RSS Feeds
Understanding ADC SFDR Specs
Jon Titus, Contributing Editor -- Design News, November 16, 2008
Past columns have discussed sampling analog signals and analog-to-digital converter (ADC) technologies. In my experience, though, engineers who simply want to use ADCs rather than design with them, may misunderstand some basic ADC specifications.
The term “dynamic range” can cause confusion because engineers interpret it several ways. Dynamic range relates to the range of input signals an ADC can handle. If you have a 14-bit ADC with an input range of 0-1V, in theory, the converter can measure a signal as small as the voltage represented by the least significant of the 14 bits. In this case 61 μV, which yields an 84-decibel (dB) dynamic range:
dynamic range (dB) = 20 * log(VLSB/Vrange) = 20 * log(2n).
Rather than use this mathematical value, you should look at the spurious-free dynamic range (SFDR) for an ADC or data-acquisition (DAQ) system. The SFDR value expresses how well a converter works in the presence of spurious (noise) signals. The SFDR of a stand-alone ADC IC might look quite good because the chip vendor used a low-noise lab circuit to make the measurement. But expect a lower SFDR in a DAQ system that can pick up small spurious signals from other inputs, clocks, switching power supplies and some components.
In essence, SFDR defines the usable measurement range above the spurious-noise floor that could interfere with or distort a signal you want to measure. The diagram, above, shows two measures of SFDR; dBc and dBFS. The lower-case “c” indicates the use of a “carrier” signal as a reference and “dBc” refers to the logarithmic ratio between the carrier signal and the spurious-signal noise floor. The “dBFS” notation refers to the log ratio between the full-scale capability of the ADC and the spurious-noise floor. Keep in mind these calculations use root-mean-square voltages.
To determine a system’s SFDR, apply a sine wave to your DAQ or ADC input and use an appropriate sample rate. (See “Sample Rates Revisited,” (DN 10.06.08); “Oversample on Purpose,” (DN 10.20.08), and “Oversample to the Extreme,” (DN 11.03.08).
Then run an FFT and examine the spectral output. You can see the fundamental frequency and “spurs” from harmonics of the test signal and system components. I recommend you make this test at several sine-wave frequencies and sample rates. You may recall that an FFT can “smear” energy at one frequency across several frequency “bins” due to the nature of processing sampled signals. Testing at several frequencies and sample rates could reveal spurs with slightly higher RMS amplitudes.
In each case, an SFDR value conveys the ratio of the spurious signal with the most energy and either an ADC’s full-scale range or a measured carrier signal. Those measurements occur from dc across the complete Nyquist bandwidth, which amounts to half the sampling rate.
Useful Links
-
Ludy, Tim, “Performance More Than Bits and Megahertz,” Evaluation Engineering
-
“Understanding Dynamic Hardware Specifications,” National Instruments
-
“Decibel,” Wikipedia
