Power supply pulling during SD card writes

[tingo]

I have a data logger which uses Teensy 3.2 to fetch data from the ADS1298 ADC and store it on an SD card.

I'm also using the 3.3V output of the LDO on the Teensy board to provide power to a miniature 9723GX Sonion microphone. The output of the microphone is sampled by the ADC at 1 KS/s. The signal I'm interested in is in the range of microvolts, but what I observe is a huge number of spikes in the recorded signal corresponding to the time instances of data being written to the SD card, see plot below.

The SD card I'm using is the Swissbit 1GB which draws 50mA during the write operation, which is well within the capabilities of the LDO. For this test Teensy was supplied by an external bench-top power supply at 4.2 V.

Any idea what could be done to get rid of this supply pulling/ground noise or mitigate it in some way?

 

[grease_lighting]

You will need to add isolation or power supply decoupling to the SD card electronics. You could try bulk capacitance at the SD card socket perhaps 20uF elco (electrolytic capacitor) or get tougher by adding a ferrite bead followed by an elco (teensy 3.3v to ferrite bead to elco) or for even better isolation especially when measuring small signals such as you are supply the SD card electronics with a seperate 3.3v regulator powered by the 4.2v supply. You may even want to use a seperate power filter network or regulator for your ADC to keep its supply ultra stable. Using multiple power supply regulators for each section of your project with a common ground point is an excellent technique to keep the digital noise out of the analog section.
Check out this app note from analog devices: http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf

 

[tingo]

You will need to add isolation or power supply decoupling to the SD card electronics. You could try bulk capacitance at the SD card socket perhaps 20uF elco (electrolytic capacitor) or get tougher by adding a ferrite bead followed by an elco (teensy 3.3v to ferrite bead to elco) or for even better isolation especially when measuring small signals such as you are supply the SD card electronics with a seperate 3.3v regulator powered by the 4.2v supply. You may even want to use a seperate power filter network or regulator for your ADC to keep its supply ultra stable. Using multiple power supply regulators for each section of your project with a common ground point is an excellent technique to keep the digital noise out of the analog section.
Check out this app note from analog devices: http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf

Thanks for the suggestions! I think I'll try a completely separate regulator for the SD card. It seems like this noise doesn't have much of an effect on the ADC, since when I short the inputs to the ADC channels and perform sampling, the noise measurement adheres to the spec quite nicely. It's only when I try to sample an external sensor which is supplied from the LDO that I get these spikes.

The data logger is quite small: it's 3 boards stacked on top of each other with Teensy in the middle, where the sizes of my boards match the size of the Teensy PCB, thus space is at a premium and putting an elco is not an option, unfortunately. Also, it seem like there's plenty of decoupling on the Teensy board itself and I also put some 10uF ceramics on my ADC board.

With regards to ferrite beads, they do seem to work somewhat, but don't solve the problem completely. What I tried doing to test this was to provide the microphone with a power supply separate from the data logger and sharing GND between microphone and the data logger. My ADC board is connected to AGND of the Teensy, while the SD card is connected to Teensy's GND; in the schematic there is ferrite bead connecting AGND and GND. The spikes were reduced, but not eliminated, which implies to me that despite the presence of the ferrite bead there was noise injection from GND to AGND.