Post Go back to editing

Hi team,

I have a question about the unity gain stability of AD8676. I am not sure if the AD8676 can be used for unity gain configuration. I am currently using it in a configuration similar to the circuit below.

Simulation Result

For AC Sweep settings I am using logarithmic from 1 to 5 Mhz with 1000 points per decade. It's a bit hard to see on this graph, but the curve starts to behave strangely before 1 Mhz. On the datasheet, it says the wide bandwidth is 10 Mhz typical. In addition, there is a graph of the closed-loop gain where the the curve starts to round off after 10 Mhz.

This leads me to guess that the AD8676 cannot be used for unity gain applications because it is not stable at unity.

I apologize if my questions seems confusing. I am still a junior at university so my understanding is limited. If you could answer my question it would be greatly appreciated.

Thank you,

Norman Guo

• Hi Norman,

AD8676 is unity gain stable. You can see the closed loop frequency response in the datasheet Figure 13 (G=1). The simulation you attached is actually G = -1V/V which should be even more stable (equivalent to G = 2V/V).

The peaking you show (assuming that's at the output?) is very minimal (50mdB) and is not normally an issue.

Regards,

Hooman

• Hi Hooman,

Thank you for answering my question. The line from the graph is my output. I am still confused on why the peaking occurs occurs at 100 Khz compared to Figure 13 (G=1) which has a -3db drop-off after10 Mhz. Shouldn't my the AC sweep be similar to the figure?

Thanks,

Norman

• Hi Norman,

Your simulation is with G = -1V/V (not G = 1V/V).

50mdB of peaking is considered subtlety and won't be adequately reflected in any macromodel.

Regards,

Hooman

• Hi Hooman,

I ran the simulation using G =1V/V and I am still getting a sharp increase after 1.0 Mhz. According to the figure

I am using this exact circuit(unity gain follower)

Shouldn't it start rolling off after 10 Mhz?.I performed the same simulations using an non-inverting amplifier for G= 10 and G=100 and they match the figure.

Thanks,

Norman

• Hi Norman,

A macromodel is an "approximation" of the device behavior. It won't be exact.

Your AD8676 LTspice simulation seems to have stopped at just above 10MHz. If you continue further in frequency, you'd see that after the ~1.8dB peaking predicted by the simulation, the output response does roll-off.

Many factors affect peaking including load resistance, layout, load capacitance, supply decoupling. The macromodel may or may not capture all these factors and again it's only a starting point approximation of the device behavior. Bench performance should be performed as the next step.

Regards,

Hooman