A Low-Noise Resonant Input Transimpedance Amplified Photodetector

Home , JFET

A low-noise resonant input transimpedance amplified photodetector W. Bowden,1, a) A. Vianello,1, 2 and R. Hobson1 1)National Physical Laboratory, Hampton Road, Teddington TW11 0LW, United Kingdom 2)Imperial College London - Department of Physics, London, SW7 2BB, United Kingdom (Dated: 29 October 2019) We present the design and characterisation of a low-noise, resonant input transimpedance amplified photodetector. The device operates at√ a resonance frequency of 90 MHz and ex- hibits an input√ referred current noise of 1.2 pA/ Hz—marginally above the the theoretical limit of 1.0 pA/ Hz set by the room temperature Johnson noise of the detector’s 16 kΩ transimpedance. As a result, the photodetector allows for shot-noise limited operation for input√ powers exceeding 14 µW at 461 nm corresponding to a noise equivalent power of 3.5 pW/ Hz. The key design feature which enables this performance is a low-noise, common-source JFET amplifier at the input which helps to reduce the input referred noise contribution of the following amplification stages.

PACS numbers: Valid PACS appear here Keywords: Suggested keywords

The task of detecting a weak signal encoded in an op- +4.5V tical ﬁeld is ubiquitous across many ﬁelds of research. It 4.7 kΩ 100 Ω 500 Ω 50 nF LMH6624 is often accomplished using a photodiode that produces a 50 nF 500 Ω photo-current proportional to the signal strength, which 4.7 kΩ is then converted to a voltage with a transimpedance CE3514M4 400 Ω

