24-Bit Sigma-Delta ADC Reduces the Requirement for an Analog Front-End (AFE) for ECG/EKG Applications

要約

This application note describes the essential workings of an electrocardiogram (ECG). It discusses factors that disrupt the ECG signals and make reliable, highly-accurate electrical characterization difficult. The industry-standard solution for ECG electrical characterization, which uses an analog front-end and ADC combination, is explained. The article then introduces the MAX11040K simultaneous-sampling, sigma-delta ADC as a compelling, highly integrated solution that eliminates the need for the AFE, and saves both space and cost for the application.

Introduction

After electrodes are placed on the skin at opposite sides of the heart, an electrocardiogram (ECG or EKG) records the electrical activity of the heart over time. The ECG displays the voltage difference between pairs of the electrodes that register heart muscle activity. This display indicates the overall rhythm of the heart and can be used by a medical professional to detect weaknesses in different parts of the heart muscle.

Actual ECG signals are in the few mV range with frequency content of no more than a few hundred hertz. ECG measurement is a challenging task because of artifacts such as: capacitive coupling of 50Hz to 60Hz from the ECG's mains, which overpowers the signal of interest; the body's skin-contact impedance and mismatch between sensor impedance, which result in offsets and lack of common mode rejection; and contact noise and other interference from electronic and magnetic sources.

The widely published approach to extract these signals uses an analog front-end (AFE) for signal amplification and filtering, followed by a 12- or 14-bit ADC for data acquisition. This application note explains essential AFE components for the ECG application. The article then presents a highly integrated device, the MAX11040K 24-bit simultaneous-sampling, sigma-delta ADC. The MAX11040K supplies all the essential specifications demanded for the application and eliminates the need for the AFE.

Essential AFE Elements

The analog front-end consists of three or four main components (Figure 1).

Figure 1. Typical ECG solutions use an AFE for signal amplification and filtering, and an ADC for data acquisition.

Instrumentation Amplifier (IA)

The main task for the instrumentation amplifier (IA) is to reject the common-mode (mainly 50Hz/60Hz) signals. The ECG application requires a common-mode rejection ratio (CMRR) of 90dB or higher to remove the 50Hz/60Hz (coupling from the mains) sufficiently before the gain stage. Even though you may have an IA with a high CMRR, other factors such as mismatch between differential (ECG) electrodes or skin contact impedance result not only in offset drifts, but also in less-than-ideal CMRR. This mismatch in impedance is primarily caused by physical skin contact, perspiration, and movement of muscles.

The secondary task for the IA is amplification (gain). When setting the gain of an IA, however, care must be taken not to amplify the gain so much that it causes clipping or saturation.

It should also be noted that the band of interest for audio is not the same as ECG. Consequently, typical audio amplifiers and sigma-delta ADCs for the audio market are not well-suited for this application, as these devices have higher input referred noise in the frequency band of interest.

The IA's input impedance specification is also important, because the ECG is measuring weak low-level signals. An IA with higher impedance is usually recommended as lower input impedance will result in higher attenuation of signals.

Highpass Filter

Although the initial signal of interest was in mV, it will rise up to 10s of mV after the gain of x5 or x10 from the IA. This gain, however, only covers a very small portion of the ADC's input range. For example, a 12-bit ADC with a ±4.096V input range will have an LSB of 2mV and not have enough resolution to distinguish between signal and sampling noise. Consequently, before the signal can be amplified again, the unwanted DC drift signals must be eliminated. The common AFE strategy uses a highpass filter to feedback the unwanted (low-frequency) signal as a negative offset to the IA.

Second Amplifier Stage

With the DC and other low-frequency signals removed by the IA and highpass filter, a second amplifier applies additional gain to the signal to match the input range of the ADC. Some designs also add a notch filter here for further 50Hz/60Hz rejection.

Lowpass/Anti-Aliasing Filter

A lowpass filter is included basically to reject higher-frequency signals. It also acts as an anti-aliasing filter (i.e., to stop any frequency content above Nyquist or f_sampling/2 of the ADC from folding back in).

To further reduce the common-mode signals at the input, ECG designs commonly have a "right leg driver" which drives the opposite of the common-mode signal back to the body. To ensure the safety of the patient, this task is typically done with an operational amplifier and a current-limiting resistor, so a very weak signal source is driving into patient's body. A shield driver is also added sometimes to reduce coupling of external noise into the signal carrying ECG probes.

To summarize, the signal of interest for the ECG application is less than 100mV, but a 2V range is needed because of offsets and common-mode signals. Therefore, the AFE must have the 2V range and distinguish down to 100s or even 10s of µV at a sampling rate around 1ksps.

The Right ADC Reduces/Eliminates the Need for an AFE

With the AFE selected, many ADCs will now match the application's requirements for resolution, speed, and input range. Consider, however, an alternate solution. There is an ADC with enough resolution, CMRR, and other exceptional features that make it very suitable for ECG-type applications.

There are several reasons why the MAX11040K simultaneous-sampling, sigma-delta ADC surpasses the minimum requirements for this application. By replacing most, if not all parts of the AFE, it also gives you better reliability, a smaller package, and a simpler design.

Several MAX11040K specifications are essential for this application:

Input range ±2.2V

Differential inputs

110dB CMRR (typ)

24 bits:

SNR > 110dB
Around 19 bits of noise-free range
Effective resolution = 2/2¹⁹ = 3.8µV

±6V input overvoltage protection

Four differential-channel, simultaneous-sampling ADCs

Cascadable for up to 32 channels of simultaneous sampling

Programmable output data rate

Serial interface (easy for safety isolation)

Programmable phase (subdata rate)

Overvoltage/fault detection

Figure 2 illustrates how the application can be simplified with the use of the MAX11040K. With differential inputs and a CMRR of 110dB to reject the 50HZ/60Hz coupling from the ECG's mains, the MAX11040K performs the first function of the IA. With its 24-bit resolution and a noise-free range of ~19 bits, the MAX11040K has enough resolution in the full ADC input range to capture signal variation down to a few µV. This effectively eliminates the need for first-stage amplification (the IA's second function), the second-stage amplifier, and the highpass filter. As an added benefit, the device's ±2.2Vinput range is also a good match for the application.

Figure 2. The same ECG application can be achieved with only the MAX11040K ADC. The reduction in components will save board space and lower overall solution cost.

Although the sampling rate for the MAX11040K is 3.072MHz (oversampling sigma-delta), the output data rate (i.e., the effective sample rate) is programmable from 64ksps down to 250sps for added system flexibility. The device has smooth errors (i.e., INL error vs. input compared to SAR devices) for small signals; as a sigma-delta ADC, it also eliminates the requirement for anti-aliasing filters. For added information on sigma-delta ADCs, please see application note 1870, "Demystifying Sigma-Delta ADCs."

Two other MAX11040K features are also ideally suited for the ECG application: simultaneous sampling and programmable phase delay. In the world of ECG when in-depth analysis is needed, a 12-sensor ECG is preferred and importance is placed on maintaining phase integrity. Each MAX11040K provides four differential channels, which is equivalent to eight probes. In fact up to eight MAX11040K devices can be cascaded together to provide 64 simultaneous-sampling probes. Not only can the channels sample simultaneously, but each phase is also programmable with the resolution (0 to 333µs in 1.33µs steps) finer than the output data rate (up to 64ksps) of a single MAX11040K device.

A companion device, the 16-bit MAX11046, also features ±6V input protection and sends out an overvoltage (FAULT detection) alert whenever the signal goes beyond 88% of the input range. A Serial Peripheral Interface (SPI) reduces the requirement for multiple optoisolators and eliminates the need for an extra power regulator because it runs both digital and analog off the same supply.

Test Results for the MAX11046 Solution

Figure 3. The ECG application was tested on a MAX11040K EV kit.

Figure 3 is the block diagram of the MAX11040K evaluation (EV) kit used to test the application design. The kit contains two MAX11040K devices, which are set up to run as a single 8-channel simultaneous acquisition system. The EV kit plugs onto the PC with USB and also contains memory and DSP for advanced project development.

The experiment was conducted by just adding copper plates to conduct ECG signals, and with 22kΩ series resistors between the ADC input and the user's hands. With an input impedance of 130kΩ (based on a X_IN clock frequency of 24.567MHz), the signal should be attenuated by 75% (Figure 4). Test results are seen in Figure 5.

Figure 4. Effective input impedance of the MAX11040K.

More detailed image (PDF, 745kB)
Figure 5. Collected EKG data.

Conclusion

At the time of this publication, the MAX11040K ADC is the only device available in the market that meets the requirements for ECG measurement and provides other highly desirable features at no extra cost. The MAX11040K will reduce your R&D budget, design time, board area, and the design's component count. It will also increase solution performance and reliability.