Fullscreen HTML5Med Res (??MB)HTML4

Circuit Surgery Regular clinic by Ian Bell Rail-to-rail and single-supply op amps T his month’s topic is inspired by a question posted on the EEWeb Forum by Alexandru Radu, who wrote: ‘I want to design a rail-to-rail input and output op amp and I want to be sure I understand what it really does, so I will give a few examples. First of all, from what I understood, a regular op amp can’t really reach the maximum swing, but a rail-to-rail op amp can, even surpass it, but for simplicity let’s say it is exactly the supply voltage. Let VDD = 5V and VSS = −5V, and set up my op amp as an inverting amplifier. If my input signal is 1V, R2 = 5kΩ and R1 = 1kΩ, then my output will be exactly 5V, right? For a usual op amp it could be something like... 4.9 or 4.95... Is this correct? Thank you!’ Alexandru’s overview is basically along the right lines but there are plenty of details to discuss – we will look at railto-rail op amps and at the related topic of single-supply op amp circuits. Rails and swings Before getting into the details it is worth defining our terms and context. The word ‘rail’ refers to the power supply connection to a circuit, or the voltage of that power supply. Modern electronic systems often have many different supply voltages for different subsystems, but when we focus on a particular circuit within a system, such as an amplifier, we typically have two supply rails. A key aspect of some circuit designs is the relationship between the signal-voltage range and the supply voltages. The signal voltage range handled by a circuit is often referred to as the voltage ‘swing’, and is commonly described with respect to the supplies; for example, ‘the output can swing to within 1V of the rails’. Many circuits can only work with input signal voltages which fall inside the range defined by the two supply voltages and are also only able to output voltages within that range. If inputs exceed the supply range (in either direction) the circuit may not operate correctly or may be damaged, although it is not uncommon for op amps to work with inputs 42 a few hundred millivolts outside the supply range. For outputs, it is often physically impossible for the circuit to output voltages outside the supply range. There are of course exceptions, for example DC-DC converters output higher voltages, but our discussion here is focused on analogue signal-processing circuits built with op amps (such as amplifiers and filters) where the within-supply-range restrictions usually apply to output voltages. Fig.1. Simulation schematic to obtain example waveforms. The As Alexandru in- LT1001 is a ‘Precision Op Amp’ and the LT1366 is a ‘Precision d i c a t e s , t h e t e r m Rail-to-Rail Input and Output Op Amp’. ‘rail-to-rail’ can apply supply than a 15V one. Furthermore, as is to both the input and output voltages of often the case in engineering, improving op amps. These are separate capabilione aspect of a device may degrade ties – rail-to-rail input does not imply some other characteristics. If rail-torail-to-rail output, but many op amps rail is not a requirement then a suitable branded as rail-to-rail cover both input standard range op amp may provide and output. These voltages apply to the better performance. op amp itself – the input/output voltages of the circuit as a whole may be different. For op amp inputs it is the Rail-to-rail outputs common-mode voltage which is of imAlexandru mentions some example portance – more on this later. output voltages – although 4.9V on a In many situations the limited signal 5V supply is more rail-to-rail than a range of standard op amps is not an ‘usual’ op amp. In general, rail-to-rail issue because the signals never go to the output does not mean fully to the supply levels which cause problems. However, voltages – it is more of a marketing if operation close to the supplies is term to indicate capabilities beyond required then a special effort must the standard, not an exact specification. be made to design circuits which can Typically, for BJT (bipolar junction operate over a wider signal – so we have transistor) op amps, rail-to-rail outputs rail-to-rail op amps. A question that may can go to within a collector-emitter occur at this point is, why aren’t all op saturation voltage (VCEsat) of the supply. amps designed to be rail-to-rail from the VCEsat is dependent on the transistor’s start? One answer is that it is simply more collector current and hence the op amp’s difficult, and, in the past, it was less output current. For moderate currents likely to be an issue. Over the last two (in the mA range) VCEsat is typically 100 or three decades, supply voltages have to 300mV, so that is around 4.7 to 4.9V tended to reduce due to the effects of maximum for 5V a supply. For non-railadvances in semiconductor technology. to-rail op amps the output limits are If an op amp is limited to signals to typically 1V to 2V away from the supply. within 1V of the supply this is much For example, the outputs swing for the more likely to be a problem with a 3.3V venerable LM741 is ±12 to ±14V on a Practical Electronics | October | 2020 Fig.2. Positive peak of output waveforms from circuit