Digitally controlled amplifiers redefine analog design

Digital designs have been encroaching on the analog domain for a long time with firmware-controlled internal timers, comparators and input/output gates, in conjunction with external resistors and capacitors.

Some of the analog functions that are easily realized in this environment are the digital-to-analog converters (DACs), analog-to-digital converters (ADCs) or integrators. As primitive as the implementation of these functions can be, they do get the job done for applications for which the accuracy and fidelity of high-precision analog functions are just not needed.

But, now, the engineering community is asking that these compromises be eliminated. For every technological step forward, a paradigm is always challenged. The old paradigm professes that precision analog cannot co-exist on the same chip with digital controllers or processors. This limitation has been a process issue. The new paradigm on the horizon is that precision analog can be integrated onto the digital chip. With this paradigm shift come three changes. These stepping-stones bring the two domains together on the same chip with the creation of precision analog functions that are microprocessor- and microcontroller-friendly. The result is looking very much like a system on a chip (SoC).

Changing the process from bipolar to CMOS is the first and most important shift for this new paradigm. There are more and more analog circuits migrating from bipolar processes into the CMOS world.

In addition, the performance of these CMOS devices is improving. For instance, operational amplifiers that were designed with CMOS processes were known to have substandard common-mode rejection, power-supply rejection and offset voltage. Now these specifications are more within required ranges. Successive approximation register (SAR) ADCs have migrated from an R-2R-ladder topology to the input capacitive-array arrangement.

This topology change reduced the silicon size and was synergistic with the CMOS process. The majority of the silicon in a delta-sigma ADC is dedicated to the digital circuitry and has always been designed in CMOS processes. This high-precision device is a perfect candidate for the controller/processor chip. In conjunction with this migration of analog functions from bipolar to CMOS, the CMOS processes are tightening up and the IC designers are continuing to design with innovative improvements.

Another controller-friendly feature is the programmable-analog device. This is not the classical analog definition of programmability, where a resistor is changed in the hardware. This breed of analog programmability is achieved with on-chip non-volatile or volatile memory. With this change in perspective, the nichrome resistor and zener-zapping analog networks are abandoned in favor of the digital RAM, EPROM and EEPROM cells. Non-volatile digital memory "burns" in permanent changes electronically during the final manufacturing step.

This approach reduces the overhead costs of laser trim equipment and yield losses at the wafer level. Additionally, changes can be implemented on the fly (using volatile memory) during system operation, which ultimately produces a device that works in a larger number of applications.

The third — and most critical — shift in this paradigm change is that digital houses are now building their analog muscles by including analog content in their established digital product lines. At first glance, this migration does not seem to be that difficult; however, when you take into account the differences between analog and digital disciplines, there is a significant culture re-alignment that occurs on both sides.

An example where a standard amplifier can be added into the digital circuit is shown (see figure). In days prior to using digital memory in analog circuits, the operational amplifier was designed in hardware to one bandwidth, one quiescent current and one offset voltage. Although the operational amplifier is known for its flexibility in a variety of applications, the rigidity of these performance specifications in fact locks a single amplifier into a small set of applications. Now that the operational amplifier has come into its own in the single-supply CMOS process, these functions can be manipulated with a simple stroke of a key: firmware programming.

A CMOS operational amplifier can be designed to take advantage of nonvolatile fuses.

An example of the ingenuity that is being applied to this challenge is illustrated in the accompanying figure. This simple example of an operational amplifier uses the CMOS process in conjunction with non-volatile EPROM switches.

The switches illustrated in this figure are used in the active load of the differential input pair of the amplifier. The offset voltage of the amplifier is adjusted by using the switches to steer current through one or the other side of the differential input-pair. These switches can be electrically accessed in a test mode during final test. The benefits that have been gained because of this approach are higher yields, tighter specifications and on-the-fly programmability.

Until now, the analog approach changes these currents by adjusting the load using nichrome laser trimming or zener-zapping. These analog processes can damage the passivation area of the silicon chip, and they are not synergistic with the CMOS digital process. EPROM switches do not compromise the integrity of the chip. Their reliability is established by the work done over the years with memory devices and controllers/processors. These types of switches are also used to implement amplifier bandwidth or quiescent current changes, to name a few.

Try as we may, we are finding that analog will never go away. Standing at the hardware/firmware wall, it appears possible for complex digital circuits to co-exist with high-performance analog functions. The advantage of the digital device has always been "one size fits all." With a simple adjustment of the code, these devices can be applied to dramatically different markets.

Now, we are entertaining analog circuits that also have that "one size fits all" feature. These analog devices must be economical, efficient, compact and streamlined for multiple applications.

Historically, the microcontroller and microprocessor devices catered to horizontal markets and analog to vertical markets. These two domains are now positioned to come together because of the power of digital programmability, making SoCs more attractive. The only thing missing is to determine what analog functions are needed and when will these functions come together under one roof?

Bonnie C. Baker is analog/mixed-signal applications engineering manager at Microchip Technology Inc. (Chandler, Ariz.).