AD9022SQ データシートの表示（PDF） - Analog Devices

部品番号

コンポーネント説明

メーカー

AD9022SQ

12-Bit 20 MSPS Monolithic A/D Converter

Analog Devices 

AD9022

This will be true only for converters in which perfect quantiza-

tion noise dominates. There may be an upper sample rate,

above which the thermal noise of the converter is the dominant

source of noise. In this case, normalization would be based on

the noise bandwidth of the ADC. For an AD9022 with a typical

SNR of 64 dB and a sample rate of 20 Msps, the normalized

SNR is equal to 134 dB (64 + 70). Both thermal and quantiza-

tion noise contribute to this number.

The SNR of the input is assumed to be limited by the thermal

noise of the input resistance, or –174 dBm/Hz. The input signal

level is +10 dBm (2 V p-p into 50 Ω). Noise figure of the ADC

can be calculated by:

AD9022 NOISE PERFORMANCE

High speed, wide bandwidth ADCs such as the AD9022 are op-

timized for dynamic performance over a wide range of analog

input frequencies. However, there are many applications (Imag-

ing, Instrumentation, etc.) where dc precision is also important.

Due to the wide input bandwidth of the AD9022 for a given in-

put voltage, there will be a range of output codes which may oc-

cur. This is caused by unavoidable circuit noise within the

wideband circuits in the ADC. If a dc signal is applied to the

ADC and several thousand outputs are recorded, a distribution

of codes such as that shown in the histogram below may result.

–76

NF = SNR (in) – SNR (out) = [+10 – (174)] – 134 = 50 dB

–75

Most ADCs detect input voltage levels, not power. Conse-

+25°C

quently, the input SNR can be determined more accurately by

–74

ROOM

OBSOLETE determining the ratio of the signal voltage to the noise voltage of

the terminating resistor. However, both the input signal and

noise voltage delivered to the ADC are also a function of the

source impedance. The dependence of NF on sample rate, lin-

earity, source and terminating impedances, and the number of

assumptions that are required highlight the weakness of using

NF as a figure of merit for an ADC. The rather large number

that results bolsters this belief by indicating the ADC is often

the weakest link in the signal processing path.

Linearity

The Third Order intercept point for a linear device (with some

nonlinearity) is a good way to predict 3rd order spurious signals

as a function of input signal level. For an ADC, however, this in

an invalid concept except with signals near full scale. As the in-

put signal is reduced, the performance burden shifts from the

input track-and-hold (T/H) to the encoder. This creates a non-

linear function, as contrasted with the third order intercept be-

havior, which predicts an improvement in dynamic range as the

–73

–55°C

–72

–71

–70

+125°C

–69

–68

0

12 3 4567 8

ANALOG INPUT FREQUENCY – MHz

9 10

Figure 11. ADC Equivalent Input Noise

The correct code appears most of the time, but adjacent codes

also appear with reduced probability. If a normal probability

density curve is fitted to this Gaussian distribution of codes, the

standard deviation will be equal to the equivalent input rms

noise of the ADC. The rms noise may also be approximated by

signal level is decreased.

converting the SNR, as measured by a low frequency FFT, to

an equivalent input noise. This method is accurate only if the

For signals near full scale, the intercept point is calculated the

SNR performance is dominated by random thermal noise (the

same as any device:

low frequency SNR without harmonics is the best measure).

Intercept Point = [Harmonic Suppression/(N –1)] + Input Power Sixty-three dB equates to 1 LSB rms for a 2 V p-p (0.707 V

where N = the order of the IMD (3 in this case)

rms) input signal. The AD9022 has approximately 0.5 LSB of

AD9022 Intercept Point = 80/2 + 3 dBm (7 dBm below full scale)

= 43 dBm

For signals below this level, the spurious free dynamic range

(SFDR) curves shown in the data sheet are a more accurate pre-

dictor of dynamic range. The SFDR curve is generated by mea-

suring the ratio of the signal (either tone in the two-tone

measurement) to the worst spurious signal which is observed as

the analog input signal amplitude is swept.

The worst spurious signal is usually the second harmonic or 3rd

order IMD. Actual results are shown on several plots. The

straightline with a slope of one is constructed at the point where

the worst SFDR touches the line. This line, extrapolated to full

scale, gives the SFDR of the ADC. This value can then be used

to predict the dynamic range by simply subtracting the input

level from the SFDR.

It should be noted that all SFDR lines are constructed to be

valid only below a certain level below full scale. Above these

points, the linearity of the device is dominated by the

rms noise or a noise limited SNR of 69 dB, indicating that noise

alone does not limit the SNR performance of the device (quanti-

zation noise and linearity are also major contributors).

This thermal noise may come from several sources. The drive

source impedance should be kept low to minimize resistor

thermal noise. Some of the internal ADC noise is generated in

the wideband T/H. Sampling ADCs generally have input band-

widths which exceed the Nyquist frequency of one-half the

sampling rate. (The AD9022 has an input bandwidth of over

100 MHz, even though the sampling rate is limited to 20 Msps.)

Wide bandwidth is required to minimize gain and phase distor-

tion and to permit adequate settling times in the internal ampli-

fiers and T/Hs. But a certain amount of unavoidable noise is

generated in the T/H and other wideband circuits within the

ADC; this causes variation in output codes for dc inputs. Good

layout, grounding, and decoupling techniques are essential to

prevent external noise from coupling into the ADC and further

corrupting performance.

nonlinearities of the front end and best predicted by the inter-

cept point.

–8–

REV. A

Share Link: