A flexible, open-source radio-frequency driver for acousto-optic and electro-optic devices
D.S. Barker, N.C. Pisenti, A. Restelli, J. Scherschligt, J.A. Fedchak, G.K. Campbell, S. Eckel
AA flexible, open-source radio-frequency driver for acousto-optic andelectro-optic devices
D.S. Barker, a) N.C. Pisenti, b) A. Restelli, c) J. Scherschligt, J.A. Fedchak, G.K. Campbell, and S. Eckel Joint Quantum Institute, University of Maryland and National Institute of Standards and TechnologyCollege Park, MD 20742, USA Sensor Science Division, National Institute of Standards and TechnologyGaithersburg, MD 20899, USA (Dated: 7 August 2019)
We present a design for a radio-frequency driver that leverages telecom amplifiers to achieve highpower output and wide bandwidth. The design consists of two compact printed circuit boards (totalarea <
255 cm ), which incorporate power (turn-on) and thermal management to prevent accidentaldamage to the amplifier circuitry. Our driver provides > . ≥ ≈
70 kHz bandwidth), as well as digitalpower switching ( >
30 dB of extinction within 40 ns and final extinction >
90 dB). The radio-frequency source can also be digitally switched between an external input and an integrated voltage-controlled oscillator. Our design is motivated by the need for flexible, inexpensive drivers of opticallyactive devices, such as acousto-optic and electro-optic modulators.
I. INTRODUCTION
Active optical devices are a ubiquitous tool in experimen-tal optics. In particular, electro-optic (EO) and acousto-optic (AO) devices are used for amplitude, frequency, andphase modulation, as well as beam deflection and polar-ization control. AO and EO modulators have found ap-plications in a variety of fields, which include ultrafastlaser physics, microscopy, quantum computing, and spec-troscopy. These applications often require analog and digi-tal control of both the amplitude and frequency of the modu-lator’s radio-frequency drive. For example, analog frequencymodulation is necessary for several spectroscopy techniques, fast digital amplitude control is crucial for quantum logic op-erations, and analog amplitude control is required for stableoptical dipole traps. AO and EO modulators typically operate in the radio-frequency (RF) range and require (cid:38) (cid:38)
100 MHz,power control with (cid:38)
20 dB of dynamic range, and switch-ing times on the order of 10 ns with (cid:38)
60 dB extinction ratio.Commercially-available drivers for AO/EO modulators oftenhave a narrow RF bandwidth and typically offer only a subsetof the desired RF controls. Researchers (such as ourselves)often attempt to circumvent the limitations of commercially-available drivers by combining off-the-shelf RF componentsto produce a custom AO/EO driver.
However, both com-mercial and custom RF drivers are typically only suitable forspecific modulators or applications.We present a design for a high-flexibility AO/EO devicedriver that has high output power over a wide frequency band-width. Our circuit uses a custom power sequencing architec-ture to leverage telecom amplifiers, which allow us to achieveRF signal gain (cid:38)
30 dB and output power (cid:38) a) [email protected] b) Present address: IonQ Inc., College Park, MD 20740, USA c) [email protected] to 1 . <
255 cm ).Careful PCB layout keeps the channel temperature of thehigh-power amplifier well below its maximum rating, even inthe absence of forced-air cooling. Our design incorporates anarray of analog and digital controls, which make it compatiblewith a variety of applications. The RF driver’s output poweris continuously tunable with a dynamic range of ≈
30 dBand an analog modulation bandwidth of ≈
70 kHz. Duringdigital switching, the RF power falls by 10 dB in ≈
25 ns,and by >
30 dB within 40 ns, with a final extinction ratio >
90 dB. The driver also allows digital switching betweenan integrated voltage-controlled oscillator (VCO), which pro-vides analog frequency tunability, and an external RF source,such as a direct digital synthesizer (DDS) or software-definedradio (SDR).The RF driver is packaged in a 2U 19” rack enclosure,with the size of the case limited by the driver’s linear powersupply. We discuss the circuit design in Section II. Designfiles for both PCBs, including schematics, board layouts, andbills of materials, are available online.
