|Elliott Sound Products||AN-001|
Rod Elliott (ESP)
The two rules are as follows ...
- An opamp will attempt to make both inputs exactly the same voltage (via the feedback path)
- If it cannot achieve #1, the output will assume the polarity of the most positive input
These two rules describe everything an opamp does in any circuit, with no exceptions ... provided that the opamp is operating within its normal parameters. This means power supply voltage(s) must be within specifications, signal voltage is within the allowable range, and load impedance is equal to or greater than the minimum specified. The signal frequency must also be low enough to ensure that the opamp can perform normally for the chosen gain. For most cheap opamps, a gain of 100 with a frequency of 1kHz should be considered the maximum allowable, since the opamp's open loop gain may not be high enough to accommodate higher gain or frequency.
Armed with these rules and a basic understanding of Ohm's Law and analogue circuitry, it is possible to figure out what any opamp circuit will do under all normal operating conditions. Needless to say, the rules no longer apply if the opamp itself is faulty, or is operating outside its normal parameters (as discussed briefly above).
Typically, the precision rectifier is not commonly used to drive analogue meter movements, as there are usually much simpler methods to drive floating loads such as meters. Precision rectifiers are more common where there is some degree of post processing needed, feeding the side chain of compressors and limiters, or to drive digital meters.
There are several different types of precision rectifier, but before we look any further, it is necessary to explain what a precision rectifier actually is. In its simplest form, a half wave precision rectifier is implemented using an opamp, and includes the diode in the feedback loop. This effectively cancels the forward voltage drop of the diode, so very low level signals (well below the diode's forward voltage) can still be rectified with minimal error.
The most basic form is shown in Figure 1, and while it does work, it has some serious limitations. The main one is speed - it will not work well with high frequency signals. To understand the reason, we need to examine the circuit closely. This knowledge applies to all subsequent circuits, and explains the reason for the apparent complexity.
Figure 1 - Basic Precision Half Wave Rectifier
For a low frequency positive input signal, 100% negative feedback is applied when the diode conducts. The forward voltage is effectively removed by the feedback, and the inverting input follows the positive half of the input signal almost perfectly. When the input signal becomes negative, the opamp has no feedback at all, so the output pin of the opamp swings negative as far as it can. Assuming 15V supplies, that means perhaps -14V on the opamp output.
When the input signal becomes positive again, the opamp's output voltage will take a finite time to swing back to zero, then to forward bias the diode and produce an output. This time is determined by the opamp's slew rate, and even a very fast opamp will be limited to low frequencies - especially for low input levels. The test voltage for the waveforms shown was 20mV at 1kHz. Although the circuit does work very well, it is limited to relatively low frequencies (less than 10kHz) and only becomes acceptably linear above 10mV or so (opamp dependent).
Note the oscillation at the rectified output. This is (more or less) real, and was confirmed with an actual (as opposed to simulated) circuit. This is the result of the opamp becoming open-loop with negative inputs. In most cases it is not actually a problem.
Figure 2 - Rectified Output and Opamp Output
Figure 2 shows the output waveform (left) and the waveform at the opamp output (right). The recovery time is obvious on the rectified signal, but the real source of the problem is quite apparent from the huge voltage swing before the diode. While this is of little consequence for high level signals, it causes considerable nonlinearity for low levels, such as the 20mV signal used in these examples.
The circuit is improved by reconfiguration, as shown in Figure 3. The additional diode prevents the opamp's output from swinging to the negative supply rail, and low level linearity is improved dramatically. A 2mV (peak) signal is rectified with reasonably good linearity. Although the opamp still operates open-loop at the point where the input swings from positive to negative or vice versa, the range is limited by the diode and resistor. Recovery is therefore a great deal faster.
Figure 3 - Improved Precision Half Wave Rectifier
This circuit also has its limitations. The input impedance is now determined by the input resistor, and of course it is more complicated than the basic version. It must be driven from a low impedance source. Not quite as apparent, the Figure 3 circuit also has a defined output load resistance (equal to R2), so if this circuit were to be used for charging a capacitor, the cap will discharge through R2. Although it would seem that the same problem exists with the simple version as well, R2 (in Figure 1) can actually be omitted, thus preventing capacitor discharge. Likewise, the input resistor (R1) shown in Figure 1 is also optional, and is needed only if there is no DC path to ground.
Figure 4 - Precision Full Wave Rectifier
This circuit is sensitive to source impedance, so it is important to ensure that it is driven from a low impedance, such as an opamp buffer stage. Input impedance as shown is 6.66k, and any additional series resistance at the input will cause errors in the output signal. The input impedance is linear. As shown, and using TL072 opamps, the circuit of Figure 4 has good linearity down to a couple of mV at low frequencies, but has a limited high frequency response. Use of high speed diodes, lower resistance values and faster opamps is recommended if you need greater sensitivity and/ or higher frequencies.