in Fig.1 with Vin=0.96V. The green trace is ideal (−5 times the input voltage); the red trace is a standard op amp; the magenta trace is a rail-to-rail op amp swinging to 200mV below the 5V supply without distorting the signal. The rail-to-rail op amp can get closer to the supply than 200mV in the example circuit – but again, at the expense of distortion. This is shown in Fig.4, where the input has been increased to 1.0V (with the loads at the original 2.0kΩ). Ideally, the output should have a 5V peak – exactly at the supply. The railto-rail device clips the signal less than 100mV below the supply voltage, but the large distortion would be unacceptable in many situations. The waveforms in Fig.2 to Fig.4 indicates that that a simple figure of maximum output voltage may be insufficient when considering usable output range. The distortion produced by rail-to-rail op amps increases significantly as output levels reach a few hundred millivolts from the supply. At moderate loads, a limit of around 0.5V below the supply should avoid excessive distortion as a rule of thumb for many rail-to-rail devices, but it depends on output current and should be checked carefully if low distortion is important. Rail-to-rail inputs Fig.3. Same simulation as in Fig.2 but with the load resistors reduced to 500Ω. Both outputs peak at a lower voltage and there is now significant distortion from the rail-to-rail op amp. Fig.4. The same simulation as in Fig.1 but with Vin=1.0V. Compare with Fig.2, the rail-to-rail device’s output is within 100mV of the supply, but the waveform is distorted. ±15V supply with a 10kΩ load. Some op amps have built-in DC-DC converter circuits to internally generate higher voltages than the supply to the chip. Such devices can produce fully railto-rail outputs. Fig.1 is an LTspice simulation schematic for generating some illustrative waveforms. The circuit has two inverting amplifiers with a gain of 5 driven from the same input and operating on a ±5V supply. One amplifier uses an LT1001 ‘Precision Op Amp’ and the other an LT1366 ‘Precision Rail-to-Rail Input and Output Op Amp’ (these are arbitrary representative of each type). With the Practical Electronics | October | 2020 input at 0.96V peak, the output should be 4.8V peak. Fig.2 shows the response of the two circuits and an idealised output obtained by directly plotting −5 × vin. The rail-torail device successfully outputs the signal – the peak is 200mV from the supply, but the standard op amp clips the signal at over 1V below the supply. Fig.3 shows the effect of increasing output current. This is the same situation except with the load resistors (RL1 and RL2 in Fig.1) reduced to 500Ω. Here we see that both outputs limit at a lower voltage, resulting in a significant increase in distortion from the rail-to-rail op amp. Op amps have differential inputs and amplify the voltage difference between their two inputs (the inverting and noninverting inputs). In last month’s Circuit Surgery, we discussed the basic BJT differential amplifier – this circuit forms the basis of the first stage of BJT op amps, although there are many refinements and variations in commercial op amp designs. As discussed last month, when dealing with differential signals and amplifiers we must also consider the commonmode input voltage. Given the two input voltages are V1 and V2, the differential signal (which is amplified) is V1 − V2 and the common-mode voltage is the average voltage at the inputs (V1 + V2)/2. Op amps have very high gain and therefore in normal operation, in circuits such as amplifiers and filters, their differential input voltage is very small. Last month, we saw that the differential amplifier only provides linear amplification for differential inputs up to a few tens of millivolts. Large differential inputs of a few volts may damage the op amp by causing the base-emitter junctions of the input transistors to go into reverse breakdown (like a Zener diode). Some op amps have maximum differential input voltages, around 600700mV (or multiples thereof) due to the use of protection diodes to limit differential input voltage. Damage can still occur if currents through the protection diodes are not limited. Although exceeding a maximum differential input voltage of 600-700mV may seem difficult to avoid, it is actually unlikely to occur in standard op amp circuits with negative feedback 43 +Vsupply R 1 R 2 O ut In Q 1 In Q 2 Ibias –Vsupply Fig.5. Differential amplifier created with a pair of NPN BJTs. and it is often not an issue. However, some op amps have built-in resistors to limit the current and can withstand much larger differential inputs (eg, ±30V). The maximum differential input is not directly related to the supply voltage. When op amps are described as ‘having rail-to-rail input capability’ this refers to the common-mode input signal. As we saw last month, the differential amplifier is a symmetrical circuit whose ‘balance’, and hence differential output, is not affected by changing the common-mode input voltage. However, this assumes a common-mode voltage somewhere in the middle of the supply range. If the common-mode voltage gets close to the supplies the operation of the circuit may fundamentally change and the differential amplifier either stops working or delivers much reduced performance. Last month, we looked at the basic differential amplifier, as shown in Fig.5. In the context of this month’s discussion it is worth considering the common-mode input voltage range over which it will operate. In order for the input transistors to be conducting they need a base-emitter voltage (VBE) in the 0.6 to 0.7V range. If we assume the current source is the basic one we looked at last month, then there +Vsupply Ibias In Q 1 In Q 2 O ut R 3 R 4 –Vsupply Fig.6. Differential amplifier created with a pair of PNP BJTs. 44 is a single output transistor, +Vsupply whose collector-emitter voltage must be above the saturation voltage (VCEsat). R 1 R 2 Ibias1 The minimum voltage above the negative supply is the O ut sum of these two voltages, In Q 1 Q 2 In so it is typically around 1V. Q 4 Q 3 If more complex circuits, such as higher-performance current sources are used, the ‘stack’ of transistors between O ut the negative supply and input may be larger and R 3 R 4 have a bigger minimum voltage drop. To find the maximum Ibias2 common-mode input voltage –Vsupply (with no differential signal) we have to find the condition for the input transistors Fig.7. Rail-to-rail differential amplifier input stage – note the going into saturation (we use of both NPN and PNP BJTs. may change with common-mode input need their collector-emitter voltages to voltage depending on which differential be larger than VCEsat for acceptable circuit amplifier is active. performance). Working through the circuit from the input to Q1, its emitter is at Vin – VBE. The collector must be at least Single-supply op amps VCEsat above this at Vin – VBE + VCEsat, and Voltages are measured with respect the supply is at the voltage dropped by R1 to ground (0V) and often one of the supply rails will be ground – this is (IbiasR1) above the collector voltage. The particularly likely in digital circuits. supply is fixed, so we can write: However, analogue signals are often bipolar – they can take on both negative VSupply = Vin – VBE + VCEsat + IbiasR1 and positive values – which means that if one of the supplies is ground then part Rearranging this we can find the maximum of the signal (typically the negative half Vin is given by: with a positive supply) will be outside the supply range. For this reason, it is Vin = VSupply + VBE – VCEsat – IbiasR1 common for analogue signal-processing circuits to have a split power supply; Using a suitable choice of bias and R1, that is, two supply rails of equal and the voltage drop across R1 can be in (say) opposite voltages (for example +5V and the 0.2 to 0.3V range, which with VBE = – 5V, as in Fig.1). With split supplies, 0.7V, and VCEsat also in the 0.2 to 0.3V the signal, which is varying around 0V, range, means that Vin can be around 0.1 is at a voltage which is in the middle of to 0.3V above the supply voltage before the supply range. the transistor saturates. So, this circuit Split supplies provide a more can operate with common-mode voltages straightforward design scenario for the up to and just beyond the upper supply internal circuitry of op amps, but single rail, but not all the way to the lower rail. supplies have significant advantages in The same differential amplifier circuit terms of size and complexity of the power can be implemented using PNP transistors, circuits. Along with reduction in supply as shown in Fig.6. This reverses the voltages over the years, there has been relationship between the supplies and input – the input can go to, or beyond, the lower rail, but not to the upper rail. This +Vsupply makes the PNP circuit more suitable for R f single-supply op amps, where operation with the input at the lower rail (at 0V) C1 is often needed – more on these circuits R i Vin shortly. Rail-to-rail input op amps can be – Vout implemented using both NPN and PNP + differential amplifiers in the same circuit – together they cover the entire supply C2 range. The basic idea is shown in Fig.7, –Vsupply but this is a simplification. Circuits are needed to combine the output signals for the differential amplifiers. The op Fig.8. Op amp inverting amplifier with a amp’s characteristics (eg, offset voltage) split supply. Practical Electronics | October | 2020 +Vsupply Vin R f C1 R i – Vout + Fig.9. Op amp inverting amplifier with single supply. +Vsupply R R Vin C1 R 1 f C3 i – Vout + Virtual ground R 2 C2 Fig.10. Single-supply op amp inverting amplifier with pseudo/virtual ground and capacitive coupling. an increased use of single supplies. For circuits with a single supply it is not uncommon for it to be necessary for the circuit to operate with the input and/or output at 0V – this is equal to one of the supplies, so rail-to-rail type circuitry may be required for correct operation (at least for the ground side). It follows that not all op amps are suited to use with single supplies – at least if the applications are not limited to those only handling mid-range voltages. For this reason, op amp manufactures often make a point of stating when devices are suitable for single-supply operation. Fig.8 and Fig.9 show the same op amp circuit – a standard inverting amplifier of gain –Rf/Ri with two different power supply arrangements. Note the supply decoupling capacitors – you should always consult device datasheets for the specifics of what to use. Fig.8 shows a split supply with voltages of ±VSupply and Fig.9 shows a single supply of ground (0V) and +Vsupply R R 1 R Virtual ground A solution to the problem with the circuit in Fig.9 is to create a virtual or pseudo ground at half the single-supply voltage +Vsupply R f – R 1 Vout + Virtual ground +VSupply. With Ri = 1kΩ and Rf = 5kΩ the gain is 5 (as in Alexandru’s example and Fig.1). If we have an input signal which is a 0.5V peak sinewave centred on 0V then the circuit in Fig.8 will happily output a 0.5 × 5 = 2.5V peak sinewave, also centred on 0V. Unfortunately, the circuit in Fig.9 will not work because the amplifier is inverting, so a positive input signal should result in a negative output voltage – which is outside the supply range and not possible. Even a rail-to-rail op amp will not help here as the output would be required to go well outside the supply range. Op amps with negative feedback, such as the circuit in Fig.9, control their output voltage such that the input voltage difference is as close to zero as possible. Their very high gain means that the input voltage difference is very small and a ‘close to zero’ input difference is achieved. With negative feedback active, the op amp’s inputs behave almost like they are shorted together (a virtual short circuit). For the circuit in Fig.9 the noninverting input is wired to ground, so the inverting input will also be at 0V if the op amp is acting as a linear amplifier. In general, it is the signal at the inputs of the op amp itself – not of the whole circuit – which is what matters in terms of rail-to-rail capability. As discussed above, this will be very small, so if the op amp common-mode voltage is within the usable range the signal should not cause a problem. Therefore, if the op amp has rail-to-rail input capability, the circuit in Fig.9 may be able to operate correctly in terms of the input, but as already noted it will not be able to output the full waveform of an AC waveform centred on 0V. If the op-amp is not designed for single-supply use with commonmode input at the negative rail then the circuit will not work at all. The exact non-functional behaviour will vary with different op amp types. 2 Fig.11. DC equivalent circuit for Fig.10. Practical Electronics | October | 2020 – f Vout + Virtual ground R 2 R i Fig.12. DC equivalent circuit for Fig.10 with C1 removed (DC input) and capacitively couple the input signal. This is shown in Fig.10, in which the virtual ground is produced by the two equal resistors R1 and R2 and the capacitor C2. R1 and R2 typically have values in the tens to hundreds of kilohms range. Higher values reduce the drain on the supply and the op amp should not need to take large currents from the divider, so low values should not be needed. C2 plays a similar role to a supply decoupling capacitor and reduces noise on the virtual ground. Virtual grounds like this can degrade performance and more sophisticated circuits can be used – for example, buffering the potential divider with a unity-gain op amp amplifier. Texas instruments make a ‘rail splitter’ precision virtual ground IC, the TLE2426. C1 in the circuit in Fig.10 removes the DC component from the input (typically 0V) and just allows the AC signal through to the op amp. Another effect of C 1 blocking DC is in the relationship between the virtual ground voltage and the output. Fig.11 shows the DC equivalent circuit for Fig.10 – this is obtained by replacing the capacitors with open circuits. With C1 open, Ri is disconnected at one end, so is completely removed. For Fig.11, we see that as far as DC is concerned the op amp has 100% negative feedback. The circuit effectively has the virtual ground connected to a unity-gain buffer. Thus, the DC output of the op amp is 1× the virtual ground voltage. Using equal resistors (as indicated above) will give half the supply as both the virtual ground and the output with no signal present. If an output centred on 0V is required, then an output coupling capacitor can be used. AC coupling the input makes the virtual ground straightforward to use. If we don’t use C1 in Fig.10 then we can amplify a DC input, but the virtual ground setup is more complex. A DC equivalent circuit of Fig.10 with C1 removed and the input connected to a source with 0V DC under no signal conditions is shown in Fig.12. The circuit is now a non-inverting amplifier as far as the virtual ground is concerned. With an inverting gain of −5 the non-inverting gain is 6, so the virtual ground needs to be at 417mV to give 2.5V DC at the output with no signal – hence different resistor values are needed in the potential divider. This illustrates the fact that single-supply op amp circuits can be more difficult to work with than traditional split-supply versions. Simulation files Most, but not every month, LTSpice is used to support descriptions and analysis in Circuit Surgery. The examples and files are available for download from the PE website. 45