LIGO-T040052-00-K
11 March, 2004
Technical Note on the Strathclyde OSEM Development
Infrared source
The infrared LED source used in this work was an OD-50L, with a peak wavelength of 880
nm. Other emitters were tried, but none was found that offered the features of the OD-50L—
its sustainable high infrared power output, in particular.
This LED emitter was separated from the dual photodiode detector described below by a fixed
face-to-face distance of 39 mm. The LED was then moved laterally relative to the dual
detector in order to find the displacement-to-voltage transfer function.
The OD-50L infrared LED was driven by the circuit below, using feedback from a ‘monitor’
photodiode mounted close (10 mm) to the LED but off-axis, so that it did not obscure the dual
detector. The output of the transconductance amplifier built around the AD820 was compared
with the modulation voltage input signal, the difference between them being forced to zero by
the feedback action, as seen in Figure 1(b). Thus the infrared output power of the OD-50L was
a close facsimile of the modulation input voltage.
Fig.1(a). Modulation circuit for the OD-50L infrared LED. The monitor output voltage was
available at pin 5 of IC1B.
1
LIGO-T040052-00-K
11 March, 2004
Fig.1(b). Upper trace: modulation input voltage. Lower trace: monitor output voltage.
Infrared detector
Normally, the detector would be mounted in a hermetically-sealed can with an unfiltered glass
window, but such a configuration was not available at short notice from CENTRONIC.
Instead a de-capped dual photodiode detector was used in the OSEM design (CENTRONIC
LD2(15)), each element being a PIN photodiode 4 mm x 4 mm, the two elements being
separated by 0.5 mm. They were connected in the common-cathode configuration.
Fig.2. CENTRONIC LD2(15) dual photodiode detector in its mounting block.
The differential photodiode signal was detected using the circuit in Figure 3. The output of
each detector was amplified by a separate transconductance amplifier, the two outputs of these
amplifiers being differenced subsequently by a differential unity-gain amplifier. The
differential output was then phase-sensitively detected by an AD630 JN lock-in amplifier IC,
before being filtered by a 2-pole Bessel filter with a –3 dB point at 60 Hz, in order to produce
the final demodulated output. A digital phase-shift circuit in the ‘Ref in’ path allowed the
demodulated output signal to be maximised.
2
LIGO-T040052-00-K
11 March, 2004
Differential AC photodiode preamplifier
BNC in
Ref. in
+15 V*
100n
0
Lock-in
+12 V,
reg’d.
Amplifier
11
1n5 polystyrene
100n
1
Channel A
+15 V*
+15 V*
8
K
TP
1/2 x OP270
AD630 JN
or
AD630 BD
6
−15 V*
Amp. A
A+
+
A-
−
12
NC
13
o/p
17 R IN_B
−
3
10k0
+
BNC out
4
18
NC
19
−15 V*
4k3
+
Amp. B
B+
+
B-
−
-V
RB 14
10k
RF 15
1n5 polystyrene
10k0
3n3
polystyrene
10k
RF
SS
2k5
TP
screened lowcapacitance
twisted-pair
= (i A − i B ).RF
7
OP-O7
4
red
1/2 x CENTRONIC
LD2(15)
−
1k
−
2
2
10k0
1
2
NC
20
Vout_diff.
1k
A
+V
2k5
+
3
black
R IN_A
10k0
RA 16
5k
comp.
9
SEL_B
10
SEL_A
−
channelstatus
B/ A
+
-V
diff. offset adj.
3n3
polystyrene
3
diff. adj. 4
7
NC
SS
8
5
NC
6
NC
10k0
4k3
screened lowcapacitance
twisted-pair
1/2 x CENTRONIC
red
LD2(15)
RF
−15 V*
−
6
A
1/2 x OP270
K
12-turn
cermet pot.
5
yellow
7
+
100n polyester
Bessel LP Filter, gain = x3.5.
+15 V*
18k
*Note: all ICs decoupled
as shown close to chip
+ 10µ
10µ
0µ1
ceramic
ceramic
35V
Tant.
+ 35V
Tant.
7
OP-07
polyester
0µ 1
ceramic
0µ1
2
150n
−15 V
+15 V
36k
+
+12 V,
reg’d.
10µ
+15 V*
35V Tant.
OUT IN
COMM
0µ33
3
L4931CZ120
4
6
Vout_demod.
