Biopotential and Transducer Data
Magnitude and Frequency Ranges of Biopotential Signals
Human Body Model
Noise Voltage and Current
Common Mode Rejection Ratio
Common Mode Input Impedance
Common Mode Bootstrapping
Differential Input Impedance
Input Rectification and RFI
Interference (Notch or Comb) Filters
To call something “living” is partly a recognition of observed complexity. Living things are very sophisticated and certainly complicated enough to presently defy total understanding. However, life’s processes can be observed and relationships established between one kind of system and another. Regarding human physiology, great strides have been made in characterizing the functions of organs, pathways and body systems.
We presently understand substantial details about how state information is communicated between parts of the body. We have determined that the senses convert physical stimulus of one type, such as visible light intensity and color or the air pressure waves associated with audible sound to electrical signals which travel to the brain. We see how the nervous system is mediated by aspects of body chemistry. We know that the different parts of the body play important roles as specialized components of a whole person.
People started developing understanding about the human body’s characteristics as a consequence of observation. Observation, defined as the sensing of data, is the preliminary step to understanding. Biophysical data from the body can be any kind of signal which might be perceived by the human senses. As we start to perceive something, we can better consider the relationship between this new perception and that which may be related but not presently observable. We can extrapolate beyond, and consider more, than what we presently perceive. Our senses are pretty capable and we can synthesize copious knowledge even from limited data. Historically, we have learned a great deal about human physiology just by turning our unassisted sensory focus to people.
With the advent of certain technologies, as applied to the field of medical instrumentation, our present sensory capacity has substantially expanded. We have access to types and amounts of biophysical data from the human body for which there is no historical precedent. The rate at which we are increasing our knowledge-base and understanding about the human body is radically accelerating.
Biophysical data from the body includes precise information about electrical signals and pathways, oxygen levels in the blood, metabolic function, blood flow and pressure, temperature, motion and position, blood chemistry, sensory function and cognitive function. Much information in the body is mediated via traveling electrical signals and so considerable body biophysical data is obtained simply through the use of specialized electrical signal amplifiers, called biopotential amplifiers. The electrical signals generated throughout the body are called biopotentials. Biopotentials are small voltages that are generated by chemical actions taking place at the cellular level. Other information from the body can be transformed into an electrical signal via the use of a transducer. A common transducer is a blood pressure sensor. This sensor converts a physical pressure input reading into a proportional voltage output signal. Transducers are very helpful tools because they extend the considerable capability of electrical signal amplifiers to be applied to the collection of nearly any type of biophysical data.
When biophysical data is converted into an electrical voltage of a specific high level range, it can be readily digitized by an analog to digital converter and stored in computer memory. Once in the computer, the data can be presented into forms easily perceived by people, typically in the form of a simple waveform graph or chart, indicating how the data changes as a function of time. Furthermore, it’s possible to present the data in a myriad of different ways. Some of these ways involve an orthogonal transformation from one domain into another, such as representing a time-varying electrical signal as a series of component waves, each of differing frequencies and amplitudes. One such type of conversion, known as the Fourier transformation, converts a time domain based waveform to a frequency domain based waveform. Another way of presenting data, called a histogram, presents information about the distribution of data. Histograms can point to the probability that data might lie in a specific set of ranges.
A wide variety of transformation methods and associated visualizations are typically required to interpret complex data sets. People are generally very skilled at visual pattern identification, so data interpretation strategies usually involve the transformation of specific time-series data into a collection of related visualizations. A person can examine these coupled visualizations, each of which is designed to emphasize different aspects of the source data. In this manner, many “viewpoints” can be simultaneously constructed to help people see patterns in the data.
Biophysical data includes all the different kinds of data that can be collected from the human body. A short list of measurable datatypes includes: biopotential, pressure, force, strain, temperature, flow, volume, position, velocity, acceleration, rotation angle, blood chemistry, impedance, conductance, optical transmission, inclination, gas concentration, humidity, weight, surface area and density. These datatypes can be categorized into two groups, biopotential data and transducer data. Biopotential data signals, collected from the body, are small voltages in the amplitude range of microvolts to millivolts. Transducer data signals result from the transducer’s ability to convert a physical (non-voltage) measurement into a voltage output that is easy amplified. Accordingly, very similar amplification methods can be used to record biopotential data and transducer data.
A common transducer topology is the Wheatstone bridge. This type of transducer contains four passive elements arranged in a ring with four contact points. At least one of the passive elements has a characteristic whereby the element’s impedance is influenced by a change in the monitored physical signal. The transducer is activated by supplying an excitation voltage to two of the contacts. The transducer’s output is provided on the remaining two contacts. The voltage differential between these two output contacts is proportional to the physical signal value. The transducer physical signal data, once converted into an electrical form, is usually in the millivolt range. Increasingly, transducers are designed to incorporate differential electrical signal amplifiers to provide high-level output voltages which can be directly digitized.
Transducers can be designed to provide a plethora of output types. In addition to an output voltage signal, transducers can be constructed to output a current or a frequency signal. A frequency signal is typically characterized by a voltage signal which varies in a sine or square wave manner.
Biophysical data, in addition to coming in a range of signal amplitudes, also comes in a range of signal frequencies. In the case of the biopotential signal created by the heart (ECG), the signal amplitude is in the millivolt region and the frequencies, contained within the ECG, are in the range of 0.05 to 150 Hz. A blood pressure transducer, which measures blood pressure over the range of 0 to 300 mmHg, also provides a signal output in the millivolt region, and blood pressure signal frequencies are in the range of 0 to 10 Hz.
Appendix A contains a table illustrating the amplitude and frequency ranges of biopotential and transducer signals.
The human body, as a signal source, can be modeled as an impedance mesh network. The impedances in the mesh network are phase-lagging in character. Each “Z” in the network, can be modeled as a simple combination of resistors and capacitors. In the simplest form, this combination looks like a resistor in parallel with a series resistor and capacitor combination. This model indicates that at low frequencies, the body looks primarily resistive, with high values. At intermediate frequencies, the body becomes increasingly reactive (capacitive) and at high frequency, the body looks resistive, with low values.
At the frequencies associated with surface biopotential data (DC-500Hz), the human body model indicates a resistive source impedance in the region of 5 Kohms to 30 Kohms, assuming standard skin preparation techniques and hypertonic electrode gel with Ag/AgCl electrodes with approximately 0.75 cm squared surface contact area. The following graph illustrates the typical impedance magnitude, as a function of frequency, between two locations on the arm, each location sensed with a 0.75cm squared surface area Ag/AgCl electrode.
In the context of human physiological recording, an electrode is a conductive probe which is used to sense the presence of biopotentials arising at points in the body. When biopotentials are generated in the body, they can be sensed using a differential voltage amplifier. When a biopotential source activates, electrons are moved within the volume of the source thus creating a voltage potential difference (positive to negative) across the volume of the source. Biopotential sources are continually changing in the body, so the associated differential voltage potentials change too.
Electrodes establish a equipotential area when attached to a location on or in the body. An equipotential area means that anywhere on the surface area of the electrode the potential (voltage) will be the same. This is because electrodes are substantially more conductive than the surrounding tissue. To record a biopotential signal, two electrodes are required. The pair of electrodes will measure the voltage which is impressed between the two equipotential areas created by the electrodes.
Biopotential recording electrodes are of two types, namely invasive and non-invasive. Invasive electrodes are designed to sense biopotentials inside the body, meaning they are inserted inside the body, past the skin boundary. Historically, invasive electrodes have been designed as needles or wires to better penetrate the skin. By controlling the conductive and insulating areas of the needle or wire, the electrode can be used to sense biopotentials originating at any depth in the body volume. If the invasive electrode is to be inserted into the body for a long time, then the electrode might take a different shape, such as a conductive disk which can be sutured into place.
Non-invasive electrodes are designed to sense biopotentials which manifest on the outside skin surface. These kinds of electrodes do not penetrate the skin surface layer. Because the body is a volume conductor, biopotentials originating anywhere in the body travel to the entire skin surface area. As an example, a polarized biopotential wave is generated by the heart during the course of each beat. This signal is easily detectable between any two locations on the body’s skin surface that are physically oriented to different sides of the heart.
In practice, it’s best to collect biopotential data as near to the biopotential generating site as possible for the best recording. This is because the body generates a large number of biopotential signals, sourced from a variety of excitable cells. Because the body is conductive, a myriad of these signals will superimpose at any body recording location. As the distance grows between a particular source and the differential recording location, the source signal attenuates. Sources, that are subject to a range of attenuation, will simply add to the desired signal being measured, as they are included in the volume conduction path defined by the differential biopotential recording.
When collecting biopotential data from surface electrodes, the contact area of the electrode has an impact on the frequency and amplitude components recorded in the biopotential. Because the body volume is relatively conductive, increased electrode surface area has a tendency to reduce biopotential amplitudes and suppress higher frequency components. This happens because multiple, and often uncorrelated, biopotential sources in the body can act in random opposition as the various biopotentials superimpose in the conductive body volume and then express on the skin surface. A biopotential averaging process takes place in the body tissue volume and an aspect of the averaged signal is reflected on the skin surface area. This biopotential signal averaging quality increases in scope as the size the measurement surface electrodes increases.
If a biopotential electrode is reduced in size and then inserted into the body near the location of interest, then measured signals will typically increase in amplitude and frequency components. These kinds of electrodes are called needle or wire electrodes and typically have an insulating needle shaft with a conductive tip. Despite certain benefits of needle electrodes, complications can arise when breaking the skin barrier and inserting foreign objects into a subject’s body. Needle electrodes can be difficult to use because they have smaller surface area than surface electrodes. Furthermore, smaller contact surface area always means that the electrode contact impedance is higher. Very high contact impedance, typically in excess of 1 Mohm, specifies the need for excellent lead shielding and specialized input amplifier characteristics.
Biopotential measurements from surface electrodes on the body have material harmonic content of less than 500 Hz. Nearly all of the signal power which can be collected from surface electrodes is between DC and 500 Hz. If the electrode is smaller, such as a conductive-tipped needle, and inserted through the skin surface to near the biopotential of interest, biopotential signal frequencies can exceed 10,000 Hz.
Electrodes can be made out of all kinds of conductive materials. Electrodes are made out of tin, gold, silver, stainless steel, carbonized conductive rubber and silver / silver-chloride (Ag/AgCl). With surface electrodes usually some kind of electrolyte gel is used to mediate a good conductive connection between the skin surface and electrode. The highest performing electrodes presently manufactured are Ag/AgCl electrodes. These electrodes are considered non-polarizable and they have low and stable electrode offset potentials.
Electrode to skin junction offset potentials result from the juxtaposition of three conductive materials, namely skin, electrolyte and electrode. The conductive pathway between skin and electrode is mediated by conductive ions in the electrolyte (electrode gel). Any current passing through this pathway also contributes to the total offset potential. All electrode-electrolyte-skin contact junctions create a small potential from skin to electrode. This voltage behaves similarly to the voltage produced by a battery. For biopotential measurements, it’s helpful to have electrode to skin junction offset voltages be as small and stable as possible because these offset potentials are measured in series with the biopotential of interest.
When performing biopotential measurements it’s very helpful to use identical electrode types at the recording sites. In addition, much better recording performance can be obtained if the electrodes have low offset voltages and are very stable. Because electrodes are used to sense a specific biopotential in the tissue volume, their offset potentials are effectively added to the measured biopotential. When both recording sites use the same type of stable and low offset electrodes, then the electrode offset potentials will largely cancel each other out. This is because the electrode to skin junction voltage polarities are in opposite directions when adding up all the associated potentials in any biopotential measurement loop.
Before performing biopotential measurements it’s helpful to perform electrode impedance checks between electrodes attached to the subject. Checks are performed between any two differential recording electrodes and also between those electrodes and the reference ground. Impedance checks are best performed using an alternating voltage or current source. A typical measurement frequency is around 25Hz, because this frequency is found “in-band” to most biopotential signal bandwidths and is thus representative of source impedances for those measurements.
For biopotential amplification, the recording quality improves with reduced source impedances. This improvement is a consequence of several reasons.
- Amplifier common mode rejection ratio (CMRR) reduces when source impedances are widely different. This is because the amplifier input resistance to ground, for each differential input, is non-infinite. As a consequence, common mode signals will induce the creation of a differential signal which is approximately equal to the difference in source impedances divided by the amplifier common mode input resistance to ground. However, if the amplifier incorporates common mode bootstrapping, then common mode input impedance (CMII) can be made extremely high, thus minimizing problems associated with disimilar source impedances.
- High source impedance signal conductive paths are influenced by evironmentally-sourced, interfering, displacement currents. When the impedance-to-ground of a conductive path is high, and if interfering displacement currents are permitted to flow to that path, then interfering potentials (resulting from the displacement current flow) can be large if the environmentally-sourced potential is also large. However, if the signal source impedance-to-ground is low, then impinging interfering displacement currents generate proportionally lower interfering potentials. Environmentally-sourced potentials, that can induce interfering displacement current flow, include energized mains power conductors, switching light sources, monitors, computers and motors. This problem can be minimized using shields or shielded cables to redirect the sourced interfering displacement currents to ground. In this circumstance, the interfering displacement currents are prevented from impinging upon sensitive, high source impedance, signal conductive paths. However, if the interfering signal has a strong magnetic field (B) component, electrostatic shielding is ineffective. To shield against B fields, B field conductors (high permeability) must be used to surround the source signal paths.
- As source impedance increases, amplifier noise at the input increases. This happens because the amplifier noise current is directed though the source impedance, thus creating a noise voltage. For a fixed noise current, related noise voltage increases in direct proportion to increased source resistance. This problem can be minimized by reducing the current noise associated with the amplifier input.
There are many possible configurations for differential input biopotential amplifiers. Configurations are a function of design goals and implementation methods. Design methods vary, depending on the nature of biopotential amplifier isolation from mains ground. Primarily, design methods are concerned with strategies for maintaining high common mode rejection ratio (CMRR), high differential input impedance and high common mode input impedance (CMII). Other factors to consider involve protection for transient artifact, such as from static discharge, electrosurgical or defibrillation equipment; radio frequency interference (RFI) suppression; and controls to limit patient leakage current.
- Not Isolated – In this case, the amplifier’s power supply ground reference is galvanically coupled (via a conductive pathway) to mains ground. This configuration is capable of being considered as type “B” medical equipment design according to IEC60601-1.Biopotential amplifier design methods, for this category, should consider strategies to enhance CMRR and CMII using the following methods. All of these methods will help to enhance CMRR and CMII.A) Common mode driven ground electrode to subject
B) Common mode driven amplifier reference to power supply
C) Common mode driven amplifier input bias current network
D) Driven shield lines for input signals – single-ended or common mode
- Capacitively Isolated – In this case, the amplifier’s power supply ground reference is capacitively coupled to mains ground. Capacitive isolation can vary over a wide range, subject to equipment design. This configuration is capable of being considered as type “BF” or “CF” medical equipment design according to IEC60601-1. Depending on the relative magnitude of the capacitive coupling of amplifier ground reference to mains ground, the use of these methods will have varying degrees of effectiveness, with methods C) and D) providing the most assistance.A) Common mode driven ground electrode to subject
B) Common mode driven amplifier reference to power supply
C) Common mode driven amplifier input bias current network
D) Driven shield lines for input signals – single-ended or common mode
- Fully Isolated – In this case, the amplifier’s power supply is battery-backed and the measurement system is typically worn on the subject. This is a variant of type 2), but capacitive coupling to mains ground is extremely small. This configuration is capable. of being considered as type “CF” medical equipment design according to IEC60601-1. In this configuration, design strategies C) and D) are possibly helpful to improve CMRR and CMII, depending on the length of input signal lines or physical size of input signal measurement system.C) Common mode driven amplifier input bias current network
D) Driven shield lines for input signals – single-ended or common mode
In laboratory environments, mains power lines and mains-powered equipment will source displacement currents to the body of the subject. These currents are simply the alternating currents that flow due the capacitance between the subject’s body and the mains source. The flowing of these currents through the subject’s body will manifest in two ways. Primarily, the superposition of the sum total of currents will establish a alternating voltage on the subject’s body with respect to mains ground. Secondly, the flowing of these currents through the subject’s body will create differential voltages from one part of the subject’s body to another.
