REV. B
AD807
–6–
Typical Performance Characteristics
SIGNAL DETECT LEVEL mV
200.0E+3
0.0E+0
0.0 30.05.0
R
THRESH
10.0 15.0 20.0 25.0
160.0E+3
140.0E+3
100.0E+3
80.0E+3
120.0E+3
180.0E+3
60.0E+3
40.0E+3
20.0E+3
35.0
TPC 1. Signal Detect Level vs. R
THRESH
TEMPERATURE C
35.0E3
0.0E+0
40 8020
SIGNAL DETECT LEVEL Volts
0204060
25.0E3
15.0E3
20.0E3
30.0E3
10.0E3
5.0E3
100
R
THRESH
= OPEN
R
THRESH
= 49.9k
R
THRESH
= 0
TPC 2. Signal Detect Level vs. Temperature
TEMPERATURE C
9.00
3.00
40 8020
ELECTRICAL HYSTERESIS dB
0204060
7.00
5.00
6.00
8.00
4.00
100
R
THRESH
= OPEN
R
THRESH
= 49.9k
R
THRESH
= 0
TPC 3. Signal Detect Hysteresis vs. Temperature
SUPPLY VOLTAGE Volts
35.000E3
0.000E+0
4.4 5.64.6
SIGNAL DETECT LEVEL Volts
4.8 5.0 5.2 5.4
25.000E3
20.000E3
10.000E3
5.000E3
15.000E3
30.000E3
R
THRESH
= OPEN
R
THRESH
= 49.9k
R
THRESH
= 0
TPC 4. Signal Detect Level vs. Supply Voltage
POWER SUPPLY V
8.00
0.00
4.4 5.64.6
ELECTRICAL HYSTERESIS dB
4.8 5.0 5.2 5.4
6.00
5.00
2.00
1.00
4.00
7.00
3.00
R
THRESH
= OPEN
R
THRESH
= 49.9k
R
THRESH
= 0
TPC 5. Signal Detect Hysteresis vs. Power Supply
1
2
1
2 2
S
N
)
(
erfc
10 12 14 16 18 20 22 24
S/N dB
1E1
5E2
3E2
2E2
1E2
1E3
1E4
1E5
1E6
1E8
1E10
1E12
BIT ERROR RATE
1278
NSN
1276
1279
1277
TPC 6. Bit Error Rate vs. Signal-to-Noise Ratio
REV. B
AD807
–7–
RMS JITTER Degrees
30
0
1.4 2.31.5
PERCENTAGE %
1.6 1.7 1.8 2.2
20
5
15
25
10
1.9 2.0 2.1
TEST CONDITIONS
WORST-CASE:
40C, 4.5V
TPC 7. Output Jitter Histogram
FREQUENCY Hz
1E+3
100E3
10E+0 10E+6
JITTER TOLERANCE UI
10E+0
100E+0 1E+3 10E+3 100E+3
100E+0
1E+0
1E+6
AD807
SONET MASK
TPC 8. Jitter Tolerance
NOISE V p-p @ 311MHz
3.0
0
0 0.60.1
JITTER ns p-p
0.2 0.3 0.4 0.5
2.0
1.0
1.00.7 0.8 0.9
PSR NO FILTER
CMR
PSR WITH FILTER
TPC 9. Output Jitter vs. Supply Noise and
Output Jitter vs. Common Mode Noise
THEORY OF OPERATION
Quantizer
The quantizer (comparator) has three gain stages, providing a
net gain of 350. The quantizer takes full advantage of the Extra
Fast Complementary Bipolar (XFCB) process. The input stage
uses a folded cascode architecture to virtually eliminate pulse
width distortion, and to handle input signals with common-
mode voltage as high as the positive supply. The input offset
voltage is factory trimmed and guaranteed to be less than 500 µV.
XFCB’s dielectric isolation allows the different blocks within
this mixed-signal IC to be isolated from each other, hence the
2 mV Sensitivity is achieved. Traditionally, high speed compara-
tors are plagued by crosstalk between outputs and inputs, often
resulting in oscillations when the input signal approaches 10 mV.
The AD807 quantizer toggles at ± 650 µV (1.3 mV sensitivity) at
the input without making bit errors. When the input signal is
lowered below ± 650 µV, circuit performance is dominated by
input noise, and not crosstalk.
0.1F
0.1F
0.1F
0.1F
0.1F
FERRITE BEAD
OPTIONAL FILTER
0.1F
50
309
5050
3.65k
+5V
10F
CHOKE
BIAS TEE
311MHz
NOISE
INPUT
0.1F
0.1F
500
500
13
12
14
11
6
3
PIN
NIN
AV
CC2
AV
CC1
V
CC1
V
CC2
AD807
0.1F
QUANTIZER
INPUT
Figure 6. Power Supply Noise Sensitivity Test Circuit
0.1F
0.1F
0.1F
0.1F
0.1F
50
309
5050
3.65k
+5V
10F
CHOKE
BIAS TEE
311MHz
NOISE
INPUT
0.1F
0.1F
500
500
13
12
14
11
6
3
PIN
NIN
AV
CC2
AV
CC1
V
CC1
V
CC2
AD807
0.1F
QUANTIZER
INPUT
Figure 7. Common-Mode Rejection Test Circuit
Signal Detect
The input to the signal detect circuit is taken from the first stage
of the quantizer. The input signal is first processed through a gain
stage. The output from the gain stage is fed to both a positive
and a negative peak detector. The threshold value is subtracted
from the positive peak signal and added to the negative peak signal.
