Electronics - exception synchronous rectifier

Created; 28/08/2015, Changed; 23/11/2023, 25/03/2020

Previous page; Electronics - Electronics

In the previous page I showed examples of very cautious design in order to maximise the performance of an amplifier, potentially beyond what might be expected by reading the data sheets.  But it is not always desirable or necessary to protect the input and output of an operational amplifier to that degree when doing so would prohibit or compromise something else.  The design below is of a synchronous filter amplifier with the demodulated signal tapped and then output.  The benefit of chopper stabilised amplifier can be reduced DC offset and particularly thermal noise, but the trade-off is that synchronous noise verses switching noise that will inject a charge and therefore create an offset voltage error.

THERE IS NO EXCEPTION - but there are different things that can be traded off. 

(To come clean on my sensational title "Electronics - exception synchronous rectifier")

The example below, the choice of op-amp and analogue switch results in virtually no compromise anyway.

A synchronous detector like this is used to minimise low frequency noise such as thermal drift, ambient heat or light, in a situation where simple DC amplification those would be a problem.  This synchronous detector is also a good low frequency Amplitude demodulator or a Frequency demodulator and could be used as part of a phase locked loop can be used for narrowband filtering by increasing C1 thereby limiting the capture range.  The benefit potentially could be to achieve very narrowband pass filtering with lesser temperature coefficient, different rather than have fewer screening issues compared to a conventional a number of analogue filter stages that use large size timing capacitors which have their own electrostatic coupling screening issues.

Note that some capacitors have a mark to show the hot connection where the plate is on the outside of the capacitor - this should be connected to 0V or a low impedance point where possible.  It may be marked with a solid line. You can add a "+" or other mark and make it pin 1 say to the symbol on the circuit to identify it but also add a note to avoid ambiguity.

The draft circuit above has not been optimised but put down in order to consider the design issues; 

• The Operational-Amplifier is a fast settling type.  Analog Devices make a feature of this point for applications such as input to Analogue to Digital to converters in many of their op-amps.  A fast step response is required in this application.  The highlighted 80MHz Gain Bandwidth product is not the same as the figure given by other manufacturers, consequently, and that Gain Bandwidth Product figure for example falls as the gain is increased. 

• The op-amp also has a good low input bias current and offset.  And although the noise performance is specified at a high frequency (100KHz) generally FET and MOSFET amplifiers preform very well at low frequency's.  Take care, this amplifier is not protected from output being short-circuited, unlike most op-amps.

• I have also selected another Analog Devices part for the synchronous filter.  ADG1213 analogue switch has been selected for its particularly low charge injection of <1pC.

In addition, the above bridge like configuration both adds together some of the charge injection and to some degree may cancel some of it - which in practice should improve the thermal offset drift due to changes in charge injection with temperature, but this potential improvement is not very fully quantifiable from the datasheets - But an improvement by a factor of 5 may reasonably be expected though. 

• The capacitance directly connected to the input and output and any supply of the Op-amp due to the analogue switch section is less than; 3.6pF = 1pF + 2.6pF then the switches' resistance is in series with other capacitance.  This is safely much lower value than what most operational amplifier can tolerate.  The worst figure from experience with any op-amp is likely to be 100pF, but in this case, this op-amp should be loaded with less than 35pF. The design along with the resistive output stage should not be any problem, provide the PCB design is optimised for minimal track lengths.

• The instrument amplifier was selected for its low cost, but it has been necessary to select one that works with a lower supply voltage than is optimal for the op-amp and the analogue switch. It turns out that there is no design penalty - the input series resistors are required for filtering, but they are also provide input protection. 

• During the time that the output is switching there is a transient injected into the input; C1 + the resistance of the switch (2x 80 ohms) but no figure is given, but I have made a guess of 80R set a time constant that limit this.  During this transition, the amplitude of the pulse on the input will put both the amplifier input outside its linear range and would also reflect a pulse back into the input device (photo-diode) that will rectify it and create an offset error. But because the pulse is very short, the error will be a small percentage of the time.  This amplifier AD8034 is better than many, being fast at 80V/uS, 1pA Ib, 135nS overload recovery.

• The instrument amplifier has a gain of 10 so the AC signal will be small in the AC circuit around the op-amp.  Working with a small signal will give the best performance from the op-amp because the small signal performance of an op-amp is always better.  Additionally, component impedance can be larger and the transients smaller these things are all helpful to keep the time when the op-amp's characteristic's during the switching transition which are not controllable short to minimise the variability of the design.

The circuit is based on the circuit on the previous page with the photo-diode capacitance decoupled from the input/output by the series resistor R4.  The value of 470R is based on the output loading and effect on distortion given in the datasheet and is the lowest value quoted. 

A small photo-cell illuminated by a LED and involving some sort of light path and optical components may generate 1uA of signal. For best noise performance, once again the design operates with zero bias and is in any case <10mV. Significantly to develop 10mV across R4 would require 20uA signal and ambient light current and in this case although R3 this is unlikely to occur in this case.  Therefore, if it is useful, R3 could be relocated to across C1 or left as it is but C8 removed and linked out.

