Electrocardiogram (ECG) lead detection in wearable devices
Resources
Hello and welcome to the next video of the series. In this video, we will talk about ECG lead detection in wearable devices. We will start with an overview of lead detection, beginning with the motivation to detect leads on and leads off conditions. Next, we will talk about DC lead detection, showing the ways to bias the leads to detect leads on and leads off conditions. We will show a design example to illustrate how DC lead detection works. We will then present the principle of operation of AC lead detection and give a design example to illustrate how AC lead detection works.
Let us consider a wrist-borne wearable device that has electrodes for ECG signal acquisition. Here, we consider a case where two of the electrodes on the bottom phase of the watch make contact with the left rest of the user when the watch is worn. We call these the LA, or left arm, and RLD, or right leg drive, electrodes. Whenever the signal is needed to be acquired, the user touches an electrode on the top face of the watch. We refer to this as the RA, or right arm, electrode.
As you can see from the Illustrations, there are three possible conditions of connection to the electrodes. When the watch is off the wrist, none of the electrodes are in contact with the user. We call this the all leads off configuration. When the user wears the watch on their wrist, the bottom two electrodes make contact with the user. We call this the wrist leads on configuration.
When the user touches the top electrode, all three electrodes make contact with the user. We call this the all leads on configuration. In this configuration, the ECG signal can be acquired. Depending on the number of electrodes and their position relative to the bottom face and top face of the watch, the possible lead configuration could be slightly different as compared to the ones illustrated on this slide.
The utility of being able to distinguish between the different configurations should be apparent. For example, in the all leads off and wrist leads on configurations, ECG signal acquisition is not possible. Therefore, a detection of these two configurations could lead to a powering down of the ECG signal chain in the analog front-end to save battery power.
On the other hand, whenever an all leads on configuration is detected, the ECG signal chain can be powered on. And the signal acquisition can be started. In addition to determining leads on and leads off configurations, it might also be useful to determine the strength of the contact to the leads. Such a determination could, for example, be used to alert the user to make better contact to the electrodes.
Lead detection requires biasing the leads to a potential, which could be high or low. Such biasing could be done either using a resistor or a current source. When a resistor is used for biasing, the other end of the resistor can be connected to a high side potential, V high, or a low side potential, V low.
When the lead is open, the pin gets pulled by the resistor to V high or V low. When the current source is used for biasing, the polarity of the current source either, source or sink, determines in which direction the pin gets pulled to when the lead is open. In addition to lead biasing, a mechanism is also required to detect the leads on and leads off condition. A pair of high side and low side comparators can be used to compare the voltage at the pin with a high and low threshold voltage respectively.
Lead biasing using the resistor or current source bias causes the potential in the leads to be different between a case of leads on and leads off. And lead detection using the comparators enables detecting the potential and determining whether the leads are connected or not. The mechanism of lead detection described in this slide is referred to as DC lead detection, since the bias is applied to the leads in a DC manner.
To understand how DC lead detection works to distinguish between the three configurations of lead connections described earlier, we will start with a case where all the leads are off. Let us consider the case of resistors biasing the leads. Here, we show the bias resistor R lead biasing both INP and INM to VSS, or ground.
Since the leads are open and there is no other mechanism to set the bias on the pins, the voltages at the INP and INM pins are solely determined by the lead bias potential. Therefore, the voltage at both INP and INM pins is equal to 0 volts. By setting the high and low threshold voltages for the comparators to voltages within the range of VDD and VSS, a leads off condition of both INP and INM leads can be determined. In this case, the low side comparators on both INP and INM read 1.
Next, we consider the case where the watch is worn on the wrist and the wrist leads are in contact with the user, but the electrode on the top face of the watch is not contacted by the figure on the right arm. Therefore, the left arm and RLD leads are on, but the right arm lead is off.
In this case, the INP pin is not connected. And so the R bias resistor on the INP pulls the pin to 0 volts. The RLD feedback loop adjusts the voltage on the RLD pin such that the common mode voltage at the INP and INM pins goes close to the target common mode voltage for the loop.
Let us assume the case where the target common mode voltage for the RLD loop is exactly equal to mid supply. Since INP is at VSS, the RLD feedback loop action would tend to drive the body to a potential close to VDD so that the INP pins would be close to VDD. And the average voltage of INP and INM would be at roughly mid supply.
The RLD buffer may or may not have the required range to drive the body all the way up to VDD. However, by adjusting the high side threshold voltage to an appropriate value, the high side comparator on the INM pin can be made to output a 1 in this configuration. Therefore, a wrist leads on configuration can be discerned when a 1 is output on the low side comparator on INP and on the high side comparator on INM.
Next, we consider the case where the watch is worn on the wrist and the user also makes contact to the right arm lead. In this case, all pins are contacted. The RLD feedback loop adjusts the voltage on the RLD pin such that the body is driven to the target common mode voltage, which in this case is equal to mid supply, or VDD by 2.
The VDD by 2 potential on the body in turn causes the common mode voltage in the INP and INM pins to be close to VDD by 2. Since the voltage on the input pins is well inside the range of the high and low comparator thresholds, all four comparators output a 0. Therefore, an all leads on configuration can be distinguished by all four comparator outputs reading 0.
Let us look at a design example that illustrates the choices of the comparator thresholds and lead detection currents and the implications of these choices on the threshold of contact impedance between a leads on and leads off condition. We take a case where the supply voltage, VDD, is 1.8 volts. VCM, the target voltage of the input common mode voltage for the RLD loop, is assumed to be VDD by 2, which is 0.9 volts.
We consider the current source-based lead biasing for this design example and assume a lead bias current of 30 nanoamperes. To illustrate both the high and low side through the same design example, we assume a sink lead biased current on the INP side and a source lead bias current on the INM side. The high and low threshold voltages for the comparators are set to 1.5 volts and 0.3 volts respectively. It is important to choose these threshold voltages outside the normal operating range of the ECG signal chain to avoid a false detection of leads off condition.
We start our analysis with assuming that all of the three leads are connected and the contact impedance of the electrodes is small. In this condition, the RLD loop drives the body to a potential low 0.9 volts. Since the lead bias currents flowing through the low electrode contact impedances result in a negligibly small voltage drop, the voltages at the INP INM pins can be assumed to be roughly 0.9 volts. Since this voltage is well within the high and low threshold, all the detection capacitors read 0.
Next, let us worsen the electrode contact, making the contact impedance progressively higher. This results in higher and higher voltage to be dropped by the lead bias current flowing across the contact impedances, causing the bias voltage at INP to drop and the bias voltage at INM to rise. When each contact impedance reaches a value of 20 megaohms, the bias voltage at the INP and INM pins reach values of 0.3 volts and 1.5 volts respectively. This causes the low comparator at INP and the high comparator at INM to read 1.
For the configuration described in the slide, a contact impedance of roughly 20 megaohms represents the threshold between a leads on and leads off condition. A more realistic case might involve contact impedance at the INP and INM sides to be of different values. Also, each electrode may have some DC offset. And this changes the exact value of R constant at which the leads off thresholds are reached. The basic principles of the operation outlined here can be extended to cover such cases.
Next, let's look at the concept of AC lead detection. In AC lead detection, the lead bias is applied as a switching, or AC, signal. An AC lead bias could be realized by having both source and sink lead bias current sources at the INP and INM pins and by alternately switching their polarity at a certain frequency.
In the illustration shown here, the polarity of the current source on the INP in switches from source to sink. And the corresponding polarity of the current source on the INM switches from sink to source. The AC lead current flows through an AC impedance network and develops a differential AC voltage at the ECG input pins that is superimposed on top of the signal.
The AC lead potential goes through the signal chain and appears at the output of the ADC. By appropriate digital signal processing of the ADC data, the strength of the tone at the AC lead switching frequency can be extracted, thereby giving a measure of the AC contact impedance of the leads. Note that in addition to extracting the strength of the tone at the AC lead switching frequency, additional filtering might be required to remove the tone at the AC lead switching frequency from the ECG output.
The capacitance, C parallel, shown in the figure represents the equivalent differential capacitance between the input pins when the leads are off. Such a capacitance could be a combination of the input capacitance of the ECG pins, the PCB trace capacitance, as well as the capacitance resulting from the electrode metal plates. The parallel R contact and C constant represent a model for the AC contact impedance. And Z body is a model of the impedance of the body between the two electrodes. For simplicity, we assume Z body to be negligibly small compared to the electrode contact impedance.
To illustrate how AC lead detection works, we will start with a case where the leads are off. Consider a case where the AC lead switching frequency is 1 kilohertz and the AC lead detect current has an amplitude of 10 nanoamperes. If C parallel is 50 picofarads, the AC impedance of C parallel at the AC lead switching frequency of 1 kilohertz is about 3 megohms. The 10 nanoampere lead detect current therefore generates about 30 millivolts of AC signal at the ECG inputs.
Depending on the ionate gain and the frequency response of the channel, this signal would appear at the ADC output as a tone of 1 kilohertz with some amplitude. When contact is made to the electrodes, a series combination of two Z contact impedances appears in shunt across Z parallel, thereby reducing the differential AC impedance across which the AC lead current flows. This causes the AC voltage at the ECG input pins to be at a much lower value as compared to the leads off case.
If we assume a case where Z contact has a value of 50 kiloohms at the AC lead switching frequency of 1 kilohertz, then the differential AC impedance across the ECG input pins is roughly 100 kiloohms. And the voltage swing at the AC lead switching frequency is reduced to about 1 millivolt at the ECG pushpins. Therefore, in a leads on case, the strength of the 1 kilohertz tone at the ADC output is significantly lower than for the leads off case. And this information can be used to determine the strength of the contact of the leads.
In this presentation, we have looked at the concept of DC and AC lead detection in wearable ECG systems. Lead detection is an important function in an ECG signal acquisition system. Determining the lead configuration can have many benefits, including a determination of which mode to operate the electronics in.
For example, when all leads are contacted, the ECG signal chain can be automatically enabled. And ECG signal acquisition can be started. Similarly, when contact to one or more electrodes is lost or becomes weak, either the user can be alerted to make better contact or the electronics can be shut off to save power.
Certain implementations of lead detection allow it to be active, even when the rest of the ECG signal chain is off. DC lead detection comprises lead biasing, using either resistors or current sources, to bias the leads to a high or low potential and comparators to distinguish between leads on and leads off conditions. The transition between leads on and leads off happens at a threshold value of electrode contact resistance. This threshold is dependent on the lead bias resistance, or current source value, and the detection threshold.
AC lead detection can be realized through switching current sources at the leads that generate an AC lead bias signal. The signal strength at the AC switching frequency appearing at the signal chain output can be used to determine the contact impedance of the lead. Additional signal processing may be required to remove the AC lead switching frequency tone from the ADC output.
This concludes our overview of ECG lead detection in wearable devices. For more information, visit TI.com/medial. Thanks for watching.