Figure 5 - Original Analog Devices Circuit
It was pointed out in the original application note that the forward voltage drop for D2 (the FET) must be less than that for D1, although no reason was given. As it turns out, this may make a difference for very low level signals, but appears to make little or no difference for sensible levels (above 20mV or so).Simplified Alternative
Figure 6 - Simplified Version of the AD Circuit
It is virtually impossible to make a full wave precision rectifier any simpler, and the circuit shown will satisfy the majority of low frequency applications. Where very low levels are to be rectified, it is recommended that the signal be amplified first. While the use of Schottky (or germanium) diodes will improve low level and/or high frequency performance, it is unreasonable to expect perfect linearity from any rectifier circuit at extremely low levels. Operation up to 100kHz or more is possible by using fast opamps and diodes. R1 is optional, and is only needed if the source is AC coupled, so extremely high input impedance (with no non-linearity) is possible.
The simplified version shown above (Figure 6) is also found in a Burr-Brown application note .Another Version
Figure 6A - Another Version of the AD Circuit
While it initially looks completely different, that's simply because of the way it's drawn (I copied the drawing layout of the original). This version is interesting, in that the input is not only inverting, but provides the opportunity for the rectifier to have gain. The inverting input is of no consequence (it is a full wave rectifier after all), but it does mean that the input impedance is lower than normal ... although you could make all resistor values higher of course. Input impedance is equal to the value of R1, and is linear as long as the opamp is working well within its limits.
If R1 is made lower than R2 and R3, the circuit has gain. If R1 is higher than R2 and R3, the circuit can accept higher input voltages because it acts as an attenuator. For example, if R1 is 1k, the circuit has a gain of 10, and if 100k, the gain is 0.1 (an attenuation of 10). All normal opamp restrictions apply, so if a high gain is used frequency response will be affected.
One interesting result of using the inverting topology is that it enables the circuit to sum multiple inputs. R1 can be duplicated to give another input, and this can be extended. The original SSL circuit used two of these rectifiers with four inputs each. Remember that this is the same as operating the first opamp with a gain of four, so high frequency response may be affected.
Figure 7 - Original Intersil Precision Rectifier Circuit
During the positive cycle of the input, the signal is directly fed through the feedback network to the output. R3 actually consists of R3 itself, plus the set value of VR2. The nominal value of the pair is 15k, and VR2 can be usually be dispensed with if precision resistors are used (R3 and VR2 are replaced by a single 15k resistor).
This gives a transfer function of ...
Gain = 1 / ( 1 + (( R1 + R2 ) / R3 )) ... 0.5 with the values shown above
1V input will therefore give an output voltage of 0.5V. During this positive half-cycle of the input, the diode disconnects the op-amp output, which is at (or near) zero volts. Note that the application note shows a different gain equation which is incorrect. The equation shown above works.
During a negative half-cycle of the input signal, the CA3140 functions as a normal inverting amplifier with a gain equal to -( R2 / R1 ) ... 0.5 as shown. Since the inverting input is a virtual earth point, during a negative input it remains at or very near to zero volts. When the two gain equations are equal, the full wave output is symmetrical. Note that the output is not buffered, so the output should be connected only to high impedance stage, with an impedance much higher than R3.
Figure 8 - Modified Intersil Circuit Using Common Opamp
Where a simple, low output impedance precision rectifier is needed for low frequency signals (up to perhaps 10kHz as an upper limit), the simplified version above will do the job nicely. It does require an input voltage of at least 100mV, because there is no DC offset compensation. Because the LM358 is a dual opamp, the second half can be used as a buffer, providing a low output impedance. Minimum suggested input voltage is around 1V peak (710mV), which will give an average output voltage of 320mV. Higher input voltages will provide greater accuracy, but the maximum is a little under 10V RMS with a 15V DC supply as shown. The LM358 is not especially fast, but is readily available at low cost.
Limitations: Note that the input impedance of this rectifier topology is non-linear. The impedance presented to the driving circuit is 30k for positive half cycles, but only 10k for negative half-cycles. This means that it must be driven from a low impedance source - typically another opamp. This increases the overall complexity of the final circuit.
The circuit is interesting for a number of reasons, not the least being that it uses a completely different approach from most of the others shown. The rectifier is not in the main feedback loop like all the others shown, but uses an ideal diode (created by U1B and D1) at the non-inverting input, and this is outside the feedback loop.
