We found the noise performance of the majority of the API op amps in this class that we evaluated to be abysmal; measuring in at several nv√Hz in most designs. Some of the better devices specified in the 1 to 1.5 nv√Hz range, but they all seemed to achieve this noise performance by putting shunt inductors across the RE resistors in the input stage which “shorts out” the RE resistors and eliminates their noise contribution. While this technique yields lower noise, it does so at the expense of input stage linearity since the linearizing affect of the RE resistors are lost by the shorting inductors. RE resistors in the input stage are a good thing, and will linearize the input stage by swamping the non-linear Rbb of the input stage transistors. On the SS2590 discrete audio op amp, we opted to keep the input stage RE resistors in order to reap their associated linearity benefit, and we reduced the noise back down by paralleling the input transistors. It’s the best of both worlds, with the only drawback being a slight increase in input stage complexity.
The DC precision of the devices that we evaluated in this class left something to be desired as well. We saw offsets in the double digit mV range with warm-up drifts of several mV more. The SS2590 has a built in POT to allow precise offset adjustment and it has a warm up drift of less than 1mV. Input bias current induced offsets are greatly reduced by using on-board current sources to supply the input bias current for the device. This keeps the SS2590 discrete audio op amp from pulling input bias current from the outside world and creating offsets against the resistances seen by the input pins.
To put it another way, we did every thing on the SS2590 that we wanted to do on the SS3601 / SS3602 but were not able to due to its small size and more limited PCB space.