Typical capacitance values from mains power hot to a person in a room are in the range of 1-10pF. The person’s capacitance to mains ground is typically on the order of 10-100pF, or perhaps 10 to 100 times larger. Assuming a ratio of 1/100, and a mains voltage of 120VAC, the voltage impressed on the person will be about 120VAC/100 or 1.2VAC. This impressed voltage is many thousands of times greater then the biopotentials generated by the person, accordingly the measurement of biopotentials requires the use of differential amplifiers with very high common mode rejection ratio, typically 80dB (10,000) or higher.
The second manifestation of mains-initiated, displacement currents is problematic for biopotential measurements. In this case, even if the biopotential amplifier has arbitrarily high common mode rejection ratio, mains interference will be recorded as part of the differential signal. There are a number of methods to overcome this interfering signal. It can be filtered out or be synchronously removed. Alternatively, conductive shielding can be used to redirect offending displacement currents. Finally, because the subject’s body has relatively low impedance, differential biopotential measurements over a smaller area on the subject’s body will have a smaller interfering signal.
Problematic interfering signals, that can be introduced on the subject’s body, are sourced from alternating magnetic fields. In typical laboratory environments, alternating magnetic fields strong enough to cause interference are sourced from transformers and rotating machinery. In both of these cases, large currents are used to create powerful magnetic fields in equipment. If these magnetic fields are close to the subject, the alternating magnetic field will induce currents in conductive “loops” associated with the subject’s body. These currents move through the subject’s body, or attached electrode leads, thus inducing a proportional voltage which may be amplified as an interfering signal. It’s difficult to shield against magnetic fields, so usually the biopotential recording site is moved well away from any magnetic field generators. In practice, this is usually manageable because the interfering magnetic field strength drops off proportionally to the inverse-square of the distance moved away.
All electrical signal amplifiers amplify some amount of noise along with the signal of interest. Noise is defined as a random signal. If a signal source has any internal resistance then the source will produce a specific amount of electrical noise. This kind of noise is called Boltzmann noise. Boltzmann noise is defined as En= SQRT (4KTRdF). This type of noise increases as the square root of an increase in temperature, resistance and measuring bandwidth.
Amplifiers, just by themselves, also contribute noise to the measurement. This is because any amplifier has some amount of noise voltage and noise current. These two noise sources are associated with the transistor characteristics at the amplifier input. The noise current from the amplifier is directed through the impedance of the source thus causing a noise voltage. This noise voltage is vectorially added to the original source’s Boltzmann noise and also to the amplifier’s noise voltage.
It’s always helpful to reduce any of these noise sources, but if one can’t be reduced beyond a certain point, then it’s increasingly unnecessary to reduce the other noise sources in relation because the largest noise source will dominate vectorially as per the the equation Vnt = SQRT (vn1^2 + vn2^2 + vn3^2 + etc).
Generally, for low source impedances, the noise voltage of the input amplifier is dominant. For these cases, it’s best to use amplifiers with input bipolar junction transistors (BJTs). For large source impedances, total noise can increase dramatically due to the contributions of the source’s Boltzmann noise and the impact of amplifier noise current multiplied by the source impedance. In the case of high source impedances, it’s usually best to use amplifiers with input field effect transistors, typically jFETs.
In order to choose the proper amplifier type, first determine the nominal impedance of the signal source (Zs). Then consider the ratio of the amplifier’s input noise voltage (Vn) to the input noise current (In). Optimum performance, for any given amplifier, results when Vn/In = Zs. In this case, the noise contributions from the amplifier’s noise voltage and the noise current are balanced and collectively minimized.
There are many sources of common mode interference. Common mode interference signals can range from close to zero (DC) to very high frequency. Some common sources:
Mains power sources and associated conductive networks –
These are the primary sources of power for any laboratory environment. These signals are relatively high voltage, with magnitudes ranging from 120 to 240 VAC, and frequencies from 50 to 60Hz. In addition, harmonics of these signals are present, thus creating interfering signals at integer multiples of the fundamental.
Computer monitors and light sources: The most problematic common mode sources are those which involve signal scanning or switching. Computer monitors which employ raster scanning methods to present data will radiate common mode displacement currents at all the fundamental scanning frequencies, such as pixel to pixel, line to line and screen to screen refresh. Light sources, such as fluorescent or LED, will radiate displacement currents linked to switching frequencies. Fluorescent lights, due to high switching voltages, can be notorious noise sources.
Broadcast signals: These common mode sources include AM, FM radio and analog and digital sources such as those associated with wireless local area networks, television and general radio communications. The most common interfering signals are in the range of 540kHz to 2.5GHz, however man-made sources cover the range of 3Hz to 300GHz.
Generally, for these above mentioned sources, one can assume that the common mode interfering signals are referenced to mains earth ground. Common mode signals can impinge and travel a variety of ways in electrical networks. When measuring small electrical signals, one should try to keep interfering signals away from the measurement location. However, in practice, this can be quite difficult. For low frequency interfering common mode signals, it’s often less complicated to tolerate the interference and employ special methods to record data using common mode rejection methods.
In addition, or alternatively, shielding methods can be used to shield sensitive input lines from common mode sources. An important point to remember though, is that common mode sources can affect the measurement circuit in a few different ways. Common mode currents can flow from sources like mains power networks or computer monitors to the measurement circuit. Alternatively, common mode currents can flow from mains connected power lines, into measurement equipment, through the measurement amplifier and then out through the sensitive input measurement lines. Accordingly, shielding against common mode currents on both power and signal input sides is important.
Increasingly, in contemporary electrical instrumentation, a very common and problematic common mode source is equipments’ internal AC/DC-DC converters. AC/DC-DC converters are used to move energy around efficiently. They are used to convert mains 120-240 VAC @ 50/60 Hz power to voltages used to power instrumentation, such as +1.8, +3.0, +3.3, +5, +/- 12vdc. All AC/DC-DC converters employ internal high frequency switching to perform energy conversion. Usually, these frequencies are in the range of 50kHz to 500kHz. These interfering signals can be very disruptive to sensitive measurements. These AC/DC-DC common mode signals are typically referenced to either mains ground or instrumentation ground or some combination of the two. Subject to design methods, the equipment’s AC/DC-DC converter will typically primarily reference to mains ground. In this situation, if the converter’s common mode generated output current is not routed back to mains ground, then this current will seek out alternative paths, such as through the measurement input lines. A solution to this problem is to employ common mode chokes to create high impedance paths for common mode currents as they attempt to flow through the measurement inputs. Alternatively, low impedance paths can be provided to direct common mode currents back to mains ground. Usually, the best solution to constrain AC/DC-DC converter noise is to introduce a high impedance common mode path, at the converter switching frequencies and harmonics, to the input or output of the converter. Then a low impedance common mode path is created to shunt the now high source impedance converter noise to ground.
There are several potential problems with common mode signals being present on the input measurement lines. The first problem is that a common mode signal may become slightly unbalanced when impacting differential input measurement lines due to subtle differences in impedance of those lines to common mode ground. This circumstance can arise when input lines have unequal shielding or no shielding. A related problem is when the input lines have unequal source impedance. In either of these situations, a common mode signal creates a differential signal that is amplified by the front-end instrumentation amplifier. These two situations are a consequence of finite amplifier common mode input impedance (CMII). Another problem is that, even if the common mode signals are perfectly balanced, the input amplifier has finite common mode rejection ratio (CMRR). Finite CMII and CMRR ensure that some portion of the common mode signal will always get amplified, assuming the signal is within the amplifier measurement bandwidth. Also, as the frequency of the common mode signal increases, the amplifier’s CMRR reduces. If the common mode signal is very high frequency, another possibility is that the common mode signal may be rectified by the input amplifier front end. This situation readily occurs if the common mode signal has a frequency that is substantially higher than the amplifier bandwidth. At high frequencies, instrumentation amplifiers behave non-linearly, primarily due to slew rate limiting behaviors and other factors. These non-linear processing characteristic of instrumentation amplifiers may result in the appearance of sporadic level shifts and blips in the amplified signal. These jumps in signal baseline might easily be the non-symmetrical slew rate limited, rectified and amplified consequence of sporadic changes in high frequency, common mode, signals present at the amplifier inputs.
The common mode signal, while performing a biopotential measurement, is the signal component that is identically present on both inputs (Vin+ and Vin-) of the differential input biopotential amplifier. A common mode signal can be generated endogenously – from internal physiological processes, or exogenously – from outside sources. Common mode signals generated by internal physiological processes will be of the same order of magnitude of the differential signal of interest. Common mode signals from outside sources are generated in all laboratory environments. These sources include mains power wiring (120/240 VAC at 50/60 Hz), computer monitors, florescent lights, rotating machines and radiated electromagnetic interference from switching circuits. Typically, the largest contributor to common mode signals is activated mains power wiring. Because these source voltages are much higher than biopotential voltages, the common mode signals generated by these source voltages will usually be many hundreds or thousands times larger than the biopotential signal of interest.
By definition, a differential, biopotential amplifier should measure just the voltage signal sensed between the amplifiers’s inputs. However, an amplifier requires a supply voltage to operate and that supply voltage generator has a coupling impedance to mains ground. An exogenous common mode signal generator will source a signal referenced to mains ground. Accordingly, an externally-sourced common mode signal will reference to a biopotential amplifier’s inputs and the amplifier must rely on it’s common mode rejection ratio (CMRR) performance to remove the interfering common mode signal from the differential signal of interest. A critical ability of a biopotential amplifier is its performance in removing the common mode signal present on the amplifier inputs. This specified ability is defined as the CMRR of the amplifier. CMRR is typically specified in voltage decibels (dB) and is 20 times the log ratio of the common mode input voltage over the gain normalized common mode signal present at the amplifier output.
Despite manufacturer claims of extremely high biopotential amplifier CMRR, sometimes to 120dB, the practical biopotential measurement reality is typically on the order of 60-70dB. This practical measurement is known as the Interference Rejection Ratio (IRR). By controlling all aspects of the CMRR measurement, including: identical source impedances at input, extremely short input leads and controlled input signal signals, CMRR can be ideally measured to be extremely high. However, this type of measurement is unrealistic and practically meaningless. A manufacturer can provide a slightly better estimate of expected maximum CMRR if they include impacts associated with typical source impedance imbalances, however, in real-world conditions, differential source impedances will be unbalanced and there will be capacitive coupling pathways between input leads, source and the external environment. Furthermore, a primary interference factor related to biopotential measurement results from the fact that common mode voltages will create corresponding differential signals in the region of measurement interest, due to the myriad of capacitive coupling pathways in physical environments. For real-world measurements, the IIR will often be 60-70dB, despite a biopotential amplifier CMRR specification above 80dB. In effect, any amplifier CMRR specification in excess of 80dB is realistically meaningless for improving biopotential measurement performance.
Biopotential amplifiers typically operate one of two ways. One type of amplifier will “sense” the common mode signal and simply reject it as well as possible. In this case, the amplifier ground is passive and acts as a common mode reference signal. The other type of amplifier will “sense” the common mode signal and feed it back – inverted – to the subject’s body to remove the common mode signal to the extent possible. In this situation, the first amplifier’s ground attachment (to the subject) is replaced with an inverted common mode signal drive. With contemporary integrated instrumentation amplifier technology, there is no material advantage of one method over the other, when considering single isolated biopotential amplifiers operating in standard laboratory environments. However, when connecting multiple biopotential amplifiers (each employing common mode drive ability) to the same subject, the circumstances can be problematic because:
- the amplifiers will measure different common mode signals
- extra electrode attachments are needed for the common mode drives
- the subject is exposed to actively-driven voltage sources, creating possibility for exceeding safety current thresholds
In these cases, it’s typically less problematic to employ biopotential amplifiers with passive and identical grounds. Additionally, in this case, only one ground lead is required for any number of biopotential amplifiers.
A single biopotential amplifier typically makes three electrical connection points to the subject, two inputs which establish a differential input voltage vector and a reference ground. The two inputs consist of a positive (Vin+) and negative input (Vin-). The amplifier amplifies the voltage difference between Vin+ and Vin-. If the amplifier has a gain of K, then the amplifier output will be K * (Vin+ – Vin-). The ground connection is only important to establish a common mode reference. A common mode reference is typically important to obtain high quality signals when recording biopotentials, especially so when the amplifier is non-isolated from mains ground.
The common mode input impedance (CMII) is the impedance seen by a common mode signal that is simultaneously applied to the differential inputs of an amplifier. This is the impedance of the common mode signal to the common mode reference, otherwise called “ground”. Common mode rejection ratio (CMRR) and common mode input impedance are related concepts. If a common mode signal is present on the differential amplifier inputs, and if the two inputs have different source impedances, then common mode signals will convert partially to a differential signal. This happens because either source impedance (SI) and the common mode input impedance act to create a voltage divider at the amplifier input. The voltage divider has a formulaic expression of CMII/(CMII + SI). As the common mode input impedance increases, the created voltage divider will have a value closer to 1. If there is a source impedance mismatch, between the differential inputs, then the voltage dividers acting on the common mode signal will be mismatched. However, if the CMII is very high, then larger SI mismatches can be tolerated while still maintaining high CMRR.
Because it’s difficult to control the values of source impedances in biopotential recordings, it’s best to use an amplifier with high common mode input impedance. Most commercially available instrumentation amplifiers have very high common mode input impedance. In practice, its often important to supply bias currents to the amplifier without routing them through the source impedance. This practice is critical when high source impedances are present or when the source impedances are capacitive. Because bias currents typically source from common mode reference or “ground”, it can be problematic to provide bias without degrading common mode input impedance. However, it’s possible to maintain very high common mode input impedance, while still providing amplifier input bias currents, using a technique called “boot-strapping”. Boot-strapping is a design method which employs the derived common mode signal to source a bias impedance network which is connected to the differential inputs. If the same common-mode signal appears on both ends of the bias impedance network, then the effective current through the bias impedance is zero, and so very high common mode input impedance is maintained.
Another method of boot-strapping employs the derived common mode signal to bias the power supplies of the instrumentation amplifier. Essentially, the input instrumentation amplifier’s ground reference is driven by the sensed common mode signal. In this configuration the power supply voltages track the common mode signal, thus effectively removing the presence of a common mode signal at the amplifier inputs.
The amplifier differential input impedance is the impedance from one differential input to the other. If this impedance is significantly larger than the signal’s source impedance than the signal level will be relatively unaffected. The higher the differential input impedance, the less impact amplifier connection loading will have on the signal level. If the differential input impedance is on the order of 100 times larger than the source impedance then the signal will be 99% of actual. Differential input impedance is usually modeled as the parallel combination of the amplifier’s differential input resistance and capacitance.
The amplifier’s input impedance to ground is the same as the common mode input impedance. Ideally, this impedance should be as high as possible. For unequal source resistances, different voltage dividers will be created between the differential inputs, for a non-infinite single-ended impedance to ground. The higher the input impedance to ground, the higher the CMRR for unequal source resistances. It’s possible to create very high common mode input impedance through the use of bootstrapping. In this case, the amplifier input bias and offset currents are supplied through an input impedance network that is driven by the common mode signal. It’s generally necessary to have very high common mode input impedance at any interfering signal frequency. Interfering frequencies are usually the power mains frequencies (50/60 Hz) and their harmonics. However, they may also be associated with computer monitor scanning rates, high frequency fluorescent lights and electric motors.
A driven shield is a conductive shield surrounding one or both of the amplifier inputs that is held either to the common mode voltage of the differential inputs or the identical voltage of a single input. A driven shield is very helpful in that it reduces the coupling capacitance of the input (center conductor) to the shield. Because the shield is driven by a signal nearly identical to the center conductor being shielded, there is no created potential that could drive a displacement current through the capacitance between center conductor and shield. Accordingly, when signals are recorded from high source impedances, a driven shield can act to intercept a potentially interfering displacement current without band-limiting the input signal. Potentially interfering displacement currents are typically sourced from mains power lines in the laboratory environment. A driven shield is used as a compromise solution to recording from high input impedance sources because, even though it can protect the input signal conductors from displacement currents, the shield drivers can also add some noise back into the input lines. However, in the case of differential inputs where each line is shielded by the same driven shield, noise addition will be significantly reduced due to the CMRR of the amplifier.
A grounded shield is simply a shield for an amplifier input where the shield is attached to amplifier ground. This type of shield will result in a capacitance between the shield and the input (center) conductor. Accordingly, for source impedance “R” and shield capacitance “C”, the shield will act to lowpass filter the source signal as per the expression Fo = 1/(2*Pi*R*C). If the source impedance and shield capacitance remain small-valued, then the impact on input signal bandwidth will be minimal. A grounded shield has an advantage in that it will not add any noise to the input signal conductor.
Shields should be configured to avoid the carrying of signal current, when employed in a cabling system for small signal transmission. This is because the shield will be an anchor for displacement currents, and because the shield has finite impedance, these alternating currents will induce voltages along the length of the shield. If the shield is series connected to the circuit under measurement then these induced voltages will be measured along with the desired signal. For small signal transmission it’s much better to simply tie the cable shield to either end of the cable. The optimal end to tie the shield to is determined by evaluating the desired direction of common mode current flow.
For a passive transducer, a shield line tied to transducer excitation (-), at the transducer side, will result in shield displacement currents being directed back along the transducer excitation (-) line to the location of the excitation voltage source. This displacement current flow behavior happens if the excitation voltage source’s impedance to ground is much lower than the transducer’s impedance to ground. In this case, any displacement currents will effectively flow through the excitation supply and signal lines to the transducer, thus increasing the common mode interfering voltage seen by the differential amplifier. To improve this situation, it’s better to attached the shield line to the (-) on the excitation voltage source side, versus the transducer side. In this case, impinging displacement currents will not flow through any excitation or signal lines attached to the transducer.
The input capacitance for a typical amplifier has two primary contributors. The biggest contributor is usually the capacitance of input conductor to other conductors present in the environment. A secondary contributor is the input capacitance of the amplifier. Ideally, the input capacitance should be as low as possible. In addition to reducing the bandwidth of the input source signal, input capacitance acts to limit CMRR of the amplifier. CMRR will be constrained in the case of unequal source resistances or impedances, because differing voltage dividers will manifest between the differential inputs.
All instrumentation amplifiers will have a bandwidth that is determined by the associated construction of the basis transistors and the internal design of the amplifier. In particular, the input transistors and related circuitry of the instrumentation amplifier will typically have a construction which results in the highest CMRR, lowest noise, highest input impedance and largest frequency response possible. Different types of instrumentation amplifiers are designed to establish compromises between these and related variables in order to best match specific, system-level, design constraints. In cases where instrumentation amplifiers are amplifying signals that may be corrupted by very high frequency signals, such as radio frequency interference (RFI), it’s usually required to filter out the RFI to prevent it from being rectified by the input staging of the instrumentation amplifier. This filtering can often be performed using series resistors and shunt capacitors on each differential input. RFI filters can be implemented to filter both differential or common mode RFI. For reducing common mode RFI, series resistances and/or common mode chokes can be used with shunt capacitors to amplifier ground. For reducing differential RFI, employ series resistances and/or inductances with differentially-placed capacitors. Best performing RFI filters will usually employ a combination of RFI common-mode and differential filtering methods.
A displacement current is the current which “flows” through a capacitor. Capacitance exists between any two conductive surfaces that are separated by a dielectric. For two parallel and identical conductive plates, capacitance is defined as [eA/D], where e is the permittivity of the dielectric between the conductive plates, A is the effective area of a plate and D is the distance between the plates. For any conductive surfaces, the capacitance between them will be proportional to their mutually effective areas and the permittivity of the dielectric between them. The capacitance will also be inversely proportional to the the distance between the conductive surfaces. In laboratory environments and nearly anywhere alternating mains power can be found, there will typically be significant displacement currents flowing between conductive elements and mains wiring. A usual case involves the displacement currents flowing between mains wiring, subjects and grounded conductive surfaces. A capacitance is established between the subject, mains wiring and all other conductive surfaces. The capacitance is primarily a function of the distance between these various elements. Accordingly, for mains power alternating voltage between 120 and 240 VAC, a corresponding and proportional AC voltage will appear on the subject’s body. If the subject is connected directly to mains ground, then the voltage on the subject’s body, with respect to ground, will be very small or zero. If the subject is well-insulated from mains ground and is physically close to mains wiring, then the voltage on the subject’s body becomes increasingly close to the value of the mains voltage. This result happens when the capacitance from subject to main wiring increases in relation to the capacitance of subject to mains ground.
