Design of medical imaging system based on ADI 20-bit DAC

Healthcare is a hot topic today. The high performance AD5791 digital-to-analog converter (DAC) can improve application performance without compromise.

Governments around the world invest heavily in healthcare research and system development to protect the well-being of their people and ensure physical and mental health. This requires proactive prevention of disease rather than reactive response, as well as the correct diagnosis of some conditions.

Medical imaging systems are playing an important role in this situation. With the help of images, doctors can observe patients in more detail and understand the disease status without surgery. In addition, images can help surgeons study cases before starting surgery.

There are many different imaging methods available today, such as computed tomography, X-rays, ultrasound, and magnetic resonance imaging. Each system has its own advantages and disadvantages and can be used to produce still images of a body part or organ, or to produce dynamic images so that doctors can verify or study the movement of an organ. Dynamic images are also used in some surgeries.

The imaging capabilities of different systems also vary. X-ray technology is well suited for diagnosing bone diseases. Ultrasound uses sound waves to monitor the fetus and can image organs and blood flow in the atria, ventricles, and vessels. MRI is suitable for imaging soft tissue. For each of these medical imaging systems, Analog Devices has corresponding professional technical solutions. This article focuses on a new high-resolution DAC developed for high-performance applications such as magnetic resonance imaging (MRI).

Magnetic resonance imaging

MRI is mainly used to produce high-quality images of the inside of the human body. It can be used to detect diseases and distinguish tumors from normal tissue. 70% of the human body is fat and water, both of which contain hydrogen atoms. MRI uses the magnetic properties of hydrogen atoms to form images.

A strong homogeneous magnetic field is required to perform an MRI. The unit of magnetic field strength is Tesla (T). 1 Tesla is equal to 10,000 Gauss, and the Earth's magnetic field strength is about 0.5 Gauss. Current MRI systems use magnetic field strengths of 1.5 T to 3 T, and sometimes even up to 7 T. Such a strong magnetic field is generated by a superconducting coil magnet, and the patient is placed in the magnetic field. Figure 1 shows the position of the patient in relation to the MRI scanner coil.

Figure 1. The position relationship between the patient and the MRI coil

For a 1.5T system, the applied frequency is about 64 MHz, and for a 3T system it is 128 MHz. This causes the proton spins inside the body to be in a low-energy or high-energy state, either parallel or antiparallel to the direction of the magnetic field. The higher the magnetic field, the greater the energy difference between these two spin states. After the applied magnetic field is removed, the protons forward the magnetic energy, which is measured by a receiving coil or antenna. These antennas are designed with sensitive preamplifiers, gain blocks, and high-resolution ADCs to meet the overall dynamic range requirement of 120 dB to 140 dB. Since we are only interested in imaging small slices of the body, a gradient is added to this homogeneous magnetic field.

Figure 2. High-resolution gradient control loop

This gradient signal (magnetization vector) is transmitted using a large coil to provide a response from the single slice of interest. Figure 2 shows the gradient control loop implemented in an MRI system. The signal sent to the gradient coil is generated by an amplifier with an output power of several megawatts. The frequency range is quite low, so the key requirements are stability, high linearity, and low drift. This is exactly what the AD5791 20-bit DAC provides.

Why use a 20-bit DAC?

As mentioned above, the power required to drive the gradient coils of an MRI system is measured in megawatts. If a 2 MW amplifier is driven with only 16-bit accuracy, 1 LSB will equate to a step size of 30 W minimum! This is why a higher resolution DAC is needed. If designed properly, a 20-bit DAC can enable system performance to reach an accuracy level of 2 W/LSB.

The frequency of the gradient signal is only a few hundred Hz, so high stability, low short-term drift, and low noise are necessary to meet the overall requirements. To design an ultra-low noise, low-frequency system, it is necessary to carefully examine the components used. The filter will add noise and phase shift, so the selected signal chain components must be able to achieve good DC performance and low noise in the low frequency band close to DC. The AD5791 combines high resolution, high stability, and low noise, making it the best choice for this application.

A Closer Look at the AD5791

The AD5791 is a single-channel, 20-bit, voltage-output DAC. To achieve a high dynamic range, the device must operate with a high supply voltage, because the higher the supply voltage, the easier it is to get away from the noise floor. This is not a problem for the AD5791, whose supply voltage VDD ranges from 7.5 V to 16.5 V and VCC ranges from –7.5 V to –16.5 V.

The architecture of this DAC consists of a calibrated voltage-mode R2R ladder network. The thin-film resistors used to build the core of the converter provide excellent matching and stability. To achieve high linearity, the R2R ladder is divided into two sections. A 14-bit R2R ladder network produces the lower 14 bits (S0 to S13). The remaining upper 6 bits of the 20-bit digital code are used to drive an independent 6-bit DAC, which controls the reference voltage for the lower 14 bits. Together, these two sections form the body of an excellent multiplying DAC. Figure 3 shows the R2R ladder structure implemented in this device.

