5.2 Coherent Input Frequency Selection
www.ti.com
ADC Evaluation
Typical ADC analysis requires users to collect the resulting time-domain data and perform a Fouriertransform to analyze the data in the frequency domain. A stipulation of the Fourier transform is that thesignal must be continuous-time; however, this is impractical when looking at a finite set of ADC samples,usually collected from a logic analyzer. Consequently, users typically apply a window function to minimizethe time-domain discontinuities that arise when analyzing a finite set of samples. For ADC analysis,window functions have their own frequency signatures or lobes that distort both SNR and SFDRmeasurements of the ADC.
TI uses the concept of coherent sampling to work around the use of a window function. The centralpremise of coherent sampling entails that the input signal into the ADC is carefully chosen such that whena continuous-time signal is reconstructed from a finite sample set, no time-domain discontinuities exist. Toachieve this, the input frequency must be an integer multiple of the ratio of the ADC sample rate (f
s
) andthe number of samples collected from the logic analyzer (N
s
). The ratio of f
s
to N
s
is typically referred to asthe fundamental frequency (f
f
). Determining the ADC input frequency is a two-step process. First, theusers select the frequency of interest for evaluating the ADC; then, they divide this by the fundamentalfrequency. This typically yields a non-integer value, which should be rounded to the nearest odd,preferably prime, integer. Once that integer, or frequency bin (f
bin
), has been determined, users multiplythis with the fundamental frequency to obtain a coherent frequency to program into their ADC input signalgenerator. The procedure is summarized as follows.
f
f
= f
s
/N
s
f
bin
= Odd_round(f
desired
/f
f
)
Coherent frequency = f
f
×f
bin
SLAU206B – September 2007 – Revised April 2008 15Submit Documentation Feedback