feature

True RMS to DC conversion; 200 mV full scale; high precision laser trimming; maximum error 0.5% (AD636K); maximum error 1.0% (AD636J); wide range of response capabilities: calculate the rms value of AC and DC signals; 1 MHz–3 dB bandwidth: V rms > 100 mV; signal crest factor of 6 for 0.5% error; decibel output in 50 dB range; low power consumption: 800 mA quiescent current; single-supply or dual-supply operation; monolithic integration circuit; available in chip form at low cost.

Product Description

The AD636 is a low power monolithic integrated circuit for rms to dc conversion of low level signals. It provides performance at a higher cost than hybrid and modular converters that are comparable or superior. The AD636 is designed for a signal range of 0 mV to 200 mV rms. Crest factors up to 6 cans accommodate less than 0.5% additional error, allowing accurate measurement of complex input waveforms.

The AD636's low supply current requirement (typically 800 microamps) allows it to be used in battery-operated portable equipment. A variety of power supplies can be used, from ±2.5 V to ±16.5 V or a single +5 V to +24 V supply. The input and output terminals are fully protected; the input signal can exceed the power supply without damaging the device (allows the input signal to exist in the absence of supply voltage) and the output buffer amplifier is short-circuit protected.

The AD636 includes an auxiliary DB output. This signal is derived from an internal circuit point representing the logarithm of the rms output. The 0db reference level is supplied by the external current and is user selectable corresponding to from 0 dBm (774.6 mV) to –20 dBm (77.46 mV). The frequency response range is 1.2 MHz at 0 dBm level and over 10 kHz at -50 dBm. The AD636 is designed for ease of use. The device is factory trimmed at the wafer level for positive input and output offset negative waveform symmetry (DC reversal error) with a full-scale accuracy of 200 mV rms. Therefore, no exterior trim is required to achieve full rated accuracy. AD636 has two accuracy grades; AD636J has a total reading error of ±0.5 mV ±0.06%, AD636K

Accurate within ±0.2 mV to ±0.3% of reading. Both versions are specified over a temperature range of 0°C to +70°C and are available in sealed 14-pin dip or 10-pin lead to -100 metal cans. There are also chips.

Product Highlights

1. AD636 calculates the true root mean square of complex AC (or AC+DC) input signals and gives the equivalent DC output level. The true rms value of a waveform is a more useful quantity than the average rectified value because it is a measure of signal power. The rms value of an AC-coupled signal is also its standard deviation.

2. The AD636's 200-mV full-scale range is compatible with many popular display-oriented analog-to-digital converters. Low supply current requirements allow use in battery powered handheld instruments.

3. The only external component required to measure to the full specified accuracy is the averaging capacitor. The value of this capacitor can be selected based on requirements for low frequency accuracy, ripple and settling time.

4. On-chip buffer amplifiers can be used to buffer input or output. As an input buffer, it provides the precise performance of a standard 10 MΩ input attenuator. As an output buffer, it can provide up to 5mA of output current.

5. The AD636 will operate over a wide range of supply voltages, including single supply +5 V to +24 V or split supply ±2.5 V to ±16.5 V. A standard 9V battery will provide several hundred hours of continuous operation.

Standard connection

The AD636 is simple to connect for most high-accuracy rms measurements, requiring only an external capacitor to set the averaging time constant. Standard connections are shown in Figure 1. In this configuration, the AD636 will measure the rms value of the ac and dc levels at the input, but will show the error on the low frequency input as a function of the filter capacitor CAV, as shown in Figure 5. Therefore, if a 4µf capacitor is used, the additional average error is 0.1% at 10 Hz and 1% at 3 Hz. Accuracy at higher frequencies should be within specification. If a DC input needs to be rejected, a capacitor is placed in series with the input, as shown in Figure 3; the capacitor must be non-polar. If the AD636 is being driven by a supply with considerable high frequency ripple, it is recommended to bypass both supplies to ground as close as possible to the 0.1µf ceramic disk of the device. cf is an optional output ripple filter as described elsewhere in this datasheet.

Application of AD636

Input and output signal ranges are a function of supply voltage as detailed in the specification. The AD636 can also be used in unbuffered voltage output mode by disconnecting the input to the buffer. The output then appears unbuffered across a 10 kΩ resistor. Buffer amplifiers can be used for other purposes. Additionally, the AD636 can be used in current output mode by disconnecting the 10 kΩ resistor from ground. The output current is available at pin 8 (pin 10 on the H package) and is nominally scaled to 100 microamps per volt rms input, positive output.

High precision optional fine-tuning

If the accuracy of the AD636 needs to be improved, the trim shown in Figure 2 can be added. r4 is used to trim the offset. The scale factor is trimmed using r1 as shown. Insertion of r2 allows r1 to increase or decrease the scale factor by ±1.5%.

The trimming steps are as follows:

1. Ground the input signal, vehicle identification number (vin), and adjust r4 so that the output voltage of pin 6 is zero. Alternatively, r4 can be adjusted to give the correct output with the lowest expected value vin.

2. Connect the desired full-scale input level to the VIN (VIN), either a DC signal or a calibrated AC signal (1 kHz is the optimum frequency); then fine-tune r1 to give out the correct output from pin 6, i.e. 200 mV DC input should give 200 mV DC output. Of course, a ±200mV peak-to-peak sine wave should provide a DC output of 141.4mV. The residual errors given in the specification are due to nonlinearities.

Single power connection

The applications in Figures 1 and 2 assume dual power supplies. As shown in Figure 3, the AD636 can also be used with only one positive supply below +5 volts. Figure 3 is optimized for a 9-volt battery. The main limitation of this connection is that only AC signals can be measured because the input stage must be off ground to function properly. This biasing is done at pin 10; therefore, it is critical that no extraneous signals are coupled to this point. Biasing can be accomplished by using a resistive divider between +vs and ground. To reduce power dissipation, the resistor value can be increased since only 1µA of current flows into pin 10 (pin 2 on the "H" package). Alternatively, the COM pins of some CMOS ADCs provide a suitable artificial ground for the AD636. As shown, only capacitor C2 is required for AC input coupling; no DC link is required since a DC link is provided internally. C2 was chosen as an appropriate low frequency disconnect point, with an input resistance of 6.7 kΩ; for a disconnect at 10 Hz, C2 should be 3.3 μF. The signal range in this connection is slightly more limited than that in the dual power connection. The load resistor rl is necessary to provide the current sinking capability.

Choose Average Time Constant

The AD636 will calculate the rms of the ac and dc signals. If the input is a slowly varying DC voltage, the output of the AD636 will track the input accurately. At higher frequencies, the average output of the AD636 will approach the rms value of the input signal. As shown in Figure 4, the actual output of the AD636 will differ from the ideal output due to dc (or average) error and a certain amount of ripple.

The DC error depends on the frequency and cav value of the input signal. Figure 5 can be used to determine the minimum value of CAV that will produce a given % DC error above a given frequency using a standard rms connection.

The AC component of the output signal is ripple. There are two ways to reduce ripple. The first method is to use a large cav value. Since ripple is inversely proportional to cav, a 10x increase in capacitance will reduce the ripple by a factor of 10. When measuring waveforms with high crest factors (such as low duty cycle pulse trains), the average time constant should be at least 10 times the signal period. For example, a pulse rate of 100 Hz requires a time constant of 100 ms, which corresponds to a 4 μf capacitor (time constant = 25 ms/μf).

Standard rms connection

The main disadvantage of using large cavs to eliminate ripple is that the settling time for a step change in input level increases proportionally. Figure 5 shows the relationship between CAV and 1% settling time, which is 115 ms for each CAV microfarad. The settling time is twice as long for both decreasing and increasing signals (values in Figure 5 are for decreasing signals). As shown in Figure 6, the settling time also increases for low signal levels.

A better way to reduce output ripple is to use a "post filter". Figure 7 shows a suggested circuit. If a single-pole filter is used (C3 is removed, Rx is shorted), C2 is about 5 times the value of CAV, the ripple decreases, as shown in Figure 8, and the settling time increases. For example, when cav = 1µf and c2 = 4.7µf, the ripple on a 60Hz input is reduced from 10% of reading to about 0.3% of reading. However, the settling time is increased by about a factor of 3. Therefore, the values of cav and c2 can be lowered to allow faster settling times while still providing substantial ripple reduction.

A two-pole post filter uses an active filter stage to provide greater ripple reduction without significantly increasing settling time on circuits with single-pole filters. The values of cav, c2, and c3 can then be lowered to allow for extremely fast settling times of constant amount of ripple. Care should be taken when choosing the value of CAV because the DC error depends on this value and is independent of the post filter.

rms-to-dc conversion application guide, 2nd edition, available at:

RMS measurement

AD636 working principle

The AD636 contains an implicit solution to the rms equation, overcoming the dynamic range and other limitations inherent in directly computing the rms. The actual calculation performed by the AD636 follows the equation:

Figure 9 is a simplified schematic of the AD636; it is subdivided into four main parts: the absolute value circuit (active rectifier), the square/divider, the current mirror, and the buffer amplifier. The input voltage vin can be ac or dc, and is converted into a unipolar current i1 by the active rectifiers a1, a2. I1 drives one input of the square/divider, which has a transfer function:

The output current i4 of the squarer/divider drives the current mirror through a low pass filter formed by r1 and an externally connected capacitor cav. If the r1,cav time constant is much larger than the longest period of the input signal, then the effective average i4. The current mirror returns a current i3 equal to avg[i4], back to the square/divider to complete the implicit rms calculation. therefore:

The current mirror also produces an output current iout, which is equal to 2i4. IOUT can be used directly or converted to voltage with R2 and buffered with A4 to provide a low impedance voltage output. Therefore, the transfer function of the AD636 yields:

Since the voltage at this point is proportional to –log vin, the db output comes from the emitter of q3. The emitter follower, q5, buffer and level shift this voltage so that when the emitter current (iref) externally supplied to q5 is close to i3, the db output voltage is zero.

AD636 Buffer Amplifier

The buffer amplifier in the AD636 provides the user with additional application flexibility. It is important to understand some of the characteristics of this amplifier for optimum performance. Figure 10 shows a simplified schematic of the buffer.

Since the output of the rms-to-dc converter is always positive, it is not necessary to use a traditional complementary class AB output stage. In the AD636 buffer, use a class A transmitter follower instead. In addition to the excellent positive output voltage swing, this configuration allows the output to swing fully down to ground in single-supply applications without the problems associated with most integrated op amps.

When used as an input buffer amplifier driving load resistors related to ground, steps must be taken to ensure sufficient negative voltage swing. For negative outputs, current will flow from the load resistor through the 40 kΩ emitter resistor, setting a voltage divider between –vs and ground. This reduces the effective addition of an external resistor while re-changing the voltage divider, increasing the possibility of negative swing.

Figure 11 shows the rexternal values for a specific ratio of vpeak to –vs for some rload values. In addition, Rexternal increases the buffer amplifier's quiescent current equivalent to Rext/–vs. At –vs=–5v, the nominal buffer quiescent current without Rexternal is 30 µA.

Frequency response

The AD636 uses a logarithmic circuit when performing an implicit rms calculation. As with any logarithmic circuit, the bandwidth is proportional to the signal level. The solid line in the figure below represents the frequency response of the AD636 at input levels from 1 millivolt to 1 volt rms. The dashed lines indicate the upper frequency limits for additional errors of 1%, 10% and ±3db of reading. For example, note that a 1 volt rms signal will produce an additional error of less than 1% of the reading up to 220 kHz. A 10 mV signal can be measured with an additional error of 1% of reading (100 microvolts) up to 14 kHz.

AC Measurement Accuracy and Crest Factor

Crest factor is often overlooked when determining the accuracy of AC measurements. The crest factor is the ratio of the peak signal amplitude to the rms value of the signal (cf=vp/v rms). Most common waveforms, such as sine and triangle waves, have relatively low crest factors (<2). Waveforms resembling low duty cycle pulse trains, such as those found in switching power supplies and thyristor circuits, have high crest factors. For example, a rectangular pulse train with a duty cycle of 1% has a crest factor of 10 (cf=1η).

Figure 13 is a plot of the reading error for the AD636 with a 200 mV rms input signal, with a crest factor of 1 to 7. Since the rectangular pulse train (pulse width 200 μs) is the worst waveform measured by the rms (all energy is contained in the peak value), the rectangular pulse train is used for this experiment. The duty cycle and peak amplitude were varied while holding the 200 mV rms input amplitude constant to produce a crest factor from 1 to 7.

Full AC Digital Voltmeter

Figure 14 shows the design of a complete AD636-based low-power AC digital voltmeter circuit. The 10 MΩ input attenuator allows full-scale ranges of 200 mV, 2 V, 20 V, and 200 V rms. The signal is capacitively coupled to the AD636 buffer amplifier, which is connected in an AC bootstrap configuration to minimize loading. The buffer then drives the AD636's 6.7 kΩ input impedance. The COM terminal of the ADC chip provides the false ground required for the AD636's single-supply operation. The AD589 1.2-volt reference diode is used to provide a stable 100-mV reference to the ADC in linear rms mode; in DB mode, a 1N4148 diode is inserted in series to provide correction for the temperature coefficient of the DB scale factor. Calibration of the meter is done by first adjusting offset tank R17 for a correct zero reading, then adjusting R13 for an accurate full-scale reading.

Calibration of the db range is done by adjusting r9 for the desired 0db reference point and then r14 for the desired db scale factor (a scale of 10 counts per db is convenient). Using the Model 7106 ADC, the total supply current for this circuit is typically 2.8 mA.

Development of a Low Power Consumption and High Input Impedance Decibel Meter

The portable decibel meter circuit here combines the functions of an AD636 rms converter, AD589 voltage reference, and a μA776 low-power op amp. This meter offers excellent bandwidth and excellent high and low level accuracy while consuming the minimum power of a standard 9 volt transistor radio battery.

In this circuit, the built-in buffer amplifier of the AD636 is used

Circuit Description

The input voltage vin is AC coupled by c4, while resistor r8 provides high input voltage protection along with diodes d1 and d2.

The output pin 6 of the buffer is AC coupled to the input of the rms converter (pin 1) through capacitor C2. Resistor R9 is connected between the output of the buffer, the Class A output stage, and the negative output swing. Resistor r1 is the "bootstrap" resistor for the amplifier.

Using this circuit, single supply operation is possible by placing "ground" at a point between the positive and negative battery terminals. This is accomplished by sending 250 microamps from the positive battery terminal through resistor R2, then through the 1.2 volt AD589 bandgap reference, and finally back to the negative battery through resistor R10. This sets the ground to 1.2 volts + 3.18 volts (250 microamps x 12.7 kΩ) = 4.4 volts below battery positive and 5.0 volts on battery negative (250 microamps x 20 kΩ). Bypass capacitors C3 and C5 keep both sides of the battery low AC impedance to ground. The AD589 bandgap reference establishes a regulated reference voltage of 1.2 volts, which together with resistor R3 and trimmer potentiometer R4 sets a zero decibel reference current, IREF.

calibration

1. First, calibrate the zero decibel reference level by applying a 1khzsine wave from an audio oscillator at the desired zero decibel amplitude. This can be anywhere from zero dBm (770 mV rms – 2.2 volts pp) to –20 dBm (77 mV rms 220 mV – pp). Adjust the IREF calibration trimmer so that it shows zero on the analog meter.

2. The final step is to calibrate the meter scale factor or gain. Apply a 40dB input signal at the set zero-dB reference and adjust the scale factor calibration trimmer so that it reads 40µA on the analog meter.