Figure 3. Main body of the R2R resistor ladder used in the AD5791.

The reference input voltage is selectable over a ±10 V range. With such a wide reference voltage range, the LSB level can be as high as 20 µV. This helps the converter maintain 20-bit (1ppm) integral and differential nonlinearity (INL and DNL), as shown in Figure 4a/b.

Figure 4a. AD5791 integral nonlinearity < ±0.6LSB.

Figure 4b. AD5791 differential nonlinearity < ±0.5LSB.

In addition to excellent linearity performance, other key features include 7.5nV/ѴHz voltage noise density, 0.6µVp-p noise (0.1 Hz to 10 Hz frequency range), and 0.05ppm/°C temperature stability.

What else should I consider for the MRI loop?

In an MRI gradient control system, the coils are driven with high precision and the response is measured by a high performance receive channel. Often, the weakest part of the loop determines the ultimate performance of the system. Previous systems were designed using multiple high resolution DACs in parallel, and averaging the DAC outputs reduced errors and improved absolute performance. The AD5791 provides a high precision 1 ppm DAC function in a single device, so no other tricks are needed to achieve the accuracy target. However, the DAC is not the only device in the signal chain, so attention must be paid to the other devices in the loop.

The DAC provides an unbuffered voltage output with a 3.4kΩ DAC resistor. The Johnson noise of the resistor ladder is the dominant component of the 7.5nV/ѴHz voltage noise density. To buffer the DAC output, an amplifier is needed to ultimately drive the high voltage power stage of the gradient amplifier in the system. High frequency noise is easily removed by RC filters, but filtering low frequency noise (typically represented by 1/f noise of 0.1 Hz to 10 Hz) necessarily affects the DC performance of the system. The most effective way to minimize low frequency noise is to use a circuit that never introduces this low frequency noise component. The guideline for the maximum allowable low frequency noise error of the entire system is 0.1 x the desired LSB level. For this particular application, the maximum error is 2 µVp-p based on an LSB level of 20µV. The most suitable amplifier is the AD8671, a successor to the OP27/37 that has an excellent 1/f noise of only 77 nVp-p and contributes very little to the noise of the entire signal chain. When using the AD8671 as a buffer amplifier at the DAC reference input and as a buffer for the DAC output stage, only 220 nVp-p of noise is added to the system. This, combined with the DAC’s 0.8 µVp-p noise contribution, yields a noise level well below the desired maximum level of 2.0 µVp-p.

Another important characteristic of this application is the drift performance of the system. Since the signal is measured and controlled at low frequencies, the drift is considered low frequency noise. The single-channel AD8671 and dual-channel AD8672 are also recommended amplifiers that can keep the drift performance within the required range. The single-channel AD8671 has a maximum drift of 0.5 µV/°C, which will contribute an additional output drift of 0.025 ppm/°C, resulting in a total drift of 0.125 ppm/°C. The dual-channel amplifier AD8672 has a slightly increased drift due to different heat dissipation conditions in the package and greater power dissipation. If additional gain adjustment is required, low temperature drift metal foil resistors are recommended. Last but not least, the accuracy of the system cannot be better than the accuracy of its reference voltage. References with built-in ovens are now available, which can maintain temperature stability and eliminate temperature drift. When the ultimate goal of the system is the highest performance, this reference should be considered. Figure 5 shows the circuit diagram of the entire output stage of the AD5791.

Figure 5. AD5791 and required amplifiers.

Although this article focuses on the high-resolution output stage for gradient control in MRI systems, the ADC signal chain in this loop is equally important to meet the overall performance requirements. Analog Devices offers a range of 24-bit Σ-Δ converters that combine high performance with high output data rates. The companion chip for the AD5791 is the AD776x family, which has an output data rate range of 312kSPS to 2.5MSPS and a dynamic performance of nearly 120 dB, complementing the DAC output.

Summarize

The chip industry is in a general trend of reducing supply voltage, power consumption, and package size, which is mainly driven by the market demand of consumer electronics, and portable and battery-powered systems require small size and low power consumption. This trend, combined with the growth of demand, has forced chip manufacturers to consider where to invest their resources. But as shown in this article, there are exceptions. Healthcare, industrial, military, and aerospace applications still pursue high performance and innovative technology. Analog Devices has shown with the high-resolution, high-performance 20-bit DAC AD5791 that high integration and miniaturization are possible without compromising technical specifications. The AD5791 belongs to a new series of digital-to-analog converters, and its introduction once again proves Analog Devices' global leadership in this market.