Electronic Design
 
Product Locator Letail
Product Locator >> Semiconductors >> Amplifiers >> Instrumentation Amplifiers

Reprints   Email this Article    RSS        Font Size     What's This?


Basics and Pitfalls of Difference Amplifiers and IAs
August 2003 Issue
Published Date: August 01, 2003

By Gert N. Helles, Maxim Integrated Products

The term instrumentation amplifier (IA) denotes a high-gain, dc-coupled amplifier with differential inputs, single-ended output, high input impedance, and high common-mode rejection (CMR)—by its description, an ideal building block (Figure 1). The following article covers the basics of IAs and difference amplifiers and explores the pitfalls you may encounter in applying IAs based on one, two, or three op amps.

Difference amplifiers and instrumentation amplifiers use the same basic building blocks (op amps), but the performance of IAs and "diff amps" differs greatly from that of standard op amps. Op amps are single-ended devices that operate in applications for which the function is determined mainly by the feedback network. Difference amplifiers and IAs, on the other hand, amplify small differential voltages in the presence of common-mode signals, provide very high CMR, and, normally, do not employ external feedback from output to input.

If you ever consider building your own IA from discrete components, be prepared to spend a lot of time and effort on the design. The easy, convenient, and proven alternative is an integrated IA or differential amplifier. For now, consider the IA as a black box that amplifies small differential inputs while providing high common-mode rejection.

To better understand IAs and why CMR is important, consider the common Wheatstone Bridge transducer of Figure 2, in which R1 = R2 = R3 = R4 = 5 KW, and the excitation voltage (Vex) = 10V.

Evaluating the bridge under no-load conditions leads to:

V1 = Vex[R2/(R2+R1)] >> V1 = 5V,

V2 = Vex[R3/(R3+R4)] >> V2 = 5V, therefore

ΔV = V1-V2 = 5V-5V = 0.

The transducer output is the voltage difference (ΔV) across the bridge. Assume a stimulus is applied to the four active arms of the transducer, in which R1 and R4 increase in value while R2 and R3 decrease in value. Also assume that R1 = R4 = 5001 and R2 = R3 = 4999.

Here, Vex = 10V >> V1 = 5.001V, V2 = 4.999V and the signal of interest (ΔV) = V1-V2 = 2 mV. A calculation of CMV shows that even though we unbalanced the transducer, CMV still equals (V1+V2)/2 = 5V. Ideally, this application's output is Vo = ΔV x gain.

The equations above illustrate very well the difficult task of measuring a small ΔV in the presence of relatively large common-mode signals. The ΔV (in millivolts) is measured by subtracting one large voltage (V1) from another (V2), both measured in volts.

Errors
The first "ratiometric" measurements were performed using galvanometers. Because those instruments floated at the common-mode voltage of the Wheatstone bridge, they (unlike an IA) did not suffer from common-mode voltages. The example of Figure 2 is illustrative, but does not include error sources. Significant system errors in the real world include CMR, PSR, Vos, Ib, and Ios (see Figure 3):

Common Mode Rejection (CMR) describes a shift in the amplifier's input offset voltage due to the applied CMV, with CMR defined as being 20log(CMV/ΔVos). Consider this: What level of CMR is required to obtain an error smaller than 1% on the differential signal from the Wheatstone bridge in Figure 2?

VCM = 5V, ΔV = 2mV >> ΔVos = 1% of ΔV = 0.02 mV.

CMR = 20log(CMV/ΔVos) >>

CMR = 108 dB.

A CMV below 0.1% requires a CMR of 128 dB!

With discrete op amps and careful trimming, you can achieve CMRs of 100 dB or better. Unfortunately, the inductance of wirewound resistors required to obtain such high stability causes degradation of the CMR over frequency.

Power Supply Rejection (PSR) describes a shift in amplifier input offset due to a change in the applied power supply voltage, with PSR defined as being 20log(ΔVos/ΔVsupply).

Vos, ΔVos/Δtemp, Ib, and Ios errors are handled in the same way as for a traditional op amp. As a minimum, one must consider for every error analysis the temperature range, source resistance, power supply regulation, common-mode input voltage, gain, Vos, ΔVos, Ib, and Ios. For worst-case analyses, you should always use min/max values from the data sheet, but in some cases you can use typical values, because all parameters are unlikely to assume their limiting values at the same time.

Consider an error analysis based on Figure 3, in which the following worst-case conditions apply to the differential instrumentation amplifier:

CMR = 74 dB

PSR = 90 dB

Vos = 500 mV

ΔV/ΔT = 20 mV/°C,

The following conditions apply to the application:

Common mode voltage: 2.5V

Power supply: 5V ±10%

Temperature range: 0°C-50°C.

To calculate errors, we apply the superposition approach:

CMR error = 20log(ΔVos/VCM) >> 200 mV

PSR Error = 20log(ΔVos/ΔVsupply) >> 16 mV

Vos = 500 mV (can easily be trimmed to a smaller value)

ΔVos/ΔT (25°C) = 5 mV x 25°C = 125 mV

The total error budget is therefore 200 mV + 16 mV + 500 mV + 12 5mV = 841 mV. Keep in mind that source resistance also influences the differential amplifier's CMR. The main contribution, however, is Vos (500 mV). And that error can easily be trimmed out, thereby leaving a total error of 341 mV. Again, any impedance added to the REF pin degrades the CMR!

IA topologies
Having reviewed IA basics, we now consider the three main topologies with which IAs are configured: the differential amplifier, the 3-op-amp IA, and the 2-op-amp IA. The simplest IA is the difference amplifier, which in its simplest form consists of one op amp and four resistors.

As an example, the MAX4198 and MAX4199 are precision, low-power differential amplifiers with factory-trimmed internal gain-setting resistors. The MAX4198 is trimmed to a gain of +1V/V and the MAX4199 is trimmed to a gain of +10V/V.

The functional diagram for these ICs shows their internal structure (Figure 4). Although the resistors are precision matched, their absolute values may vary by ±25%.

For the MAX4198, typical input impedances are 50 KW for the non-inverting input and 25 KW for the inverting input. Typical input impedances for the MAX4199 are 275 KW for the non-inverting input and 25 KW for the inverting input. Common-mode inputs for the internal op amp can range from VEE to (VCC − 1.1V).

The internal op amp inputs are not rail-to-rail, but internal resistors for the MAX4198 form a voltage divider that extends the input common-mode range beyond the rails. With VCC = 5V, the MAX4198 input range extends 100 mV beyond the rails without causing an adverse effect on common-mode rejection or phase reversals from input to output. The MAX4199 input common-mode range extends from 100 mV below the negative rail to (VCC − 1V).

The simplified equation for a standard difference amplifier is Vo = Vb−Va, which requires that R1 = R2 = R3 = R4. Any mismatch in these resistor values causes a degradation in the CMR. Table 1 shows the CMR achievable for a given tolerance on the resistors.

Though widely used, the discrete difference amplifier circuit has major disadvantages:

  • Input resistance is equal to R1 and is relatively low;
  • Input resistances usually differ greatly;
  • Resistors must be very well matched to obtain an acceptable common-mode rejection ratio;
  • Input impedance differences degrade the CMR at higher frequencies;
  • CMR is affected by the signal source impedance.

The Classic 3-Op-Amp IA
The work-around to remove limitations of the signal-source impedance (while also providing a convenient way to change the gain setting) is to add two buffers and three resistors to the difference amplifier. Figures 5 and 6 illustrate two variants of this topology. CMR still depends on the matching of R1, R2, R3, and R4 (as for the differential amplifier), but it does not depend on the matching of R1 and RG. The input stage (A1, A2) amplifies only the ΔV difference, with a gain of (1+ R1/RG).

The MAX4194­MAX4197 family of low-power IAs implements a 3-amplifier topology (Figure 5). The input stage consists of two op amps, which together provide a fixed-gain differential and a unity common-mode gain. The output stage is a conventional differential amplifier that provides an overall common-mode rejection of 115 dB (G = +10V/V). The MAX4194's gain is externally set between +1V/V and +10,000V/V (see Table 2). MAX4195/MAX4196/MAX4197 devices have internal gain-setting resistors (Figure 6), with fixed gains of +1V/V, +10V/V, and +100V/V, respectively.

The common-mode input range for all of these amplifiers is (VEE+0.2V) to (VCC−1.1V). Ideally, an IA (Figure 7) responds only to differential voltages applied to the inputs IN+ and IN-. With both inputs at the same voltage, the output is VREF. A differential voltage at IN+ (VIN+) and IN- (VIN-) develops an identical voltage across the gain-setting resistor, thereby causing a current (IG) to flow. This current also flows through the feedback resistors of the two input amplifiers A1 and A2, generating a differential voltage of:

VOUT2 − VOUT1 = IG x (R1 + RG + R1) where VOUT1 and VOUT2 are the output voltages of A1 and A2, RG is the gain-setting resistor (internal or external to the part), and R1 is the feedback resistor of the input amplifiers. IG is determined by the following equation:

IG = (VIN+ −VIN−)/RG.

The IA output voltage (VOUT) is expressed in the following equation:

VOUT = (VIN+ −VIN−) x (2R1)/RG + 1)

The common-mode input range is a function of the supply voltage and the amplifier's output voltage. With a power supply of VCC, the largest output-signal swing can be obtained with REF tied to VCC/2, which produces an output voltage swing of ±VCC/2. An output voltage swing less than full scale increases the common-mode input range.

If a designer does not pay attention to his choice of IA, he or she may get an unpleasant surprise when the common-mode input range is affected by the supply voltage, the gain and the reference-pin voltage. The reason for this is that various internal nodes can reach voltages for which the amplifier no longer operates in the linear region—it then goes into saturation.

Figure 8 illustrates the typical MAX4194 common-mode input-voltage range vs. output-voltage swing at unity gain (operating with a single supply voltage of VCC = +5V, with pins 1 and 8 floating), and a bias reference voltage of VREF = VCC/2 = +2.5V. Points A and D show the full input voltage range of the input amplifiers, i.e. (VEE + 0.2V) to (VCC − 1.1V). With a +2.5V output, the input differential swing is zero.