BNC out
−15 V*
1k
2k5
ceramic
N.A. Lockerbie
9 January, 2004
Fig.3. Detection circuit for the CENTRONIC dual-photodiode sensor.
Output voltage linearity with AC modulation amplitude
The modulation input to the LED driving circuit of Figure 1 was made up of a constant DC
bias, or ‘offset’, actually equal to 2.16 volts, DC, plus a superimposed sinusoidal signal at 1
kHz. As can be seen from Figure 4 the output of the detection-side demodulator was very
linear with increasing AC amplitude, this continuing beyond the amplitude at which the
displacement sensitivity and noise measurements for the OSEM were made (‘1.9 V’). In
order for the LED not to be cut-off over some region of the AC signal the AC modulation
amplitude was kept to a value that was always less than the DC bias.
3
LIGO-T040052-00-K
11 March, 2004
15
Demodulator output (V)
Offset = 2.16 V DC (measured).
Sinewave modulation at 1 kHz.
Intercept of best-fit line to data = -0.07 V.
10
Noise and displacementsensitivity measurements
were made with this
amplitude of modulation.
5
0
0.0
0.5
1.0
1.5
Measured amplitude of 1 kHz LED modulation signal (V)
2.0
Fig.4. Modulation-to-detection linearity for the LED source/dual-photodiode detector.
Output voltage vs Displacement
A trade-off was made between span and linearity, and although a 4 mm span was arrived at, as
shown in Figure 5, this could have been increased to 6 mm, albeit with a rather poorer linearity.
The slope sensitivity over the central 1.5 mm of the span = 7.58 kV.m−1.
15
Final demodulation output (volts)
10
5
0
-5
-10
-15
-3
-2
-1
0
1
Relative position of OD-50L LED (mm)
2
3
Fig.5. Displacement sensitivity (LED displaced laterally with respect to the dual-detector).
The average slope sensitivity of the line in the figure is 7.00 kV.m−1, but over the central 1.5
mm it is equal to 7.58 kV.m−1.
4
LIGO-T040052-00-K
11 March, 2004
Low-frequency noise performance
The low-frequency noise performance of –130.625 dBVrms at 10 Hz, taken together with the
displacement slope-sensitivity, gave an overall direct displacement sensitivity of 3.88 × 10-11
m/rt-Hz. Therefore, by using the displacement-doubling prism described below a displacement
sensitivity of 1.94 × 10-11 m/rt-Hz should be attainable.
Fig.6. Output noise performance over the frequency range 0–12.5 Hz. The origin of the peak
in the noise spectrum at 5.3 Hz has not yet been identified.
Displacement-doubling Prism
The concept of the displacement-doubling prism is shown in the diagram below. In the lower
part of the figure a fixed input (infrared) beam enters from below into a special prism mounted
in an OSEM’s flag, exiting from the prism in the direction of the dashed line. If the flag +
prism are then displaced horizontally by a certain amount, as indicated by the blue arrow, then
the output infrared beam will be displaced parallel to itself by double that amount. The optical
path length through the prism remains constant, so either an unfocused beam, as shown, or a
focused beam may be used.
5
LIGO-T040052-00-K
11 March, 2004
Fig.7. Displacement-doubling prism.
Noise as a function of displacement
A feature of the OSEM used in this work was that the noise at its null point (zero displacement)
was lower than at any other point in its ±2 mm span. At the extremities of the span the noise
was some 30 dB higher than at the null point. This excess noise was investigated over a range
of ±100µm about the null point, and the results are shown in Figure 8. Clearly, the excess
noise appears to be proportional to the size of the detected (differential) signal. No mechanism
for the production of this excess noise has yet been determined.
Over a range of approximately ±125 µm the sensitivity of this displacement detector is at a
level of ≤ 5 × 10-11 m/rt-Hz.
6
LIGO-T040052-00-K
11 March, 2004
Fig.8. Excess noise as a function of displacement away from the null point.
Discussion
The simple LED/dual-photodiode detector, together with a suitable displacement-doubling
prism, has come very close to—within a factor 2 of—the target of 10-11 m/rt-Hz. However,
unless or until the source of the excess noise can be determined and mitigated, this type of
sensor cannot maintain such a level of sensitivity over the desired span of ±1.5 mm.
7