Your circuit is oscillating, almost certainly because of how you've constructed it; I would never, EVER attempt to build a circuit with such a fast op amp on a solderless breadboard due to the substantial stray capacitances and inductances inherent in that method of construction.

The attached application note from Jim Williams of Linear Technology, AN47-- High Speed Amplifier Techniques, has a lot of good information on how to design and build high speed circuits, including breadboarding techniques.

 

Your circuit is oscillating, almost certainly because of how you've constructed it; I would never, EVER attempt to build a circuit with such a fast op amp on a solderless breadboard due to the substantial stray capacitances and inductances inherent in that method of construction.

The attached application note from Jim Williams of Linear Technology, AN47-- High Speed Amplifier Techniques, has a lot of good information on how to design and build high speed circuits, including breadboarding techniques.

I have often said that anyone doing analog design should read everything Jim Williams ever wrote. Back when it was new, I actually made a point of getting a second paper copy of AN47 out of fear some evil fate might befall one.

Having reasonably (well, perhaps, sort-of, maybe) concluded that decoupling is not at the root of this problem, parasitics are pretty much all that remain. It is interesting that the oscillation is so voltage sensitive, but then then loop gain must be exactly one when loop phase shift is exactly 360 degrees, so if the gain changes ever so slightly with supply voltage, that might account for it.

I note that another person very recently struggled with inexplicable behavior and weird apparent "DC through a capacitor" only to discover flaky connections in the breadboard. Solderless breadboards are wonderful things, but like every other component it is necessary to recognize their characteristics and limitations.

 

Having reasonably (well, perhaps, sort-of, maybe) concluded that decoupling is not at the root of this problem, parasitics are pretty much all that remain. It is interesting that the oscillation is so voltage sensitive, but then then loop gain must be exactly one when loop phase shift is exactly 360 degrees, so if the gain changes ever so slightly with supply voltage, that might account for it.

Loop gain must be exactly one to obtain sustained, constant-amplitude oscillations at the 360 degree frequency; if the gain is >1 there will still be oscillations, but they will grow in amplitude until limiting occurs.

As for the voltage sensitivity, I've seen it before in some op amp circuits where it occurs only under certain conditions of supply voltage, output voltage and load impedance.

 

I built the circuit first on a breadboard, and it was working. Then I designed a PCB with SMD type components. The oscillation was found when I was testing the PCB. Then I went back to check the design on the breadboard, same issue appears... I guess I will have to read through the High Speed Amplifier Techniques to see if I can find anything that is useful.

 

A picture of your PCB would be helpful. Maybe somone can spot a problem and supply a way to fix it.

 

The moment I saw the oscillation of the high frequency opamp then I knew it was built on a solderless breadboard with stray capacitances all over the place.

 

The moment I saw the oscillation of the high frequency opamp then I knew it was built on a solderless breadboard with stray capacitances all over the place.

Except that he has now tried it on a PCB and the problem remains, though there is no note of frequency of oscillation. I would be extremely surprised if were the same as on the breadboard.

One of the most notorious problems with high speed amps is capacitance to ground on the summing node (inverting input). With very fast amps, even a picofarad or two can cause problems. Especially on multilayer or thin PCB laminate it is necessary to remove copper from the back side of the board in the vicinity of the inverting input pin and its connections.

A small capacitor (a few picofarads) across the feedback resistor is a common "fix" for this problem with voltage feedback amplifiers. The datasheet recommends 510 ohms for the feedback resistor. The 1k used isn't likely to change things by a huge amount, but would make the effects of capacitance at the inverting input worse.

Figure 35 of the datasheet show the effect of supply voltage at a gain of +2. I suspect the effect is reduced but not entirely eliminated at +11, so the oscillation's sensitivity to supply voltage is not unexpected.

 

I have rebuilt the circuit on the breadboard, separated out the components, lower down the gain to 4.3 ( 1+ 330ohms/100ohms), and it is working, the oscillation is gone. However, then I modified my PCB (4 layers PCB, with the whole layer of GND Panel and Power Panel in the middle), removed the 1k ohms SMD resistor that makes 11 gain, replaced with two wires and connected the wires to a 330ohms resistor. It seems to work but not really. when I put my hand on top of the PCB (5cm higher), the oscillation appears again. The oscillation also occurs when I bend the wires in certain directions (not all directions). And then I replaced the wires with 330 ohms SMD resistor, the oscillation is still there.

 

See

 

Just found another problem. On my PCB, I have a whole layer of copper as GND Panel, and I have a few header ports that connected to the GND panel. If I only connect one of the GND Ports with multi-core wire to the GND of the power supply, there is oscillation. However, if I connect two of the GND ports with two multi-core wires to the same GND of the power supply, the oscillation is gone. If I change the multi-core wire into a single core wire, I only need to connect one GND port to the GND of the power supply to get rid of the oscillation. What is the theory behind this?

 

Sounds like power supply lead inductance is creating an oscillator. That is a very high speed, high gain amplifier. You need both component and zone decoupling. Suggest 0.1 uF ceramic and 10 uF ceramic or electrolytic across the power input connector pins on the pc board, shortest possible leads. Same two caps across the opamp power pins. Twist the power and ground leads to the power supply.

ak