J. Instrum. Soc. India 38(1) 63-70 63 DESIGN OF A HIGH FREQUENCY SMPS POWERING LASER DIODE A. Sharma, C.B. Panwar, R. Arya and A.K. Nath Solid State Laser Division, Raja Ramanna Centre for Advanced Technology, Indore 452 013 ABSTRACT This paper describes the design of a high frequency SMPS capable of driving semiconductor laser diodes upto 120 W rating in CW (continuous) mode or pulsed mode upto 100 Hz. The duty cycle of the pulsed operation can be varied. The SMPS takes input from 230 V 1 utility mains and delivers a regulated pulsed current to the diode. The current in CW or pulsed mode is programmable upto 40A (rms ripple < 0.15% and line & load regulation better than 0.1%) supporting a maximum diode voltage of 3V The SMPS is a MOSFET based double switch forward converter operated in peak current mode control with a two loop feedback. Peak Current mode control implements the inner loop and it effectively reduces the order the system by one and also introduces a feed forward path of the input line voltage that prevents its variations from reaching the output. Hall based sensor from LEM senses the diode load current which is used both for metering purpose and for closing the outer loop in average current mode. The high switching frequency of 250 kHz has advantages like small size, reduced weight, low ripple and faster dynamic response. There are suitably designed differential and common mode filters in the output stage and the load terminals are galvanically isolated from the utility mains as well as from the common of sensing and control circuit to prevent a possible EMI and static charge based damage of the diode. All power and control components are housed in one PCB. Such a diode driver find applications in many fields including ophthalmologic laser, laser marker etc. 1. INTRODUCTION Laser diode (LD) pumped solid state lasers have progressed spectacularly over last decade[l]. Applications range from its use in laboratory by a research scholar to its commercial use in a diode laser marker to health care in laser imaging and ophthalmology. In health care, lasers and optics has already affected the life of many patients by offering a new and better solution to major health problems. It may be pointed out here that amongst various branches of surgery, ophthalmology has adopted and incorporated laser techniques into clinical practice more rapidly because the interior of the eye, otherwise difficult to access by a tool is easily accessible by the laser beam. For ophthalmological surgery, solid state lasers pumped by laser 64 A. Sharma, C.B. Panwar, R. Arya and A.K. Nath diodes are the preferred ones. Important reasons of using laser diode as a pump source for solid state laser are: (1) Electrical to optical efficiency is high and can be in the region of 10 to 30 %. Optical to optical conversion efficiency from pump source to useful output power is in the region of 25-60%. (2) The MTBF is 5x104 hours for CW diodes and 1010 pulses for pulsed LD. (3) The optical power per unit volume of LD as pump source is the highest. (4 ) LD as a pump source is selective and can pump only a single absortion band in contrast to a flash lamp. This paper describes the considerations for selection of the topology of the switching converter, the control methods and the filter requirements with their design for a diode driver capable of driving upto 40A at 3V. 2. LOAD REQUIREMENTS Electrically, a semiconductor laser behaves as a diode with a turn on knee voltage greater the 1 V. When the laser diode is fed with threshold value of current Ith, it just begins to lase. A further increase in the injected current (called Ipower ) results in a proportional increase in laser optical output. So laser diode is a nonlinear load electrically. For the particular application described in the paper, the diode has a threshold current, Ith, of 8A and a corresponding knee voltage of 1.4 V .It can be operated in CW mode or pulsed mode ( upto 100 Hz) with a rise time of about 5 ms . In the pulsed mode, the laser current is modulated between the two preset values Ith and Ipower. Ipower can be varied from 8A to 40 A by a current reference generated by microcontroller and applied to the SMPS driver through a serial DAC. For 4 W optical power, the value for Ipower is 40A at a diode voltage of 2.8V. Before considering switching converter design, it must be stressed that that the laser diode is a very sensitive load. Catastrophic optical damage may occur due to an unintentional current spike which may drive the diode beyond its rated electrical or optical limits leading to localized melting and termination of laser action [1]. Therefore the current supplied to a laser diode should be properly regulated and filtered to remove 100 Hz mains ripple as well as switching ripple and other possible transients. Also the antistatic precautions should be taken by keeping the LD terminals shorted when not powered. The laser driver is fitted with a thermo-electric cooler so as to maintain constant laser chip temperature so that optical power and wavelength remains constant at constant current. Optical power sensing is done for metering and display purpose by using a PIN photodiode operated in photovoltaic mode. The details of both are not considered here Specifications of Laser Diode driver Input 230 V, 50 Hz ac mains Output current source, 40 A max, 3V 10-90 % rise and fall time 5 ms Converter switching frequency 250 kHz Laser current ripple 80 mA pk-pk @ 40A Line and load regulation < 0.1 % Laser diode ON resistance < 10 mOhms Laser diode ON voltage < 3 V @ 40A Design of a high frequency SMPS powering laser diode 65 3. CIRCUIT DESCRIPTION (a) Power circuit configuration : Low power current regulated power supplies based on SMPS can utilize many schemes; two simple schemes are compared. In one, high DC link voltage of ~ 300 V is obtained from rectification of AC mains voltage and first SMPS stage converts it to a low isolated voltage level of say 12 V or 5 V. The second stage employs MOSFETs or Power BJTs in linear region to generate a current source, pulsed or DC. This is a lossy method but becomes indispensable when requirement is of extremely low ripple in the current. The other scheme uses only one stage of SMPS with an increased order of output filter stage such that switching ripple is sufficiently attenuated. This has higher reliability owing to low active component count and has high efficiency also. The LD driver described in this paper utilizes the second scheme. Its SMPS is based on double switch forward converter’ with a output filter designed to attenuate switching ripple as per requirement. Current control mode is used as it greatly attenuates line ripple reaching the output and thus lessens the requirement of bulk capacitor that supports input DC link voltage. Half bridge isolated converter was not considered as it does not provide any benefit over forward converter. On the flip side, half bridge is a pole structure and always carries a chance of shoot through fault, making it less reliable. Also it is difficult to implement peak current control mode in half bridge circuit. The schematic of single stage SMPS driver is shown is shown in figure[l]. Fig. 1 : Schematic of the Laser diode driver As indicated in the figure, supply is taken from AC mains, common mode and differential mode noise due to SMPS is filtered to attenuate before it enters the AC mains and vice-versa. A DC link voltage of 250 V to 350 V, for low line to high line AC inputs, is obtained by rectification by diode bridge and is maintained by the bulk capacitor. This DC link voltage delivers power to the forward converter whose control and design details are given in the following sub-section. (b) Control Circuit: Referring to fig (1) which has two nested control loops, we start with an assumption that the forward converter of fig. (1) operates in continuous conduction mode (CCM) such that pk-pk ripple of inductor L1 is very less so that peak current is nearly 66 A. Sharma, C.B. Panwar, R. Arya and A.K. Nath equal to the average current and also that the transformer magnetizing current is a small fraction of current through inductor L1 reflected on the primary side, and can be neglected. Then the MOSFET switch current during on time is equal to the reflected current of inductor L1. A clock signal initiates turn on pulse for the MOSFET switch. The termination of each pulse occurs when an analog (ILl.RS1/N1.N2) of inductor current peak becomes equal to’ control level signal (Vc) coming from the compensator Gc which is a part of the slow outer loop. In this way the peak value of inductor current is controlled directly whereas the pulse-width is not. The outer loop senses LD current, and compares it with a set reference. The error is compensated through Opamp based PI compensator Gc and its output becomes the reference current value for the inner loop. The conventional control of switching converters is an output feedback single loop control where the compensated error signal is compared with a fixed amplitude ramp signal and directly controls the pulse width of PWM signal. The control has no feed-forward correction and uses average control mode which is sluggish. Peak current mode provides several performance advantages, which have all been favorably exploited in the design of this LD driver. First and foremost is its inherent feed forward path, which corrects for input voltage variations within one time period, T, without using any of the dynamic range of the outer loop compensator. Analytically, this translates into a small signal line to output transfer function, which is sufficiently small in magnitude [2]. Thus the line regulation is excellent, and outer loop compensator Gc can dedicatedly work to correct load variations exclusively. To explain further, refer to fig (1) and fig(2). Suppose the input dc link voltage is subjected to a deliberate step increment. The slope of the MOSFET switch current (Vin/L1N12) goes up instantly. When the peak value of the switch current or the referred inductor current or reaches the control level signal V c coming from the outer loop compensator Gc (same as error amplifier), the switch is turned off. Therefore its turn on time is reduced, but the peak current remains the same. This correction takes place in one time