S
server
Guest
Hi all,
My question is about precision transimpedance amplifiers and, specifically, what properties of the op-amp dictate the drift characteristics of the TIAâs output signal.
Suppose we take as an example a TIA made with the OPA847, which is not a standard high precision amp and is instead chosen for its GBW. The gain resistor is taken as 10k, and we have 100uA of photocurrent, for a voltage output of 1V. Letâs also assume for a minute that the tempco of the feedback resistor is negligible. Iâm interested in understanding the op-amp specifications that are important for minimizing the drift of this 1V output.
The usual specification of interest for high precision amps is the offset voltage, which is 100uV and exhibits a drift of 0.25uV/deg. The offset voltage is of course gained up by the circuitâs noise gain. However, considering that the shunt resistance of the photodiode is infinite compared to the 10k gain resistor, the noise gain is unity at DC. Thus, we have a 0.25uV/deg drift on the output. This works out to be 0.25ppm/deg, which is very good. Note that I donât care about the output accuracy but rather that it doesnât change.
As noted in Jerald Graemeâs book, the offset voltage and its associated drift also appear across the photodiode and tickle its reverse biasing.. This changes the dark current that is amplified that is amplified along with the photocurrent of interest. At a typical reverse bias of -5V, the dark current of my standard photodiode is ~40pA and possesses a linearized bias sensitivity of ~5pA/V. Assuming 100uA of photocurrent and the offset voltage drift of 0.25uV/deg, this effect contributes at less than the ppt level and is insignificant.
It seems to be that offset voltage stability is not an issue, and the relative contribution of offset voltage drift can always be cured with additional transimpedance gain.
The next issue would be the input bias currents of the TIA. Since the OPA847 has a bipolar frontend, it exhibits larger bias currents and drifts. The non-inverting terminal of the amp is connected to ground via a resistor matching the gain resistor of the TIA, so we should only have to consider input current offsets (0.1nA/deg). However, even if we ignore this matching and assume a worst-case scenario for the input bias current (19uA) and drift (15nA/deg), the drift contributes at the 15nA/deg/100uA = 150ppm/deg level..
It seems to me that the op-amp properties of bias current drifts are only germane to low-light applications, i.e., photocurrents on the order of the bias current offset drifts. Again, Iâm not concerned about an overall 100uV offset of the output signal. I just want to make sure that it does not change over time. The effects of voltage offset drift and bias current offset drift seem to be smaller than the tempco of the feedback resistor itself, e.g., 100ppm/deg for a ~1% resistor.
That being said, why do I always see app notes and articles in the scientific literature where people jump through all sorts of hoops to have TIA frontends specifically designed to minimize this drift, including the use of chopper-stabilized and zero-drift amps. I feel like my analysis is either wrong or incomplete. What am I missing ?
Thanks in advance for your help and input !
-Jon
My question is about precision transimpedance amplifiers and, specifically, what properties of the op-amp dictate the drift characteristics of the TIAâs output signal.
Suppose we take as an example a TIA made with the OPA847, which is not a standard high precision amp and is instead chosen for its GBW. The gain resistor is taken as 10k, and we have 100uA of photocurrent, for a voltage output of 1V. Letâs also assume for a minute that the tempco of the feedback resistor is negligible. Iâm interested in understanding the op-amp specifications that are important for minimizing the drift of this 1V output.
The usual specification of interest for high precision amps is the offset voltage, which is 100uV and exhibits a drift of 0.25uV/deg. The offset voltage is of course gained up by the circuitâs noise gain. However, considering that the shunt resistance of the photodiode is infinite compared to the 10k gain resistor, the noise gain is unity at DC. Thus, we have a 0.25uV/deg drift on the output. This works out to be 0.25ppm/deg, which is very good. Note that I donât care about the output accuracy but rather that it doesnât change.
As noted in Jerald Graemeâs book, the offset voltage and its associated drift also appear across the photodiode and tickle its reverse biasing.. This changes the dark current that is amplified that is amplified along with the photocurrent of interest. At a typical reverse bias of -5V, the dark current of my standard photodiode is ~40pA and possesses a linearized bias sensitivity of ~5pA/V. Assuming 100uA of photocurrent and the offset voltage drift of 0.25uV/deg, this effect contributes at less than the ppt level and is insignificant.
It seems to be that offset voltage stability is not an issue, and the relative contribution of offset voltage drift can always be cured with additional transimpedance gain.
The next issue would be the input bias currents of the TIA. Since the OPA847 has a bipolar frontend, it exhibits larger bias currents and drifts. The non-inverting terminal of the amp is connected to ground via a resistor matching the gain resistor of the TIA, so we should only have to consider input current offsets (0.1nA/deg). However, even if we ignore this matching and assume a worst-case scenario for the input bias current (19uA) and drift (15nA/deg), the drift contributes at the 15nA/deg/100uA = 150ppm/deg level..
It seems to me that the op-amp properties of bias current drifts are only germane to low-light applications, i.e., photocurrents on the order of the bias current offset drifts. Again, Iâm not concerned about an overall 100uV offset of the output signal. I just want to make sure that it does not change over time. The effects of voltage offset drift and bias current offset drift seem to be smaller than the tempco of the feedback resistor itself, e.g., 100ppm/deg for a ~1% resistor.
That being said, why do I always see app notes and articles in the scientific literature where people jump through all sorts of hoops to have TIA frontends specifically designed to minimize this drift, including the use of chopper-stabilized and zero-drift amps. I feel like my analysis is either wrong or incomplete. What am I missing ?
Thanks in advance for your help and input !
-Jon