OPA827: OPA827 voltage follower exhibits input oscillation with capacitative source

Hi peter,

The 28 - 40 MHz oscillation is in the range of the OPA827 unity gain bandwidth frequency (22 MHz typical, but can be higher) so this suggests that the phase margin has become zero with the circuit configuration. Your exact circuit configuration isn't one I've encountered before so please provide us with an illustration of the circuit. I want to be sure where and how the "open" cable is connected in the circuit. Also, what the output load is in the circuit. Please include any other information you think will be helpful. If you have any DSO images of the oscillation I would like to see them.

Regards, Thomas

Precision Amplifiers Applications Engineering 

Hi Thomas,

Thank you for the reply. I've taken some scope photos.

I attach an extract from the schematic of the board I am using. It shows the input of one channel with the OPA827 used as a voltage follower. Also shown are the EMC and protection components, but if I remove and link the ferrite L14 and remove the protection diode pair D27, there is no difference in oscillation behaviour. The output of the buffer drives the input of an ADG1208 multiplexer, but in these tests, the multiplexer enable input is held low, so the amplifier is driving only the 1.5pF input capacitance.

I connected a standard 50 Ohm BNC cable (approx 210pF) to the input connector leaving the remote end unconnected. I used a high-impedance scope probe (10M+3pF) attached to the rear of the BNC connector to observe the waveform there.

I attach a scope photo showing the waveform. This shows a stable 32MHz oscillation with 300mV pk-pk amplitude, but, maybe significantly, a +136mV d.c. component. I emphasise that this oscillation is coming out of the non-inverting input of the amplifier. The amplifier output (and therefore inverting input) pins show a very similar waveform with the same amplitude but lagging 90 degrees wrt the non-inverting input.

With this particular BNC cable, I can prevent the oscillation happening by changing R63 from a link (0R0) to a physical 220R resistor. However, this does not necessarily prevent oscillation wih other types and lengths of cable.

Hi Peter,

Thank you for the additional information and clarifications. For oscillation there has to be a gain of 1 V/V, or higher, and enough phase shift around a loop to result in positive feedback. I do know of instances where a local loop inside an op amp oscillates as the result of a unique external impedance condition, but those cases were not with the OPA827. Do note that when an op amp is oscillating that the dc behaviors can be very different than what is observed when it is operating correctly as an amplifier. 

Let me give this some thought and I'll see if any of my colleagues have any ideas.

Regards, Thomas

Precision Amplifiers Applications Engineering

Hi Peter,

Would you be willing to try a small experiment? If so, try disconnecting the external 50 Ohm cable and in its place connect a 200 pF, or 220 pF capacitor across J14. Use the O-scope as you have previously and see if the OPA827 still oscillates under this condition.

You mention in your earlier response, "Also shown are the EMC and protection components, but if I remove and link the ferrite L14 and remove the protection diode pair D27, there is no difference in oscillation behaviour." Please confirm my understanding that this means the common-mode filter is removed and the OPA827 still oscillates as before when the cable is attached to J14.

Is the cable coaxial cable such as RG-58, or RG-174? If not, what kind is it and what is its length?

Regards, Thomas

Precision Amplifiers Applications Engineering 

Hi Thomas,

It's not as easy to provoke the oscillation with just a capacitor and no BNC cable. If I measure the capacitance of a cable that causes oscillation at  210 pF and then put a 220pF polystyrene capacitor in place instead of the cable, there is no oscillation. However, significantly increasing the physical capacitor value does cause the problem, so the picture below shows the result with 10nF across the input BNC connector. Interestingly, the oscillation frequency does not vary inversely with the capacitance as one might have suspected, the 10nF generating a higher frequency (40 MHz) than the 210pF cable (37 MHz). Note that this is a different OPA827 sample  than the one I used for my previous photos.

You were correct about removal of the ferrite and protection diodes.

Hi Peter,

Thanks for the info regarding the input capacitor test. Please provide me the specifics of the input cable; type, length, etc.

Regards, Thomas

Precision Amplifiers Applications Engineering

The cable that most reliably causes the oscillation problem is one we have that came supplied with another piece of equipment. The cable is 2m in length with moulded-on BNC plugs, and has "Sparkomatic Taiwan RG-58/U" printed on it.

Peter

Hi Peter,

I suspect that the oscillation may be the result of what the coaxial appears as to the input of the OPA827 circuit. A 2 meter, open-ended length of RG-58/U transmission line appears not as a capacitance to the OPA827 non-inverting input, but instead an impedance of Z = 1.57 + j24.8 ohms. At 32 MHz this appears as a 1.57 ohm resistor in series with 123.3 nH inductor. That is much different than the 210 pF capacitor which exhibits a Z = -j23.7 ohms at 32 MHz. I suspect that in conjunction with circuit stray capacitances the conditions are such that there is enough phase shift to support oscillation at 32 MHz. 

I have attempted to model the OPA827 circuit with the equivalent circuit for the transmission line at the input. The OPA827 model is an older style model so I am not sure we are getting the exact ac behavior from it, and the stray capacitances of yur PC board and layout are not known. Nonetheless, when I model the circuit as shown below it shows very extensive ringing which indicates the phase margin is very low and at best marginally stable. In an actual circuit the actual amplifier may go into oscillation. Note that I used a 10 mV pulse excitation to test for ringing.

There are two ways to damp the circuit; reduce the Q. The first is adding a small resistance such as 100 ohms in series with the line going to the OPA827 non-inverting input. The result is shown here.

If the application can't handle the divider formed by the 100 ohm and 1 Meg resistors, the input to the OPA827 can be loaded with a series RC that becomes effective after the GBW of the OPA827. Adding a series RC combination of 100 ohms and 72 pF (75 pF) from the non-inverting input line to ground damps the ringing. The result is nearly identical to that received by the series 100 ohm input resistor.

Regards, Thomas

Precision Amplifiers Applications Engineering

Hi Thomas,

Thank you for the interesting analysis. I'm sure your explanation is very close to representing the actual circuit performance in our equipment. However, what I find extraordinary is that a voltage follower formed from an OPA827 JFET amplifier exhibits stability problems with certain source impedances that would be perfectly normal to encounter in everyday laboratory usage.

I wired up a separate OPA827 as a voltage follower with its non-inverting input shunted by 470K//10nF to ground. It worked excellently with low-impedance external sources, but when I unplugged the input BNC cable from the amplifier, it sat there oscillating at close to 40MHz. Substituting an OPA140 (the only other JFET amplifier I could easily find in the MSOP package) gave very stable results with any source impedance, albeit at half the bandwidth.

I note your suggestion that we could continue to use the OPA827 either by putting a 100R series resistor or a 100R-75pF shunt combination at the OPA827's non-inverting input, but this would involve changes to the PCB layout that are not practicable at this stage of the production cycle. Unless I can find another amplifier in the MSOP package that has similar bandwidth to the OPA827, I will have to fit OPA140s in this version of the product and suffer the reduction in input bandwidth.

Thanks for all your help with this matter. I do feel that other prospective users of the OPA827 should be warned of potential source impedance problems when using it as a voltage follower.

Peter