OPA657: TIA output deformation with less bandwidth than expected

Hi Follin,

Follin Ju said:

Unit I extend the light pulse to around 20us than the square wave shows up.

Are you generating the ligh pulse by the help of a LED? Keep in mind that a LED may be too slow to generate a fast light pulse.

Kai

I used a specific light pulse generator and the duration can be set precisely. I got the cutoff frequency esstimation fomula from photodiode supplier 1/2πcR so currently I guess the the most likely explanationt is the diodes bandwidth limitation....

Hi Follin,

what photodiode do you use?

Kai

Hi Kai, I used the Si APD

Hi Follin,

are you sure the feedback capacitance is 100fF and not higher?

Kai

So Follin

That simple equation for BW only applies if the feedback pole has been set to 0.707X that frequency, that would be at 2MHz which then solves to a 0.05pF feedback C - not easy to get that - most SMD R's have 0.18pF parasitic. You might want to consider a feedback T network where the first R will be low enough to reach your BW and then make up the gain in the T. 

Hi Michael, thanks for your suggestions. Before I started this design I also considered the T network but it seems if I want to get best SNR I should use the single large resistor. But since this BW issue I'll try the T network as a backup design.

Hi Kai, Yes I used a 0.1pF±0.0p5f capacitor

Hi Follin,

Follin Ju said:

Yes I used a 0.1pF±0.0p5f capacitor

But you still have to add the parasitic parallel capacitance of 1M5 resistor which is in the range of some 100fF.

If the parasitic parallel capacitance of 1M5 resistor is really that high, you can try to remove the 100fF (C2) cap all the way or at least decrease its value.

The simplest remedy is to decrease the 1M5 resistor. This allows you to use a bigger feedback capacitance and your circuit would suffer way less from the unpredicted parasitic parallel capacitance of feedback resistor.

Kai

Hi Kai, yes so I used 0805 resistor to minimize the parasitic capacitance. I will try to move the 100fP to see. The actual bandwidth may not get close to the calculation value.

Many thanks!