Figure 9 - Burr-Brown Circuit Using Suggested Opamp
For a positive-going input signal, the opamp (U1A) can only function as a unity gain buffer, since both inputs are driven positive. Both the non-inverting and inverting inputs have an identical signal, a condition that can only be achieved if the output is also identical. If the output signal attempted to differ, that would cause an offset at the inverting input which the opamp will correct. It is worth remembering my opamp rules described at the beginning of this app. note.
For a negative-going input signal, The ideal diode (D1 and U2B) prevents the non-inverting input from being pulled below zero volts. Should this happen, the opamp can no longer function normally, because input voltages are outside normal operating conditions. The opamp (U1A) now functions as a unity gain inverting buffer, with the inverting input maintained at zero volts by the feedback loop. If -10uA flows in R1, the opamp will ensure that +10uA flows through R2, thereby maintaining the inverting input at 0V as required.
Limitations: Input impedance is non-linear, having an almost infinite impedance for positive half-cycles, and a 5k input impedance for negative half-cycles. The input must be driven from an earth (ground) referenced low impedance source. Capacitor coupled sources are especially problematical, because of the widely differing impedances for positive and negative going signals. The maximum resistance for a capacitor-coupled signal input is 100 ohms for the circuit as shown (one hundredth of the resistor values used for the circuit), and preferably less. The capacitance is selected for the lowest frequency of interest.
A simulation using TL072 opamps indicates that even with a tiny 5mV peak input signal (3.5mV RMS) the frequency response extends well past 10kHz but for low level signals serious amplitude non-linearity can be seen. The original article didn't even mention the rectifier, and no details were given at all. However, I have been able to determine the strengths and weaknesses by simulation. Additional weaknesses may show up in use of course.
Figure 10 - Simple Precision Full Wave Rectifier
One thing that is absolutely critical to the sensible operation of the circuit at low signal levels is that all diodes must be matched, and in excellent thermal contact with each other. The actual forward voltage of the diodes doesn't matter, but all must be identical. The lower signal level limit is determined by how well you match the diodes and how well they track each other with temperature changes.
The first stage allows the rectifier to have a high input impedance (R1 is 10k as an example only). Nominal gain as shown is 2 (short R3 for unity gain), but actual gain is reduced slightly by an amount that changes with level. 1V peak input gives 1.96V peak output, and 100mV gives 172mV out. The difference in each case is small and diminishes with reducing input level, but it's enough to mean that the circuit can't really be called a precision rectifier.
Limitations: Linearity is not very good. The circuit requires closely matched diodes for low level use because the diode voltage drops in the first stage (D1 & D2) are used to offset the voltage drops of D3 & D4. At input voltages of more than a volt or so, the n non-linearities are unlikely to cause a problem, but diode matching is still essential (IMO).
Simple capacitor smoothing cannot be used at the output because the output is direct from an opamp, so a separate integrator is needed to get a smooth DC output. This applies to most of the other circuits shown here as well and isn't a serious limitation.
One thing that became very apparent is that the circuit is very intolerant of stray capacitance, including capacitive loading at the output. Construction is therefore fairly critical, although adding a small cap (as shown in Figure 5) will help to some extent. I don't know why this circuit has not overtaken the 'standard' version in Figure 4, but that standard implementation still seems to be the default, despite its many limitations. Chief among these are the number of parts and the requirement for a low impedance source, which typically means another opamp. The impedance limitation does not exist in the alternative version, and it is far simpler.
The Intersil and Burr-Brown alternatives are useful, but both have low (and non-linear) input impedance. They do have the advantage of using a single supply, making both more suitable for battery operated equipment or along with logic circuitry. Remember that all versions (Figures 7, 8 & 9) must be driven from a low impedance source, and the Figure 7 circuit must also be followed by a buffer because it has a high output impedance.
In all, the Figure 6 circuit is the most useful. It is simple, has a very high (and linear) input impedance, low output impedance, and good linearity within the frequency limits of the opamps. The Figure 6A version is also useful, but has a lower input impedance and requires 2 additional resistors (R1 in Figure 6 is not needed if the signal is earth referenced).
The above circuits show just how many different circuits can be applied to perform (essentially) the same task. Each has advantages and limitations, and it is the responsibility of the designer to choose the topology that best suits the application. Not shown here, but just as real and important, is a software version. Digital signal processors (DSPs) are capable of rectification, conversion to RMS and almost anything else you may want to achieve, but are only applicable in a predominantly digital system.
|Copyright Notice.This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is Copyright © 2004 - 2009. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro-mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference while constructing the project. Commercial use is prohibited without express written authorisation from Rod Elliott. Referenced material is Copyright - see original material for details.|