OFFSET VOLTAGE (mV) 0.4 VDD = 5V VSS = -5V OFFSET VOLTAGE (mV) 0.5 0.4 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 OUTPUT COMMON MODE (V) 1966 G04 Gain and Offsets vs Temperature 0.5 0.5 0 1966 G05 GAIN ERROR (%) -0.5 OFFSET VOLTAGE (mV) 0.3 0.2 -1.0 Gain and Offsets vs Output Common Mode OFFSET VOLTAGE (mV) 0.4 VDD = 5V VSS = -5V OFFSET VOLTAGE (mV) 0.5 0.4 0.4 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 INPUT COMMON MODE (V) 1966 G01 Gain and Offsets vs Output Common Mode 0.5 0.5 0 1966 G02 Gain and Offsets vs Output Common Mode GAIN ERROR (%) 0.5 0.4 -0.3 -0.5 GAIN ERROR (%) 0.5 OFFSET VOLTAGE (mV) 0.3 VDD = 5V VSS = GND Gain and Offsets vs Input Common Mode OFFSET VOLTAGE (mV) 0.3 0.2 OFFSET VOLTAGE (mV) GAIN ERROR (%) 0.4 GAIN ERROR (%) 0.5 Gain and Offsets vs Input Common Mode GAIN ERROR (%) Gain and Offsets vs Input Common Mode -1.0 -0.5 -60 -40 -20 0 20 40 60 80 100 120 140 TEMPERATURE (C) 1966 G07 1966fb 5 LTC1966 Typical Performance Characteristics VDD = 5V 0.1 VIOS VOOS GAIN ERROR 0.3 0.3 0.2 0.1 0 0 - 0.1 NOMINAL SPECIFIED CONDITIONS -0.1 -0.2 -0.3 -0.4 -0.5 0.4 -6 -5 - 0.2 -2 -3 VSS (V) -4 0 -1 VSS = GND 0.6 VIOS 0.2 0.4 0.2 0.1 GAIN ERROR 0 -0.1 0 - 0.2 VOOS -0.2 - 0.4 - 0.3 -0.3 - 0.6 - 0.4 -0.4 - 0.8 -0.5 -0.5 2.5 3.0 3.5 4.0 VDD (V) 4.5 5.0 200 250Hz 100Hz 190 180 200mVRMS SCR WAVEFORMS = 4.7F C 160 VAVE= 5V DD 5%/DIV 150 6 2 3 5 4 1 10 0.15 0.10 0.05 0 -0.20 200 0.10 CAVE = 1F 0.08 VIN2 = GND -0.08 -0.10 -500 -300 100 -100 VIN1 (mV) 300 500 1966 G14 -10 AC INPUT VDD = 5V -1% ERROR DC INPUT VDD = 5V -15 AC INPUT VDD = 3V 0 0.5 1 1.5 VIN1 (VRMS) 2.5 2 Shutdown Currents vs ENABLE Voltage 250 150 125 100 75 50 25 0 -25 1 2 5 3 4 VDD SUPPLY VOLTAGE (V) IDD 150 100 500 IEN 50 250 ISS 0 0 -50 ISS 0 VDD = 5V 200 IDD SUPPLY CURRENT (A) SUPPLY CURRENT (A) EFFECT OF OFFSETS MAY BE POSITIVE OR NEGATIVE -0.06 -5 6 1966 G16 -100 -250 0 1 4 3 5 2 ENABLE PIN VOLTAGE (V) 6 -500 ENABLE PIN CURRENT (nA) VOUTDC - |VINDC| (mV) 0.06 -0.04 AC INPUTS = 60Hz SINEWAVES VIN2 = GND 1% ERROR 1966 G24 VSS = GND 175 5.0 0 50 100 150 200 250 300 350 400 450 500 VIN1 (mV ACRMS) Quiescent Supply Currents vs Supply Voltage -0.02 4.5 1966 G13 DC Linearity 0 2.5 3.0 3.5 4.0 CREST FACTOR 5 -20 0 1966 G12 0.02 2.0 Output Accuracy vs Signal Amplitude 60Hz SINEWAVES CAVE = 1F VIN2 = GND CREST FACTOR 0.04 1.5 1966 G15 -0.15 8 100Hz 200.2 199.8 1.0 -1.0 5.5 -0.10 7 60Hz 200.0 -0.05 170 20Hz 200.4 VOUT (mV DC) - VIN (mVRMS) 60Hz VOUT (mV DC) - VIN (mV ACRMS) OUTPUT VOLTAGE (mV DC) 0.20 20Hz 200.6 AC Linearity FUNDAMENTAL FREQUENCY 210 200mVRMS SCR WAVEFORMS CAVE = 10F 200.8 VDD = 5V O.1%/DIV 1966 G10 Performance vs Large Crest Factors 220 201.0 0.8 1966 G11 230 Performance vs Crest Factor 1 OFFSET VOLTAGE (mV) 0.2 0.4 OFFSET VOLTAGE (mV) GAIN ERROR (%) 0.3 0.5 GAIN ERROR (%) 0.4 Gain and Offsets vs VDD Supply 0.5 OUTPUT VOLTAGE (mV DC) Gain and Offsets vs VSS Supply 0.5 1966 G18 1966fb 6 LTC1966 Typical Performance Characteristics Quiescent Supply Currents vs Temperature Input Signal Bandwidth 150 VDD = 5V, VSS = GND 30 VDD = 2.7V, VSS = GND 25 20 130 120 15 VDD = 5V, VSS = -5V 110 100 VDD = 5V, VSS = GND 0.1% ERROR 35 VDD = 2.7V, VSS = GND 10 1% ERROR 10% ERROR -3dB 10 1 100 10K 100K 1K INPUT SIGNAL FREQUENCY (Hz) 194 192 190 188 186 1M 30 0.5%/DIV CAVE = 47F 25 Common Mode Rejection Ratio vs Frequency 110 VIN2 = GND THREE REPRESENTITIVE UNITS 20 200 199 198 197 15 10 5 0 196 -5 0 10 20 30 40 50 60 70 80 90 100 INPUT FREQUENCY (kHz) 1966 G21 -10 -20 -15 -10 0 5 -5 VIN1 (mV DC) 10 15 1000 1966 G20 DC Transfer Function Near Zero VOUT (mV DC) OUTPUT DC VOLTAGE (mV) 196 184 1%/DIV CAVE = 2.2F 182 10 100 1 INPUT FREQUENCY (kHz) COMMON MODE REJECTION RATIO (dB) Bandwidth to 100kHz 195 200 198 1966 G19 1966 G17 201 202 5 0 90 - 60 -40 -20 0 20 40 60 80 100 120 140 TEMPERATURE (C) 202 Input Signal Bandwidth 100 ISS (A) IDD (A) 140 VDD = 5V, VSS = -5V OUTPUT DC VOLTAGE (mV) 160 1000 OUTPUT DC VOLTAGE (mV) 40 170 20 1966 G22 VDD = 5V VSS = -5V 5V INPUT CONVERSION TO DC OUTPUT 100 90 80 70 60 50 40 30 20 10 100 1k 10k FREQUENCY (Hz) 100k 1M 1966 G23 1966fb 7 LTC1966 Pin Functions GND (Pin 1): Ground. A power return pin. VSS (Pin 4): Negative Voltage Supply. GND to - 5.5V. OUT RTN (Pin 6): Output Return. The output voltage is created relative to this pin. The VOUT and OUT RTN pins are not balanced and this pin should be tied to a low impedance, both AC and DC. Although it is typically tied to GND, it can be tied to any arbitrary voltage, VSS < OUT RTN < (VDD - Max Output). Best results are obtained when OUT RTN = GND. VOUT (Pin 5): Output Voltage. This is high impedance. The RMS averaging is accomplished with a single shunt capacitor from this node to OUT RTN. The transfer function is given by: ENABLE (Pin 8): An Active Low Enable Input. LTC1966 is debiased if open circuited or driven to VDD. For normal operation, pull to GND, a logic low or even VSS. IN1 (Pin 2): Differential Input. DC coupled (polarity is irrelevant). IN2 (Pin 3): Differential Input. DC coupled (polarity is irrelevant). ( VOUT - OUT RTN) = VDD (Pin 7): Positive Voltage Supply. 2.7V to 5.5V. 2 Average (IN2 - IN1) 1966fb 8 LTC1966 Applications Information START NOT SURE READ RMS-TO-DC CONVERSION DO YOU NEED TRUE RMS-TO-DC CONVERSION? FIND SOMEONE WHO DOES AND GIVE THEM THIS DATA SHEET NO YES CONTACT LTC BY PHONE OR AT www.linear.com AND GET SOME NOW DO YOU HAVE ANY LTC1966s YET? NO YES DID YOU ALREADY TRY OUT THE LTC1966? DO YOU WANT TO KNOW HOW TO USE THE LTC1966 FIRST? NO YES READ THE TROUBLESHOOTING GUIDE. IF NECESSARY, CALL LTC FOR APPLICATIONS SUPPORT NO YES NO DID YOUR CIRCUIT WORK? READ THE DESIGN COOKBOOK YES CONTACT LTC AND PLACE YOUR ORDER YES NOW DOES YOUR RMS CIRCUIT WORK WELL ENOUGH THAT YOU ARE READY TO BUY THE LTC1966? NO READ THE TROUBLESHOOTING GUIDE AGAIN OR CALL LTC FOR APPLICATIONS SUPPORT 1966 TA02 1966fb 9 LTC1966 Applications Information RMS-TO-DC CONVERSION Definition of RMS RMS amplitude is the consistent, fair and standard way to measure and compare dynamic signals of all shapes and sizes. Simply stated, the RMS amplitude is the heating potential of a dynamic waveform. A 1VRMS AC waveform will generate the same heat in a resistive load as will 1V DC. 1V DC + - R 1V ACRMS R 1V (AC + DC) RMS R SAME HEAT 1966 F01 Figure 1 Mathematically, RMS is the root of the mean of the square: VRMS = V2 Alternatives to RMS Other ways to quantify dynamic waveforms include peak detection and average rectification. In both cases, an average (DC) value results, but the value is only accurate at the one chosen waveform type for which it is calibrated, typically sine waves. The errors with average rectification are shown in Table 1. Peak detection is worse in all cases and is rarely used. The last two entries of Table 1 are chopped sine waves as is commonly created with thyristors such as SCRs and Triacs. Figure 2a shows a typical circuit and Figure 2b shows the resulting load voltage, switch voltage and load currents. The power delivered to the load depends on the firing angle, as well as any parasitic losses such as switch ON voltage drop. Real circuit waveforms will also typically have significant ringing at the switching transition, dependent on exact circuit parasitics. For the purposes of this data sheet, SCR waveforms refers to the ideal chopped sine wave, though the LTC1966 will do faithful RMS-to-DC conversion with real SCR waveforms as well. The case shown is for = 90, which corresponds to 50% of available power being delivered to the load. As noted in Table 1, when = 114, only 25% of the available power is being delivered to the load and the power drops quickly as approaches 180. With an average rectification scheme and the typical calibration to compensate for errors with sine waves, the RMS level of an input sine wave is properly reported; it is only with a nonsinusoidal waveform that errors occur. Because of this calibration, and the output reading in VRMS, the term true RMS got coined to denote the use of an actual RMS-to-DC converter as opposed to a calibrated average rectifier. + VLOAD - AC MAINS + VLINE - ILOAD CONTROL + - VTHY 1966 F02a Figure 2a Table 1. Errors with Average Rectification vs True RMS WAVEFORM VRMS AVERAGE RECTIFIED (V) Square Wave 1.000 1.000 11% Sine Wave 1.000 0.900 *Calibrate for 0% Error Triangle Wave 1.000 0.866 -3.8% SCR at 1/2 Power, = 90 1.000 0.637 -29.3% SCR at 1/4 Power, = 114 1.000 0.536 -40.4% VLINE ERROR* VLOAD VTHY ILOAD 1966 F02b Figure 2b 1966fb 10 LTC1966 Applications Information How an RMS-to-DC Converter Works How the LTC1966 RMS-to-DC Converter Works Monolithic RMS-to-DC converters use an implicit computation to calculate the RMS value of an input signal. The fundamental building block is an analog multiply/ divide used as shown in Figure 3. Analysis of this topology is easy and starts by identifying the inputs and the output of the lowpass filter. The input to the LPF is the calculation from the multiplier/divider; (VIN)2/VOUT. The lowpass filter will take the average of this to create the output, mathematically: The LTC1966 uses a completely new topology for RMSto-DC conversion, in which a modulator acts as the divider, and a simple polarity switch is used as the multiplier as shown in Figure 4. D VIN VOUT - REF VIN VOUT 1 ( V )2 IN = , VOUT LPF Figure 4. Topology of LTC1966 Because VOUT is DC, 2 ( V )2 ( VIN ) IN , so = VOUT VOUT VOUT The modulator has a single-bit output whose average duty cycle (D) will be proportional to the ratio of the input signal divided by the output. The is a 2nd order modulator with excellent linearity. The single bit output is used to selectively buffer or invert the input signal. Again, this is a circuit with excellent linearity, because it operates at only two points: 1 gain; the average effective multiplication over time will be on the straight line between these two points. The combination of these two elements again creates a lowpass filter input signal proportional to (VIN)2/VOUT, which, as shown above, results in RMS-to-DC conversion. ( V )2 IN = , and VOUT ( VOUT )2 = ( VIN )2, or VOUT = ( VIN )2 = RMS( VIN ) (VIN ) 2 VOUT VIN x / VOUT LPF VOUT 1966 F03 Figure 3. RMS-to-DC Converter with Implicit Computation Unlike the prior generation RMS-to-DC converters, the LTC1966 computation does NOT use log/antilog circuits, which have all the same problems, and more, of log/antilog multipliers/dividers, i.e., linearity is poor, the bandwidth changes with the signal amplitude and the gain drifts with temperature. The lowpass filter performs the averaging of the RMS function and must be a lower corner frequency than the lowest frequency of interest. For line frequency measurements, this filter is simply too large to implement on-chip, but the LTC1966 needs only one capacitor on the output to implement the lowpass filter. The user can select this capacitor depending on frequency range and settling time requirements, as will be covered in the Design Cookbook section to follow. This topology is inherently more stable and linear than log/antilog implementations primarily because all of the signal processing occurs in circuits with high gain op amps operating closed loop. 1966fb 11 LTC1966 Applications Information More detail of the LTC1966 inner workings is shown in the Simplified Schematic towards the end of this data sheet. Note that the internal scalings are such that the output duty cycle is limited to 0% or 100% only when VIN exceeds 4 * VOUT. Linearity of an RMS-to-DC Converter Linearity may seem like an odd property for a device that implements a function that includes two very nonlinear processes: squaring and square rooting. However, an RMS-to-DC converter has a transfer function, RMS volts in to DC volts out, that should ideally have a 1:1 transfer function. To the extent that the input to output transfer function does not lie on a straight line, the part is nonlinear. A more complete look at linearity uses the simple model shown in Figure 5. Here an ideal RMS core is corrupted by both input circuitry and output circuitry that have imperfect transfer functions. As noted, input offset is introduced in the input circuitry, while output offset is introduced in the output circuitry. Any nonlinearity that occurs in the output circuity will corrupt the RMS in to DC out transfer function. A nonlinearity in the input circuitry will typically corrupt that transfer function far less, simply because with an AC input, the RMS-to-DC conversion will average the nonlinearity from a whole range of input values together. INPUT INPUT CIRCUITRY * VIOS * INPUT NONLINEARITY But the input nonlinearity will still cause problems in an RMS-to-DC converter because it will corrupt the accuracy as the input signal shape changes. Although an RMS-to-DC converter will convert any input waveform to a DC output, the accuracy is not necessarily as good for all waveforms as it is with sine waves. A common way to describe dynamic signal wave shapes is crest factor. The crest factor is the ratio of the peak value relative to the RMS value of a waveform. A signal with a crest factor of 4, for instance, has a peak that is four times its RMS value. Because this peak has energy (proportional to voltage squared) that is 16 times (42) the energy of the RMS value, the peak is necessarily present for at most 6.25% (1/16) of the time. The LTC1966 performs very well with crest factors of 4 or less and will respond with reduced accuracy to signals with higher crest factors. The high performance with crest factors less than 4 is directly attributable to the high linearity throughout the LTC1966. The LTC1966 does not require an input rectifier, as is common with traditional log/antilog RMS-to-DC converters. Thus, the LTC1966 has none of the nonlinearities that are introduced by rectification. The excellent linearity of the LTC1966 allows calibration to be highly effective at reducing system errors. See System Calibration section following the Design Cookbook. IDEAL RMS-TO-DC CONVERTER OUTPUT CIRCUITRY * VOOS * OUTPUT NONLINEARITY OUTPUT 1966 F05 Figure 5. Linearity Model of an RMS-to-DC Converter 1966fb 12 LTC1966 Applications Information The LTC1966 RMS-to-DC converter makes it easy to implement a rather quirky function. For many applications all that will be needed is a single capacitor for averaging, appropriate selection of the I/O connections and power supply bypassing. Of course, the LTC1966 also requires power. A wide variety of power supply configurations are shown in the Typical Applications section towards the end of this data sheet. Capacitor Value Selection The RMS or root-mean-squared value of a signal, the root of the mean of the square, cannot be computed without some averaging to obtain the mean function. The LTC1966 true RMS-to-DC converter utilizes a single capacitor on the output to do the low frequency averaging required for RMS-to-DC conversion. To give an accurate measure of a dynamic waveform, the averaging must take place over a sufficiently long interval to average, rather than track, the lowest frequency signals of interest. For a single averaging capacitor, the accuracy at low frequencies is depicted in Figure 6. However, if the output is examined on an oscilloscope with a very low frequency input, the incomplete averaging will be seen, and this ripple will be larger than the error depicted in Figure 6. Such an output is depicted in Figure7. The ripple is at twice the frequency of the input because of the computation of the square of the input. The typical values shown, 5% peak ripple with 0.05% DC error, occur with CAVE = 1F and fINPUT = 10Hz. If the application calls for the output of the LTC1966 to feed a sampling or Nyquist A/D converter (or other circuitry that will not average out this double frequency ripple) a larger averaging capacitor can be used. This trade-off is depicted in Figure 8. The peak ripple error can also be reduced by additional lowpass filtering after the LTC1966, but the simplest solution is to use a larger averaging capacitor. 1This frequency dependent error is in addition to the static errors that affect all readings and are therefore easy to trim or calibrate out. The Error Analyses section to follow discusses the effect of static error terms. ACTUAL OUTPUT WITH RIPPLE f = 2 x fINPUT OUTPUT Design Cookbook Figure 6 depicts the so-called DC error that results at a given combination of input frequency and filter capacitor values1. It is appropriate for most applications, in which the output is fed to a circuit with an inherently band limited frequency response, such as a dual slope/integrating A/D converter, a A/D converter or even a mechanical analog meter. IDEAL OUTPUT DC ERROR (0.05%) PEAK RIPPLE (5%) PEAK ERROR = DC ERROR + PEAK RIPPLE (5.05%) DC AVERAGE OF ACTUAL OUTPUT TIME 1966 F07 Figure 7. Output Ripple Exceeds DC Error 0 -0.2 C = 4.7F -0.4 DC ERROR (%) -0.6 C = 10F C = 2.2F -0.8 C = 1.0F C = 0.47F C = 0.1F C = 0.22F -1.0 -1.2 -1.4 -1.6 -1.8 -2.0 1 10 INPUT FREQUENCY (Hz) Figure 6. DC Error vs Input Frequency 20 50 60 100 1966 F06 1966fb 13 LTC1966 Applications Information 0 -0.2 PEAK ERROR (%) -0.4 C = 100F -0.6 -0.8 C = 47F -1.0 C = 22F C = 10F C = 2.2F C = 4.7F C = 1F -1.2 -1.4 -1.6 -1.8 -2.0 1 10 INPUT FREQUENCY (Hz) 20 Figure 8. Peak Error vs Input Frequency with One Cap Averaging A 1F capacitor is a good choice for many applications. The peak error at 50Hz/60Hz will be <1% and the DC error will be <0.1% with frequencies of 10Hz or more. Note that both Figure 6 and Figure 8 assume AC-coupled waveforms with a crest factor less than 2, such as sine waves or triangle waves. For higher crest factors and/or AC + DC waveforms, a larger CAVE will generally be required. See Crest Factor and AC + DC Waveforms. Capacitor Type Selection The LTC1966 can operate with many types of capacitors. The various types offer a wide array of sizes, tolerances, parasitics, package styles and costs. Ceramic chip capacitors offer low cost and small size, but are not recommended for critical applications. The value stability over voltage and temperature is poor with many types of ceramic dielectrics. This will not cause an RMS-to-DC accuracy problem except at low frequencies, where it can aggravate the effects discussed in the previous section. If a ceramic capacitor is used, it may be necessary to use a much higher nominal value in order to assure the low frequency accuracy desired. Another parasitic of ceramic capacitors is leakage, which is again dependent on voltage and particularly temperature. If the leakage is a constant current leak, the I * R drop of the leak multiplied by the output impedance of the LTC1966 will create a constant offset of the output voltage. If the leak is Ohmic, the resistor divider formed with the LTC1966 output impedance will cause a gain error. For < 0.1% gain accuracy degradation, the parallel impedance of the 14 50 60 100 1966 F08 capacitor leakage will need to be >1000 times the LTC1966 output impedance. Accuracy at this level can be hard to achieve with a ceramic capacitor, particularly with a large value of capacitance and at high temperature. For critical applications, a film capacitor, such as metalized polyester, will be a much better choice. Although more expensive, and larger for a given value, the value stability and low leakage make metal film capacitors a trouble free choice. With any type of capacitor, the self resonance of the capacitor can be an issue with the switched capacitor LTC1966. If the self resonant frequency of the averaging capacitor is 1MHz or less, a second smaller capacitor should be added in parallel to reduce the impedance seen by the LTC1966 output stage at high frequencies. A capacitor 100 times smaller than the averaging capacitor will typically be small enough to be a low cost ceramic with a high quality dielectric such as X7R or NPO/COG. Input Connections The LTC1966 input is differential and DC coupled. The LTC1966 responds to the RMS value of the differential voltage between Pin 2 and Pin 3, including the DC portion of that difference. However, there is no DC-coupled path from the inputs to ground. Therefore, at least one of the two inputs must be connected with a DC return path to ground. Both inputs must be connected to something. If either input is left floating, a zero volt output will result. 1966fb LTC1966 Applications Information For single-ended DC-coupled applications, simply connect one of the two inputs (they are interchangeable) to the signal, and the other to ground. This will work well for dual supply configurations, but for single supply configurations it will only work well for unipolar input signals. The LTC1966 input voltage range is from railto-rail, and when the input is driven above VDD or below VSS (ground for single supply operation) the gain and offset errors will increase substantially after just a few hundred millivolts of overdrive. Fortunately, most single supply circuits measuring a DCcoupled RMS value will include some reference voltage other than ground, and the second LTC1966 input can be connected to that point. on the coupling capacitor connected to the second input to follow the DC average of the input voltage. For differential input applications, connect the two inputs to the differential signal. If AC coupling is desired, one of the two inputs can be connected through a series capacitor. In all of these connections, to choose the input coupling capacitor, CC, calculate the low frequency coupling time constant desired, and divide by the LTC1966 differential input impedance. Because the LTC1966 input impedance is about 100 times its output impedance, this capacitor is typically much smaller than the output averaging capacitor. Its requirements are also much less stringent, and a ceramic chip capacitor will usually suffice. For single-ended AC-coupled applications, Figure 9 shows three alternate topologies. The first one, shown in Figure 9a uses a coupling capacitor to one input while the other is grounded. This will remove the DC voltage difference from the input to the LTC1966, and it will therefore not be part of the resulting output voltage. Again, this connection will work well with dual supply configurations, but in single supply configurations it will be necessary to raise the voltage on the grounded input to assure that the signal at the active input stays within the range of VSS to VDD. If there is already a suitable voltage reference available, connect the second input to that point. If not, a midsupply voltage can be created with two resistors as shown in Figure 9b. Output Connections The LTC1966 output is differentially, but not symmetrically, generated. That is to say, the RMS value that the LTC1966 computes will be generated on the output (Pin 5) relative to the output return (Pin 6), but these two pins are not interchangeable. For most applications, Pin 6 will be tied to ground (Pin 1), and this will result in the best accuracy. However, Pin 6 can be tied to any voltage between VSS (Pin 4) and VDD (Pin 7) less the maximum output voltage swing desired. This last restriction keeps VOUT itself (Pin 5) within the range of VSS to VDD. If a reference level other than ground is used, it should be a low impedance, both AC and DC, for proper operation of the LTC1966. Finally, if the input voltage is known to be between VSS and VDD, it can be AC-coupled by using the configuration shown in Figure 9c. Whereas the DC return path was provided through Pin 3 in Figures 9a and 9b, in this case, the return path is provided on Pin 2, through the input signal voltages. The switched capacitor action between the two input pins of the LTC1966 will cause the voltage Use of a voltage in the range of VDD - 1V to VDD - 1.3V can lead to errors due to the switch dynamics as the NMOS transistor is cut off. For this reason, it is recommended that OUT RTN = 0V if VDD is 3V. VDD CC VIN 2 3 VDD CC LTC1966 IN1 IN2 (9a) 3 VIN VSS VDD 2 VDD LTC1966 2 IN1 IN2 VDC R1 100k 3 VIN + - R2 100k (9b) LTC1966 IN1 IN2 CC 1966 F09 VSS OR GND (9c) Figure 9. Single-Ended AC-Coupled Input Connection Alternatives 1966fb 15 LTC1966 Applications Information 0 In any configuration, the averaging capacitor should be connected between Pins 5 and 6. The LTC1966 RMS DC output will be a positive voltage created at VOUT (Pin 5) with respect to OUT RTN (Pin 6). The LTC1966 is a switched capacitor device, and large transient power supply currents will be drawn as the switching occurs. For reliable operation, standard power supply bypassing must be included. For single supply operation, a 0.01F capacitor from VDD (Pin 7) to GND (Pin1) located close to the device will suffice. For dual supplies, add a second 0.01F capacitor from VSS (Pin 4) to GND (Pin 1), located close to the device. If there is a good quality ground plane available, the capacitors can go directly to that instead. Power supply bypass capacitors can, of course, be inexpensive ceramic types. The sampling clock of the LTC1966 operates at approximately 200kHz, and most operations repeat at a rate of 100kHz. If this internal clock becomes synchronized to a multiple or submultiple of the input frequency, significant conversion error could occur. This is particularly important when frequencies exceeding 10kHz can be injected into the LTC1966 via supply or ground bounce. To minimize this possibility, capacitive bypassing is recommended on both supplies with capacitors placed immediately adjacent to the LTC1966. For best results, the bypass capacitors should be separately routed from Pin 7 to Pin 1, and from Pin 4 to Pin 1. The LTC1966 needs at least 2.7V for its power supply, more for dual supply configurations. The range of allowable negative supply voltages (VSS) vs positive supply voltages (VDD) is shown in Figure 10. Mathematically, the VSS constraint is: - 3 * (VDD - 2.7V) VSS GND The LTC1966 has internal ESD absorption devices, which are referenced to the VDD and VSS supplies. For effective in-circuit ESD immunity, the VDD and VSS pins must be connected to a low external impedance. This can be accomplished with low impedance power planes or simply with the recommended 0.01F decoupling to ground on each supply. LTC1966 OPERATES IN THIS RANGE -2 VSS (V) Power Supply Bypassing -1 -3 -4 -5 -6 2.5 3 3.5 4 4.5 VDD (V) 5 5.5 1966 F10 Figure 10. VSS Limits vs VDD Up and Running! If you have followed along this far, you should have the LTC1966 up and running by now! Don't forget to enable the device by grounding Pin 8, or driving it with a logic low. Keep in mind that the LTC1966 output impedance is fairly high, and that even the standard 10M input impedance of a digital multimeter (DMM) or a 10x scope probe will load down the output enough to degrade its typical gain error of 0.1%. In the end application circuit, either a buffer or another component with an extremely high input impedance (such as a dual slope integrating ADC) should be used. For laboratory evaluation, it may suffice to use a bench top DMM with the ability to disconnect the 10M shunt. If you are still having trouble, it may be helpful to skip ahead a few pages and review the Troubleshooting Guide. What About Response Time? With a large value averaging capacitor, the LTC1966 can easily perform RMS-to-DC conversion on low frequency signals. It compares quite favorably in this regard to prior generation products because nothing about the circuitry is temperature sensitive. So the RMS result doesn't get distorted by signal driven thermal fluctuations like a log/antilog circuit output does. However, using large value capacitors results in a slow response time. Figure 11 shows the rising and falling step responses with a 1F averaging capacitor. Although they both appear at first glance to be standard exponential 1966fb 16 LTC1966 Applications Information 120 120 CAVE = 1F 100 LTC1966 OUTPUT (mV) LTC1966 OUTPUT (mV) 100 80 60 40 80 60 40 20 20 0 CAVE = 1F 0 0.1 0.2 0.3 TIME (SEC) 0.4 0 0.5 0 0.2 0.4 0.6 TIME (SEC) 0.8 1 1966 F11b 1966 F11a Figure 11a. LTC1966 Rising Edge with CAVE = 1F Figure 11b. LTC1966 Falling Edge with CAVE = 1F SETTLING ACCURACY (%) 10 C = 0.1F C = 0.22F C = 0.47F C = 1F C = 2.2F C = 4.7F C = 10F C = 22F C = 47F C = 100F 1 0.1 0.01 0.1 1 SETTLING TIME (SEC) 10 100 1966 F12 Figure 12. LTC1966 Settling Time with One Cap Averaging decay type settling, they are not. This is due to the nonlinear nature of an RMS-to-DC calculation. Also note the change in the time scale between the two; the rising edge is more than twice as fast to settle to a given accuracy. Again this is a necessary consequence of RMS-to-DC calculation.2 Although shown with a step change between 0mV and 100mV, the same response shapes will occur with the LTC1966 for ANY step size. This is in marked contrast to prior generation log/antilog RMS-to-DC converters, whose averaging time constants are dependent on the signal level, resulting in excruciatingly long waits for the output to go to zero. The shape of the rising and falling edges will be dependent on the total percent change in the step, but for less than the 100% changes shown in Figure 11, the responses will be less distorted and more like a standard exponential decay. For example, when the input amplitude is changed from 100mV to 110mV (+10%) and back (-10%), the step responses are essentially the same as a standard exponential rise and decay between those two levels. In such cases, the time constant of the decay will be in between that of the rising edge and falling edge cases of Figure 11. Therefore, the worst case is the falling edge response as it goes to zero, and it can be used as a design guide. Figure12 shows the settling accuracy vs settling time for a variety of averaging capacitor values. If the capacitor value previously selected (based on error requirements) gives an acceptable settling time, your design is done. 2To convince oneself of this necessity, consider a pulse train of 50% duty cycle between 0mV and 100mV. At very low frequencies, the LTC1966 will essentially track the input. But as the input frequency is increased, the average result will converge to the RMS value of the input. If the rise and fall characteristics were symmetrical, the output would converge to 50mV. In fact though, the RMS value of a 100mV DC-coupled 50% duty cycle pulse train is 70.71mV, which the asymmetrical rise and fall characteristics will converge to as the input frequency is increased. 1966fb 17 LTC1966 Applications Information But with 100F, the settling time to even 10% is a full 38 seconds, which is a long time to wait. What can be done about such a design? If the reason for choosing 100F is to keep the DC error with a 75mHz input less than 0.1%, the answer is: not much. The settling time to 1% of 76 seconds is just 5.7 cycles of this extremely low frequency. Averaging very low frequency signals takes a long time. However, if the reason for choosing 100F is to keep the peak error with a 10Hz input less than 0.05%, there is another way to achieve that result with a much improved settling time. Reducing Ripple with a Post Filter The output ripple is always much larger than the DC error, so filtering out the ripple can reduce the peak error substantially, without the large settling time penalty of simply increasing the averaging capacitor. Figure 13 shows a basic 2nd order post filter, for a net 3rd order filtering of the LTC1966 RMS calculation. It uses the 85k output impedance of the LTC1966 as the first resistor of a 3rd order Sallen-Key active RC filter. This topology features a buffered output, which can be desirable depending on the application. However, there are disadvantages to this topology, the first of which is that the op amp input voltage and current errors directly degrade the effective LTC1966 VOOS. The table inset in Figure 13 shows these errors for four of Linear Technology's op amps. C1 1F 5 LTC1966 6 R1 38.3k R2 169k CAVE 1F RB - + LT1880 A second disadvantage is that the op amp output has to operate over the same range as the LTC1966 output, including ground, which in single supply applications is the negative supply. Although the LTC1966 output will function fine just millivolts from the rail, most op amp output stages (and even some input stages) will not. There are at least two ways to address this. First of all, the op amp can be operated split supply if a negative supply is available. Just the op amp would need to do so; the LTC1966 can remain single supply. A second way to address this issue is to create a signal reference voltage a half volt or so above ground. This is most attractive when the circuitry that follows has a differential input, so that the tolerance of the signal reference is not a concern. To do this, tie all three ground symbols shown in Figure 13 to the signal reference, as well as to the differential return for the circuitry that follows. Figure 14 shows an alternative 2nd order post filter, for a net 3rd order filtering of the LTC1966 RMS calculation. It also uses the 85k output impedance of the LTC1966 as the first resistor of a 3rd order active RC filter, but this topology filters without buffering so that the op amp DC error characteristics do not affect the output. Although the output impedance of the LTC1966 is increased from 85k to 285k, this is not an issue with an extremely high input impedance load, such as a dual slope integrating ADC like the ICL7106. And it allows a generic op amp to be used, such as the SOT-23 one shown. Furthermore, it easily works on a single supply rail by tying the noninverting input of the op amp to a low noise reference as optionally shown. This reference will not change the DC voltage at the circuit output, although it does become the AC ground for the filter, thus the (relatively) low noise requirement. C2 0.1F 5 LTC1966 OP AMP LTC1966 VOOS VIOS IB/OS * R TOTAL OFFSET RB VALUE ISQ LT1494 375V 73V 648V 294k 1A LT1880 LT1077 LT2050 200V 150V 60V 3V 329V 329V 27V 679V 589V 230V SHORT 294k SHORT 1.2mA 48A 750A Figure 13. Buffered Post Filter 6 R1 200k CAVE 1F C1 0.22F R2 681k C2 0.22F - OTHER REF VOLTAGE, SEE TEXT + LT1782 1966 F13 1066 F14 Figure 14. DC Accurate Post Filter 1966fb 18 LTC1966 Applications Information Step Responses with a Post Filter Both of the post filters, shown in Figures 13 and 14, are optimized for additional filtering with clean step responses. The 85k output impedance of the LTC1966 working into a 1F capacitor forms a 1st order LPF with a -3dB frequency of ~1.8Hz. The two filters have 1F at the LTC1966 output for easy comparison with a 1F only case, and both have the same relative (Bessel-like) shape. However, because of the topological differences of pole placements between the various components within the two filters, the net effective bandwidth for Figure 13 is slightly higher (1.2 * 1.8 2.1Hz) than with 1F alone, while the bandwidth for Figure 14 is somewhat lower (0.7 * 1.8 1.3Hz) than with 1F alone. To adjust the bandwidth of either of them, simply scale all the capacitors by a common multiple, and leave the resistors unchanged. an issue with input frequency bursts at 50Hz or less, and even with the overshoot, the settling to a given level of accuracy improves due to the initial speedup. As predicted by Figure 6, the DC error with 1F is well under 1mV and is not noticeable at this scale. However, as predicted by Figure 8, the peak error with the ripple from a 10Hz input is much larger, in this case about 5mV. As can be clearly seen, the post filters reduce this ripple. Even the wider bandwidth of Figure 13's filter is seen to cut the ripple down substantially (to < 1mV) while the settling to 1% happens faster. With the narrower bandwidth of Figure 14's filter, the step response is somewhat slower, but the double frequency output ripple is just 180V. The step responses of the LTC1966 with 1F only and with the two post filters are shown in Figure 15. This is the rising edge RMS output response to a 10Hz input starting at t = 0. Although the falling edge response is the worst case for settling, the rising edge illustrates the ripple that these post filters are designed to address, so the rising edge makes for a better intuitive comparison. Figure 16 shows the step response of the same three cases with a burst of 60Hz rather than 10Hz. With 60Hz, the initial portion of the step response is free of the boost seen in Figure 15 and the two post filter responses have less than 1% overshoot. The 1F only case still has noticeable 120Hz ripple, but both filters have removed all detectable ripple on this scale. This is to be expected; the first order filter will reduce the ripple about 6:1 for a 6:1 change in frequency, while the third order filters will reduce the ripple about 63:1 or 216:1 for a 6:1 change in frequency. The initial rise of the LTC1966 will have enhanced slew rates with DC and very low frequency inputs due to saturation effects in the modulator. This is seen in Figure 15 in two ways. First, the 1F only output is seen to rise very quickly in the first 40ms. The second way this effect shows up is that the post filter outputs have a modest overshoot, on the order of 3mV to 4mV, or 3% to 4%. This is only Again, the two filter topologies have the same relative shape, so the step response and ripple filtering trade-offs of the two are the same, with the same performance of each possible with the other by scaling it accordingly. Figures 17 and 18 show the peak error vs. frequency for a selection of capacitors for the two different filter topologies. To keep the clean step response, scale all three capacitors INPUT BURST 0 200mV/ DIV 0 200mV/ DIV INPUT BURST 1F ONLY FIGURE 13 FIGURE 14 1F ONLY FIGURE 13 FIGURE 14 20mV/ DIV STEP RESPONSE 100ms/DIV 1966 F15 Figure 15. Step Responses with 10Hz Burst 0 20mV/ DIV STEP RESPONSE 100ms/DIV 1966 F16 0 Figure 16. Step Responses with 60Hz Burst 1966fb 19 LTC1966 Applications Information 0 -0.2 C = 10F PEAK ERROR (%) -0.4 -0.6 C = 4.7F -0.8 C = 2.2F C = 1.0F C = 0.47F C = 0.22F C = 0.1F -1.0 -1.2 -1.4 -1.6 -1.8 -2.0 10 INPUT FREQUENCY (Hz) 1 100 1966 F17 Figure 17. Peak Error vs Input Frequency with Buffered Post Filter 0 C = 10F -0.2 PEAK ERROR (%) -0.4 C = 4.7F -0.6 C = 2.2F C = 1.0F C = 0.47F C = 0.22F C = 0.1F -0.8 -1.0 -1.2 -1.4 -1.6 -1.8 -2.0 10 INPUT FREQUENCY (Hz) 1 Figure 18. Peak Error vs Input Frequency with DC Accurate Post Filter within the filter. Scaling the buffered topology of Figure 13 is simple because the capacitors are in a 10:1:10 ratio. Scaling the DC accurate topology of Figure 14 can be done with standard value capacitors; one decade of scaling is shown in Table 2. Table 2. One Decade of Capacitor Scaling for Figure 14 with EIA Standard Values CAVE C1 = C2 = 1F 0.22F 1.5F 0.33F 2.2F 0.47F 3.3F 0.68F 4.7F 1F 6.8F 1.5F 100 1966 F18 Figures 19 and 20 show the settling time versus settling accuracy for the buffered and DC accurate post filters, respectively. The different curves represent different scalings of the filters, as indicated by the CAVE value. These are comparable to the curves in Figure 12 (single capacitor case), with somewhat less settling time for the buffered post filter, and somewhat more settling time for the DC accurate post filter. These differences are due to the change in overall bandwidth as mentioned earlier. The other difference is the settling behavior of the filters below the 1% level. Unlike the case of a 1st order filter, any 3rd order filter can have overshoot and ringing. The filter designs presented here have minimal overshoot and ringing, but are somewhat sensitive to component mismatches. Even the 12% tolerance of the LTC1966 output impedance can be enough to cause some ringing. The dashed lines indicate what can happen when 5% capacitors and 1% resistors are used. 1966fb 20 LTC1966 Applications Information SETTLING ACCURACY (%) 10 C = 0.1F C = 0.22F C = 0.47F C = 1.0F C = 2.2F C = 4.7F C = 10F C = 22F C = 47F C = 100F 1 0.1 0.01 0.1 1 SETTLING TIME (SEC) 10 100 1066 F14 Figure 19. Settling Time with Buffered Post Filter SETTLING ACCURACY (%) 10 C = 0.1F C = 0.22F C = 0.47F C = 1.0F C = 2.2F C = 4.7F C = 10F C = 22F C = 47F C = 100F 1 0.1 0.01 