amplifier (TIA). In such applications, it is desirable that 700 nH 10 µF 100 nF 1 µF the signal-to-noise ratio of the measurement is limited 50 nF A 50 Ω by the inherent shot noise present on the optical signal Ip In Req Cp L 40 Ω rather than by electronic noise introduced by the detection electronics. As such, an important figure of merit when choosing a detector is the noise-equivalent-power (NEP), which defines the minimum signal which can be FIG. 1. Circuit diagram showing the two-stage resonant resolved given then detector’s intrinsic electronic noise. TIA photodetector. Not shown is the on-board voltage ref- Along with the NEP, the other important consideration erence used to generate the supply voltage and the op-amp is detector bandwidth. Unfortunately, there is often a decoupling capacitors. Inset: the simplified circuit used for trade-off between these two parameters as low noise mo- modeling which includes the photocurrent Ip and the input tivates high transimpedance values, but this limits de- current noise In in parallel with the the effective transcon- tector response and, when combined with the parasitic ductance Req set by the gate resistance (see equation 1), the capacitance of the photodiode, leads to increased noise inductance L, and the input capacitance Cp—arising primar- gain at high frequencies1. ily from the photodiode. One approach to overcome this limitation is to adopt a resonant architecture. By adding an inductive element at the input, the effect of parasitic capacitance can be sup- op-amp stage were unable√ to achieve an input-referred 8 pressed, enabling a large transimpedance gain at a spe- current noise below 3 pA/ Hz at 50 MHz or higher . 9 cific resonance frequency. As such, they are well-suited One option would be to use a transformer to couple the to heterodyne and homodyne measurement schemes as diode to the amplifier in order to boost the photocurrent, evidenced by their use in gravitational wave detectors2, but this has the drawback of also amplifying the diode 10 cavity-stabilised lasers3, and the generation of entangled capacitance . With the JFET input, the current noise 4 depends on the drain leakage current set by the effective arXiv:1910.11941v1 [physics.ins-det] 24 Oct 2019 states . The resonant photodetector circuit, which builds on resistance Req of the gate. This is given by the real part several previous designs5–7, is shown in figure 1. The of the frequency dependent input impedance of the JFET 11 main modification presented in this work is the inclu- gate , equal to: sion of a JFET common-source pre-amplifier at the input located between the diode and op-amp. Without 1 + ω2R2 C2 this buffer amplifier, the photodetector exhibited high R = L L , (1) eq ω2g R2 C C frequency noise which we attributed to excess op-amp in- m L L gd put current noise. Initial prototype designs that had the where RL and CL are the output load resistance and ca- resonant input connected directly to the non-inverting pacitance, Cgd is the parasitic capacitance coupling the drain to the gate, and gm is the JFET transconductance. This resistance gives rise to a thermal Johnson noise cur- p rent It = 4kbT/R which sets the minimum achievable a)[email protected] NEP. Therefore, it is advantageous to choose a JFET 2 with small parasitic capacitance in order to achieve a between 10 and 100 MHz by comparing the response large transimpedance. One may also want to choose a of a photodiode in balanced vs unbalanced configura- JFET with small transconductance and low output resis- tion. However, in order to suppress the excess inten- tance, but it is important to recognize that this reduces sity noise further, the light is filtered by a Fabry-Pérot the small signal gain of the amplifier, thus increasing the cavity with a full-width at half-maximum of 20 MHz. input referred voltage noise of the op-amp. At incident powers of 60 and 200 µW, the peak output The complex transimpedance Z(s) of the photodetec- noise increased by 7.3 and 11.9 dB respectively, as shown tor is given by: in figure 2. Both measurements yield a PNEP of 14 µW corresponding√ to equivalent input referred current noise sL of 1.2 pA/ Hz. From the width of the noise spectra, Z(s) = 2 (2) the Q-factor of the resonance was measured to be 39, (1 + s LCp) + sL/Req yielding an effective transimpedance of 16 kΩ. The re- s/ω0Q = R , (3) sulting Johnson current√ noise associated with this resis- eq 1 + s/ω Q + s2/ω2 0 0 tance is 1.0 pA/ Hz, indicating√ that there is an ex- which is set by the parallel combination of the gate re- cess current noise of 0.7 pA/ Hz. The transfer function 12 sistance, the input capacitance Cp and the inductance of the photodetector was independently verified by us- L. As the values Cp and Req are not precisely known, it ing an amplitude modulated beam and measuring the is useful to express the transimpedance in terms of the circuit’s response. By changing the inductor value, the p resonant frequency (ω0 = 1/ LCp) and the Q-factor measurement was repeated for two other resonant fre- p quencies, 77 MHz (L = 1 H) and 123 MHz (L = 0.5 H). (Q = Req C/L) which can both be directly measured. µ µ The JFET used for the circuit was the CE3521M413, The resulting√ input current noise√ values were measured manufactured by California Eastern Laboratories, which to be 1.0 pA/ Hz and 1.4 pA/ Hz respectively, consis- is similar to the JFET used in other resonant tent with the fact that the effective transimpedance is photodetectors14. It was chosen based on its low noise- reduced at high frequencies. factor and high transconductance. Unfortunately, no information is provided regarding its input and reverse transfer capacitance. As a result, it is not possible to 30 verify equation 1. However, its intended application for Pi = 0 W high-frequency microwave amplifiers requires such capac- 35 Pi = 60 W itance values to be small. The op-amp used was the 11.9 dBm Pi = 200 W LMH6624, manufactured by Texas Instruments. Its high 40 7.3 dBm gain-bandwidth product, low input noise, and small capacitance makes it well-suited for this application. It 45 was observed that the NEP could be improved by oper- 50 ating the op-amp slightly below its recommended min- Power (dBm) imum supply level of 5 V. The PIN photodiode used 55 for the detector was the Hamamatsu S5973-02 which −1 has a responsivity AR equal to 0.34 A W and a mea- 60 sured capacitance of 2 pF. Care is taken to minimize further PCB related parasitic capacitance by removing the 65 50 60 70 80 90 100 110 120 130 ground planes under the signal path to ensure that the Frequency (MHz) input capacitance is dominated by the diodes contribution. The entire circuit is powered through an on-board FIG. 2. Output spectrum of the resonant photodetector voltage regulator and is enclosed in a metal case. with and without input light. The increase in noise can be The detector’s frequency response and NEP was char- attributed to the inherent amplitude shot-noise present on acterised by measuring the shot noise limited amplitude the light which can be used to determine the NEP of the noise present on an input signal of known optical power detector. Not shown is the detection noise floor, which is well Pi. The resulting shot noise on the photocurrent Is is below the measured photodetector noise. The above spectra √ are an average of 10 sweeps with a resolution bandwidth of given by 2qARPi, where q is the fundamental unit of charge. To boost the signal strength, an additional am- 30 kHz. plifier (Mini-circuits ZFL-500LN+) was added at the output. In general, if the incident light increases the power In an attempt to verify equation 1, the JFET was switched to an IF142, manufactured by InterFET, as spectral density noise by ∆ dB, the incident power PNEP which will induce a current shot noise equivalent to the the required parameters needed to calculate the gate detector current noise is given by: resistance are given in the datasheet (gm = 3 mS, Cgd = 0.6 pF). Combined with the load resistance of 400 Ω and capacitance of 2 pF set by the op-amp input Pi PNEP = (4) terminals, the estimated gate resistance at the resonant 10∆/10 − 1 frequency is 5 kΩ. Based on the measured Q-value of The light source used for the tests is an intensity sta- the detector’s response, the gate resistance is approxi- bilized external cavity diode laser. The laser was ver- mately 4√.5 kΩ. The resulting NEP was measured to be ified to have less than 1 dB of excess intensity noise 5.3 pW/ Hz which corresponds to an input referred cur- 3 √ 3 rent noise of 1.8 pA/√ Hz—in reasonable agreement with J. M. Robinson, E. Oelker, W. R. Milner, W. Zhang, T. Legero, the value of 1.5 pA/ Hz set by the Johnson noise of the D. G. Matei, F. Riehle, U. Sterr, and J. Ye, Optica 6, 240 (2019). 4T. Serikawa, J.-i. Yoshikawa, S. Takeda, H. Yonezawa, T. C. predicted gate resistance. Ralph, E. H. Huntington, and A. Furusawa, Phys. Rev. Lett. In conclusion, we have presented a simple design for 121, 143602 (2018). a low noise resonant photodetector and characterised its 5C. Chen, Z. Li, X. Jin, and Y. Zheng, Review of Scientific In- performance. We have presented a simple model which struments 87, 103114 (2016). 6T. Eberle, Realization of Finite-Size Quantum Key Distribution can be used to predict the effective transimpedance of the based on Einstein-Podolsky-Rosen Entangled Light, Ph.D. thesis, photodetector set by the JFET’s√ gate resistance. The in- Technische Informationsbibliothek und Universitätsbibliothek put current noise of 1.2 pA/ Hz at 90 MHz is near the Hannover (TIB) (2013). 7 theoretical limit imposed by the Johnson noise set by the S. Steinlechner, Quantum metrology with squeezed and entangled light for the detection of gravitational waves, Ph.D. thesis (2013). transimpedance and corresponds to shot noise limited 8Op-amp input current noise, especially for FET inputs, climbs performance for incident optical powers above 14 µW. quickly with frequency and can exceed the nominal value given The intended use of this photodetector is for cavity-based in the datasheet. Unfortunately, high frequency current noise is 15 non-destructive detection of trapped strontium atoms , rarely specified in most data-sheets except for rare√ examples, e.g. but it can be easily modified to operate at a different the LTC6268 which√ has a current noise of 7 fA/ Hz at 100 kHz, but over 5 pA/ Hz at 100 MHz. resonant frequency or wavelength, making it well suited 9C. Chen, S. Shi, and Y. Zheng, Review of Scientific Instruments to a range of applications. 88, 103101 (2017). 10 This work was funded by the UK Department for Busi- S. Potnis and A. C. Vutha, Review of Scientific Instruments 87, 076104 (2016). ness, Energy and Industrial Strategy as part of the Na- 11C. D. Motchenbacher and J. A. Connelly, Low noise electronic tional Measurement System Programme. The authors system design (Wiley, 1993). thank Jochen Kronjaeger for useful discussions and de- 12We used an air coil inductor manufactured by Coilcraft (part signing the photodiode enclosure. Figure 1 was made number 2929SQ-501GEC). It was chosen based on its high self using Circuit Diagram, an online open source program resonant frequency (485 MHz). Furthermore, the inductance can be changed by splaying out the coil windings, allowing for fine for drawing circuits. tuning of the resonance frequency by several megahertz. 13Commercial components are identified for the sake of clarity and does not imply that they are necessarily the best alternatives. 14T. Serikawa and A. Furusawa, Review of Scientific Instruments 1J. G. Graeme and J. G. Graeme, Photodiode amplifiers: op amp 89, 063120 (2018). 15 solutions (McGraw-Hill New York, 1996). G. Vallet, E. Bookjans, U. Eismann, S. Bilicki, R. L. Targat, and 2H. Grote, Review of scientific instruments 78, 054704 (2007). J. Lodewyck, New Journal of Physics 19, 083002 (2017).