Section III con-tains our measurements of the RF driver’s performance. Wesummarize the results and discuss future modifications to ourdesign in Section IV. The cost of both fully-populated PCBsis ≈ $1000 for a run of one. In a production run of 40 RFdrivers, the cost per driver, including the rack enclosure andall auxiliary components, was $1085, which is competitivewith commercially-available alternatives. II. CIRCUIT DESIGN
The central features of our design are wide RF ampli-fication bandwidth, high output power, fast digital switch-ing, and high-dynamic-range analog power tuning. The de-sign requirements were informed by the AO/EO devices com-monly employed in our laboratories, which fall into four broadclasses: low-power ( (cid:39) (cid:39) a r X i v : . [ phy s i c s . i n s - d e t ] A ug VCOExtRF -3 dB RFMon
RFOut
HMC8410
DirectionalCoupler VVA
HMC1099 S M A High PowerAmplifier
Attenuator SW1SW2SW3 SW4
FIG. 1. Schematic of radio-frequency components. The RF switch SW2 selects either the VCO or Ext RF connector as the RF signal source.Switches SW3 and SW4 reduce cross-talk between the two RF signal sources and increase the RF extinction ratio. The RF signal passesfrom SW2 into a low-noise, wide-bandwidth pre-amplifier (HMC8410), which increases the RF power by ≈
20 dB. A directional coupler(ADC-20-4) splits off ≈ ≈
30 dB. The switch SW1, in conjunction with SW3/SW4, toggles the RFoutput power. A short SMA cable connects the pre-amplified RF signal to a separate high power amplifier PCB (highlighted in green), wherethe RF signal is amplified to (cid:38) nation of the RF bandwidth and low frequency ( (cid:46)
100 MHz)power requirements pose the most significant obstacle to therealization of an RF driver compatible with all these disparatedevice classes.
A. RF Amplification and Control
Figure 1 is a diagram of our design’s RF components. Wearrange the pre-amplifier (HMC8410), directional coupler,voltage-variable attenuator (VVA), and rightmost RF switch(SW1) for better dynamic range and extinction. The com-ponent arrangement reduces the attainable RF output powerdue to the insertion losses in SW1, the VVA, and the direc-tional coupler. Our component selection represents a com-promise between the competing goals of high output powerand dynamic range. The directional coupler picks off ≈ The RF components are connectedby grounded coplanar waveguides with a designed charac-teristic impedance of 50 Ω at 1 GHz. To improve the low-frequency performance, we use 100 nF broadband capacitorsto AC-couple the RF signal between components. Two wide-bandwidth, high-gain amplifiers, an HMC1099and an HMC8410, form the core of our RF driver.
TheHMC8410 serves as a pre-amplifier, a task for which it is ide-ally suited due to its low noise figure ( (cid:46) < >
10 W and high power-added effi-ciency ( (cid:38)
50 % under our operating conditions). It provideshigh power amplification and is mounted on a separate PCB(green highlighted region in Figure 1) to increase design flexi-bility. An inline attenuator can be installed between the PCBsto restrict the maximum output power to prevent damage todelicate optical devices. Both amplifiers are internally pre-matched to 50 Ω across their full operating bandwidth, whichgreatly simplifies external impedance matching. We based the external impedance matching network, including componentselection and placement, on the typical application circuits foreach amplifier. Unfortunately, both amplifiers lack inter-nal biasing circuitry and also have strict power sequencing re-quirements. The gate voltage (V GG ), drain voltage (V DD ), andRF input must be applied sequentially to each amplifier to pre-vent damage. We leave discussion of the details of our powersequencing electronics to Section II C.Four RF switches (MASWSS0178) provide digital controlof the power output and signal source of the RF driver. Each switch has >
50 dB of isolation and a switching timeof ≈
20 ns. Two 5 V TTL logic signals, V on and V sel , governthe state of the RF switches. Switch SW1 turns the RF outputon or off depending on whether V on is high or low. SwitchSW2 sets the RF source to the VCO when V sel is high, or tothe external RF input when V sel is low ( ¬ V sel is high). Theremaining switches, SW3 and SW4, increase the extinctionratio (when V on is low) or reduce bleed-through of the unde-sired RF source onto the RF output (depending on the state ofV sel ).We use a linear-in-dB, voltage-variable attenuator (F2255)for analog adjustment of the driver’s RF output power. TheVVA has low insertion loss ( < . ≈
30 dB. In contrast to its excellent static properties, theF2255 attenuator has a comparatively slow, and poorly spec-ified, response to changes in its control voltage, V att . In ap-plications with lower dynamic range or output power require-ments, the VVA can be replaced with a pin-compatible partfrom another manufacturer, which offer up to ≈