period T. Therefore despite variations in input voltage, the peak of inductor current ILl remains constant for a particular reference voltage Vref (figs 2). In this way, the feedforward mechanism of peak current programmed control transforms the inductor Ll into a current source (fig 3). Mathematically, the inductor current ILl is eliminated as a state variable of the converter and the overall order of the closed loop converter is reduced by one. Then the small signal control to output transfer function has one less pole as compared to the single feedback voltage mode control. Derivation follows in the next section. This yields another advantage. The outer loop compensator network’ can now be designed to yield a stable closed loop converter, now of one order less, with greater bandwidth than would be possible for stabilizing a higher order system. Third, in current mode control, transfer function of the power stage in CCM is almost similar to that in DCM (a first order approximation from low to mid frequencies). [3] Fourth, by limiting the maximum value of primary switch current, MOSFET failures due to excessive switch current can be prevented. This ensures that MOSFET will turn off whenever the current through it becomes too large, on a cycle-by-cycle basis, and enhances reliability of diode driver. Fifth, benefit is automatic elimination of transformer saturation problem unlike in full bridge or push-pull isolated converter where extra circuits are needed. Design of a high frequency SMPS powering laser diode 67 Sixth, with peak current mode, many LD drivers can be operated in parallel without extra current equalization as the output is a current source of value Vref / Rs2 which can be kept same for all converters. All this makes the system control more robust. For implementation of peak current mode, two current sensing circuits are required for inner as well as outer loop. In this case both are non-contact type of measurements and thus isolation is ensured. For the inner loop, a current transformer (CT) wound on a toroidal ferrite core T-18 of material N-87 of EPCOS senses the primary switch pulsed current at a frequency of 250 kHz. For the outer loop implementation, a LEM make hall sensor L T-l OOP senses the LD current and the signal is compared with a reference Vref obtained from microcontroller (PIC16F876 of Microchip Inc.) through a 12 bit serial DAC (TL V 5618 of Texas Inc). The error is fed to the compensator Gc that decides the phase margin and 3 dB bandwidth of the outer loop, and outputs a control signal Vc which becomes the current reference of the inner loop and is compared with the CT signal by a comparator inside the Unitrode UC3844 IC. The output pulses from this IC are given to the mosfets through pulse transformer as given in fig (l). Fig. 2 : Equivalent control block diagram of the LD driver 4. EQUIVALENT CONTROL BLOCK The circuit diagram of fig (1) can be represented as an equivalent control block diagram of fig (2) which also shows the two loops. With the inductor Ll transformed into a regulated current source ILl by the inner loop circuitry, small signal model of the outer loop of fig (2) is reduced to that shown in fig (3). It can be noticed that the equivalent current source is driving a filter of order one less than the actual filter shown in fig (1) and in fig (2). Fig. 3 : Inductor L1 as a current source for outer loop 68 A. Sharma, C.B. Panwar, R. Arya and A.K. Nath Here, R = sum of resistance of inductor L2 and of laser diode. ~ iL1 is a first order perturbation in inductor current source IL1 ~ iLD is a first order pertubation in laser diode current ILD. After this simplification, the two loop control block of fig (2) is reduced to a small signal single loop model of fig (4). Fig. 4 : Control Block diagram of LD driver after reduction of inner loop Therefore with Gc = 1, the overall system now behaves as a second order one, within the subset of the assumptions given earlier. Loop transfer function from fig (4) can be derived to be Gc.N1.N2.Rs2 GH(s) = -----------------------------------(2) Rs1(1+RCs + L2Cs2) This is a first order transfer function valid upto one fifth of the switching frequency and closely matches the actually measured Bode plot at nominal input voltage and nominal load conditions. In our case, NI = 24, N2 = 50, Rs2 = 2 V / 40 A = 0.05 ohm, Rsl = 33 ohm, R =10 mohm, L2 = 3 H, C = 500 F. 1.8 Substituting these values in eqn (2), we get GH(s) = -------------------------------------- (3) (1+5x10E - 6s+1.5x10E - 9s2) The corner frequency, fo = 1/(2L2 C ) = 402kHz and Q = (L2/C)/R = 0.8 Various loop transfer functions were plotted for wide range of the line and load conditions by measurements using AP200 network analyzer for the open loop case with Gc = 1. Fig (5) gives Bode plot for Vin = 230 V and 30A at an output voltage of 2 V load. These measurements agree with the loop transfer function given in equation (3). Utilizing these open loop Bode plots, PI compensator was designed such that sufficient phase margin of atleast 40° for low load to high load was ensured. Before concluding this section, two aspects may be mentioned. (I) Firstly, it has been assumed thus far that the magnetizing current is a negligible fraction of the transformer load current. Also, in a low ripple case, peak current of inductor LI is assumed equal to its average current before treating it as a current source. However these assumptions are inaccurate at Design of a high frequency SMPS powering laser diode 69 Fig. 5 : Open loop transfer function for 30 A LD current at 2 V with 230 V AC input low load and the DC amplitude of the loop transfer function differs in practice at low load when measured experimentally. The poles, zeroes and the corner frequency however remain same. (2) Secondly, in peak current mode control, slope compensation is essential as per theory for stable operation of converters of any topology beyond 50% duty ratios [3]. In practice it is required for duty ratios above 35-40%. However, in a transformer based converter like this one, it is achieved by the magnetizing current which adds a slope of Vin / Lm to the reflected inductor current slope of Vin/N12L1 and more than compensates for a stable operation. Hence extra slope addition was not done in this case. 5. POWER CIRCUIT DESIGN This section briefly gives a design-oriented overview of some power stage elements: (1) Primary side elements: The 230V, 1 input mains is filtered through an RFI filter, and rectified and filtered by a bulk electrolytic capacitor of 450V, 2 x 100F. The two switch forward converter utilizes Mosfets IRF840 as switches and diodes BYV26C. Mosfets are hard switched with the switching loss and dv/dt limited by an RCD snubber. (2) Power Transformer: The output of transformer in fig (1) is a unipolar pulse whose average voltage can be calculated from output voltage (Vo) requirement, voltage drop across inductors L1 & L2 which can eah be assumed 2% of Vo, and voltage drop across schottky rectifier. Therefore, the average secondary voltage of transformer = V 0 + drop across L2 + drop across L1+ drop across schottky rectifier = 3 + 0.6 + 0.6 + 0.5 = 3.62 V With maximum duty cycle being 40% and voltage drop in leakage inductance assumed as 10% , the secondary output during Ton is 3.62/(0.9 x 0.4)=10 V 70 A. Sharma, C.B. Panwar, R. Arya and A.K. Nath At low AC line of 190V, DC link voltage will be about 240V = 240V So, the required turns ratio, Nl:N2 = 240/10 = 24:1 The transformer was fabricated on ETD 39 core of N-87 material from EPCOS India, with 1 secondary turn made from 100um copper foil sandwitched between 2 layers of primary of 12 turns each made from SWG 28 copper wire. Sandwitching reduces proximity loss as well as minimizes leakage inductance which would otherwise lead to significant duty cycle loss at high frequency. Maximum operating flux is chosen to be 0.13T and core loss is 6W at 250 kHz under these conditions. The leakage inductance on secondary winding is 20nH. (3) Output filter elements: To filter the unipolar pulse, a Chebyshev 0.5 dB filter was selected with a cutoff frequency of 4.8 kHz and load resistance of 10 m. At a nominal frequency of n = 1 rad/s, its elements are Rn = 1, Lln = 1.572 H. L2n= 0.932 H, and Cn = 1.518 H [4]. After scaling, these values become Ll= 5uH, L2= 3uH, and C=500uF. Inductances L1 and L2 were wound using SWG 14 enameled copper wire on EE20 core with suitable airgap. Five low ESR capacitors of 100uF, 63V connected in parallel provided the value of 500uF. Converter operates in continuous conduction mode during entire LD current range for these values of filter elements which was also verified using Pspice simulation .The filter attenuates the input ripple of 250 kHz by nearly 70 dB. At LD current of 40 A DC, the ripple current was 80mA pk-pk at 250 kHz. The ripple in LI was 2A under same conditions. 4. SUMMARY The entire circuit comprises of power stage elements and the control electronics is housed in one PCB of 12 cm x 15 cm size. Components have been selected with an emphasis on low cost and easy availability in India. The lower efficiency of 70% is characteristic of low voltage high current power supplies. The output is kept fully isolated from the mains earth as well as from the control common, as LD is a very sensitive load. The faults detected are (a) LD over-current (b) LD overvoltage and (c) any mismatch between set reference and the actual LD current. After occurrence of any of these faults, the driving gate pulses to mosfets are inhibited as well as the microprocessor turns off the SSR delivering AC input power to the diode driver. Such drivers are being used satisfactorily for about one year at our laboratory. 7. ACKNOWLEDGMENTS Expertise of Shri Vijay Bhawsar, SSLD, was used in fabrication of diode driver PCB using ORCAD software, which is acknowledged. REFERENCE 1. DW Hughes and JRM Barr “Laser Diode pumped Solid State Lasers”, J. Phys. D. Appl. Phys 25, 1992. 2. D.M. Sable, R.B. Ridley and B.H. Cho “Comparison of performance of Single loop and current injection control for PWM Converters which opertate in both continuous mode and discontinuous mode” IEEE PESC, June 1990. 3. R.W. Erickson and D. Maksimovic “Fundamentals of Power Electronics”, Edn. 1997. 4. A.I. Zverev, “Handbook of Filter Synthesis”, Wiley, 1967.