The other points (B, C, E, and F) are determined by the input voltage range of the input amplifiers minus the differential input amplitude necessary to produce the associated VOUT. For higher-gain configurations, the VCM range increases at the end points (B, C, E, and F) because smaller differential voltages are necessary for the given output voltage.

Internal vs. External Rg
The gain-setting resistor (Rg) is a vital part of the IA, and the temperature coefficient has a large impact on the amplifier's overall performance. The internal Rg is generally preferred because it is matched with respect to the TC and to temperature. The benefits of this IA type are:

  • Easy gain setting;
  • High-impedance inputs;
  • Minimum gain equal to 1;
  • Excellent CMR, even at high frequencies of 50Hz-60Hz.

Disadvantages are:

  • Three internal op amps required;
  • Limitations in the combinations of Vin, common-mode input voltage, gain, and REF-pin voltage.

Two-Op-Amp IA
In Figure 9, if R1=R11 and R2=R22, it can be proven that:

VO = (Va−Vb)(1 + R1/R11 + 2R22/Rg)

The benefits of this IA type are:

  • Only two op amps required;
  • High-impedance inputs.

Disadvantages are:

  • Minimum gain equal to 2;
  • CMR vs. frequency for the 2-op-amp IA falls off faster than for the 3-op-amp IA, as a consequence of different noise gains for the two op amps;
  • Limitations in the combinations of Vin, common-mode input voltage, gain, and REF-pin voltage.

The choice of topology (1, 2, or 3 op amps) depends on the application. Common-mode input voltage, supply voltage, gain, REF-pin voltage, and sensor impedance must all be evaluated for each case. All three topologies have a REF pin that enables an optional trimming of the output offset voltage. Ensure a low source impedance when feeding a trim voltage to the REF pin, because any additional impedance at the REF pin degrades the CMR.

Traditional High-Side Monitor
In its simplest form, a high-side monitor requires a precision op amp and a handful of precision resistors. One common approach for high-side measurements has been the classic differential amplifier, which is used as a gain amplifier and level shifter from the high side to ground (Figure 10). Though widely used, that discrete circuit has three major disadvantages:

  • Input resistance (equal to R1) is relatively low;
  • Inputs usually exhibit a large difference in input resistance;
  • Resistors must be very well matched to obtain an acceptable common-mode rejection ratio: a 0.01% deviation in any resistor value lowers the CMRR to 86 dB, a 0.1% deviation lowers it to 66 dB, and a 1% deviation lowers it to 46 dB.

When selecting an IA topology, one parameter is especially important: The input common-mode voltage range should include (as a minimum) the high-side voltage plus a safety margin, under all conditions of the amplifier's output voltage swing.

Level shifter
A basic understanding of this circuit's operation can be gained by considering the MAX4198 as a 3-input summing amplifier (Figure 11). The voltage transfer function is then Vout = Vb ­ Va + Vshift. As this relation shows, the output responds to a difference signal and algebraically adds the voltage at the "REF" input. VREF may therefore assume any arbitrary value that will not saturate the MAX4198 amplifier's output. The MAX4194 is also suitable as a precision amplifier, which can easily be configured for fixed gains of −1, 2, or ±1.

More IA Applications
The real strength of the 3-op-amp topology is its ability to make true differential measurements (an advantage of high CMR) while providing very high input impedances. That capability is useful in many applications, especially those for which the signal source has very high impedance. To minimize leakage current from the signal to ground, this example uses the guard technique, in which the shield of the signal cable is driven at a potential of Vcm+ΔV/2.

As a final thought, consider a bridge instrumentation amplifier including a Wheatstone Bridge sensor (Figure 12). Note that the bridge impedance can be moderately low without degrading the IA's CMR.

Gert N. Helles is a Field Application Engineer at Maxim Integrated Products, Denmark. He can be reached via e-mail at: Gert_Helles@maximHQ.com or by phone at: 45 8761 0857.





  Find it now on Digi-Key:
Amplifiers
Digi-Key


Top 20 Products
1) Plug-In PCB Terminal Blocks Keep Costs Low  94
2) 3G Transceiver Consumes 30% Less Power And Delivers 50% Better EVM  91
3) 100-W DC/DC Converters Serve Industrial Tasks  89
4) Vista-Compliant Audio Subsystems Save Space And Power In Notebooks  88
5) 650-W Power Supplies Meet Medical Standards  74
ALL TOP 20 >>
Search News
 
Email Newsletter
Subscribe to the EEPN Products of the Week Newsletter.



PartFinder

Find real-time pricing, stock status, same-day/next-day shipping options and more. Brought to you by Digi-Key. Go to PartFinder.    
GlobalSpec

PART SEARCH :
Powered by: GlobalSpec - The Engineering Search Engine
Sponsored Links

Electronic Design Europe Electronic Design China EEPN Power Electronics Auto Electronics Microwaves & RF
Mobile Dev & Design Schematics Find Power Products Military Electronics EE Events Related Resources