10 timeslarger amplitude modulation bandwidths. We employ a VCO as the RF driver’s on-board RF source.The VCO is controlled and frequency modulated with a tun-ing voltage, V tune . The RF bandwidth of a VCO is typicallymuch narrower than the bandwidth of the other RF compo-nents in our design. Multiple VCOs could be used in con-junction with a switching network to increase the bandwidthof the on-board RF source, but the additional componentswould increase cost and complexity. We have instead madethe RF driver compatible with all 5 V and 12 V VCOs in theCVCO55 series (from Crystek), or the JTOS and ROS series(from Minicircuits). Most VCOs have sufficient output powerto saturate the HMC8410 pre-amplifier. To reduce saturationof the pre-amplifier, and the associated harmonic distortion,we insert a 3 dB fixed attenuator between the VCO and SW4.High-frequency VCOs often have a lower output power, sothe fixed attenuator can be replaced with a short when drivinghigh-power, high-frequency devices. B. Analog and Digital Control
The RF driver allows both internal and external controlof the analog and TTL signals that control its RF output.Front-panel switches permit the user to chose internal, ex-ternal, or, for analog signals, summed (internal + external)control of V on , V sel , V att , and V tune . High-speed optocouplers(TLP2767) detect external TTL signals and provide groundisolation. We configure the optocouplers to be compatiblewith both 3 . Ω input impedance, and a 50 Ω resis-tor isolates the input reference level from circuit ground. Anon-board microcontroller supplies internal setpoints for theanalog voltages V att and V tune using a dual-channel digital-to-analog converter (DAC). Both internal setpoints can be ad-justed using a front-panel rotary encoder and LCD display.Wide-bandwidth op-amps (LM7171) buffer both the inter-nal and external analog control voltages. These op-amps mul-tiply or divide the analog signals such that the output volt-age range of both a typical external voltage source and theDAC will span the input range of the VVA or VCO. In thesummed control mode, V att = V DACatt + V extatt / tune = DACtune + V exttune ). Currently, V att and V tune are only limited bythe ±
15 V power rails of the LM7171 op-amps. It is there-fore possible to exceed the maximum rating of the VVA con-trol voltage and the tuning voltage of certain VCOs when thedriver is operated in either the summed or external controlmodes. Future versions of the RF driver may integrate Zenerdiodes on the op-amp outputs to protect the RF components.In addition to controlling the DAC, the microcontroller isconfigured to accept triggers on an auxiliary optocoupled TTLline and to control an AD9910 DDS. A DDS shield PCB(based on the AD9910 evaluation board ) can be mountedon to the RF driver, which incorporates extra voltage regula-tors to power the DDS. The current version of the microcon-troller software permits only manual adjustment of the DDSfrequency. The auxiliary TTL trigger enables tabled controlof V
DACatt , V
DACtune , and all DDS parameters to be implemented infuture microcontroller software versions. The code base andDDS shield could also be adapted to other AD99XX seriessynthesizers, some of which can generate sinusoidal wave-forms over the full RF amplification bandwidth of our RFdriver. C. Amplifier Power Sequencing and Bias Control
The performance of monolithic microwave integratedcircuit (MMIC) amplifiers, such as the HMC1099 andHMC8410, depends sensitively on the amplifier’s drain cur-rent, I DD . The drain current depends on the gate and drainvoltages (V GG and V DD ) as well as the input RF power. Ex-cessive drain current can damage the amplifier, so V GG , V DD ,and the RF signal must be applied to the amplifier sequen-tially.There are two methods for biasing I DD : constant gate volt-age and constant drain current. The constant drain currentapproach has greater long-term repeatability, but tends to re-strict maximum power output because the RF input cannotdynamically pull additional current from the V DD supply. Inthe constant gate voltage technique, V GG is set to produce aparticular quiescent drain current, I DQ , prior to activation ofthe RF input. A fixed V GG allows the amplifier to pull addi-tional drain current as the RF input power increases, which re-duces gain saturation. However, the gate voltage that producesthe desired I DQ may change as an amplifier ages. Biasing anamplifier with a constant drain current is generally preferablewhen an amplifier’s optimal I DQ is large enough thatI DQ V DD > P sat , (1)where P sat is the saturated output power. Condition (1) is validfor the HMC8410 pre-amplifier, but not the HMC1099 high-power amplifier. Consequently, the RF driver employs bothdrain current biasing procedures.We control the HMC8410 pre-amplifier using an activebias controller (HMC920LP5E), which also ensures propersequencing of V GG1 and V
DD1 during start up and shut down(see Figure 1). Our bias control circuit is an amalgamationof example circuits from the HMC920LP5E and HMC8410datasheets.
To increase the pre-amplifier’s saturated out-put power, we bias the HMC8410 with V
DD1 = . DD1 =
75 mA. Operating the HMC8410 under these biasconditions, at the high end of the permitted range, slightlydegrades its noise figure. In applications with lower power re-quirements, the noise figure can be improved by reconfiguringthe active bias control circuit.A power supply sequencer (LTC2924) governs the gate anddrain voltages for the high-power amplifier. The LTC2924can drive the gates of external N-channel MOSFETs to se-quence up to four external supply voltages. Due to its limitedoutput voltage, the LTC2924 can normally only control pos-itive power supplies with output ≤ DD2 ≥
24 V during normal operation and V
GG2 as lowas − ≈
10 V), which puts the PNP transis-tor Q0 into conduction and so pulls the gate of MOSFETQ1 to ground. As Q1 goes into conduction, the bufferingop-amp U1 pulls V
GG2 down to − GG2 into a pos-itive voltage, which the power sequencer can sense at IN1.The LTC2924 waits until V
GG2 reaches a threshold value,
SS1, NC U1U2IN1 OUT1 OUT3IN3 I DD Mon
U3 IN2OUT2
RFEnable
OUT4IN4 SS2, NOR trim
R2 R1Q1Q0 Q3 Q2Q4
FIG. 2. Simplified schematic of the I DD bias controller for the high-power amplifier. Blue-filled (red-filled) circles denote the output (in-put) pins of a LTC2924 power sequencer. Purple-filled circles markthe gate and drain voltage of the HMC1099 amplifier (see Figure 1).The power sequencer consecutively activates OUT1 through OUT4to put transistors Q1 through Q4 into conduction. As each transistorturns on, the LTC2924 waits for the associated input (IN1 to IN4)to reach a resistively programmed threshold value before activatingthe next output (the threshold setting voltage dividers are omitted forclarity). The transistors Q1 to Q3 successively set V GG2 to − DD2 to 24 V; and switch V
GG2 to a value determined by R trim .The I DD Mon output allows us to adjust R trim to obtain a quiescentdrain current I
DQ2 =
100 mA. Transistor Q4 actuates an optocoupledsolid-state relay on the pre-amplifier PCB via the RF Enable output.The solid-state relay locks V on low until both amplifiers’ gate anddrain voltages have stabilized. V th = − .
61 V × ( R / R DD2 up to 24 V.Pin IN2 senses the HMC1099 drain voltage through a resis-tive divider (not shown in Figure 2). Once V
DD2 stabilizes,the power sequencer raises OUT3 to open the normally-closed(NC) relay SS1. The gate voltage then switches from − trim . The op-amp U3 monitors the drain current, I DD2 , and allows us to ad-just R trim to achieve the recommended quiescent drain current,I
DQ2 =
100 mA.A front-panel rocker switch enables both the pre-amplifier’sactive bias controller and high-power amplifier’s power se-quencer. To prevent RF-induced damage to both amplifiersduring power sequencing, two optocoupled solid-state relayslock V on low when the rocker switch is off. Once both ampli-fiers are on, the status output of the active bias controller andOUT4 of the power sequencer switch the two relays, whichpermits the RF signal to be enabled. Both controllers reversetheir power sequences when the rocker switch is flipped off,so both amplifiers are fully protected from damage.During normal operation, the rocker switch should be in theoff position whenever the driver’s AC mains are switched onor off. We also tested the operation of the power sequencerswhen the AC mains are improperly toggled while the rockerswitch was turned on. To increase the potential for damageto the amplifiers, we applied 1 mW to the external RF inputand set V on high during the test. After 100 improper powersequencing cycles there was no degradation of the RF driver’sperformance. Note that the amplifiers are not auto-terminated,so they can still be damaged if the power sequencers are acti- vated when no output load is connected. D. Thermal Management
Even though the high-power amplifier has high power-added efficiency, it still generates (cid:39) . × . × . K ≈ . − K − ). The area beneath the heatsink has nosolder mask and the only traces in this region, on any layer ofthe PCB, are power or signal connections to the HMC1099.More than 100 vias provide heat conduction from the bottomlayer of the PCB, where the amplifier is attached, to the toplayer, with the BGA heatsink. The amplifier PCB is mountedhorizontally in the RF driver box with the fan attached the sideof the box and ≈ ◦ C of ambient regardless of the RF input power (seeSection III, Figure 3).The acoustic noise of the DC fan may be unacceptable incertain laboratory environments. With the fan disabled, theBGA heatsink equilibrates at ≈ ◦ C above ambient with noRF input and I
DQ2 =
100 mA. When we apply 10 mW tothe external RF input, the BGA heatsink stabilizes at ≈ ◦ Cabove ambient, which corresponds to a heat load of ≈ i.e. we neglect the presence of the vias) and esti-mate that the thermal conductivity of FR4 is 0 .
