Hello together,
I recently bought the ADS1274EVM-PDK from Mouser in order to evaluate the ADS1274. With its four simoultanesouly sampling channels, the ADS1274 seems to be the right ADC for my application in which the phase angle between two AC signals has to be measured with the lowest possible measurement uncertainty.
Since the noise of the ADC and its analog front end is crucial in such an application, my intention is to use the ADS1274 (or in later stage of development the ADS1278) in "High Resolution Mode" with clock frequency of f= 27Mhz (= 6.75 MHz modulator frequency) and a resulting sampling rate of 52.7 kSps.
After installation of the ADCpro software on my PC and setup of the hardware (correct jumper settings etc.), I started to perform noise measurement using the ADCpro Software. Much to my surprise the RMS and peak-to-peak values are substantially higher in "Hi-Resolution"-Mode than in "Hi-Speed"-Mode when measured at 27.0 MHz clock frequency.
The following settings were used for noise evaluation:
- all buffer amps powered down; ADC input switches on "HDR" position (direct input from header); each input directly shorted using jumpers on the header pins.
- TDM fixed Mode (Frame Sync due to connection to MMB0-board)
- wall plug power supply
- clock generated by PLL
Using the histogram function (block size: 32768 samples, 8 codes/bin = 2.4 mikrovolts) I measured the folowing values at a clock rate of f= 27MHz:
High speed mode: standard deviation: 26.5 codes = 7.9 mikrovolts RMS , peak-to-peak: approx. 210 codes = 62.6 mikrovolts
High Res. mode: standard deviation: 48 codes =14.3 mikrovolts RMS , peak-to-peak: approx. 550 codes = 164 mikrovolts
While the noise figures for the "high speed mode" are better than specified (< 8.5 mikrovolts), the noise figures for "high resolution mode" are 2.5times the normal specified value of 5.5 mikrovolts.
I tried the following steps to reduce the noise in high resolution mode:
- using the buffer amps
- DC power from a laboratory grade DC supply (linear regulation)
- ferrite beads on the power supply cable and the USB cable
-additional ceramic and tantalum caps between AGND and +5VA
-referencing the ADC inputs directly to AGND
-using the on-board 27MHz XTAL oscillator
None of these measures did really improve the noise figures in "high res. mode". Meanwhile the noise figures in "high speed mode" remained happily as specified.
In high speed mode the shape of the histogram look almost perfectly like a gaussian distribution, but in high resolution mode the the peak-to-peak values become excessive.
I tried the "mulitFFT" to look for particular signals that might interfere with the ADC, but found nothing. However using the ADCpro "multiscope" function I observed unusally large peaks, normally not seen in white or 1/f noise. Again in "high speed mode" these peaks didn't show up.
Finally I reduced the clock frequency to f=22.00 MHz (PLL) and then the noise figures improved significantly in high resolution mode: 17.4 codes standard deviation = 5.2 mikrovolts RMS, peak-to-peak: 140 codes.
I measured the noise figures at various clock frequencies. At more than 23 MHz the noise figures more than 7.0 mikrovolts RMS, and above 25 MHz more than 10 mikrovolts RMS.
My question: What is the cause for this strange phenomenon ? Why does the "hi speed mode" deliver noise figures as specified (also using the buffer amps) at various frequencies up to 32.7MHz, but operation in "high resolution mode" show excessive noise values above 22.0MHz clock ?
I would really appreciate your help. I could use the ADS1274 with high speed mode at 27MHz and 105.5 kSps, but would really prefer running the ADS1274 in high resolution mode at 52.7 kSps and (hopefully) better SNR.
Best regards,
Martin Damev