The positive and negative peak signals are then compared. If the
positive peak, POS, is more positive than the negative peak,
NEG, the signal amplitude is greater than the threshold, and the
output, SDOUT, will indicate the presence of signal by remain-
ing low. When POS becomes more negative than NEG, the
signal amplitude has fallen below the threshold, and SDOUT
will indicate a loss of signal (LOS) by going high. The circuit
provides hysteresis by adjusting the threshold level higher by a
factor of two when the low signal level is detected. This means
that the input data amplitude needs to reach twice the set LOS
threshold before SDOUT will signal that the data is again valid.
This corresponds to a 3 dB optical hysteresis.
REV. B
AD807
–8–
AD807
COMPARATOR
STAGES AND
CLOCK RECOVERY
PLL
PIN
NIN
THRESHOLD
BIAS
+
+
IHYS
ITHR
SDOUT
POSITIVE
PEAK
DETECTOR
NEGATIVE
PEAK
DETECTOR
LEVEL-
SHIFT
UP
LEVEL-
SHIFT
DOWN
Figure 8. Signal Level Detect Circuit Block Diagram
Phase-Locked Loop
The phase-locked loop recovers clock and retimes data from
NRZ data. The architecture uses a frequency detector to aid
initial frequency acquisition; refer to Figure 9 for a block diagram.
Note the frequency detector is always in the circuit. When the
PLL is locked, the frequency error is zero and the frequency
detector has no further effect. Since the frequency detector is
always in the circuit, no control functions are needed to initiate
acquisition or change mode after acquisition.
DET
F
DET
DATA
INPUT
S + 1
RETIMING
DEVICE
1
S
VCO
RECOVERED CLOCK
OUTPUT
RETIMED DATA
OUTPUT
Figure 9. PLL Block Diagram
The frequency detector delivers pulses of current to the charge
pump to either raise or lower the frequency of the VCO. During
the frequency acquisition process the frequency detector output
is a series of pulses of width equal to the period of the VCO.
These pulses occur on the cycle slips between the data frequency
and the VCO frequency. With a maximum density data pattern
(1010 . . . ), every cycle slip will produce a pulse at the frequency
detector output. However, with random data, not every cycle
slip produces a pulse. The density of pulses at the frequency
detector output increases with the density of data transitions. The
probability that a cycle slip will produce a pulse increases as the
frequency error approaches zero. After the frequency error has
been reduced to zero, the frequency detector output will have
no further pulses. At this point the PLL begins the process of phase
acquisition, with a settling time of roughly 2000 bit periods.
Jitter caused by variations of density of data transitions (pattern
jitter) is virtually eliminated by use of a new phase detector
(patented). Briefly, the measurement of zero phase error does
not cause the VCO phase to increase to above the average run
rate set by the data frequency. The jitter created by a 2
7
–1
pseudorandom code is 1/2 degree, and this is small compared to
random jitter.
The jitter bandwidth for the PLL is 0.06% of the center fre-
quency. This figure is chosen so that sinusoidal input jitter at
92 kHz will be attenuated by 3 dB.
The damping ratio of the PLL is user programmable with a
single external capacitor. At 155 MHz, a damping ratio of 5
is obtained with a 0.15 µF capacitor. More generally, the damp-
ing ratio scales as (f
DATA
× C
D
)
1/2
.
A lower damping ratio allows a faster frequency acquisition;
generally the acquisition time scales directly with the capacitor
value. However, at damping ratios approaching one, the acquisi-
tion time no longer scales directly with capacitor value. The
acquisition time has two components: frequency acquisition and
phase acquisition. The frequency acquisition always scales with
capacitance, but the phase acquisition is set by the loop band-
width of the PLL and is independent of the damping ratio.
Thus, the 0.06% fractional loop bandwidth sets a minimum
acquisition time of 2000 bit periods. Note the acquisition time
for a damping factor of one is 15,000 bit periods. This comprises
13,000 bit periods for frequency acquisition and 2,000 bit peri-
ods for phase acquisition. Compare this to the 400,000 bit
periods acquisition time specified for a damping ratio of 5; this
consists entirely of frequency acquisition, and the 2,000 bit
periods of phase acquisition is negligible.
While a lower damping ratio affords faster acquisition, it also
allows more peaking in the jitter transfer response (jitter peaking).
For example, with a damping ratio of 10, the jitter peaking is
0.02 dB, but with a damping ratio of 1, the peaking is 2 dB.
Center Frequency Clamp (Figure 10)
An N-channel FET circuit can be used to bring the AD807 VCO
center frequency to within ±10% of 155 MHz when SDOUT
indicates a Loss of Signal (LOS). This effectively reduces the
frequency acquisition time by reducing the frequency error
between the VCO frequency and the input data frequency at
clamp release. The N-FET can have “on” resistance as high as
1 k and still attain effective clamping. However, the chosen
N-FET should have greater than 10 M “off” resistance and
less than 100 nA leakage current (source and drain) so as not to
alter normal PLL performance.
16
15
14
13
12
11
10
9
1
2
3
4
5
6
7
8
AD807
DATAOUTN
DATAOUTP
V
CC2
CLKOUTN
CLKOUTP
V
CC1
CF1
CF2
V
EE
SDOUT
AV
CC2
PIN
NIN
AV
CC1
THRADJ
AV
EE
N_FET
C
D
Figure 10. Center Frequency Clamp Schematic
FREQUENCY Hz
10 20k
0.02dB/DIV
100 1k 10k
C
D
PEAK
0.1
0.15
0.22
0.33
0.12
0.08
0.06
0.04
Figure 11. Jitter Transfer vs. C
D

AD807A-155BRRL7

Mfr. #:
Manufacturer:
Analog Devices Inc.
Description:
IC FIBER OPTIC RCVR 16-SOIC
Lifecycle:
New from this manufacturer.
Delivery:
DHL FedEx Ups TNT EMS
Payment:
T/T Paypal Visa MoneyGram Western Union