Decoupling is with 1nF capacitors in which are very good in the VHF range in which some of the switching current will reside.  There would also need to be lower frequency decoupling of say 100nF, but this does not have to be close to each IC's pins. 

C7 should be selected empirically for optimum amplifier stability - slightly damped as opposed to ringing at the switching transitions.  Significantly, that the output of the amplifier does not saturate at the supply rail.  Overload recovery is very good with this amplifier, but it also causes more time when the amplifier is not controlled by passive components and would be temperature sensitive. The optimum value of C7 will also depend on circuit layout and photo current.

The diode within the photo diode could cause adverse distortion in any switching transients reflected back into the photo diode.  The purpose of C7 is to eliminate that issue.  But adding another simple amplifier stage prior to this stage may be necessary and could be a disadvantaged if any switching noise were coupled in.   C7's value should be as low as possible without, but be high enough to keep the circuit both stable and not suffer big voltage spikes during the chopper switching times.   If the switches were implemented using discrete transistors, then including gate series resistors would be beneficial. 

The next diagram is based on the circuit above but rationalised but has a few more compromises, and they may be trivial;

Photo synchronous chopper amplifier demodulator (AL-0003-02C)

The above design is abstract because there is no real design objective, so there is no real worst case design carried out.  The optimal chopping frequency would be a trade-off between using a low frequency where the switching transients would be a small proportion of the output signal.  The effect of thermal noise when chopping frequency is 50Hz to 5KHz is likely to be in the region of best all over performance.  But this is a region whether mains frequency pick up is likely to be an issue. 

300Hz for example would filter both 50Hz or 60Hz but would suffer a low frequency beat if the mains frequency and the synchronising clock were mismatched.  But 330Hz would create a beat of about 5Hz this may be acceptable.  But the most important aspect of this design is that the chopped and most sensitive part of the circuit is very small in PCB area because the design is simple.  To maintain this advantage take care with layout of the wiring loom keep the chopped light signal away from the photodetector and also keep both wired signals away from the circuit board.

The chopped light would be best driven from a resistor in series with a simple push-pull driver such as a gate or CCD clock driver from a well regulated supply.  If you need to use an active current source (op-amp transistor switch and LED) this could introduce the biggest accuracy issue because of the circuit's settling time.  I have successfully used such an active current source to illuminate a CCD because the LED light intensity was the most important parameter and needed to be automatically adjusted.  The chopping in that case was fine and the CCD which worked at 8-10 bits performance anyway was not compromised.

Using the second op-amp as a buffer is not optimal as shown, and a series resistance must be added to the output for stability because this op-amp's output should not be loaded by more than 35pF.  By comparison, the instrumentation amplifier in the first circuit is stable with up to 1nF of output capacitance and remains fairly stable up to 10nF of output capacitance.

The circuits above on this page were drawn using CADSTAR on Windows, Output as PDF then imported to Gimp for Linux, cropped and exported as Jpeg.

--------------------------------------------------------------------------------

These circuits are a low accuracy, cheap implementation of what using an instrument amplifier IC would do much better.

Capacitive loading of an output and when it is necessary can be done by adding output series resistance;

• • Even though a general purpose op-amp has some transconductance and may be stabilised similarly by adding a 100uF capacitor (I have tried it with an op-amp similar to LM6142, and it does reduce susceptibility to noise similarly).  This should not be done under any circumstances, doing so would make a circuits' behaviour unpredictable.

  • To compensate a very low power op-amp, many of these become unstable with nearly impractically low capacitance on the output but become stable again with an output R+C network or simply higher value C on the output. ST.COM op-amp types such as; TSU101-2-4 and TSV711-2-4 are good but Microchip's cheaper op-amp types have good application notes explanation of the characteristic and how to work with it.

Using increased compensation is done in operational amplifiers used is power switch mode IC's where magnetic and electrostatic fields are high, and the circuit may not conveniently be screened from such fields.  That is why it is usually better to using suitable power supply controller ICs, and simple transistor amplification in which you can decouple every electrode of the transistor with NPO/COG capacitors.  The TL431 programmable reference is often used in such applications and is well compensated for such use without adding anything.

• Voltage regulators and references are all different and some have a minimum series resistance (decoupling capacitor's ESR for stability)

• A band gap voltage reference is stable with a capacitance of <10nF or >4uF depending on current in the case of TL431

To restate the points made in the previous page; 

I have repeated the point because I have seen that when these very precautions not taken that I speculate; 

• A washing machine sometimes stops before completion,

• A train toilet door that unlocks when the flush is pushed.

In these latter cases, a choke or a common mode choke or resistors and COG/NPO capacitors at the PCB boundary connection is most likely to solve the issue and with minimal cost. 

To discuss these electronics pages see; Blog page Electronics 

NEXT PAGE; Electronics - Noise Cancellation