Noise or grounding problem in a two-stage Transimpedance amplifier circuit.

Hey all, been busy with exams for a while, thanks so much for your comments on this, just seen them all, a bit overwhelming to say the truth, I will consider them all as much as possible, and hopefully have a positive feedback.

I just want to add one thing, actually the first sign of trouble I saw in the circuit was while I was taking measurements, I noticed that the voltage reading of the ouputs of the summing and subtracting amps (the second stage) was changing with the multimeter I was using, and changing with the number of multimeters, the more hooked to the same node at the same time the higher the potential reading was, which meant the introduction of a multimeter itself was influencing the circuit, so none of those measurements are true. In these pictures all measurements are made on the same node (output of the subtracting amp, if I am not mistaken) under the same conditions, later when I used the oscilloscope it had the same effect but it showed the significant oscillations as well. If that gives you a hint, that's great. That was on another breadboard, the connections of which (grounding especially) were triple-checked carefully, to no use, I reconstructed it on another, seemingly better, breadboard, got the same results, that's when I came to consult you guys.

I think John is on to it here - Vbias is meant to be 5V DC (there should be no AC here at all and it should be the same at any physical point on the node.).

Check your ground connections (and all other connections).

Try just shorting Vbias to ground.

Try simulating the circuit in LTSpice (its free and will be illuminating).

Ceramic capacitors (100nF - 1uF) from the op amp supplies to ground for every chip are essential with high speed parts.

A few of good rules of thumb are:

• Simulate first
• The effective operating frequency of anything with op amps is the gain bandwidth product of the op amp.
• The chances of a breadboard lash-up working are inversely proportional to the effective operating frequency.
• Plug in prototype boards don't work above 1MHz
• If the position of the scope probe (a capacitance to earth)  on a node affects what happens then either you don't have the circuit you thought or the inductance of the wires is significant.
• If attaching the scope probe to a node anywhere affects the circuit then think why a 10 - 100pF cap is doing this (check cap figures for your scope probes on x1 or x10)
• For frequencies above 1MHz and more than 10 parts go straight to pcb
• For switch mode PSUs go straight to pcb

Like all such rules they are very far from absolute and if your prototype and analogue skills are good you may be able to break many or all of them - but you should know why you are breaking them.

MK

Michael's suggestion to short the inputs is a good one.

These op-amps may be just able to oscillate at 20 to 30 MHz without external compensation, but your 100pf caps should prevent oscillations at this frequency. If doubling these caps doesn't affect the noise, then it might not be an op-amp oscillation.

If altering the caps affects the frequency and amplitude of the noise, look for positive feedback from the output of the second op-amp to the input of the first op-amp. You can short the inputs of the second op-amp to see what effect it has.

Another thing to check is the power supply noise. These op-amps have no power supply rejection at 30 MHz, so any power supply noise may not be attenuated.

Also, be sure the scope probe arrangement is not the culprit. It is easy to find fictitious noise if the scope probe ground is too far from the signal.

When building something like this I always ground my input to see what kind of self noise I have in the circuit. Sometimes it reveals a noise source I never considered.

Balancing the input Z on the opamps will minimize thermal drift of your front-ends.  As to your 'not shown' biasing potentiometers, have you tied your unused ends to the wiper, degenerating them into variable resistors?  If the free ends float, they can act as little antennas, as it states in Art of Electronics.

What is the rational of your second stage?  It looks like a pair of unity-gain inverters, but then you have this summing thing happening.  Looks kind of confusing to me.

That looks like the kind of oscillation you get with an unwanted positive feed back path. There is a lowish impedance path from the output of your opamp to the + pin via the diode; it may help to have some significant resistance in the + pin connections to c_gnd, 10k for example. Or there could be stray capacitance from the breadboard, and there is no ground plane. It may be better to build it on a prototype pcb with a ground plane. Or try a change in opamp to one with a ,low bandwidth.

I'm not an expert in analogue design either.

If you didn't need the performance, then it's a liability because it means the devices could oscillate. It also leaves them more susceptable to interference.

You mentioned DC on the output, so it sounded like your system might have been very low frequency in its operation. In your case, however, you do need a reasonable performance since you've got signals up to several tens of kHz and if you drop the GBW product too far you won't have enough open-loop gain to drive the loop properly.

By form of construction, I meant the breadboard. Your layout is neat and tidy and the layout is sensible, but for all that it can't be as good as a PCB.

Just noticed, looking at the picture of the circuit board, that you could do with some bulk decoupling on your 15V and 5V rails. A 10uF or 22uF electrolytic (or tant) on each rail where the power comes on to the board would probably do.

I don't understand the "Vbias...at the PS lead" and the "Vbias .. at diodes' common cathode lead" traces. One is a 531mV sinewave on 4.8V and the other is 1.62V. Aren't they effectively the same point, circuit-wise? And aren't they supplied by a 5V PSU? Nothing in the circuit is capable of driving the output of a PSU like that. Are you sure all the C_GND connections are actually connected together?

Early op-amps had a capacitor to slow down the amp to stabilize it in low-gain situations.  The types are 'compensated.' stable at all frequencies, 'un-compensated,' and 'de-compensated,' which  was stable for gains greater than, like, five.  People were often too thick to use it, so I think it kinda fell from favor.

How much dark signal do you get, what is the product of your leakage and your amplifier gain?

At least, from what I know, having a high Opamp frequency response is likely to pick up more high frequency signals and amplify them, and thus what you are seeing might possibly be an oscillation of sorts which might not happen with an Opamp with lower GBP mainly because it's just physically incapable of producing such signals. These tend to be more picky about supply quality and need good bypassing to prevent oscillation as well. The amount of inductance and capacitance with breadboard traces probably provide just enough coupling to cause these sorts of oscillation.

The other thing I can think of, having made a simple transimpedance amplifier myself (with ~10Hz response) is that the photodiode inputs should be as shielded from external noise as possible. External RF noise and impulse noise can get amplified, and most transimpedance amplifier design notes tend to recommend using a PCB and surrounding the photodiode traces with a ground guard ring to sink external stray fields.

Unfortunately, aside from that, I can't claim to be any expert, so I'll let the others contribute their ideas.

- Gough

Would you please elaborate on the liabilities of using an op amp with such ratings?
By form of construction do you mean the breadboard type of construction? or the circuit layout itself?
I'd appreciate more information, I am learning here after all