Usually, the capacitance of the subject to mains ground is much greater (by a factor of 100 or more) than the capacitance of the subject to mains wiring. This is because of the relative areas of mains wiring to grounded conductive surfaces in a room. Consequently, voltages measured on a subject in such a room will be roughly 1/100 of the mains power voltages – namely a few volts or less. The voltages that are impressed on the subject, are usually expressed as common mode voltages, in that they appear everywhere on the subject’s body.
When performing biopotential measurements, these common mode voltages are often a thousand times or larger than the biopotentials that require measurement. Measuring a small biopotential, in the presence of a much larger common mode signal, requires the use of a very high common mode rejection ratio (CMRR) differential amplifier. If the common mode rejection ratio of the differential amplifier is 80 dB, this means the amplifier will effectively reduce a common mode signal by 10,000 (20*log10,000 = 80dB) times in comparison to the differential signal amplified. Practically considered, there is no significant benefit in differential amplifier CMRR in excess of 80dB. This is because mains power induced signals on the body will not manifest as exactly equal potentials all over the body. The body has finite impedance and there will be multiple capacitive pathways from the body to various points of mains wiring and grounded conductive surfaces. Accordingly, the impressed mains voltage on the body will vary slightly from one part of the body to another. In effect, the common mode voltage from the mains wiring expresses largely as a proportional common mode voltage on the body with minute variations in amplitude and phase from one part of the body to another. In practice, it’s common for these voltages to be as close as 99% to 99.99% equivalent. If the common mode voltages between two parts of the body that are are few inches apart are 99.99% equivalent and each is approximately 1 VAC, then the differential voltage between these locations will be 1/10,000 or 100uV. This interfering differential signal will simply be measured along with whatever biopotential is also present between the sites.
Assuming 3 pF coupling capacitance from mains power to person and 30 pF coupling from person to mains ground and 120VAC mains voltage, then voltage impressed on the person will be:
120vac * 3pF/(3pF + 30pF) = 10.9vac
Assuming a body resistance of 10,000 ohms (can range from 1000 to 100000 ohms at 60Hz), when ratiometrically compared to the reactance of 30pF at 60Hz:
1/(2*Pi*60*30E-12) = 88,400,000
88,400,000 ohms / 10,000 ohms is 8,840 or roughly 10,000.
This ratio provides a rough indication of the conversion of impressed common mode signals to differential signals on the human body. For an impressed voltage of 1vac, differential voltages established between locations on the body surface may often be on the order of 1vac/10,000 or 100uV. Naturally, as the physical distance between sensing locations is reduced, this differential voltage will also be reduced. However, assuming standard biopotential measurement locations, it’s reasonable to expect common mode mains voltages express as differential voltages on the order of 1-10uV. Given typical biopotential signals on the order of 1mV (p-p), this expressed differential voltage is typically no less than 1/1000 of the biopotential of interest. Accordingly, increasing amplifier CMRR beyond 80dB (10,000) is unlikely to be proportionally helpful to reduce the effects of common mode interference for differential biopotential measurements. This is why biopotential amplifiers, with CMRRs ranging from 80dB to 120dB, will provide similar results given typical mains interference on the body.
Amplifier Common Mode Input Impedance (CMII) is equally important to consider for a high quality biopotential measurement. If the common mode signal is 1vac, and if the difference between the differential input source impedances is 1,000 ohms, and if the amplifier’s CMII is 10 Mohms, then the common mode signal will be converted into a differential signal of approximately 1vac * (1,000/10,000,000) or 100 uV. Accordingly, common mode input impedances should be very high, at mains interference frequencies (50/60Hz), to reduce the impact of source impedance mismatch. Considering the previous example, an amplifier CMII increase to 1000 Mohms will reduce the size of the mains-induced differential signal to 1 uV. This is also why it’s nearly always helpful to reduce electrode to skin impedance for biopotential recordings, as this effort will reduce the difference between source impedances.
A critical aspect of any amplifier used for biopotential measurements is the ability to perform selective (notch) interference filtering. As indicated previously, significant mains power interference can result in any room outfitted with mains wiring. A biopotential amplification system requires the ability to reject signals associated with the mains power or other interfering common mode signal sources. It’s important to note that any interfering signal is likely to have significant harmonics, thus requiring the amplifier’s ability to reject common mode signals over a frequency range, while still effectively amplifying the biopotential signal of interest. However, given practical measurement realities, an interference rejection ratio (IIR) above 80dB is unrealistic, despite a CMRR specification in excess of 80dB. Accordingly, interference filters are typically required in a wide range of biopotential measurements
A common type of interference filter is known as a notch filter. Notch filters simply remove any frequency which happens to be the same as the interfering frequency. Notch filters can also be constructed as “notch-comb” filters in that they can remove frequency components associated with the interference fundamental and its harmonics. Simple filters of this kind are adequate if the biopotential signal of interest does not have useful components at the notch frequencies. The potential transient effects of notch filters can be mitigated by employing notch filters in conjunction with reasonably phase linear lowpass or highpass filters, where the notch filter is placed well into the attenuation band of the lowpass or highpass filter.
Another type of interference filter operates by synchronously removing interfering signals from measured biopotentials by monitoring the room’s main power wiring simultaneously. Signals in the measured biopotentials are stripped of any interfering signals that are synchronous to the mains power voltage. A similar methodology is employed by the algorithm-based interference “Subtraction” method. Both of these methods preserve the frequency spectrum of the measured biopotential signal and thus are optimal methods for removing synchronously induced interference.
Future configuration of biopotential recording systems, high channel count:
Removal of Power Line Interference: