Voltage Divider Impedance for STM32F103/STM32F405 ADC Input at 1Msps Sampling

I have an opamp which has designed as an amplifier.
it amplifies signals which could have around 10.8V amplitude at the output pin. This voltage is not good to be injected directly to the STM32 ADC. The first thought would be a resistor divider. I selected 2.7K and 1.2K, which convert 10.8V to around  3.3V.
But by reading the STM32 datasheet, high ADC speeds need a very low impedance at the ADC input (they give a formula for it). for the 1Msps speed, I think it should be less than 400Ohm. I have already set up this circuit and I see some ADC sampling noises on the signal when I examine the pin by an oscilloscope, something like saw teeth on the signal.
I intend to use either STM32F103 or STM32F405, which suits better.
the questions are
1) is this a good method to do this? I mean using a resistor divider? I don't have access to the 3.3V rail to rail opamps (at least 15MHz GBW and 10V/us Slew rate)
2) is this normal to see that ADC noise on the signal?

Resistive dividers can work, but you have already discovered one issue, that of low drive impedance to the ADC. With a high divider impedance, the band width may be limited. Additionally, your op amp has to be able to drive the output resistive network at the bandwidth you require. You don't say which op amp you are using, check the data sheet, there is normally a graph showing  bandwidth versus resistive load. This can be quite low (relatively) if using rail to rail op amps, for example. Putting that issue aside, there should also be a graph of output swing versus frequency as well.Haven't checked you micros data sheet, but most micros use  a sampling capacitor on the ADC input as a sample and hold. This is switched between the input pin and the ADC. I am guessing that what you are seeing will be related to charge redistribution, and will not affect the quality of the conversion, it's all in the timing. With a static input voltage check the data output.
It would help if you could post a schematic of the setup, an op amp can be reconfigured for example, to have less than a gain of one, or an attenuator may be placed on the input side, etc.,cheers,Richard

You could follow the resistive divider with another op-amp configured as a unity gain buffer which would give you a low impedance output.
I have seen ADCs feed back switching noise. Certain  technologies are more prone to it than others, but my memory is not what it was.For what it is worth,I did a design idea that will allow you to change almost any input voltage to any output voltage using suitable resistors and supplies. Still has the impedance problem, but I hope someone finds it usefulhttps://www.electronicdesign.com/analog/use-excel-calculate-d-level-shifter-resistor-values You can find the Excel worksheet here- but you will have to register with Dropbox

The Thevinin or equivalent resistance seen by your ADC, is the resistance of 2.7Kohms in parallel with 1.2Kohms: 2.7Kohms * 1.2Kohms / (2.7Kohms + 1.2Kohms) = 831 ohms. If you halved the 2.7Kohms and 1.2Kohms resistances, you would have 416 ohms, close enough to the ADC spec.Anternatively, you could place a small capacitor across the 1.2Kohms resistor, to provide the sudden charge needed when the ADC samples its input. Choose a capacitor small enough so that the capacitor does not significantly attenuate the highest-frequency component of your signal.How big is the ADC sampling noise compared to one least-significant bit in the ADC?On the digital side, are you going to be using the wideband signal directly, or are you going to extract a narrowband component of the signal?

The opamp is CA3130I decided to use STM32F303 chip which has an internal opamp which I can use as a buffer to fix the impedance issue. it seems F3 series have mode focus on the ADC than others.Besides, I realized that I can reduce the supply voltage of the OpAMp. therefore I can use lower dividing resistors