25 W m − K − ,then the thermal resistance from the heatsink to the ampli-fier’s thermal pad is ≈ . ◦ C W − . The junction-to-pad ther-mal resistance of the HMC1099 is 11 . ◦ C W − , so the junc-tion temperature is (cid:46) ◦ C under passive cooling (the ab-solute maximum junction temperature for the HMC1099 is225 ◦ C). However, the absolute maximum case temperaturefor the HMC1099 is only 85 ◦ C. Our measurements sug-gest that the amplifier case temperature could approach themaximum rating in environments with ambient temperatures > ◦ C. Laboratories with poor temperature stability shouldeither use forced-air cooling (as we suggest above) or a cus-tom copper heatsink to improve the thermal performance.
III. RESULTS
We show the RF power output of our design in Figure 3. Wemeasured the RF power in the 1st, 2nd, and 3rd harmonic asfunction of frequency using a Rohde & Schwarz FSV scalarnetwork analyzer with a controllable SMB100A tracking gen-erator. The RF output of the tracking generator was connectedto the external RF input of a test driver, which was attachedto the network analyzer through a 30 dB attenuator and a DCblock. The transmission losses of the external RF componentsand cables were calibrated out by the network analyzer. Fora 1 mW input power, our driver outputs > . > . ≥ Carrier Frequency (MHz) − − − − − P o w e r O u t pu t( W ) − − − − − P o w e r O u t pu t( W ) − − − − − P o w e r O u t pu t( W ) (a)(b)(c) FIG. 3. Power output in the 1st, 2nd and 3rd harmonics (green solid,purple dash-dotted, and gold dashed lines, respectively) as a functionof carrier frequency. The RF power applied to the external input is {
10 mW , . , } in subplot { (a), (b), (c) } . frequency range. The increase in carrier output power is ac-companied by an increase in harmonic distortion, particularlyat frequencies (cid:46)
400 MHz and in the 3rd harmonic. The har-monic distortion (referred to the carrier) is always < −
10 dBinside the RF driver’s operating bandwidth and agrees withthe 2nd harmonic data in the HMC1099 datasheet.
Commercially-available RF drivers that specify harmonicdistortion typically report harmonic content (referred to thecarrier) < −
15 dB or < −
20 dB. Our driver meets the less-stringent specification at low input power (Figure 3 (c)) exceptfor frequencies (cid:46)
30 MHz and in the range from 400 MHz to525 MHz. The driver also meets the more-stringent specifi-cation at all input powers for frequencies >
700 MHz. Whendriving narrow-bandwidth AO or EO devices, the harmonicdistortion is unlikely to be problematic because higher har-monics will be filtered by the device itself. External filteringmay be beneficial for applications involving wide-bandwidthdevices, such as fiber EO modulators, especially when thebandwidth of the device exceeds the bandwidth of the RFdriver. However, broadband AO/EO devices typically requireless RF drive power and have non-linear response to weakRF drives. These device characteristics tend to reduce boththe amount of harmonic distortion and its effect. We estimatethat, for a broadband fiber phase modulator driven at a typicalhalf-wave voltage ( (cid:39) −
13 dB, the fractional power in the first-order sidebands associated with the dominant harmonic will −
10 0 10 20 30 40 50 60 70 80
Time (ns) − . . . V o l t ag e ( V ) − − P o w e r( d B ) − . . . V o l t ag e ( V ) − − P o w e r( d B ) − . . . V o l t ag e ( V ) − − P o w e r( d B ) − . . . V o l t ag e ( V ) − − P o w e r( d B ) (a)(b)(c)(d) FIG. 4. Digital switching performance of the RF driver at input fre-quencies of 1 GHz (a), 500 MHz (b), 100 MHz (c), and 50 MHz (d).Each subplot shows the RF waveform (blue), RF envelope (gold),and TTL signal (red) on the left axis. The relative peak-to-peakpower, P pp , (see Equation 2) is plotted against the right axis in pur-ple. The TTL signal is vertically offset for clarity. The verticaldashed line marks the initial response of the RF output and the ver-tical { dotted, dash-dotted, dash-dot-dotted } lines show the time atwhich the RF output falls to {
90 %, 10 %, < . } of its initialvalue. For all these data, the VVA was set for minimum attenuation.The harmonic distortion that is apparent in each subplot is consistentwith the measurements in Figure 3. be <
10 %. For lower modulations depths ( i.e. when onlythe first-order sidebands contain appreciable power), the frac-tional power in the first-order sidebands associated with thedominant harmonic will be (cid:46) upper ( t ) and V lower ( t )), we compute therelative peak-to-peak power P pp ( t ) = (V upper ( t ) − V lower ( t )) (V upper (0) − V lower (0)) , (2)which we plot on the righthand axis in Figure 4. The peak-to-peak power begins to decrease ≈
15 ns after the arrival ofthe TTL signal edge (vertical dashed line in Figure 4). The
Carrier Frequency (MHz) − − − − − − − − E x t i n c t i o n ( d B ) FIG. 5. Extinction of the RF output as a function of frequency at1 mW input power. Each trace is the difference between the carrieroutput power with V on low and V on high (the solid green curve inFigure 3 (c)). The gold dashed (purple dash-dotted) curves show theextinction with the minimum (maximum) VVA attenuation setting.The green solid line represents the noise floor of the measurement.We acquired the V on low portion of the data for the noise floor mea-surement with the RF input to the driver terminated, but with thetracking generator turned on and the VVA attenuation maximized. propagation delay is consistent with the specifications of theoptocouplers and TTL logic chips that drive SW1 throughSW3 (the delay due to mismatched cable lengths is ≤ P pp diminishes by10 dB within 25 ns (vertical dash-dotted line in Figure 4) andit falls below the noise floor of our measurement technique, ≤ −
30 dB, within 40 ns (vertical dash-dot-dotted line in Fig-ure 4). The switching time is consistent with the specifica-tion of the RF switches and is comparable to rise/fall times ofcommercially-available drivers (the fall time of our driver isthe distance between the vertical dotted and dash-dotted linesin Figure 4). Changing from a 5 V to a 3 . P pp , so the switching times that we quoterepresent an upper bound on the performance of the RF driverin laboratory applications. The full turn-on time for the RFoutput is approximately a factor of two longer than the turn-off time (the positive voltage envelope responds less quicklyduring turn-on), which limits the maximum digital modulationfrequency to (cid:46)
