LTC1060 Просмотр технического описания (PDF) - Linear Technology

Номер в каталоге

Компоненты Описание

Список матч

LTC1060

Universal Dual Filter Building Block

Linear Technology 

LTC1060

APPLICATIO S I FOR ATIO

internal op amps, as well as the reference point of all the

internal switches are connected to the AGND pin. Because

of this, a “clean” ground is recommended.

fCLK/f0 Ratio

The fCLK/f0 reference of 100:1 or 50:1 is derived from the

filter center frequency measured in mode 1, with a Q = 10

and VS = ±5V. The clock frequencies are, respectively,

500kHz/250kHz for the 100:1/150:1 measurement. All the

curves shown in the Typical Performance Characteristics

section are normalized to the above references.

Graphs 1 and 2 in the Typical Performance Characteristics

show the (fCLK/f0) variation versus values of ideal Q. The

LTC1060 is a sampled data filter and it only approximates

continuous time filters. In this data sheet, the LTC1060 is

treated in the frequency domain because this approxima-

tion is good enough for most filter applications. The

LTC1060 deviates from its ideal continuous filter model

when the (fCLK/f0) ratio decreases and when the Q’s are

low. Since low Q filters are not selective, the frequency

domain approximation is well justified. In Graph 15 the

LTC1060 is connected in mode 3 and its ( fCLK/f0) ratio is

adjusted to 200:1 and 500:1. Under these conditions, the

filter is over-sampled and the (fCLK/f0) curves are nearly

independent of the Q values. In mode 3, the ( fCLK/f0) ratio

typically deviates from the tested one in mode 1 by ±0.1%.

f0 x Q Product Ratio

This is a figure of merit of general purpose active filter

building blocks. The f0 x Q product of the LTC1060

depends on the clock frequency, the power supply volt-

ages, the junction temperature and the mode of operation.

At 25°C ambient temperature for ±5V supplies, and

for clock frequencies below 1MHz, in mode 1 and its

derivatives, the f0 x Q product is mainly limited by the

desired f0 and Q accuracy. For instance,from

Graph 4 at 50:1 and for fCLK below 800kHz, a predictable

ideal Q of 400 can be obtained. Under this condition, a

respectable f0 x Q product of 6.4MHz is achieved. The

16kHz center frequency will be about 0.22% off from the

tested value at 250kHz clock (see Graph 1). For the same

clock frequency of 800kHz and for the same Q value of

400, the f0 x Q product can be further increased if the

8

clock-to-center frequency is lowered below 50:1. In mode

1c with R6 = 0 and R6 = ∞, the (fCLK/f0) ratio is 50/√2. The

f0 x Q product can now be increased to 9MHz since, with

the same clock frequency and same Q value, the filter can

handle a center frequency of 16kHz x √2.

For clock frequencies above 1MHz, the f0 x Q product is

limited by the clock frequency itself. From Graph 4 at

±7.5V supply, 50:1 and 1.4MHz clock, a Q of 5 has about

8% error; the measured 28kHz center frequency was

skewed by 0.8% with respect to the guaranteed value at

250kHz clock. Under these conditions, the f0 x Q product

is only 140kHz but the filter can handle higher input signal

frequencies than the 800kHz clock frequency, very high Q

case described above.

Mode 3, Figure 11, and the modes of operation where R4

is finite, are “slower” than the basic mode 1. This is shown

in Graph 16 and 17. The resistor R4 places the input op

amp inside the resonant loop. The finite GBW of this op

amp creates an additional phase shift and enhances the Q

value at high clock frequencies. Graph 16 was drawn with

a small capacitor, CC, placed across R4 and as such, at VS

= ±5V, the (1/2πR4CC) = 2MHz. With VS = ±2.5V the (1/

2πR4CC) should be equal to 1.4MHz. This allows the Q

curve to be slightly “flatter” over a wider range of clock

frequencies. If, at ±5V supply, the clock is below 900kHz

(or 400kHz for VS = ±2.5V), this capacitor, CC, is not needed.

For Graph 25, the clock-to-center frequency ratios are

altered to 70.7:1 and 35.35:1. This is done by using mode

1c with R5 = 0, Figure 7, or mode 2 with R2 = R4 = 10kΩ.

The mode 1c, where the input op amp is outside the main

loop, is much faster. Mode 2, however, is more versatile.

At 50:1, and for TA = 25°C the mode 1c can be tuned for

center frequencies up to 30kHz.

Output Noise

The wideband RMS noise of the LTC1060 outputs is nearly

independent from the clock frequency, provided that the

clock itself does not become part of the noise. The LTC1060

noise slightly decreases with ±2.5V supply. The noise at

the BP and LP outputs increases for high Q’s. Table 2

shows typical values of wideband RMS noise. The num-

bers in parentheses are the noise measurement in mode 1

with the SA/B pin shorted to V– as shown in Figure 25.

1060fb

Share Link: