OPA838: Stability problem by feedback resistor

The feedback cap is shaping the noise gain to 1 at higher Freq. the device is not unity gain stable. 

Thanks for explanation. How to solve this problem, change the gain or remove the capacitor?

This figure suggests a path, a more complete discussion is in the OPA847 datasheet under inverting compensation.

Hello Michael

I redo the simulation for AC sweep and no pole found now. But I don't know whether I could trust the result now.

Well perhaps the TI guys could chip in here and do a LG phase margin sim for you??

Thanks!

It is perfect if TI could support the simulation

Thanks a lot!

Hello

I make the circuit. However, it is not stable. Then I remove Cf, it become stable. But I need Cf to compensate the inductance on the shunt.

I don't understand why the feedback resistor impact the stability

Hi,

Could you share the circuit you are simulating that is not stable? Could you also clarify what you remove from the circuit to make it stable.

Best Regards,

Ignacio

Hello

I make the circuit with following schematic

However, this circuit is not stable.

Then I change feedback capacitor(C2) to 100pf.

Then I remove the feedback capacitor, the circuit seems stable.

But I need the feedback capacitor to compensate the inductance of the shunt

In my opinion, the feedback capacitor will help the circuit to be stable. But in this application, it seems it is not the case.

Hi,

Like Michael mentioned previously, the feedback capacitor shapes the noise gain at higher frequencies which may result in instability as this part is not unity gain stable. However, to account for this we add the capacitor to ground and tune the circuit like was mentioned. The addition of this capacitor should help stabilize the circuit. Can you share the board you designed the circuit on? The layout and/or board image will be good to see if maybe layout has something to do with the instability as the addition of the capacitor to ground should help stabilize the circuit.

Best Regards,

Ignacio

Hello

Thanks for the explanation, I understand it is mainly a noise problem , not a stability problem. From your explanation, if I want a feedback capacitor, especially the value is not so small(200pf). A not unify gain stable Op Amp is not suitable for this kind of application. Can you confirm that?

Hello

Below is the schematic and layout. I hope it helps.

Hi,

If your design requires a feedback capacitor, we avoid using decompensated devices like the OPA838 unless we adjust the circuit at the higher frequencies with the added capacitor at the input. This technique of adding the capacitor to shape the gain of the circuit at higher frequencies is a common practice and should work. I was able to look at the layout and it is a little confusing when looking at the feedback network specifically. It looks like the feedback might be clipped from the image. Could you share the schematic used to create the layout as the component labels don't seem to line up, so it is hard to understand which components in the layout correspond to the components in the schematic you shared.

Best Regards,

Ignacio

R28 in the top side, C35 in the bottom side, Op Amp in the bottom side.

R28 & C35 are almost in the same position

Now I realize OPA838 is not a good choice. I would like to try OPA2626, this part could be better

Hi,

I was looking over the layout and there are a few concerns from a layout standpoint that in theory will affect the performance of the circuit. It is hard to say this is directly what is interfering with the performance and overall stability of the circuit; however, it will definitely make things harder to debug. A couple concerns from what I can see is that the circuits feedback paths include vias, this will add inductance and is overall not recommended for our faster devices. One other concern is that the layers under the inverting and output pins are not removed. The reason we remove layers under these pins is to reduce the amount of parasitic capacitance in the feedback path where stability can be affected. Additionally, we always recommend decoupling capacitors as close as possible to the supply pins, in this case Vs+. In rare instances, we have seen unwanted oscillations at the output of a device due to improper decoupling. I would strongly recommend looking over our OPA838 EVM design as a guide for the best layout practices. Could you confirm the package you are using for the OPA838? If you are willing to do a quick device swap with a unity gain stable device and adjust the circuit, we might be able to understand what is causing the stability issue you are seeing. I would like to add most devices that are pin to pin compatible will likely not have the bandwidth in a gain of 10V/V like the OPA838. The OPA2626 will make things difficult if the bandwidth requirement is in the range of the OPA838.

Best Regards,

Ignacio

At the beginning, I though a high speed amplifier is necessary. Now I don't really think a 300MHz amplifier is necessary. I would like to try OPA625 with 100MHz bandwidth. But due to its mode control, I have to do a little modification on the layout.

Thanks!

Hello

I have run simulation for the circuit below, when I use OPA350, the result is as expected

But if I switch to OPA626, the result is not as expected

I wonder why?

Hi,

One device that is pin to pin compatible with the OPA838 and is unity gain stable is the OPA837. This device could be an option as a quick drop-in replacement without reworking the board. There is also the OPA810 that comes in this package that is unity gain stable, however this device requires a minimum supply of 4.75V. As for the simulation, I was not able to capture exactly what you were seeing on my end, just as a quick test could you try simulating this with a +/-2.5V split supply configuration and ground the 1.5V reference.

Best Regards,

Ignacio

Yes

I have applied OPA837. And I will do some test on it. I hope it works.