10 MHz (at 50 % duty cycle). The RF switcheshave a symmetric switching time specification, so the asym-metry of the rise and fall times of the RF driver is likely causedby asymmetric response of the high-power amplifier.We investigated the final RF extinction ratio using the FSVnetwork analyzer. The tracking generator sent 1 mW of RFpower to the driver and the network analyzer recorded the RFoutput with V on held low. We measured the RF power out-put with the VVA set for both minimal and maximal attenu-ation. The extinction is the difference between the measuredRF output with V on low and V on high. To calculate an extinc-tion noise floor, we also acquired the RF output power withthe driver’s RF input terminated, V on held low, the VVA setfor maximum attenuation, and the tracking generator turned Modulation Frequency (kHz) − − − − − M ag n i t ud e ( d B ) FIG. 6. Relative RF output amplitude modulation as a functionof modulation frequency. The green solid, gold dashed, and pur-ple dash-dotted lines show the relative magnitude of the RF outputresponse for sinusoidal peak-to-peak modulation depths of 100 mV,200 mV, and 400 mV, respectively. The {
100 mV, 200 mV, 400 mV } peak-to-peak modulation depth corresponds to a peak-to-peak RFpower variation of { } at DC. on. Figure 5 shows the results of our extinction measure-ments. Regardless of the VVA attenuation setting, we observeextinction ratios >
90 dB over the full RF bandwidth of the RFdriver, and >
110 dB for frequencies <
400 MHz. The noisefloor for our measurement varies with frequency, but alwayscorresponds to an extinction ratio >
125 dB. The extinctionratios that we observe, to the best of our knowledge, surpassthe specification of any commercially-available driver by atleast 20 dB.The modulation bandwidth of the VVA is not well speci-fied and RF amplitude modulation bandwidth is important inseveral applications ( e.g. laser intensity stabilization ). Tostudy the AM bandwidth of the RF driver, we set V DACatt tothe center of the VVA’s linear-in-dB range and drove V extatt with a sinusoid. The network analyzer recorded the ampli-tude of the modulation of the RF output as a function of themodulation frequency. Because the FSV is a scalar networkanalyzer, we cannot measure the full amplitude modulationtransfer function of the RF driver. Figure 6 shows the rela-tive magnitude of the amplitude modulation response for sev-eral modulation depths. The amplitude modulation bandwidthincreases slightly with increasing modulation depth and the3 dB point for the lowest modulation depth is ≈
70 kHz. Our measurement is in reasonable agreement with the small-signal modulation bandwidth suggested by the F2255 VVAdatasheet ( ≈
65 kHz). In applications where wider modula-tion bandwidths are necessary, the VVA can be replaced witha pin-compatible part with ≈
10 times faster response. IV. CONCLUSION
We have designed and characterized an RF driver for AOand EO devices. The driver achieves high output power andwide RF bandwidth using telecom amplifiers, which are fullyprotected during power cycles by a combination of custom
Frequency (MHz) . . . P o w e r O u t pu t( W ) . . . . . . FIG. 7. Output power for RF drivers from the initial production run.The external RF input power for these data is 1 mW. The output pow-ers are binned by frequency and colored according to the normally-distributed probability of occurrence for a given output power. Thecolor scale is normalized to the most probable output power in eachbin. Black dots show the output power for the test driver that weused to acquire the data in Section III. The black dashed horizontalline is a guide to the eye. The device-to-device variation in the outputpower is within the specification of the amplifiers and is exaggeratedby our measurement’s resolution bandwidth, which is 10 kHz. and commercial power sequencing electronics. The driver in-cludes flexible controls for AM, FM, and digital switching ofthe RF output. In addition to the test data on a single driverthat we relate in Section III, we also have more limited datafrom a production run of 40 drivers. We present the poweroutput performance of drivers from the production run in Fig-ure 7 and Table I. Our design could be modified to increasedigital switching speed, analog AM bandwidth, or extinctionratio at the expense of power output. Our power sequencingand bias control electronics can be adapted for other amplifiersand applications. All design materials are available online forothers to use.
TABLE I. Maximum power output of the production run RF driverswhen using the integrated VCO as the RF source. Voltage-controlledoscillators with nominal frequencies of {
40 MHz, 80 MHz, 200 MHz,400 MHz, 800 MHz } were installed in {
