Hi,
I'm working on a new design where the THS4631 would make an ideal input front-end for a user-provided analog signal. Ideally, this would be AC coupled in using a 1 uF and a 1 Mohm resistor, giving a nice low 1s time constant. The circuit is a simple non-inverting Gain=2 configuration after the AC coupling stage. Power supplies are +/- 15 V.
The problem I see is that, when the input gets up around 1 MHz and >3 Vpp, a *BIG* dc bias current starts flowing out of the non-inverting input of the THS4631. For example, at 1 MHz and 3.75 Vpp input, I measure 1 microamp of bias current through the 1 Mohm resistor to ground. The effect seems to have a strong threshold for input amplitudes above ~ 3Vpp; below that the bias current is much closer to the spec sheet 100 pA.
Is this a known "feature" of the THS4631? It seems like some input protection diodes are lighting up. I've seen the same behavior on 2 parts, and it shows up in unity gain as well as G=+2 configuration. If I short out the 1uF AC coupling capacitor, then the offsets completely go away (my sig-gen is 50 ohms, and the microamp DC bias doesn't move it much).
Thanks in advance,
--Matt
Matt,
This is a known issue with the THS4631. The root cause is that the ESD diodes used to protect the input from differential over voltage are not symmetrical. This causes a small net current to flow when an AC signal is applied and this current increases with frequency. This current causes an offset voltage to develop on the external capacitor which is determined by the RC time constant.
We have tried several things to mitigate this effect, but have not seen good success. There was some promise by adding external diodes to the input pin, but I do not believe this prooved to be successful.
The only alternatives include THS4601 if > +/- 6V is needed, or if lower supplies can be used then OPA659, OPA657, and OPA656 may be an alternative.
Cheers,
Randy