AD9262EBZ datasheet(17/33 Pages) AD

17 / 33 page

AD9262

Rev. A | Page 16 of 32

THEORY OF OPERATION

The AD9262 uses a continuous time Σ-Δ modulator to convert

the analog input to a digital word. The digital word is processed

by the decimation filter and rate-adjusted by the sample rate

converter (see Figure 37). The modulator consists of a continuous

time loop filter preceding a quantizer that samples at fMOD =

640 MSPS. This produces an oversampling ratio (OSR) of 32 for

a 10 MHz input bandwidth. The output of the quantizer is fed

back to a DAC that ideally cancels the input signal. The incom-

plete input cancellation residue is filtered by the loop filter and

is used to form the next quantizer sample.

H(f)

LOOP FILTER

QUANTIZER

–

ADC

MODULATOR

DECIMATION

FILTER

SAMPLE RATE

CONVERTER

SRC

Figure 37. Σ-Δ Modulator Overview

The quantizer produces a nine-level digital word. The quantization

noise is spread uniformly over the Nyquist band (see Figure 38),

but the feedback loop causes the quantization noise present in

the nine-level output to have a nonuniform spectral shape. This

noise-shaping technique (see Figure 39) pushes the in-band

noise out of band; therefore, the amount of quantization noise

in the frequency band of interest is minimal.

The digital decimation filter that follows the modulator removes

the large out-of-band quantization noise (see Figure 40), while

also reducing the data rate from fMOD to fMOD/16. If the internal

PLL is enabled, the sample rate converter generates samples at

the same frequency as the input clock frequency. If the internal

PLL is disabled, the sample rate converter can be programmed

to give an output frequency that is a divide ratio of the modulator

clock. The sample rate converter is designed to attenuate images

outside the band of interest (see Figure 41).

QUANTIZATION NOISE

fMOD/2

BAND OF INTEREST

Figure 38. Quantization Noise

NOISE SHAPING

BAND OF INTEREST

fMOD/2

Figure 39. Noise Shaping

fMOD/32

fMOD/16

BAND OF INTEREST

DIGITAL FILTER CUTOFF FREQUENCY

Figure 40. Digital Filter Cutoff Frequency

fMOD/16

fOUT/2

fOUT

BAND OF INTEREST

Figure 41. Sample Rate Converter

ANALOG INPUT CONSIDERATIONS

The continuous time modulator removes the need for an anti-

alias filter at the input to the AD9262. A discrete time converter

aliases signals around the sample clock frequency and its multiples

to the band of interest (see Figure 42). Therefore, an external

antialias filter is needed to reject these signals.

DESIRED

INPUT

UNDESIRED

SIGNAL

ADC

fS/2

Figure 42. Discrete Time Converter

In contrast, the continuous time Σ-Δ modulator used within the

AD9262 has inherent antialiasing. The antialiasing property

results from sampling occurring at the output of the loop filter

(see Figure 43), and thus aliasing occurs at the same point in the

loop as quantization noise is injected; aliases are shaped by the

same mechanism as quantization noise. The quantization noise

transfer function, NTF(f), has zeros in the band of interest and in

all alias bands because NTF(f) is a discrete time transfer function,

whereas the loop filter transfer function, LF(f), is a continuous

time transfer function, which introduces poles only in the band

of interest. The signal transfer function, being the product of

NTF(f) and LF(f), only has zeros in alias bands and therefore

suppresses all aliases.

LF(f)

H(z)

QUANTIZATION

NOISE

INP UT

OUTPUT

LOOP FILTER

fMOD

NTF(f)

L F(f)

Figure 43. Continuous Time Converter

AD9262EBZ Datasheet(PDF) 17 Page - Analog Devices

AD9262EBZ Datasheet(HTML) 17 Page - Analog Devices

Html Pages

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Datasheet Download

Link URL