unusual voltages at GPIO6 to GPIO 11 Pins

[naitik]

we are using esp32 wroom 32d. when we measure voltages while running the firmware using oscilloscope between GPIO6 to GPIO11 We get Voltages that ranges between 4.2v ~ -1.17v .
However when we flash empty project with simple while loop voltages between those GPIOs stays around 3.44~3.12v. the board is powered by PC. and its unclear how ESP32 is Outputting around 4 volt max between those GPIO when voltage across 3v3-GND always stays around 3.59~3.04v. I know that GPIO6-GPIO11 is used for interfacing Flash but is it common to get this much voltage at those pins?
here is voltage values with and empty project(Left hand side) running and Firmware is running (Right Hand side)

[ESP_Sprite]

How are you measuring this? Given the fact that the ESP32 is trying to measure this, there will be AC'ish waveforms on these pins. If you throw an instrument expecting DC at this, you'll get weirdness.

[naitik]

I am measuring this with Oscilloscope. and yes i am aware of this pins carries the AC'ish waveform. but since this is Digital 3v3 LVCMOS Standard Line I expected to see the switching value between 0v-3.3v but in reality the it goes upto 4.14v max as it is seen in picture below. The development board is powered up by PC. so is this normal ?

[ESP_Sprite]

What does the signal look like when zoomed in? And what is your measuring setup, as in, how do you connect the probes?

[naitik]

After changing the code yesterday, the behaviour of the out is altered so max voltage is bit different but its still more than 3.3v. the way I measure is to set the GND to the MCU GND pin and place the probe right on the pin where I want to measure.
and here is zoomed in waveform :

[ESP_Sprite]

That still looks like an artifact of your measuring method; in this case I think it's a sin(x)/x interpolation artifact; you can see that as the waveform starts to dip below GND *before* it goes up. (Ref e.g. here)

Edit: Your signal looks like 30KHz and your sample rate is 50Ksps. Not only are SPI pulses way faster than that, you're also almost guaranteed to get sin(x)/x artfacts with the sample rate and (measured) waveform frequency so close.