OPA657 Transimpedance amplifier design help

Ankur Shah

Other Parts Discussed in Thread: OPA657

Hi,

We have a device in production which has a single stage Transimpedance Amplifier OPA657 in Die form. (Feed back resistor Rf = 10k ohm)

The question is:

If I connect the output of the device to a

1) Oscilloscope: The power supply draws +/-10mA of current irrespective of the Load (50 ohm or 1M)

2) Spectrum Analyzer GSP810: Same observation as DSO

3) DSA 3561A: The power supply draws +40mA/-10mA of current as soon as I connect my device to the Input.

Also, if I change Rf from 10k to 50k, the DSA 3561A draws +/-10mA. We dont understand why?

Any suggestions will help. We use DSA3561A to measure Noise density for our device.

We have to deliver this assignment y next week and hence, it is very critical.

Thanks,

Ankur

over 12 years ago

0 Paul Grohe over 12 years ago

TI__Guru 61796 points

Hello Ankur,

By "DSA 3561A", are you referring to a Hewlett Packard 3561A Signal Analyzer (scope-size with a green screen?).

If so, the output has an internal current source designed to power certain censors. Make sure the "ICP Current" light is off. If it is on, then the analyzer is trying to source current through the output. There is a menu setting to turn ICP current on and off.

If that is not the problem, make sure the "ground" switch (under the input connector) is not set to "Floating". This can increase the input capacitance.

Otherwise, the input capacitance of the 3561 is about 15pF - which may be enough to cause instability. Try adding a 1K resistor in series to isolate the input capacitance.

Here is some info about the nV/Rt/Hz on the 3561 (it’s well hidden):

To run the analyzer in Vrms–per–root–Hz, not Vrms (dBv). You have to dig for the V/RtHz setting – it is a math function called “Noise Density”.

DEFINE TRACE (hard key) -> MATH FCN SELECT -> SPECTRAL DENSITY -> ENTER.

The space above the graph should now read “Math: SQRT(MAG^2 / BW)”.

Then put the vertical axis in Linear.

VERT SCALE (hard key) -> LINEAR

And I recommend setting the horizontal units to LOG scale – so it looks more like a noise graph – and you can see the details you can’t see in linear.

Also… keep DC out of the analyzer. The auto-scale counts DC as signal. If you apply a 2.5V input, the auto-range will range to the 10V scale – and you will waste all the bits on the DC signal.

Also II….Use a clean power supply. The typical bench supply (digital or analog) has a lot of low frequency noise <100Hz (in the 10’s of uV/Hz range). Even if you have a 60-80dB PSRR- there is still enough left-over to effect the noise results. Use a battery or precision voltage source if possible.

Watch out for Vs/2 divider resistors or output biasing resistors in a signal path – supply noise will be injected into that node.

Regards,