Tuesday, February 12, 2008

Amplifier Design Challenges

There are two major constraints for designing an OLPC peripheral: cost and power. The laptop project has prescribed a $10 cap for all devices, and of course any design must be low power so as not to drain battery life. Right now I am designing the ECG amplifier for the XO, and I am at the stage where some decisions must be made regarding the trade-offs between performance and the cost and power constraints. The rest of this post consists of technical details, so read on only if you are interested in the engineering design aspects of the project.

The ECG will consist of an amplifier box with an output plug going to the microphone jack and a set of two or three reusable electrodes as inputs (electrodes are an important technical consideration that we will leave for another post). The ECG application will be an extension of an existing oscilloscope program which can visualize audio jack input. The box will have a separate jack for power supply, probably from the laptop USB port.

This design is in contrast to the other possibility of microcontroller-based device, which we felt presents too many challenges in power consumption and cost. However, we support the idea of different groups working simultaneously on different approaches to the problem, and would welcome discussion with other teams that share our overall goal.

Unless a separate power supply is designed, the USB port of the laptop is the most likely candidate. I knew that mic jacks supply a bias voltage, but a quick check confirmed that the voltage is too low and output resistance is way too high to think about using the bias to power a circuit.

The USB port provides 5V at a maximum of 1 amp (over all ports). Of course we want to minimize the power and so won't come close to that kind of supply current. The bigger question, though, is how (and whether) to split the supply for dual-supply op-amps. It is easy to split to +/- 2.5V with a voltage divider, but for highest CMR (common-mode rejection), the reference terminals on the instrumentation amp require low impedance, requiring an additional op-amp to buffer the new reference voltage. Although if sources of interference (i.e. 50/60 Hz power lines) are sparse or our signal processing is good then this might not be required. Single-supply op-amps can be used, but single-chip instrumentation amplifiers (discussed next) generally require a split supply.

Single-chip instrumentation amps (in-amps) are probably the best bet for cost, power and performance. These chips have a full 3 op-amp circuit inside, with a single external resistor setting the gain. All the resistor trimming is taken care of on the chip so component tolerance is not a big problem. The most common in-amp chip is the AD620, which provides a nice combination of cost, power consumption, and performance. However, I am looking at others, such as the INA126, which has 2 internal op-amps and is cheaper with comparable performance. Luckily they have the same pin-out and we can continue with the rest of the design independent of this choice.

Noise/interference suppression
Some additional modifications can be made to improve CMR, the primary example being a driven right leg (DRL) circuit. One op-amp can be used to tap the average of the two signal electrodes and drive the ground electrode with the inverse of this average. This is an effective way to subtract common-mode interference before it has the chance to be amplified. However, the question in this case is whether interference will be a huge problem in the environments the laptop is meant to be used. I am leaning toward including the DRL circuit, since it is hard to predict what the noise levels of the environment will be. This also applies to the question of whether to buffer the split supply, mentioned above. Again my inkling is yes on this as well.

Second stage
Is it necessary to provide a second stage of amplification? The best performance can be had using a low gain on the in-amp chip, then adding a high gain second stage after the signal is high-pass filtered. However, the mic jack provides a coupling capacitor and the audio card has up to 30dB of preamplification on the mic input, so this might be an opportunity to cut costs and let the laptop do the work, freeing up money and board space for better noise performance.

These are only a few general questions for now. In the future we will have more technical details on this as well as electrode design, modularity of the system, and software interface.

No comments: