Early in 2011, Analog Devices and Texas Instruments introduced electrocardiogram (ECG) analog front ends (AFEs). TI’s ADS1298R and ADI’s ADAS1000 both illustrate the trend toward integrating the analog signal chain in mixed-signal applications and reflect the incredibly complex level of application-specific understanding that must be accounted for, along with the raw performance characteristics of the silicon.
The AFE for an ECG represents a non-trivial design effort. My old family doctor (retired now) had an office ECG that rolled around on its own cart. When he used it, he would have to turn off the fluorescent room lighting and the air conditioning to minimize noise pickup. Then he would carefully check each pair of leads for crosstalk and have me lie absolutely still while he ran a strip chart.
In contrast, the ADS1298R and ADAS1000 are designed to make ECGs as simple to use and inexpensive as possible. Similarly, they’re intended to make the design and manufacture of the instrument as easy (and as hard to get wrong) as possible. In that regard, recall that both TI and ADI make DSPs. Reference designs include full functionality, including power supplies and display drivers (see the figure).
This reflects the broader new paradigm for design and manufacturing. TI and ADI expect most of their design wins will be with Chinese OEMs and ODMs and that the largest part of the user base will be paramedics, not just in fire and ambulance services in the developed world, but also in places where university-trained medical doctors are few and far between.
In other words, these ICs represent potential large sales volumes for their creators—far larger than they would see if the ECGs were only destined for hospitals and group practices in developed countries—plus the opportunity to extend lifesaving medical diagnoses and life-extension practices to millions more people than have access to it today.
Even in developed countries, wider dispersion of advanced diagnostics may turn hospitals into less attractive targets for terrorist attacks.
Let’s look at the engineering. As my former GP’s precautions illustrate, getting an ECG AFE right, when you’re designing it for use by relatively unskilled users in unpredictable field environments, presents some design challenges.
A beating heart generates pulses with amplitudes of a few millivolts that cover a spectrum from 0.05 to 100 Hz. Interestingly, the differential voltages across different parts of the body are not simply manifestations of the same pulse at different amplitudes. Taken in the aggregate, they provide a multi-dimensional picture of what the heart is doing. It can be likened to looking at the beating heart from multiple viewing angles.
Doctors look at them as a set of parallel traces in the time domain. In short, an ECG is a multi-channel oscilloscope with multiple differential vertical amplifiers, plus a memory. Inputs represent the voltages between electrodes taped to different parts of the body. These can be widely separated, so the pairs cannot be twisted together or run through a shield. And to make things more complicated, some electrodes are used for more than one signal. For example, the maximum number of signals, which medical personnel call “leads,” is usually 12. In general, a “12-lead” ECG has 10 electrodes taped to various parts of the subject’s body.
The electrode on the right leg provides a reference. Others are placed on the right leg, the right and left arm, and at six places on the chest. The two arm and left leg electrodes form a common side for the measurements of the six chest electrodes, but there are three more separate leads displayed on the ECG representing the voltage between each arm and the right leg and the average of the other two limb electrodes.
In addition, a higher-end ECG will have other input signals, such as the “pace” signal, which is picked up from a pacemaker, if the patient has one. From a front-end design standpoint, it represents another common-mode noise source, along with the room lighting and air conditioner.
Another front-end design complication arises because all those electrodes are held on to somebody’s skin by little bits of adhesive tape. Not infrequently, one comes loose. This necessitates fault detection. To detect such a fault, the ECG measures the impedance between each of the differential-sensing electrodes and a separate “lead-off” electrode. Sometimes, this impedance measurement provides an input for respiration rate, which is measurable because thoracic impedance changes as the lungs fill and empty.
In the past, whenever designers set out to design an ECG, they had to factor all this into the design, along with signal conditioning, the interface to the analog-to-digital converters (ADCs), which are included the TI and ADI ICs, and the interface to any signal processing in the digital domain. That amounts to a very large chunk of development and test time, which must be amortized across a relatively low sales volume for the actual ECGs.
In contrast, all that nonrecurring engineering effort has been accomplished, optimized, and frozen in silicon by TI and ADI—long may they compete—with the potential for further incremental improvements that need not start with a blank slate.