0.1 1 SETTLING TIME (SEC) 10 Figure 20. Settling Time with DC Accurate Post Filter Although the settling times for the post filtered configurations shown on Figures 19 and 20 are not that much different from those with a single capacitor, the point of using a post filter is that the settling times are far better for a given level peak error. The filters dramatically reduce the low frequency averaging ripple with far less impact on settling time. Crest Factor and AC + DC Waveforms In the preceding discussion, the waveform was assumed to be AC-coupled, with a modest crest factor. Both assumptions ease the requirements for the averaging capacitor. With an AC-coupled sine wave, the calculation engine squares the input, so the averaging filter that follows is required to filter twice the input frequency, making its job easier. But with a sinewave that includes DC offset, the square of the input has frequency content 100 1066 F20 at the input frequency and the filter must average out that lower frequency. So with AC + DC waveforms, the required value for CAVE should be based on half of the lowest input frequency, using the same design curves presented in Figures 6, 8, 17 and 18. Crest factor, which is the peak to RMS ratio of a dynamic signal, also effects the required CAVE value. With a higher crest factor, more of the energy in the signal is concentrated into a smaller portion of the waveform, and the averaging has to ride out the long lull in signal activity. For busy waveforms, such as a sum of sine waves, ECG traces or SCR chopped sine waves, the required value for CAVE should be based on the lowest fundamental input frequency divided as such: fDESIGN = fINPUT(MIN) 3 * CF - 2 1966fb 21 LTC1966 Applications Information using the same design curves presented in Figures 6, 8, 17 and 18. For the worst-case of square top pulse trains, that are always either zero volts or the peak voltage, base the selection on the lowest fundamental input frequency divided by twice as much: fINPUT(MIN) fDESIGN = 6 * CF - 2 The effects of crest factor and DC offsets are cumulative. So for example, a 10% duty cycle pulse train from 0VPEAK to 1VPEAK (CF = 10 = 3.16) repeating at 16.67ms (60Hz) input is effectively only 30Hz due to the DC asymmetry and is effectively only: fDESIGN = 30 6 * 3.16 - 2 = 3.78Hz for the purposes of Figures 6, 8, 17 and 18. Obviously, the effect of crest factor is somewhat simplified above given the factor of 2 difference based on a subjective description of the waveform type. The results will vary somewhat based on actual crest factor and waveform dynamics and the type of filtering used. The above method is conservative for some cases and about right for others. The LTC1966 works well with signals whose crest factor is 4 or less. At higher crest factors, the internal modulator will saturate, and results will vary depending on the exact frequency, shape and (to a lesser extent) amplitude of the input waveform. The output voltage could be higher or lower than the actual RMS of the input signal. The modulator may also saturate when signals with crest factors less than 4 are used with insufficient averaging. This will only occur when the output droops to less than 1/4 of the input voltage peak. For instance, a DC-coupled pulse train with a crest factor of 4 has a duty cycle of 6.25% and a 1VPEAK input is 250mVRMS. If this input is 50Hz, repeating every 20ms, and CAVE = 1F, the output will droop during the inactive 93.75% of the waveform. This droop is calculated as: VMIN = INACTIVE TIME - VRMS 2 * ZOUT * CAVE e 1- 2 For the LTC1966, whose output impedance (ZOUT) is 85k, this droop works out to - 5.22%, so the output would be reduced to 237mV at the end of the inactive portion of the input. When the input signal again climbs to 1VPEAK, the peak/output ratio is 4.22. With CAVE = 10F, the droop is only -0.548% to 248.6mV and the peak/output ratio is just 4.022, which the LTC1966 has enough margin to handle without error. For crest factors less than 3.5, the selection of CAVE as previously described should be sufficient to avoid this droop and modulator saturation effect. But with crest factors above 3.5, the droop should also be checked for each design. Error Analyses Once the RMS-to-DC conversion circuit is working, it is time to take a step back and do an analysis of the accuracy of that conversion. The LTC1966 specifications include three basic static error terms, VOOS, VIOS and GAIN. The output offset is an error that simply adds to (or subtracts from) the voltage at the output. The conversion gain of the LTC1966 is nominally 1.000 VDCOUT/VRMSIN and the gain error reflects the extent to which this conversion gain is not perfectly unity. Both of these affect the results in a fairly obvious way. Input offset on the other hand, despite its conceptual simplicity, effects the output in a nonobvious way. As its name implies, it is a constant error voltage that adds directly with the input. And it is the sum of the input and VIOS that is RMS converted. This means that the effect of VIOS is warped by the nonlinear RMS conversion. With 0.2mV (typ) VIOS, and a 200mVRMS AC input, the RMS calculation will add the DC and AC terms in an RMS fashion and the effect is negligible: VOUT = (200mV AC)2 + (0.2mV DC)2 = 200.0001mV = 200mV + 1/2ppm 1966fb 22 LTC1966 Applications Information But with 10x less AC input, the error caused by VIOS is 100x larger: VOUT = (20mV AC)2 + (0.2mV DC)2 = 20.001mV = 20mV + 50ppm This phenomena, although small, is one source of the LTC1966's residual nonlinearity. On the other hand, if the input is DC-coupled, the input offset voltage adds directly. With +200mV and a +0.2mV VIOS, a 200.2mV output will result, an error of 0.1% or 1000ppm. With DC inputs, the error caused by VIOS can be positive or negative depending if the two have the same or opposing polarity. The total conversion error with a sine wave input using the typical values of the LTC1966 static errors is computed as follows: VOUT = ((500mV AC)2 + (0.2mV DC)2) * 1.001 + 0.1mV = 500.600mV = 500mV + 0.120% VOUT = ((50mV AC)2 + (0.2mV DC)2) * 1.001 + 0.1mV = 50.150mV = 50mV + 0.301% VOUT = ((5mV AC)2 + (0.2mV DC)2) * 1.001 + 0.1mV = 5.109mV = 5mV + 2.18% As can be seen, the gain term dominates with large inputs, while the offset terms become significant with smaller inputs. In fact, 5mV is the minimum RMS level needed to keep the LTC1966 calculation core functioning normally, so this represents the worst-case of usable input levels. Using the worst-case values of the LTC1966 static errors, the total conversion error is: VOUT = ((500mV AC)2 + (0.8mV DC)2) * 1.003 + 0.2mV = 501.70mV = 500mV + 0.340% VOUT = ((5mV AC)2 + (0.8mV DC)2) * 1.003 + 0.2mV = 5.279mV = 5mV + 5.57% These static error terms are in addition to dynamic error terms that depend on the input signal. See the Design Cookbook for a discussion of the DC conversion error with low frequency AC inputs. The LTC1966 bandwidth limitations cause additional errors with high frequency inputs. Another dynamic error is due to crest factor. The LTC1966 performance versus crest factor is shown in the Typical Performance Characteristics. Monotonicity and Linearity The LTC1966, like all implicit RMS-to-DC convertors (Figure 3), has a division with the output in the denominator. This works fine most of the time, but when the output is zero or near zero this becomes problematic. The LTC1966 has multiple switched capacitor amplifier stages, and depending on the different offsets and their polarity, the DC transfer curve near zero input can take a few different forms, as shown in the Typical Performance Characteristics graph titled DC Transfer Function Near Zero. Some units (about 1 of every 16) will even be well behaved with a transfer function that is the upper half of a unit rectangular hyperbola with a focal point on the y-axis of a few millivolts.3 For AC inputs, these units will have a monotonic transfer function all the way down to zero input. The LTC1966 is trimmed for offsets as small as practical, and the resulting behavior is the best statistical linearity provided the zero region troubles are avoided. It is possible, and even easy, to force the zero region to be well behaved at the price of additional (though predictable) VOOS and some linearity error. For large enough input signals, this linearity error may be negligible. 3In general, every LTC1966 will have a DC transfer function that is essentially a unit rectangular hyperbola (the gain is not always exactly unity, but the gain error is small) with an X- and Y- offset equal to VIOS and VOOS, respectively, until the inputs are small enough that the delta sigma section gets confused. While some units will be the north half of a north south pair, other units will have two upper halfs of the conjugate, east west, hyperbolas. The circuit of Figure 23 will assure a continuous transfer function. VOUT = ((50mV AC)2 + (0.8mV DC)2) * 1.003 + 0.2mV = 50.356mV = 50mV + 0.713% 1966fb 23 LTC1966 Applications Information LTC1966 IN2 10 5mV MIN ASYMPTOTES SHIFTED +2.5mV 5 OUTPUT 170k RMS TO DC CONVERSION 15 IINJECT DC CHARGE PUMP IN1 20 VOUT (mV DC) To do this, inject current into the output. As shown in Figure 21, the charge pump output impedance is 170k, with the computational feedback cutting the closed loop output impedance to the 85k specification. By injecting 30nA of current into this 170, with zero input, a 5mV offset IDEAL 0 CAVE 5 0 Figure 21. Behavioral Block Diagram of LTC1966 is created at the output feedback point, which is sufficient to overcome the 5mV minimum signal level. With large enough input signals, the computational feedback cuts the output impedance to 85k so the transfer function asymptotes will have an output offset of 2.5mV, as shown in Figure 22. This is the additional, predictable, VOOS that is added, and should be subtracted from the RMS results, either digitally, or by an analog means. 1966 F22b Figure 23 shows an analog implementation of this with the offset and gain errors corrected; only the slight, but necessary, degradation in nonlinearity remains. The circuit works by creating approximately 300mV of bias at the junction of the 10M resistors when the LTC1966's input/output are zero. The 10M resistor to the LTC1966 output therefore feeds in 30nA. The loading of this resistor causes a slight reduction in gain which is corrected, as is the nominal 2.5mV offset, by the LT1494 op amp. 20 5V 5V 750k 10M LTC1966 15 10M 85k VDD + OUT IN1 IN2 5mV MIN 10 - VSS GND EN 84.5k 5 ASYMPTOTES SHIFTED +2.5mV IDEAL -15 -10 0 5 -5 VIN (mV DC) 10 15 20 VOUT LT1494 LT1494 OUTRTN -5V 5V 10M 1F 0 -20 20 15 Figure 22b. AC Transfer Function with IINJECT = 30nA 1966 F21 VOUT (mV DC) 10 VIN (mV AC) 100pF -5V 1966 F23 Figure 23. Monotonic AC Response with Offset and Gain Corrected 1966 F22a Figure 22a. DC Transfer Function with IINJECT = 30nA 1966fb 24 LTC1966 Applications Information The two 10M resistors not connected to the supply can be any value as long as they match and the feed voltage is changed for 30nA injection. The op amp gain is only 1.00845, so the output is dominated by the LTC1966 RMS results, which keeps errors low. With the values shown, the resistors can be 2% and only introduce 170ppm of gain error. The 84.5k resistor is the closest match in the 1% EIA values but if the 2% EIA value of 82k were used instead, the gain would only be reduced by 248ppm. or more, IBIAS is usually only specified for maximum and this circuit needs a minimum of 30nA, therefore such an approach may not always work. This low error sensitivity is important because the LTC1966 output impedance is 85k 11.8%, which can create a gain error of 0.1%; enough to degrade the overall gain accuracy somewhat. This gain variation term is increased with lower value feed resistors, and decreased with higher value feed resistors. As mentioned in the Design Cookbook, the LTC1966 performs very well with low frequency and very low frequency inputs, provided a large enough averaging capacitor is used. A bigger error caused by the variation of the LTC1966 output impedance is imperfect cancelation of the output offset introduced by the injected current. The offset correction provided by the LT1494 will be based on a consistent 84.5k times the injected current, while the LTC1966 output impedance will vary enough that the output offset will have a 300V range about the nominal 2.5mV. If this level of output offset is not acceptable, either system calibration or a potentiometer in the LT1494 feedback may be needed. If the two 10M feed resistors to the LT1494 have significant mismatch, cancellation of the 2.5mV offset would be further impacted, so it is probably worth paying an extra penny or so for 1% resistors or even the better temperature stability of thin film devices. The 300mV feed voltage is not particularly critical because it is nominally cancelled, but the offset errors due to these resistance mismatches is scaled by that voltage. Note that the input bias current of the op amp used in Figure 23 is also nominally cancelled, but it will add or subtract to the total current injected into the LTC1966 output. With the 1nA IBIAS of the LT1494 this is negligible. While it is possible to eliminate the feed resistors by using an op amp with a PNP input stage whose IBIAS is 30nA Because the circuit of Figure 23 subtracts the offset created by the injected current, the LT1494 output with zero LTC1966 input will rest at +2.5mV, nominal before offsets, rather then the 5mV seen in Figure 22. Output Errors Versus Frequency However, the LTC1966 will have additional dynamic errors as the input frequency is increased. The LTC1966 is designed for high accuracy RMS-to-DC conversion of signals into the audible range. The input sampling amplifiers have a - 3dB frequency of 800kHz or so. However, the switched capacitor circuitry samples the inputs at a modest 100kHz nominal. The response versus frequency is depicted in the Typical Performance Characteristics titled Input Signal Bandwidth. Although there is a pattern to the response versus frequency that repeats every sample frequency, the errors are not overwhelming. This is because LTC1966 RMS calculation is inherently wideband, operating properly with minimal oversampling, or even undersampling, using several proprietary techniques to exploit the fact that the RMS value of an aliased signal is the same as the RMS value of the original signal. However, a fundamental feature of the modulator is that sample estimation noise is shaped such that minimal noise occurs with input frequencies much less than the sampling frequency, but such noise peaks when input frequency reaches half the sampling frequency. Fortunately the LTC1966 output averaging filter greatly reduces this error, but the RMS-to-DC topology frequency shifts the noise to low (baseband) frequencies. So with input frequencies above 5kHz to 10kHz, the output will slowly wander around a few percent. 1966fb 25 LTC1966 Applications Information Input Impedance The LTC1966 true RMS-to-DC converter utilizes a 2.5pF capacitor to sample the input at a nominal 100kHz sample frequency. This accounts for the 8M input impedance. See Figure 24 for the equivalent analog input circuit. Note however, that the 8M input impedance does not directly affect the input sampling accuracy. For instance, if a 100k source resistance is used to drive the LTC1966, the sampling action of the input stage will drag down the voltage seen at the input pins with small spikes at every sample clock edge as the sample capacitor is connected to be charged. The time constant of this combination is small, 2.5pF * 100k = 250ns, and during the 2.5s period devoted to sampling, ten time constants elapse. This allows each sample to settle to within 46ppm and it is these samples that are used to compute the RMS value. VDD IIN1 RSW (TYP) 6k IN1 CEQ 2.5pF (TYP) VDD VSS IIN2 RSW (TYP) 6k IN2 VSS ( )AVG = VIN1R-EQVIN2 I IN1 - VIN1 ( )AVG = VIN2REQ I IN2 REQ = 8M CEQ 2.5pF (TYP) 1966 F24 Figure 24. LTC1966 Equivalent Analog Input Circuit This is a much higher accuracy than the LTC1966 conversion limits, and far better than the accuracy computed via the simplistic resistive divider model: This resistive divider calculation does give the correct model of what voltage is seen at the input terminals by a parallel load averaged over a several clock cycles, which is what a large shunt capacitor will do--average the current spikes over several clock cycles. When high source impedances are used, care must be taken to minimize shunt capacitance at the LTC1966 input so as not to increase the settling time. Shunt capacitance of just 2.5pF will double the input settling time constant and the error in the above example grows from 46ppm to 0.67% (6700ppm). A 13pF scope probe will increase the error to almost 20%. As a consequence, it is important to not try to filter the input with large input capacitances unless driven by a low impedance. Keep time constant << 2.5s. When the LTC1966 is driven by op amp outputs, whose low DC impedance can be compromised by sharp capacitive load switching, a small series resistor may be added. A 10k resistor will easily settle with the 2.5pF input sampling capacitor to within 1ppm. These are important points to consider both during design and debug. During lab debug, and even production testing, a high value series resistor to any test point is advisable. Output Impedance The LTC1966 output impedance during operation is similarly due to a switched capacitor action. In this case, 59pF of on-chip capacitance operating at 100kHz translates into 170k. The closed loop RMS-to-DC calculation cuts that in half to the nominal 85k specified. In order to create a DC result, a large averaging capacitor is required. Capacitive loading and time constants are not an issue on the output. RIN RIN + RSOURCE 8M = VSOURCE 8M + 100k = VSOURCE - 1.25% VIN = VSOURCE 26 1966fb LTC1966 Applications Information However, resistive loading is an issue and the 10M impedance of a DMM or 10x scope probe will drag the output down by -0.85% typ. During shutdown, the switching action is halted and a fixed 30k resistor shunts VOUT to OUT RTN so that CAVE is discharged. Guard Ringing the Output The LTC1966's combination of precision and high output impedance can present challenges that make the use of a guard ring around the output a good idea for many applications. As mentioned above, a 10M resistive loading to ground will drag down the gain far more than the specificed gain tolerance. On a printed circuit board, contaminants from solder flux residue to finger grime can create parasitic resistances, which may be very high impedance, but can have deleterious effects on the realized accuracy. As an example, if the output (Pin 5) is routed near VSS (Pin 4) in a 5V application, a parasitic resistance of 1G (1,000M) is enough to introduce a -425V output offset error, more than the specified limit of the LTC1966 itself. Use of a guard ring, wherein the LTC1966 output node is completely surrounded by a low impedance voltage, can reduce leakage related errors substantially. The ground ring can be tied to OUTRTN (Pin 6) and should encircle the output (Pin 5), the averaging capacitor terminal, and the destination terminal at the ADC, filter op amp, or whatever else may be next. Figure 24a shows a sample PCB layout for the circuit of Figure 13, wherein the guard ring trace encloses R1, R2, and the terminals of C1, C2, and the op amp input connected to the high impedance LTC1966 Output. For the circuit of figure 14, the guard ring should enclose R1 and the terminals of C1 and C2, as well as the terminal at the ultimate destination. Figure 24b shows a sample PCB layout for the circuit of Figure 23. The summing node of the LT1494 has the same high impedance and high accuracy as the LTC1966 output, so here the guard ring encircles both of them. Any leakage between them is benign because the LT1494 forces them to the same nominal voltage. LTC1966 MS8 0.1F CAVE LTC1966 MS8 CAVE 1F LT1880 SO8 0.1F 1F LT1494 SO8 1966 F24b Figure 24b. PCB Layout of Figure 23 with Guard Ring 1966 F24a Figure 24a. PCB Layout of Figure 13 with Guard Ring 1966fb 27 LTC1966 Applications Information Interfacing with an ADC The LTC1966 output impedance and the RMS averaging ripple need to be considered when using an analog-todigital converter (ADC) to digitize the LTC1966 RMS result. The simplest configuration is to connect the LTC1966 directly to the input of a type 7106/7136 ADC as shown in Figure 25a. These devices are designed specifically for DVM/DPM use and include display drivers for a 3 1/2 digit LCD segmented display. Using a dual slope conversion, the input is sampled over a long integration window, which results in rejection of line frequency ripple when integration time is an integer number of line cycles. Finally, these parts have an input impedance in the G range, with specified input leakage of 10pA to 20pA. Such a leakage, combined with the LTC1966 output impedance, results in just 1V to 2V of additional output offset voltage. LTC1966 7106 TYPE OUTPUT OUT RTN 5 31 6 CAVE 30 IN HI IN LO 1966 F25a Figure 25a. Interfacing to DVM/DPM ADC LTC1966 LTC2420 1V OUTPUT OUT RTN 5 6 2 3 CAVE 4 VREF VIN SDO GND SCK CS 1966 F25b SERIAL DATA DIGITALLY CORRECT LOADING ERRORS Figure 25b. Interfacing to LTC2420 Another type of ADC that has inherent rejection of RMS averaging ripple is an oversampling . With most, but not all, of these devices, it is possible to connect the LTC1966 output directly to the converter input. Issues to look out for are the input impedance, and any input sampling currents. The input sampling currents drawn by ADCs often have large spikes of current with short durations that can confuse some op amps, but with the large CAVE needed by the LTC1966 these are not an issue. The average current is important, as it can create LTC1966 errors; if it is constant it will create an offset, while average currents that change with the voltage level create gain errors. Some converters run continuously, others only sample upon demand, and this will change the results in ways that need to be understood. The LTC1966 output impedance has a loose tolerance relative to the usual resistors and the same can be true for the input impedance of ADC, resulting in gain errors from part-to-part. The system calibration techniques described in the following section should be used in applications that demand tight tolerances. One example of driving an oversampling ADC is shown in Figure 25b. In this circuit, the LTC2420 is used with a 1V VREF. Since the LTC1966 output voltage range is about 1V, and the LTC2420 has a 12.5% extended input range, this configuration matches the two ranges with room to spare. The LTC2420 has an input impedance of 16.6M, resulting in a gain error of -0.4% to -0.6%. In fact, the LTC2420 DC input current is not zero at 0V, but rather at one half its reference, so both an output offset and a gain error will result. These errors will vary from part to part, but with a specific LTC1966 and LTC2420 combination, the errors will be fixed, varying less than 0.05% over temperature. So a system that has digital calibration can be quite accurate despite the nominal gain and offset error. With 20 bits of resolution, this part is more accurate than the LTC1966, but the extra resolution is helpful because it reduces nonlinearity at the LSB transitions as a digital gain correction is made. Furthermore, its small size and ease of use make it attractive. 1966fb 28 LTC1966 Applications Information As is shown in Figure 25b, where the LTC2420 is set to continuously convert by grounding the CS pin. The gain error will be less if CS is driven at a slower rate, however, the rate should either be consistent or at a rate low enough that the LTC1966 and its output capacitor have fully settled by the beginning of each conversion, so that the loading errors are consistent. Note that in this circuit, the input current of the LTC2420 is being used to assure monotonicity. The LTC2420 ZIN of 16.6M is effectively connected to half the reference voltage, so when the LTC1966 has zero signal, 500mV/16.6M = 30nA is provided. Alternatively, a 5V VREF can be used, but in this case the LTC1966 output span will only use 20% of the LTC2420's input voltage range. Furthermore, if the OUTRTN remains grounded, the injected current with zero signal will be 150nA, resulting in 5x the offset error and nonlinearity shown in Figure 22. In both of the circuits of Figure 25, a guard ring only has to encircle three terminals, the LTC1966 output, the top of the averaging capacitor, and the ADC input. Figure 26 shows the top copper patterns for example PCB layouts of each. The low power consumption of the LTC1966 makes it well suited for battery powered applications, and its slow output (DC) makes it an ideal candidate for a micropower ADC. LTC1966 MS8 ICL7106 MQFP CAVE 1f 1966 F26a LTC2420 SO8 LTC1966 MS8 CAVE 1f 1966 F26b Figure 26b. PCB Layout of Figure 25b with Guard Ring Figure 10 in Application Note 75, for instance, details a 10-bit ADC with a 35ms conversion time that uses just 29A of supply current. Such an ADC may also be of use within a 4mA to 20mA loop. Other types of ADCs sample the input signal once and perform a conversion on that one sample. With these ADCs (Nyquist ADCs), a post filter will be needed in most cases to reduce the peak error with low input frequencies. The DC accurate filter of Figure 14 is attractive from an error standpoint, but it increases the impedance at the ADC input. In most cases, the buffered post filter of Figure13 will be more appropriate for use with Nyquist analog-todigital converters. System Calibration The LTC1966 static accuracy can be improved with end system calibration. Traditionally, calibration has been done at the factory, or at a service depot only, typically using manually adjusted potentiometers. Increasingly, systems are being designed for electronic calibration where the accuracy corrections are implemented in digital code wherever possible, and with calibration DACs where necessary. Additionally, many systems are now designed for self calibration, in which the calibration occurs inside the machine, automatically without user intervention. Figure 26a. PCB Layout of Figure 25a with Guard Ring 1966fb 29 LTC1966 Applications Information Whatever calibration scheme is used, the linearity of the LTC1966 will improve the calibrated accuracy over that achievable with older log/antilog RMS-to-DC converters. Additionally, calibration using DC reference voltages are essentially as accurate with the LTC1966 as those using AC reference voltages. Older log/antilog RMS-to-DC converters required nonlinear input stages (rectifiers) whose linearity would typically render DC based calibration unworkable. The following are four suggested calibration methods. Implementations of the suggested adjustments are dependent on the system design, but in many cases, gain and output offset can be corrected in the digital domain, and will include the effect of all gains and offsets from the LTC1966 output through the ADC. Input offset voltage, on the other hand, will have to be corrected with adjustment to the actual analog input to the LTC1966. The methods below assume the unaltered linearity of the LTC1966, i.e. without the monotonicity fix of Figure 21. If this is present, the VOOS shift it introduces should be taken out before using either method for which VOOS is not calibrated. Also, the nonlinearity it introduces will increase the 20mV readings discussed below by 0.78% but increase the 200mV readings only 78ppm. There are a variety of ways to deal with these errors, including possibly ignoring them, but the specifics will depend on system requirements. Designers are cautioned to avoid the temptation to digitally take out the hyperbolic transfer function introduced because if the offsets are not exactly the nominals assumed, the system will end up right back where it began with a potential discontinuity with zero input, either from a divide by zero or from a square root of a negative number in the calculations to undo the hyperobic transfer function. An adaptive algorithm would most likely be necessary to safely take out more than half of the introduced nonlinearity. If a 5V reference is used in the connection of Figure 25b, the VOOS and nonlinearity created would be even larger, and will no doubt be more tempting to correct for. Designers are likewise cautioned against correcting for all of the nonlinearity. AC-Only, 1 Point The dominant error at full-scale will be caused by the gain error, and by applying a full-scale sine wave input, this error can be measured and corrected for. Unlike older log/antilog RMS-to-DC converters, the correction should be made for zero error at full scale to minimize errors throughout the dynamic range. The best frequency for the calibration signal is roughly ten times the - 0.1% DC error frequency. For 1F, -0.1% DC error occurs at 8Hz, so 80Hz is a good calibration frequency, although anywhere from 60Hz to 100Hz should suffice. The trade-off here is that on the one hand, the DC error is input frequency dependent, so a calibration signal frequency high enough to make the DC error negligible should be used. On the other hand, as low a frequency as can be used is best to avoid attenuation of the calibrated AC signal, either from parasitic RC loading or insufficient op amp gain. For instance, with a 1kHz calibration signal, a 1MHz op amp will typically only have 60dB of open loop gain, so it could attenuate the calibration signal a full 0.1%. AC-Only, 2 Point The next most significant error for AC-coupled applications will be the effect of output offset voltage, noticeable at the bottom end of the input scale. This too can be calibrated out if two measurements are made, one with a full-scale sine wave input and a second with a sine wave input (of the same frequency) at 10% of full-scale. The trade-off in selecting this second level is that it should be small enough that the gain error effect becomes small compared to the gain error effect at full-scale, while on the other hand, not using so small an input that the input offset voltage becomes an issue. 1966fb 30 LTC1966 Applications Information The calculations of the error terms for a 200mV full-scale case are: Gain = Reading at 200mV - Reading at 20mV 180mV Output Offset = Reading at 20mV - 20mV Gain DC, 2 Point DC based calibration is preferable in many cases because a DC voltage of known, good accuracy is easier to generate than such an AC calibration voltage. The only down side is that the LTC1966 input offset voltage plays a role. It is therefore suggested that a DC based calibration scheme check at least two points: full-scale. Applying the -fullscale input can be done by physically inverting the voltage or by applying the same +full-scale input to the opposite LTC1966 input. For an otherwise AC-coupled application, only the gain term may be worth correcting for, but for DC-coupled applications, the input offset voltage can also be calculated and corrected for. Note: Calculation of and correction for input offset voltage are the only way in which the two LTC1966 inputs (IN1, IN2) are distinguishable from each other. The calculation above assumes the standard definition of offset; that a positive offset is the case of a positive voltage error inside the device that must be corrected by applying a like negative voltage outside. The offset is referred to whichever pin is driven positive for the +full-scale reading. DC, 3 Point One more point is needed with a DC calibration scheme to determine output offset voltage: +10% of full scale. The calculation of the input offset is the same as for the 2point calibration above, while the gain and output offset are calculated for a 200mV full-scale case as: Gain = Reading at 200mV - Reading at 20mV 180mV Output Offset = Reading at 200mV +Reading at - 200mV - 400mV * Gain 2 The calculations of the error terms for a 200mV full-scale case are: Gain = Reading at 200mV + Reading at - 200mV 400mV Input Offset = Reading at - 200mV - Reading at 200mV 2 *Gain 1966fb 31 LTC1966 Applications Information Troubleshooting Guide Top Ten LTC1966 Application Mistakes 1.Circuit won't work-Dead On Arrival-no power drawn. - Probably forgot to enable the LTC1966 by pulling Pin8 low. 4.Gain is low by a few percent, along with other screwy results. - Probably tried to use output in a floating, differential manner. Solution: Tie Pin 6 to a low impedance. See Output Connections in the Design Cookbook. Solution: Tie Pin 8 to Pin 1. 2.Circuit won't work, but draws power. Zero or very little output, single-ended input application. - Probably didn't connect both input pins. Solution: Tie both inputs to something. See Input Connections in the Design Cookbook. IN1 3 IN2 1966 TS02 3.Screwy results, particularly with respect to linearity or high crest factors; differential input application. - Probably AC-coupled both input pins. Solution: Make at least one input DC-coupled. See Input Connections in the Design Cookbook. DC CONNECT ONE INPUT 2 LTC1966 3 IN2 OUT RTN 5 31 6 30 TYPE 7136 ADC HI LO Solution: Measure V IOS/VOOS by extrapolating readings > 5mVDC. 6. Linearity perceived to be out of specification particularly with small input signals. - This could again be due to using 0V in as one of the measurement points. Solution: Check Linearity from 5mVRMS to 500mVRMS. - The input offset voltage can cause small AC linearity errors at low input amplitudes as well. See Error Analyses section. DC CONNECT ONE INPUT 2 IN1 VOUT 5. Offsets perceived to be out of specification because 0V in 0V out. - The offsets are not specified at 0V in. No RMS-toDC converter works well at 0 due to a divide-by-zero calculation. LTC1966 NC LTC1966 1966 TS04 CONNECT PIN 3 2 GROUND PIN 6 Possible Solution: Include a trim for input offset. IN1 LTC1966 3 IN2 1966 TS03 1966fb 32 LTC1966 Applications Information 7.Output is noisy with >10kHz inputs. - This is a fundamental characteristic of this topology. The LTC1966 is designed to work very well with inputs of 1kHz or less. It works okay as high as 1MHz, but it is limited by aliased noise. Solution: Bandwidth limit the input or digitally filter the resulting output. 8.Large errors occur at crest factors approaching, but less than 4. - Insufficient averaging. Solution: Increase CAVE. See Crest Factor and AC + DC Waveforms section for discussion of output droop. 10.Gain is low by 1% or more, no other problems. - Probably due to circuit loading. With a DMM or a 10x scope probe, ZIN = 10M. The LTC1966 output is 85k, resulting in - 0.85% gain error. Output impedance is higher with the DC accurate post filter. Solution: Remove the shunt loading or buffer the output. - Loading can also be caused by cheap averaging capacitors. Solution: Use a high quality metal film capacitor for CAVE. 9. Screwy results, errors > spec limits, typically 1% to 5%. - High impedance (85k) and high accuracy (0.1%) require clean boards! Flux residue, finger grime, etc. all wreak havoc at this level. LOADING DRAGS DOWN GAIN Solution: Wash the board. LTC1966 Helpful Hint: Sensitivity to leakages can be reduced significantly through the use of guard traces. VOUT 85k OUT RTN mV 5 6 DCV 10M DMM 200mVRMS IN -0.85% KEEP BOARD CLEAN 1966 TS10 LTC1966 1966fb 33 LTC1966 Typical Applications 5V Supplies, Differential, DC-Coupled RMS-to-DC Converter 5V Single Supply, Differential, AC-Coupled RMS-to-DC Converter 5V 5V DC + AC INPUTS (1VPEAK DIFFERENTIAL) VDD VDD LTC1966 LTC1966 IN1 VOUT CAVE 1F IN2 OUT RTN AC INPUTS (1VPEAK DIFFERENTIAL) DC OUTPUT VSS GND EN IN1 VOUT IN2 OUT RTN CC 0.1F 1966 TA05 2.5V Supplies, Single Ended, DC-Coupled RMS-to-DC Converter with Shutdown 2.7V Single Supply, Single Ended, AC-Coupled RMS-to-DC Converter with Shutdown 2.7V/3V CMOS OFF ON 2V EN VDD OFF ON CC 0.1F VSS -2.5V -2V EN VDD LTC1966 IN1 0.1F X7R 2.5V 2.7V AC INPUT (1VPEAK) DC OUTPUT VSS GND EN 1966 TA03 -5V CAVE 1F VOUT CAVE 1F IN2 OUT RTN DC OUTPUT DC + AC INPUT (1VPEAK) GND LTC1966 IN1 VOUT CAVE 1F IN2 OUT RTN VSS GND -2.5V -2.5V DC OUTPUT 1966 TA04 1966 TA06 Battery Powered Single-Ended AC-Coupled RMS-to-DC Converter AC INPUT (1VPEAK) CC 0.1F 9V VDD LTC1966 IN1 GND LT1175CS8-5 SHDN VIN 0.1F X7R VOUT IN2 OUT RTN DC CAVE OUTPUT 1F VSS GND EN OUT SENSE 1966 TA07 1966fb 34 LTC1966 Simplified Schematic VDD C12 GND VSS C1 Y1 Y2 C2 IN1 2nd ORDER MODULATOR IN2 C3 C5 C7 + C9 + A1 C4 OUTPUT - A2 C8 CAVE C11 - OUT RTN 1966 SS C6 EN TO BIAS CONTROL C10 CLOSED DURING SHUTDOWN 30k BLEED RESISTOR FOR CAVE 1966fb 35 LTC1966 Package Description MS8 Package 8-Lead Plastic MSOP (Reference LTC DWG # 05-08-1660 Rev F) 3.00 0.102 (.118 .004) (NOTE 3) 0.889 0.127 (.035 .005) 5.23 (.206) MIN 0.254 (.010) 7 6 5 0.52 (.0205) REF 3.00 0.102 (.118 .004) (NOTE 4) 4.90 0.152 (.193 .006) DETAIL "A" 0 - 6 TYP GAUGE PLANE 3.20 - 3.45 (.126 - .136) 0.53 0.152 (.021 .006) DETAIL "A" 0.42 0.038 (.0165 .0015) TYP 8 0.65 (.0256) BSC 1 1.10 (.043) MAX 2 3 4 0.86 (.034) REF 0.18 (.007) RECOMMENDED SOLDER PAD LAYOUT NOTE: 1. DIMENSIONS IN MILLIMETER/(INCH) 2. DRAWING NOT TO SCALE 3. DIMENSION DOES NOT INCLUDE MOLD FLASH, PROTRUSIONS OR GATE BURRS. MOLD FLASH, PROTRUSIONS OR GATE BURRS SHALL NOT EXCEED 0.152mm (.006") PER SIDE 4. DIMENSION DOES NOT INCLUDE INTERLEAD FLASH OR PROTRUSIONS. INTERLEAD FLASH OR PROTRUSIONS SHALL NOT EXCEED 0.152mm (.006") PER SIDE 5. LEAD COPLANARITY (BOTTOM OF LEADS AFTER FORMING) SHALL BE 0.102mm (.004") MAX SEATING PLANE 0.22 - 0.38 (.009 - .015) TYP 0.65 (.0256) BSC 0.1016 0.0508 (.004 .002) MSOP (MS8) 0307 REV F 1966fb 36 LTC1966 Revision History (Revision history begins at Rev B) REV DATE DESCRIPTION B 5/11 Revised entire data sheet to add H- and MP- grades PAGE NUMBER 1 to 38 1966fb Information furnished by Linear Technology Corporation is believed to be accurate and reliable. However, no responsibility is assumed for its use. Linear Technology Corporation makes no representation that the interconnection of its circuits as described herein will not infringe on existing patent rights. 37 LTC1966 Typical Application RMS Noise Measurement 5V VOLTAGE NOISE IN 5V VDD + 100 10k 1/2 LTC6203 IN1 - 1mVDC 1VRMS NOISE VOUT CAVE 1F IN2 OUT RTN -5V 100 VSS GND EN 0.1F -5V 100k 1966 TA10 BW 1kHz TO 100kHz INPUT SENSITIVITY = 1VRMS TYP 1.5F AC CURRENT 71.2A MAX 50Hz TO 400Hz VOUT = LTC1966 70A Current Measurement Single Supply RMS Current Measurement 5V V+ LTC1966 IN1 T1 VOUT 10 CAVE 1F IN2 OUT RTN VOUT 4mVDC/ARMS AC CURRENT 71.2A MAX 50Hz TO 400Hz LTC1966 IN1 T1 CAVE 1F IN2 OUT RTN VSS GND EN T1: CR MAGNETICS CR8348-2500-N www.crmagnetics.com VOUT 10 VOUT = 4mVDC/ARMS 100k VSS GND EN -5V 1966 TA09 0.1F 1966 TA08 100k T1: CR MAGNETICS CR8348-2500-N www.crmagnetics.com Related Parts PART NUMBER DESCRIPTION COMMENTS LT(R)1077 Micropower, Single Supply Precision Op Amp 48A ISY, 60V VOS(MAX), 450pA IOS(MAX) LT1175-5 Negative, -5V Fixed, Micropower LDO Regulator 45A IQ, Available in SO-8 or SOT-223 LT1494 1.5A Max, Precision Rail-to-Rail I/O Op Amp 375V VOS(MAX), 100pA IOS(MAX) LT1782 General Purpose SOT-23 Rail-to-Rail Op Amp 40A ISY, 800V VOS(MAX), 2nA IOS(MAX) LT1880 SOT-23 Rail-to-Rail Output Precision Op Amp 1.2mA ISY, 150V VOS(MAX), 900pA IOS(MAX) LTC1967 Precision, Extended Bandwidth RMS to DC Converter 330A ISY, RMS Conversion to 4MHz LTC1968 Precision, Wide Bandwidth RMS to DC Converter 2.3mA ISY, RMS Conversion to 15MHz LTC2050 Zero Drift Op Amp in SOT-23 750A ISY, 3V VOS(MAX), 75pA IB(MAX) LT2178/LT2178A 17A Max, Single Supply Precision Dual Op Amp 14A ISY, 120V VOS(MAX), 350pA IOS(MAX) LTC2402 2-Channel, 24-bit, Micropower, No Latency TM ADC 200A ISY, 4ppm INL, 10ppm TUE LTC2420 20-bit, Micropower, No Latency ADC in SO-8 200A ISY, 8ppm INL, 16ppm TUE LTC2422 2-Channel, 20-bit, Micropower, No Latency ADC Dual Channel Version of LTC2420 1966fb 38 Linear Technology Corporation LT 0511 REV B * PRINTED IN USA 1630 McCarthy Blvd., Milpitas, CA 95035-7417 (408) 432-1900 FAX: (408) 434-0507 www.linear.com LINEAR TECHNOLOGY CORPORATION 2001