In this installment of the Joys of Low Voltage Audio series, I want to present some considerations and techniques to help you work with one of the biggest quirks found in RRIO opamps.
As I covered in a previous post, most opamps with rail-to-rail inputs use complementary NMOS/PMOS pairs in the input stage, with one handing off to the other depending on the input voltage. This leads to the possibility of crossover artifacts—just like crossover artifacts in class-B or class-AB output stages, but involving potentially more fragile input signals.
In many RRIO designs, both N and P pairs operate over a transition region, making them class-AB input stages. And while you might hope that the class-AB-ness would be enough to render the crossover effects negligible, apparently it’s not:
This transition region varies modestly with process variation. Within this region PSRR, CMRR, offset voltage, offset drift, noise, and THD performance are degraded compared to operation outside this region.
OPA2156 spec sheet, p. 19
So, let’s try to avoid the transition region, OK?
What follows is a case study using the OPA2156 for the design constraints presented by a low-voltage analog-to-digital converter project I’ve been working on. However the content is easily adaptable to other devices and contexts.
Transition region bounds
The transition region for the OPA2156 is characterized as being around (V+ – 2.25V) to (V+ – 1.25V). So if you power the device from a single 5V rail, the transition region is 2.75V to 3.75V. Further, if this circuit is biased at 2.5V, midway between ground and the rail, you’ll remain outside transition region as long as the audio signal remains below 0.25Vpeak. This is not unreasonable for some applications but certainly not all.
Avoiding the transition region
In my application, the opamp will be working as a unity gain, non-inverting filter, biased at 2.5V output, and with an in-band maximum signal amplitude of 2.8Vpp. The easiest way to achieve the needed 2.5V output bias is to bias the input by the same amount. Under these conditions, the input signal will have a range of 1.1V to 3.9V. This easily encroaches on the transition region, which suggests we should explore some ways to address this.
Rebiasing the input
One approach you might take to avoiding the transition region is to re-bias the input stage. If you bias the input at 1.65V instead of 2.5V, the input will span 0.25V (which is the device’s minimum output voltage) to 3.05V. This keeps the signal out of the transition region for most of the range.
This may be an improvement, but it comes dangerously close to the output stage limits, and it will require additional circuitry to achieve the 2.5V output bias required in the application. Still, this can be an easy way to move the signal into a less sensitive area, especially if the signal amplitude is modest and the output bias voltage isn’t important.
Adjusting the rail voltage
Another approach to avoiding the transition region is to increase the device’s rail voltage.1Many rail-to-rail opamps have limited maximum rails. Be sure to stay within the device’s limits!
Given Vtlo is the lower bound of the transition region and Vimax is the maximum input voltage, we want:
Vimax < Vtlo
Further, let
Vtlo = Vs – V1, where Vs is the supply rail and V1 is voltage between Vs and the lower bound of the transition region (2.25V in the case of the OPA2156), and
Vimax = Vb + Vip, where Vb is the input bias voltage (2.5V in this case) and Vip is the peak amplitude of the input signal (1.4V in this case).
Substituting yields:
Vb + Vip < Vs – V1,
and solving for Vs yields:
Vs > Vb + Vip + V1
For Vb = 2.5, Vip = 1.4, and V1 = 2.25,
Vs > 2.5 + 1.4 + 2.25 = 6.15V.
Note that this technique also lets you use RRO devices whose input range extends to to low rail.
The downside to this approach is that it will add complexity to the power supply design if the product is intended to operate off a USB input, a 5VDC adapter, or similar LV power source.
Conclusion
Computer scientists often talk about “leaky abstractions” in software design. The concept applies to the real world as well, and new real-world technologies tend to come with a different set of leaks than established ones.
In the case of RRIO opamps, one leak to be aware of is input stage crossover effects. Depending on the application, it may be possible to avoid regions where the input stage is crossing over by rebiasing the input or adjusting rail voltage(s).
Copyright © 2020 Mithat Konar. All rights reserved.