preventing SPI noise on ADC

AFAIK I had similar problems and i fixed it with TL431 Voltage ref on ADS Vref pin. Assume that power lines are noiseless, Vcc logic and Vcc analog should be physicaly separate with two PCB tracks connected to CP power (100uF Tantalium)

Vlogic -----------------------------------| Vanal ---/\/\/\-------| ----- | 10 ----- ----- C2 Power --- ----- C1 Power | --- 100nF | | | | AGND | -------------------- ----- DGND

With LARGE tracks on AGND, DGND and Vlogic. a simple 1order RC filter on Vanal may improve noise immunity of analog path.

Mux COM input as close as possible to AGND

Finally i placed a high stability 2.5V ref on input and the average of

Good Luck, Habib.

Probably not necessary.

Probably not necessary.

If you look at the data sheet, most SPI ADC converters will show a sampling method that has the SPI lines quiescent during the actual conversion and then active again after the conversion to transfer the data. There is quite often a faster method which clocks out the data as the conversion takes place and may actually use the SPI clock as the conversion clock. This may result in more conversion noise.

The best way to get good noise results is to keep the SPI lines stable during the conversion as much as possible, and use an internal clock as the conversion clock. This may reduce your maximum conversion rate, but life is full of such tradeoffs.

The ADS8344 data sheet discusses these tradeoffs on Page 14.

Mark Borgerson

Hi Glenn,

You got some good advice here. Just one more comment: Splitting digital and analog ground may often not be such a good idea. In theory a split might work but at several MHz into the noise spectrum which your SPI and other clock transitions will certainly create, such a separation vanishes quickly. Then it becomes an uncontrolled matter where these grounds really meet and by how much.

I never use split grounds unless there is an isolation requirement for safety reasons, such as in medical gear.

Regards, Joerg

To start with, when the SPI is running to read out the values, then ADC already has a conversion finished. So the SPI signals to the ADC to rerad it out are of no convern. However, the signals on the SPI lines that are not caused by your intent to read out the ADC are of concern.

I went as fas as trying the magnetic isolators from Analog devices, ADum1400 and such, but they are somewhat overkill.

Yes, series resistors on the SPI lines dampen the reflections.

Rene

And no SPI lines under the ADC device will also help.

Gerard