Hobbs Electro Optics QL01 Series User manual

QL01 Series
Quantum-Limited Photoreceivers
User’s Guide And Reference
Rev. 1.22

QL01 Series
Quantum-Limited Photoreceivers
Rev. 1.22
Philip C. D. Hobbs
Hobbs ElectroOptics/ElectroOptical Innovations, LLC.

i
LEGAL NOTICE AND WARRANTY
HOBBS ELECTROOPTICS/ELECTROOPTICAL INNOVATIONS, LLC. PROVIDES THIS
PUBLICATION “AS IS” WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. (Disclaimer
of expressed or implied warranties in certain transactions is not allowed in some states.
Therefore, the above statement may not apply to you.)
This manual may contain technical inaccuracies and/or typographical errors. Changes are
periodically made to this manual which are incorporated in later editions. Hobbs Elec-
troOptics/ElectroOptical Innovations, LLC. may make changes and improvements to the
product(s) and/or programs described in this publication at any time without notice. The
QL01 has non-zero failure rates associated with its hardware, firmware, design, and docu-
mentation. Do not use the product in applications where a failure or defect in the instrument
may result in injury, loss of life, or property damage.
IN NO EVENT WILL HOBBS ELECTROOPTICS/ELECTROOPTICAL INNOVATIONS,
LLC. OR ITS MEMBERS OR EMPLOYEES BE LIABLE FOR DAMAGES, INCLUDING
LOST PROFITS, LOST SAVINGS OR OTHER INCIDENTAL OR CONSEQUENTIAL
DAMAGES ARISING OUT OF THE USE OF OR INABILITY TO USE SUCH PRODUCT,
EVEN IF HOBBS ELECTROOPTICS/ELECTROOPTICAL INNOVATIONS, LLC. OR
AN APPROVED RESELLER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
DAMAGES, OR FOR ANY CLAIM BY ANY OTHER PARTY.

ii
Contents
Contents ii
1 Introduction 1
2 Specifications 2
3 Photos 5
4 Quick Guide To Using The QL01 7
5 Theory of Operation 9
5.1 Noise In Photoreceivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2 Shot noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.3 Transimpedance Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4 Bootstraps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.5 The QL01 Difference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6 Performance Verification 16
6.1 Pulse Response and Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . 16
6.2 Measuring The Noise Floor Accurately . . . . . . . . . . . . . . . . . . . . . 16
6.6 Finding the Shot Noise Limit . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.3 Transient Response (Model A and B) . . . . . . . . . . . . . . . . . . . . . . 19
6.4 Model A Noise Floor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
6.5 Model B Noise Floor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
References 24
Table of Symbols 25

1
1 Introduction
Making good optical measurements is rarely simple, but it is particularly difficult in very
low light, as anyone knows who’s ever tried it.
Of course, it’s easier if you can get more light somehow, maybe with a brighter laser or
a larger collection aperture. That isn’t always possible, unfortunately. Chemical, biological,
and microelectronic samples are often easily damaged by cranking the illumination power
too high, for instance, and large collecting apertures are bulky and expensive.
Low light naturally requires large feedback resistors in the transimpedance amplifier in
order to reduce thermal (Johnson) noise. Decent-sized photodiodes have a lot of capacitance,
though, which in combination with those large resistors form slow RC time constants that
limit measurement speed, signal-to-noise ratio, or both.
The Hobbs ElectroOptics QL01 is a highly sensitive photoreceiver that achieves shot-noise
limited performance with very dim light, from 10 nanowatts up to a few microwatts, with
a previously unattainable combination of sensitivity and bandwidth. It uses a proprietary
bootstrap architecture to reduce the effect of photodiode capacitance by a factor of over
1000, with sub-nanovolt noise densities. This allows the Model B to offer shot-noise-limited
detection performance at 25 nA out to its full 1 MHz bandwidth, and at 8 nA in 100 kHz.
The Model A trades off a bit of low-light performance for a 7x increase in detector area for
more light collection and ease of alignment.
Low-light applications in solid state physics, spectroscopy, chemistry, biology, and other
fields will see an immediate improvement in their measurements when they switch to the
QL01. For applications with a bit more light available, forthcoming versions C and D will
extend the bandwidth to 3 MHz at 1 MΩ transimpedance. Fiber-coupled versions with
performance from 800-1700 nm are in the works as well.
These instruments are designed to survive the accidents that sometimes happen in a
research lab, with medical-grade power supplies, a thick aluminum box, a stainless steel
mounting thread, and a solid metal BNC output connector securely attached to the box
itself.
The QL01 series are the first of a family of advanced photoreceivers from Hobbs Elec-
troOptics. These are all-new designs based on expertise gained through our own research
and from designing dozens of advanced instruments for our consulting clients. We also do
specials and OEM products, so if you have an unusual requirement, give us a call or send an
email.
Philip C. D. Hobbs Hobbs ElectroOptics/ElectroOptical Innovations, LLC.
Principal 160 North State Road, Suite 203
[phone] Briarcliff Manor, NY 10510
[email protected] http://hobbs-eo.com

Section 2: Specifications
2 Specifications
QL01 Quantum-Limited Nanowatt Photoreceiver
FUNCTION Single-channel, low noise optical to electrical converter
INPUT Free space optical detector.
PHOTOCURRENT
RANGE
QL01-A/B: 0 - 1 µA
OUTPUT BNC connector, output impedance 50 Ω (A high-Z load is
preferred, as this prevents increased photodiode leakage
due to internal power dissipation.)
WAVELENGTH QL01-A/B: <400 nm - 1100 nm.
DETECTOR AREA QL01-A: 2.65 mm square detector, 7 mm2area
QL01-B: 1 mm square detector, 1 mm2area
TRANSIMPEDANCE 10 MΩ (10 V/µA)
RESPONSIVITY 0.6 A/W typical @ 840 nm (set by the photodiode)
QUANTUM
EFFICIENCY
0.9 typical @ 840 nm, >10% 380-1080 nm (QL01-A),
350-1080 nm (QL01-B)
BANDWIDTH DC - 1 MHz @ -3 dB
RISE / FALL TIME 400 ns typical, 10% - 90%, see Figure 6.1
QL01 2

Section 2: Specifications
QL01 Quantum-Limited Nanowatt Photoreceiver
OFFSET CURRENT ±10 nA maximum @ 25◦C
NOISE
(INPUT-REFERRED)
@ 25 ◦C
QL01-A: Noise floor <60 fA/√Hz (DC - 100 kHz):
Shot noise limited above 10 nA (DC- 100 kHz) or 45 nA
(DC - 1 MHz). See Figure 6.2
QL01-B: Noise floor <60 fA/√Hz (DC - 100 kHz):
Shot noise limited above 8 nA (DC - 100 kHz) or 25 nA
in DC - 1 MHz . See Figure 6.3
NOISE EQUIVALENT
POWER (NEP)
Optical power needed to reach SNR of 1.0. At at 840
nm, the QL01-A’s 60 fA/√Hz noise current equals 75 fW
NEP in 1 Hz.
OVERSHOOT: 3% maximum, measured with an input pulse with 100 ns
time constant (See Figure 6.1)
OPERATING
TEMPERATURE
0◦C - 60◦C
Note: Photodiode leakage increases with temperature,
leading to increased DC offset voltages and more noise.
QL01 3

Section 2: Specifications
QL01 Quantum-Limited Nanowatt Photoreceiver
SHIELDING Solid extruded aluminum enclosure with die-cast alu-
minum end plates and conductive gaskets for EMI
rejection
POWER 24 V DC @ 150 mA maximum; universal 100-240 V, 50-
60 Hz medical-grade power brick supplied
INDICATOR The green LED on the back panel lights when all power
supply voltages are normal
DIMENSIONS
AND
MOUNTING
Hammond 1457L1201EBK enclosure with removable
stainless steel flange with 1/4-20 or M6-1 tapped hole
and four self-adhesive rubber feet. Flanged end plates
with mounting slots are optional.
QL01 4

Section 3: Photos
3 Photos
Figure 3.1: Top view of the instrument
Figure 3.2: Bottom view showing 1-3/4” mounting flange with 1/4”-20 or M6 threaded
mounting hole
QL01 5

Section 3: Photos
Figure 3.3: End view showing power connector, power indicator LED, and output BNC
QL01 6

Section 4: Quick Guide To Using The QL01
4 Quick Guide To Using The QL01
The Hobbs ElectroOptics QL01 is designed to be very simple to use: basically you put it in
your optical system, connect the power supply and output cable, tweak the alignment, and
your low-level measurement troubles just got a lot less troublesome. There are ways to go
wrong though, so here’s some advice.
•Use the power brick that came with your QL01. This is a medical-grade unit
with all the relevant agency approvals, and has low conducted and radiated electro-
magnetic interference (EMI). A poor quality brick can cause all sorts of noise and
interference problems, and cheap knock-offs may even be dangerous.
•Minimize stray light. Stray light introduces shot noise, and artificial light is usually
strongly modulated by the power source. It isn’t just 100/120 Hz, either. Light from
electronic-ballast fluorescent lamps carries a lot of junk at harmonics of the switching
frequency, usually 20-40 kHz, extending up past 1 MHz. A nanowatt or two of that
can easily spoil your measurement.
•Watch out for pickup. While the QL01 is very well shielded, there has to be an
aperture so that light can reach the photodiode. Other things can get in there too,
especially if there is wiring carrying high frequency signals or fast pulses near the
photodiode. A bit of distance helps a lot; you can’t just stick a LED up next to the
QL01’s photodiode, hit it with a pulse, and expect to get a good measurement of the
QL01’s transient response. That’s why HEO’s test fixture uses a lens to image the test
source on the photodiode.
•Beware of saturation. The QL01’s maximum output is specified as 10 V. There’s
some overrange available, but linearity will be degraded if you go much beyond 11 V.
The unit is internally protected against photocurrents up to about 5 mA (far above
even direct sunlight), but damage may occur above there. This should only occur if a
QL01 7

Section 4: Quick Guide To Using The QL01
bright laser beam were inadvertently to hit the photodiode, which shouldn’t happen
in a low-light measurement system.
Short pulses with high peak power and very low duty cycles can also cause problems.
You can check this by using a few ND 0.3 or 0.5 filters in different combinations. Their
attenuations will add if the system is operating linearly. For instance, an ND 0.3 filter
should reduce the output voltage by about half (3 dB optical, 6 dB electrical), and a
second one by half again, to a quarter of the initial value. These filters are rarely that
accurate, so a more stringent test is to measure each one individually and make sure
that the effect of using the two together is the sum of the individual attenuations.
•Keep it cool. One of the techniques leading to the QL01’s high performance is
applying reverse bias to the photodiode. This reduces its capacitance by several times
and improves the high-frequency noise floor by the same factor. However, photodiode
leakage current increases strongly with temperature, so it’s best to operate the unit
below 30 ◦C to avoid increased offset voltage and shot noise due to leakage. Prolonged
operation with a DC-coupled 50-Ω load may cause a significant amount of internal
heating, so it’s best to use high-Zloads. The QL01’s output is series-terminated, so
you don’t have to worry about cable reflections causing measurement errors with an
open-circuit load at the far end.
•Don’t short the output. Because of its 50-Ω output, it’s very convenient to use the
QL01 with an external filter for narrowband applications, and this works fine. However,
some bandpass and highpass filters present a short-circuit load at DC, and these should
be avoided. While the QL01 will typically survive an output short indefinitely, the
resulting heat will cause extra noise and drift, and above half-scale output it may
cause the internal voltage regulators to shut down. If this happens, the power LED
will turn off.
QL01 8

Section 5: Theory of Operation
5 Theory of Operation
In this section we discuss the problems of low noise photoreceiver design, with specific ref-
erence to the QL01. Some of the figures are from Building Electro-Optical Systems: Making
It All Work [1], where you’ll find lots more on this and many other topics. For any missing
background on circuit design and low noise design in particular, see The Art of Electronics
by Horowitz and Hill [2], especially Chapter 8.
5.1 Noise In Photoreceivers
Photodiodes are amazingly good transducers. They’re highly linear, spatially uniform, and
work well over more than an octave of optical frequency (some over more than two octaves).
Their main drawback is capacitance.
Capacitors don’t have noise of their own, so why is capacitance a problem? The easiest
way to look at it is to consider a simple front end consisting of a photodiode and a load
resistor, as shown in Figure 5.1, which we expect to be followed by a buffer amplifier.
The bandwidth of this circuit is set by the RC time constant RLCd. The effective load
Id
CdRL
VBias
+
(a) Topology
CdRL
INth
INs
Id
(b) Noise model
Figure 5.1: The simplest front end: a load resistor
QL01 9

Section 5: Theory of Operation
impedance ZLis Rand Cin parallel,
ZL=RL
1 + j2πfCdRL
,(5.1)
and the output voltage is IphotoZL.
The output voltage rolls off by 3 dB where the real and imaginary terms become equal,
i.e.
BWRC =1
2πRLCd
.(5.2)
Above there, most of the photocurrent gets swallowed by Cdbefore it ever gets to the
external circuit, so lower capacitance is a win. The low-capacitance champions are silicon
PIN photodiodes run at large reverse bias, some of which come in as low as 40 pF/cm2,
though most are a few times higher (100-150 pF/cm2). InGaAs devices can be easily 100
times higher (4-10 nF/cm2). The ones used in the QL01A have a zero-bias capacitance of
about 70 pF. With the QL01-A’s 10-MΩ transimpedance, the 3 dB bandwidth in that case
would be
1
2π·10 MΩ ·70 pF = 227 Hz.(5.3)
which is almost a factor of 5000 slower than the QL01. The total dark noise current is just
the Johnson (thermal) noise of the load resistor,
iN=r4kT
RL
.(5.4)
Measurements in which this thermal noise dominates are said to be Johnson noise limited.
You don’t want to be in that situation if you can help it, because you worked hard for those
photons, and they’re just going to waste.
Referring to the noise model in Figure 5.1, we see that there are three current sources:
the signal photocurrent (Id), its shot noise (INs), and the Johnson noise of RL(INth). All of
these are wired in parallel, so they get treated exactly alike: the total noise voltage rolls off
along with the signal, so the signal-to-noise ratio (SNR) is frequency-independent, which is a
bit counterintuitive at first. The capacitance problem comes in when we attach an amplifier.
QL01 10

Section 5: Theory of Operation
All amplifiers have some amount of noise in both voltage and current. In the photocurrent
range of the QL01, we’d pick a FET-input amplifier, whose current noise is very small. Its
voltage noise density eNAmp will be reasonably constant above about 1 kHz, so as the signal
rolls off as 1/f in accordance with (5.1), the SNR drops until eventually the system noise is
dominated by the amplifier. The amplifier’s voltage noise density eNAmp begins to dominate
the noise of RLwhen
f0= BWRC
eNAmp
√4kT R .(5.5)
The OPA656 is a typical high performance FET op amp whose maker promotes it for use
with photodiodes. Its input-referred voltage noise density is about 7 nV/√Hz .
With our photodiode and load resistor, the amplifier noise starts to dominate at 13 kHz,
a factor of 10 (20 dB) short of the performance of the QL01. You don’t want to be amplifier
noise limited either, because once again your hard-earned signal is being seriously degraded
by circuit noise. It is important to keep the two aspects of the problem separate in our
minds, because transimpedance amplifiers can fix one (rolloff) but not the other (noise).
5.2 Shot noise
Of course, when we’re actually detecting light, there’s also shot noise to consider. Shot noise
is the quantum version of rain on a tin roof: because photoelectrons are generated at random
times (a Poisson process), there’s an irreducible amount of noise due to the √Nfluctuations
in the counting statistics. Measurements where shot noise dominates other noise sources are
said to be shot noise limited or quantum limited. Photocurrent and photodiode leakage both
have full shot noise1, so we’ll lump them together and say that the noise current spectral
1Except in very special situations (squeezed states) that don’t happen by accident—if you’re in one, you
already know about it.
QL01 11

Section 5: Theory of Operation
density is
iNshot =p2eIphoto (5.6)
where eis the charge of the electron, about 1.602·10−19 coulombs. (This formula is not
mysterious— it’s just √Nconverted to a current and quoted in the frequency domain.)
Measurements where the noise of the photocurrent dominates are said to be in the shot
noise limit. That’s where you want to be if at all possible, because it’s the counting statistics
of your actual signal electrons and not circuit problems that set the SNR. Of course the SNR
can probably still be improved, but that will involve getting more light rather than changing
the circuit. Signal averaging and lock-in detection are also useful, of course, but they’re slow.
A clean fast measurement is always better, and that’s what the QL01 is all about.
Two very useful rules of thumb follow from (5.6): first, shot noise dominates Johnson
noise when IphotoRL>52 mV at 300 K, and second, the SNR is within 1 dB of the shot
noise limit when IphotoRL>200 mV. (You can easily derive these by equating the shot noise
and Johnson noise formulas and solving for IphotoRL.) Thus if we have lots of light we can
reduce Rfconsiderably, and thereby reduce the effect of photodiode capacitance. Milliamps
are easy; nanoamps are hard.
We’ll talk more about calculating when we’re in the shot noise limit in Section 6.6.
Id
Output
Cd
D1
A1
-VBias
RF
CF
Figure 5.2: Schematic of a conventional transimpedance amplifier
QL01 12

Section 5: Theory of Operation
5.3 Transimpedance Amplifiers
We may observe that the signal swing across the photodiode capacitance is the root of the
problem—if there’s no swing, there’s no capacitive current. Thus if we apply feedback around
A1 to wiggle the far end of RLso as to keep the photodiode end still, all of the photocurrent
has to flow through RL, and ideally there’s no rolloff. This circuit is called a transimpedance
amplifier (TIA), shown in Figure 5.2. (Load resistor RLhas become feedback resistor RF
but it’s the same resistor.)
The circuit works by applying progressively higher voltage gain at high frequencies, which
of course amplifies both the signal and the noise. Thus instead of a decreasing signal and
a flat noise floor, we have a flat frequency response and a rising noise floor. In fact the
SNR of the op amp TIA is identical to that of the same amplifier used as a buffer on our
previous circuit, and the amplifier’s noise still dominates above 13 kHz. (There really is no
free lunch.) Its noise gain is the noninverting gain of the stage, which is
AV n = 1 + Rf/j2πfCd.(5.7)
(The small capacitor Cfacross feedback resistor Rfis to prevent instability and control the
noise gain at high frequency.) Another way to look at it is that the op amp imposes its
input noise eNAmp across the photodiode, whose capacitance differentiates it, leading to a
real noise current
iN= 2πfCd·eNAmp,(5.8)
which is often called ‘eNCnoise’.
In TIAs designed for high photocurrents, Rfwill be much smaller, and so the noise
gain will be much less. Also of course the large amount of shot noise will swamp the eNC
contribution most of the time.
The combination of rising eNCnoise, flat shot noise, and the finite bandwidth of the
amplifier makes the shape of the noise floor depend on the photocurrent. While the QL01’s
QL01 13

Section 5: Theory of Operation
Cd
A1
A2
D1
RF
Figure 5.3: Conceptual schematic of a bootstrapped transimpedance amplifier
advanced design minimizes this effect, it cannot be entirely avoided, as you can see in Figures
6.2 and 6.3.
5.4 Bootstraps
Another approach, shown in Figure 5.3, is to use an auxiliary amplifier A2 to apply feedback
to the ground end of the photodiode to make it follow the signal end. In this case, we want
an amplifier with a gain as close as possible to 1.00000—but not above that, or its likely
to oscillate. The advantage of a bootstrap is that it doesn’t have to be DC-accurate, so we
can use discrete devices whose noise is much lower than a FET op amp’s. Like the TIA,
the bootstrap imposes its input noise across the photodiode, which leads to an eNCnoise
contribution as in (5.8), but a much smaller one since A2 is so quiet. The main amplifier A1
doesn’t need to be nearly as good, because from its point of view the photodiode capacitance
is effectively gone.
5.5 The QL01 Difference
The QL01’s frequency response is flat to 1 MHz, and it is shot-noise limited for photocurrents
above 45 nA in its full bandwidth (above 25 nA for the B version). This represents a
QL01 14

Section 5: Theory of Operation
20 dB improvement over the op amp TIA of Figure 5.2. The QL01’s design uses two main
features to achieve its high performance: reverse biasing the photodiode, which reduces its
capacitance by more than a factor of 5, and a proprietary bootstrap circuit with sub-nanovolt
noise density and a gain of more than 0.999 over a wide bandwidth, reducing the effect of
capacitance by more than a factor of 1000.
Design Trade-offs
While it is possible to reduce capacitance a bit further by applying more bias, doing so
starts to increase the photodiode leakage current. Leakage has full shot noise and is a
strong function of temperature, increasing about 9%/◦C, so it has to be no more than a few
nanoamps at room temperature to preserve the QL01’s low-light performance.
QL01 15

Section 6: Performance Verification
6 Performance Verification
This section presents measured noise and transient response data for the QL01-A and -B
along with a discussion of how to verify this performance yourself.
6.1 Pulse Response and Bandwidth
The easiest way to verify the bandwidth is to measure the response to a square pulse. The
transient data plotted below were taken using a Highland Technology P400 digital delay
generator driving a red LED via an RC filter with a 100 ns time constant (2 nF in parallel
with the generator’s 50 Ω output) followed by a 1 kΩ series resistor. A lens of 80 mm focal
length is used to couple the light into the QL01’s photodiode. (The resulting pulses are
slightly asymmetric, accounting for the minor difference between rise and fall times shown.)
With this setup, the 10%-90% rise and fall time should be about 390 ns. Alternatively
an averaged noise floor measurement at full scale (1 µA) or a swept-sine measurement can
be used, but the former needs a lot of averaging and the latter is harder to get right because
of the nonlinearity of the LED.
6.2 Measuring The Noise Floor Accurately
Data for the noise plots were taken using another LED driven from a quiet DC power supply
(Hewlett-Packard 6112A). The strongly rising eNCnoise floor of a simple TIA such as that
in Figure 5.2 continues well beyond its useful bandwidth, with the total noise power going
as bandwidth cubed. This makes a lot of high frequency noise that is a nuisance to get rid
of, and it makes the SNR appear poorer than it really is when viewed on an unselective
instrument such as an oscilloscope or voltmeter. The QL01 largely eliminates this extra
noise with a Gaussian lowpass filter whose bandwidth is 1.2 MHz. Thus it is convenient to
use with an oscilloscope. For narrowband applications, it is simple to add additional filtering
QL01 16