CS5501-SD datasheet(24/54 Pages) CIRRUS

24 / 54 page

Anti-Alias Considerations

The digital filter in the CS5501/CS5503 does not

provide rejection around integer multiples of the

oversampling rate [(N*CLKIN)/256, where

N = 1,2,3,...]. That is, with a 4.096 MHz master

clock the noise on the analog input signal within

the narrow

±10 Hz bands around the 16 kHz,

32 kHz, 48 kHz, etc., passes unfiltered to the digi-

tal output.

Most broadband noise will be very

well filtered because the CS5501/CS5503 use a

very high oversampling ratio of 800 (16 kHz:

2x10 Hz). Broadband noise is reduced by:

eout

= ein √



−3dB ⁄ fs

eout

= 0.035 ein

where ein and eout are rms noise terms referred to

the input. Since f-3dB equals CLKIN/409,600 and

fs equals CLKIN/256, the digital filter reduces

white, broadband noise by 96.5% independent of

the CLKIN frequency. For example, a typical op-

erational amplifier’s 50

µV rms noise would be

reduced to 1.75

µV rms (0.035 LSB’s rms at the

16-bit level in the CS5501 and 0.4 LSB’s rms at

the 20-bit level in the CS5503).

Simple high frequency analog filtering in the sig-

nal conditioning circuitry can aid in removing

energy at multiples of the sampling rate.

Post Filtering

Post filtering is useful to enhance the noise per-

formance of the CS5503. With a constant input

voltage the output codes from the CS5503 will

exhibit some variation due to noise. The CS5503

has typically 1.6 LSB20 rms noise in its output

codes. Additional variation in the output codes

can arise due to noise from the input signal source

and from the voltage reference. Post filtering

(digital averaging) will be necessary to achieve

less than 1 LSB p-p noise at the 20-bit level. The

CS5503 has peak noise less than the 18-bit level

without additional filtering if care is exercised in

the design of the voltage reference and the input

signal condition circuitry. Noise in the bandwidth

from dc to 10 Hz on both the AIN and VREF

inputs should be minimized to ensure maximum

performance. As the amount of noise will be

highly system dependent, a specific recommenda-

tion for post filtering for all applications cannot be

stated. The following guidelines are helpful. Real-

ize that the digital filter in the CS5503, like any

other low pass filter, acts as an information stor-

age unit. The filter retains past information for a

period of time even after the input signal has

changed. The implication of this is that immedi-

ately sequential 20-bit updates to the serial port

contain highly correlated information. To most ef-

ficiently post filter the CS5503 output data,

uncorrelated samples should be used. Samples

which have sufficiently reduced correlation can be

obtained if the CS5503 is allowed to execute 200

filter cycles between each subsequent data word

collected for post filtering.

The character of the noise in the data will influ-

ence the post filtering requirements. As a general

rule, averaging N uncorrelated data samples will

reduce noise by 1/

√N. While this rule assumes

that the noise is white (which is true for the

CS5503 but not true for all real system signals

between dc and 10Hz), it does offer a starting

point for developing a post filtering algorithm for

removing the noise from the data. The algorithm

348,160

364,544

398,336

445,440

470,016

486,400

497,664

506,880

512,000

516,096

518,144

519,168

Filter

Cycles

340

356

389

435

459

475

486

495

500

504

506

507

CLKIN

Cycles

Bits of

Output

Accuracy

Table 5. Settling Time of the 6 Pole Low Pass Filter in

the CS5501 to 1/2 LSB Accuracy with a Full Scale

Step Input

CS5501/CS5503

DS31F2

CS5501-SD 数据表(PDF) 24 Page - Cirrus Logic

CS5501-SD 数据表(HTML) 24 Page - Cirrus Logic

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 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54

数据表下载