In the third installment of this series, I want to talk about one of the device types you’re almost guaranteed to encounter in low-voltage audio: the rail-to-rail opamp. More specifically, I want to talk about the terminology around rail-to-rail opamps and the techniques used by IC designers to achieve rail-to-rail behavior. So let’s start by answering a fundamental question.
What is a rail-to-rail opamp?
A rail-to-rail (R2R) opamp is an opamp that is able to do either or both of the following:
- Produce an operable output signal that is essentially as large as the rails at which it is operating.
- Operably handle input signals that are essentially as large as the rails at which it is operating.
By “operable output signal” I mean one that falls within some reasonable THD specification under a range of typical loads. Similarly, by “operably handle input signals” I mean that the THD remains below some reasonable threshold and input bias current and the like don’t change dramatically. “Essentially as large as the rails” is open to broad interpretation but is typically considered to be within a quarter of a volt of either rail.
This means a rail-to-rail opamp can be rail-to-rail only on the output, rail-to-rail only on the input, or both. I have personally never encountered an opamp that is rail-to-rail on the input only, and I suspect no such device exists: Since opamp circuits are run at a loop gain of unity or greater, it’s not terribly useful to have a device that can handle an input signal that reaches a rail but then clips it in subsequent stages.
This means there effectively are only two categories of rail-to-rail opamps:
- Rail-to-rail output (sometimes referred to as RRO)
- Rail-to-rail input and output (sometimes referred to as RRIO)
If you don’t need a device with rail-to-rail input capability, the range of candidate devices will increase significantly. RRO opamps will specify their usable input range to help you determine whether it suits your application.
RRO opamps commonly use any combination of typical transistor technologies: BJT, JFET, and (C)MOS. A majority of RRIO opamps on the other hand are CMOS-based.
Neither RRO nor RRIO devices have to be specified to work at low voltages (c. 5VDC), though most are. Some devices, especially RRO types, are specified to work at high voltages (c. +/-15VDC). In fact, many recently introduced opamps specified for high-performance audio are RRO devices even though they are not marketed as such.1Check the output specs for TI’s OPA1611 and OPA1642 for example.
R2R output stage design techniques
“Classic” opamps typically use an output stage based around emitter/source follower or quasi-follower topologies. This yields an output stage that inherently has low output impedance and is very linear. Because of this, the output characteristics are not widely affected by loop gain. However, because of the VBE or VGS drop in the output transistors, the maximum output level in either direction is limited to at least one VBE/VGS of the rail.
R2R output stages on the other hand typically use common emitter/source outputs. This results in an output stage with gain, which is how the rail-to-rail requirement is satisfied. But it also produces an output characteristic that is more dependent on loop gain. In many devices this is mitigated somewhat by the use of transistors having carefully designed characteristics as well as by internal feedback of some kind. Still, this remains a “feature” of R2R output stages. R2R opamp datasheets often characterize the device’s open loop output impedance over a range of frequencies to help you decide whether a particular device is appropriate for your application.
R2R input stage design techniques
I am aware of two techniques used to produce R2R input stages. One is the de facto standard, and the other is available as a proprietary technology.
The de facto standard architecture for R2R input stages involves using two complementary input stages connected in quasi-parallel. A typical PMOS differential input stage by itself will handle a wide range of input signals, but it will start to crap out when the common mode signal gets to within about 1.8V of the positive rail. Similarly, a typical NMOS input stage can handle a wide range of input signals, but it will start to crap out when the common mode signal reaches about 1.8V above the negative rail.
The de facto standard R2R input stage architecture solves this by using a complementary pair of parallel connected differential input stages and biases them such that one hands off to the other before it reaches its crap-out zone. To ensure a smooth transition, there is typically a range where both stages are active.2I suspect this is the reason CMOS is the technology used most often for RRIO opamps. CMOS processes are well developed and probably provide a reliable and inexpensive means of making the complementary pairs.
So what we are talking about here is a class-AB input stage. However, unlike typical class-AB output stages, where the crossover happens in the center of the signal, the transition between the NMOS and PMOS circuits in R2R input stages is often designed to take place as close to one of the rails as possible. This ensures that crossover effects will be limited to only high-amplitude signals.
The most pernicious crossover effect in these input stages is a change in input offset. This happens as one input circuit (with its inherent offset) hands off to the other (which has a different inherent offset). It presents itself as a signal-dependent offset, which means it’s a source of nonlinearity. The exact amount of offset change varies widely—even for different samples of the same device. So it’s hard to design around it. Processes refinements seem to be tightening things up though, so I’m hoping future generations of devices will get increasingly better in this regard.
Other crossover effects may include shifts in capacitance, which will be of concern if you are driving the device from a high source impedance or are forced to use high loop impedances.
To eliminate the crossover effects found in the input stage architecture above, TI developed what they call the Zerø-Crossover topology.3See TI’s whitepaper for a more complete description. This approach uses only one input stage circuit, based on a PMOS differential pair. To keep the input stage from crapping out as the common mode signal approaches the positive rail, the Zerø-Crossover topology places a regulated 1.8V charge pump between the positive rail and the current source driving the differential pair. This results in the input stage actually operating from 1.8V above to positive rail.
How well this works is entirely down to how good the charge pump is. Specs for the OPA365 and OPA2365 Zerø-Crossover opamps specified for audio use are decent, which suggests they’ve done a good job. (I am leaving any subjective listening assessments out of this discussion.) To my knowledge, no one else has implemented a similar design.
I hope this post has given you enough information to start selecting the best candidates for your application. While I don’t want to discuss my subjective listening impressions regarding particular devices, I do want to mention that to my ears the subjective quality of R2R opamps is even less predictable from objective measurements than is the case for regular opamps. So be sure you find as many candidate devices as you can and make subjective listening impressions at least one criterion in your product search.Copyright © 2020 Mithat Konar. All rights reserved.