SPI clock noise

[Windfisch]

The filter overlays the filtered signal with a nasty 1kHz signal. The SPI dacs update with 1kHz update rate. Changing this changes the hum.

Since SPICLK is running at 5MHz, and since touching the SPICLK pin with a large metal mass (such as a wirecutter) leads to amore pronounced hum.

This also happens with a proper 7805-7905-battery-based power supply, so likely not a virtual GND problem.

wrapping the filter in (grounded) aluminum foil strongly reduces the hum.

Seems like we have built a radio station :(

[Windfisch]

to do:

• is the buzz still there without the actual transistor ladder connected? (i.e. only U1D with R36,37 dangling). (try this on the breadboard) -> no; the transistor ladder needs to be connected at least to (-) of the OpAmp for buzzing
• does the buzz disappear when changing R36,37 and R42,43 to 100 / 10k? -> it is reduced a bit, but still not great
• consider increasing the capacitors and the current passed through the ladder
• increase Vcc/Vee to ±12V
• place capacitor parallel to the opamp's feedback resistor
• connect the GND planes. or completely disconnect them and separately route them to dgnd
• does the issue still occur when
  • connecting the blue pill to power only?
  • completely powering the blue pill externally?
• increase source termination resistor (100-1k)

[Windfisch]

Disconnect at C2:
-> buzz is visible on the oscilloscope at R21/R22.
-> buzz is also visible across C3 and C6
-> buzz is also visible at output

Disconnect R36,37
-> buzz is visible everywhere except the output

=> OpAmp isn't the issue. Transistor ladder and prescaling are both affected. wtf.