AD8042: unstable when used as G=+2 buffer

Q 

We use Analog DAC-AD5552. We use external Amp (AD8042) for a bipolar output as
written on the data-sheet (on page 9).
Without writing ant data we gote 5MHz signal at the bipolar output.
We use external resistors for bipolar the signal and we got no problem. I mean
that we use the Vout signal from AD5552 (pin number 1) and the same Amp
(AD8042) and didn't use INV pin or RFB. We use 2 resistors for feedback (28KOhm
resistors) and Vref from AD780 and got good result. We think that using the
internal feedback cause a resonance of 5MHz in the Amp.
We connect Analog ground to the DGND, AGNDF and AGNDS and the connection is
very close to each other.
Would you have any idea to fix the problem? What cause this resonance?
We changed the AD5552 part to similar one and use another company Op Amp and
get the same problem.

 

A 

The AD8042 is a high speed amplifier with unity gain bandwidth of 160MHz. High
speed amplifiers in general require fairly small values of feedback resistors
in order to remain stable.

Figure TPC 23 on the Rev.A datasheet for the AD8042 indicates that there is a
peak in frequency response when used with a gain of +2 and a feedback resistor
of 2kohms. The peak in the gain response would indicate a zero and an unstable
configuration.

I have enclosed a SPICE model output showing the frequency response of the
AD8042 with a gain of +2 for different values of feedback resistance. You can
see that for 28kohms there is a peak at about 20MHz. The fact that you see
oscillation at 5MHz could be due to board parasitic capacitance and part to
part variation.

Page 2 of the enclosed file shows the same frequency plot but with 5pF of
capacitance across the feedback resistor.

You can see that the 5pF compensation capacitor reduces the overall bandwidth
and eliminates the peak in the response – the amplifier is stable.

I suspect that when you use the AD8042 with external resistors, there is higher
parasitic on the board and this is helping to stabilize the amplifier. This
indicates a marginal design.

I would recommend one of the following:
· use the AD8042 with the internal 28k resistors of the AD5552 but add a 5 or
10pF cap between RFB and INV. A certain amount of self experimentation is
inevitable to find the optimal value of compensating cap.
· Use the AD8042 with lower value external feedback resistors - 200 ohm
feedback resistors would provide stable operation.
· Use a lower bandwidth amplifier such as the AD820. OP196, OP184.