# What is the sine wave coming from the output of an opamp?

I'm using opa657 see datasheet for my photodiode, but when I connected the photodiode as the 1st schematic in the data sheet, I got a sine wave of Vpp=2Vs at the output of the amplifier even without any input on the photodiode.

and when I put an incident light the sine wave will disapear and no signal appears to the output of the amplifier.

Where does this sine wave come from and how fix this problem?

• what frequency is it? – Brian Drummond Mar 20 '15 at 16:08
• What frequency is the wave? I bet it 50-60Hz – Eugene Sh. Mar 20 '15 at 16:08
• i am using 5 Mbps for the bit rate of the input incident light.@BrianDrummond @Eugene Sh. – optical Mar 20 '15 at 16:11
• The frequency of the OUTPUT wave – Eugene Sh. Mar 20 '15 at 16:12
• Sounds like oscillations due to the noise gain of the opamp. 1.) reduce the input capacitance.. smaller PD or reverse bias it. 2.) bigger feedback resistor (if you can .) You then need to add a bypass cap in parallel with feedback R. (Building fast PD front ends with screaming opamps takes a bit of know how... good luck. – George Herold Mar 20 '15 at 16:37

You can't run the OPA657 at unity gain - it oscillates - even when configured as a TIA you can run into trouble with it. Here's the open loop gain and phase (red): -

The op-amp's phase margin is such that at a closed loop gain of about 6 to 8dB the negative feedback will be positive feedback i.e. open loop phase is about -180 degrees at about 850MHz and the gain hasn't gone below unity.

You could bypass the photodiode with a capacitor - see figure 3 in the data sheet - it shows a PD with quite a high self capacitance - maybe you should try adding capacitance across the PD?

This ensures that at high frequencies there is gain (because negative feedback has reduced) even though that gain may not be apparent either to the untrained eye or the photodiode. If you had a signal connected to the non-inverting input that gain would be very apparent and of course there is a signal at that pin - it's the equivalent noise of the op-amp's input - this is why, on transimpedance amplifiers it's called noise gain - you add capacitance across the photodiode to add stability i.e. a reduction in feedback and you find self-oscillation has been traded for a much noisier signal. TIAs are not trivial to design.

With a feedback resistor of 200k and a capacitor of (say) 47pF, the fed back signal will have reduced by 3dB at : -

F = $\dfrac{1}{2\pi R C}$ = 16.93kHz

By 100kHz the fed back signal is way down and no longer posing an oscillation threat.

Also, you need to ensure the chip has adequate decoupling capacitors on power lines to the device - they need to be close up to the chip.

• Hi Andy, I've never run that opamp as a TIA, but I don't think unity gain stability is needed for a TIA. The capacitance in figure 3 is the photodiode capacitance. The last thing you want to do is add more C on the inverting input. (Well unless you want it to oscillate at a lower frequency :^) (I kinda feel I should down vote your answer, but maybe you will mod it.) – George Herold Mar 20 '15 at 17:16
• @GeorgeHerold adding capacitance raises the gain and will make it stabler. See the open loop gain picture I added to my answer. – Andy aka Mar 20 '15 at 17:19
• Hi Andy, Have you every built a fast PD front end? The noise gain (voltage gain from the non-inverting input) increases as you increase the capacitance... the corner frequency is given by Cdiode*Rfeedback... (mostly, assuming C diode dominates.) There are lots of good ap-notes.. here's one that is short and sweet. ti.com/lit/an/sboa055a/sboa055a.pdf – George Herold Mar 20 '15 at 17:33
• @GeorgeHerold Yes, I do build them quite often and I've come unstuck with some op-amps. I'm currently trying to design one that has a flat gain up to about 1GHz for a data comms product and, I've designed one for a heat measurement system that needs a decent speed (300kHz) and high gain (1Mohm) - we needed to use bootsrapping of course to counter the 100pF of the PD. I'll look at the stuff you have linked but the bottom line is that adding a capacitor reduces the signal that is fed back and therefore reduces the threat of self-oscillation. – Andy aka Mar 20 '15 at 17:45
• Ahh, your answer says to add more capacitance in parallel with the PD... but then you rightly suggest bootstrapping to reduce the PD capacitance. Something is wrong... fix the answer. – George Herold Mar 20 '15 at 22:27

I can delete this later. This is the line I like least,

You could bypass the photodiode with a capacitor - see figure 3 in the data sheet - it shows a PD with quite a high self capacitance - maybe you should try adding capacitance across the PD?

The noise gain of a TIA looks like this,

simulate this circuit – Schematic created using CircuitLab

The gain hits one at a freq given by 1/2*pi*tau, where tau is Rf * Cin.
The gain grows with freq, if left unchecked (by opamp GBW or C in parallel with Rf) then it will oscillate. Adding more Cin makes it worse. As do opamps with more voltage noise and bigger Rf's

The only way I've seen GHz done is with a bias voltage->PD->resistor, with a fast gain stage looking at the resistor voltage.