4, 22, 6, 1, 3 } drivers. We re-port the mean and standard deviation of the maximum output powerfor each VCO frequency class with the 3 dB fixed attenuator installedbefore the pre-amplifier (see Figure 1). The larger relative outputpower variation at 800 MHz is caused by output power differencesbetween VCOs (we used two 800 MHz VCO models, due to productdiscontinuation).VCO Frequency Output Power40 MHz 5 . . . . . ACKNOWLEDGMENTS
The authors thank J. Gardner and K. Douglass for theircareful reading of the manuscript. Our work was partiallysupported by the Office of Naval Research, and the NationalScience Foundation through the Physics Frontier Center at theJoint Quantum Institute. DSB acknowledges support from theNational Research Council Postdoctoral Research Associate-ship Program. U. Keller, K. D. Li, B. T. Khuri-Yakub, D. M. Bloom, K. J. Weingarten,and D. C. Gerstenberger, “High-frequency acousto-optic modelocker forpicosecond pulse generation,” Optics Letters , 45 (1990). H. R. Morris, C. C. Hoyt, and P. J. Treado, “Imaging Spectrometers forFluorescence and Raman Microscopy: Acousto-Optic and Liquid CrystalTunable Filters,” Applied Spectroscopy , 857 (1994). S. Debnath, N. M. Linke, C. Figgatt, K. A. Landsman, K. Wright, andC. Monroe, “Demonstration of a small programmable quantum computerwith atomic qubits,” Nature , 63 (2016). G. Camy, C. J. Bord´e, and M. Ducloy, “Heterodyne saturation spectroscopythrough frequency modulation of the saturating beam,” Optics Communi-cations , 325 (1982). G. C. Bjorklund, “Frequency-modulation spectroscopy: a new method formeasuring weak absorptions and dispersions,” Optics Letters , 15 (1980). J. L. Hall, L. Hollberg, T. Baer, and H. G. Robinson, “Optical heterodynesaturation spectroscopy,” Applied Physics Letters , 680 (1981). V. Negnevitsky and L. D. Turner, “Wideband laser locking to an atomicreference with modulation transfer spectroscopy,” Optics Express , 3103(2012). S. R. Granade, M. E. Gehm, K. M. O’Hara, and J. E. Thomas, “All-Optical Production of a Degenerate Fermi Gas,” Physical Review Letters , 120405 (2002). M. D. Barrett, J. A. Sauer, and M. S. Chapman, “All-Optical Formation ofan Atomic Bose-Einstein Condensate.” Physical Review Letters , 010404(2001). https://dcc.ligo.org/LIGO-E1400445/public . B. Fr¨ohlich, T. Lahaye, B. Kaltenh¨auser, H. K¨ubler, S. M¨uller, T. Koch,M. Fattori, and T. Pfau, “Two-frequency acousto-optic modulator driver toimprove the beam pointing stability during intensity ramps,” Review of Sci-entific Instruments , 043101 (2007). The PCBs could be adapted to fit in modular electronics crates ( e.g.
Eu-rocard, NIM). The crate’s bus would need to source (cid:39) https://github.com/JQIamo/aom-driver . https://github.com/JQIamo/GaN-Amplifier . The identification of commercial products is for information only and doesnot imply recommendation or endorsement by the National Institute ofStandards and Technology. Future versions of our design will use two directional couplers placed be-fore SW3 and SW4 (see Figure 1) to increase both flexibility and outputpower. Results from a prototype suggest that similar performance can be achievedat lower cost using 10 nF standard capacitors with a high self-resonant fre-quency. HMC1099 GaN Power Amplifier , Analog Devices (2016), Rev. A. HMC 8410 Low Noise Amplifier , Analog Devices (2016), Rev. 0. Analog Devices Technical Support, Private Communication (2017). MASWSS0178 SPDT High Isolation Terminated Switch , Macom, Rev. V4. IDTF2255NLGK Datasheet , Integrated Device Technology (2017), REV 1. Possible replacements include RFSA2033 and RFSA2113, which increasemodulation bandwidth at the expense of dynamic range and power output,respectively. A. Bhandare, A. Patnaik, D. Pommerenke, S. Sharma, and D. Fischer, “Lowcost fast frequency switching driver for Acousto-Optic Modulators used inlaser cooling,” HardwareX , e00054 (2019). TLP2767 , Toshiba (2016), Rev. 1.0. AD9910 Direct Digital Synthesizer (2016), Rev. E. https://github.com/JQIamo/ad9910-dds . AD9914 3.5 GSPS Direct Digital Synthesizer with 12-Bit DAC , Analog De-vices (2016), Rev. F. K. Kaya,
AN-1363 Application Note: Meeting the Biasing Requirements ofExternally Biased RF/Microwave Amplifiers with Active Bias Controllers ,Analog Devices (2016), Rev. 0. HMC920LP5E Active Bias Controller , Analog Devices, v07.1013. LTC2924 Quad Power Supply Sequencer , Linear Technology (2016), REVC. Our power sequencing electronics could be adapted for other high-poweramplifiers with superior 2nd harmonic distortion. These amplifiers requireV DD >
24 V, so an additional power supply would be necessary. We find that RF pickup from the tracking generator, even when it is notattached to the RF driver, is an important contribution to the measured RF spectrum. When using the RF driver to intensity stabilize laser beams, we have ob-served intensity locking bandwidths (cid:38)(cid:38)