SBOA524 July   2021 ALM2402-Q1 , ALM2402F-Q1 , ALM2403-Q1 , INA1620

 

  1.   Trademarks
  2. 1Circuit Theory
    1. 1.1 Overview of Parallel Amplifier Approaches
    2. 1.2 Considerations
    3. 1.3 Current Mismatch Equations
    4. 1.4 Stability
  3. 2Qualitative and Empirical Comparisons
    1. 2.1 Overview Comparison
    2. 2.2 Monte Carlo Comparison
  4. 3References

Monte Carlo Comparison

Additional simulations in PSPICE for TI were performed to compare the two circuit configurations. The test conditions were Vin = 5V, Cload = 20pF and Rload = 20Ω for a total Iload of 250mA. A 0.1% tolerance, 1Ω ballast resistance was utilized on each channel. However, one of the two channels was configured to have an additional 80mΩ trace resistance to the load, consistent with a ~5800mil trace of 25mil wide, 1oz copper on an outer PCB layer (or ~3100mils on an internal layer), to represent the two channels being spaced apart on the board. The offset voltage of each channel’s driving amplifier was modified to emulate a zero-mean, σ = 100uV case. Monte Carlo simulations (1000 runs) were performed using various component tolerances for the ballast resistors and gain-setting resistors, and the histograms of the current mismatch for the settled circuit were compared. The ideal mismatch value would be 0mA, meaning the load current would be evenly balanced with each driver amplifier sourcing 125mA.

As Figure 2-1 shows, it was found that while the current mismatch of the conventional buffer circuit was centered on about 9.32mA (σ = 0.43mA), the mean of the current mismatch for the parallel improved Howland pump circuit (with RfX / RinX = 0.1 V/V) was significantly closer to the desired value of 0mA.

Figure 2-1 Comparison of Current Mismatch Histograms for Circuit Approaches

In the case where 0.1% tolerance resistors were utilized for the gain setting resistances of the parallel improved Howland pump circuit (as shown in Figure 2-1), the mean was -15.6µA with a standard deviation of 1.2mA, compared to a mean of 75.6µA and standard deviation of 11.1mA when 1% tolerance resistors were used for RinX and RfX. This highlights the importance of utilizing high-quality resistors for these components when building a parallel improved Howland pump circuit, to better control the distribution of the current mismatch. The overall results support the conclusion that the parallel improved Howland pump circuit is significantly more immune to trace mismatch, and that provided well-matched components are used and the pumps are in attenuating gain, this approach can in many cases result in superior circuit performance when compared to the conventional buffer arrangement.