Application Note AN 2011-02 Application Note AN 2011-02 V1.0 February 2011 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Buck Converter: Negative Spike at Phase Node IFNA PMM SMD AMR DCDC 1 Ejury Jens 1 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Edition 2011-02-02 Published by Infineon Technologies Austria AG 9500 Villach, Austria © Infineon Technologies Austria AG 2011. All Rights Reserved. Attention please! THE INFORMATION GIVEN IN THIS APPLICATION NOTE IS GIVEN AS A HINT FOR THE IMPLEMENTATION OF THE INFINEON TECHNOLOGIES COMPONENT ONLY AND SHALL NOT BE REGARDED AS ANY DESCRIPTION OR WARRANTY OF A CERTAIN FUNCTIONALITY, CONDITION OR QUALITY OF THE INFINEON TECHNOLOGIES COMPONENT. THE RECIPIENT OF THIS APPLICATION NOTE MUST VERIFY ANY FUNCTION DESCRIBED HEREIN IN THE REAL APPLICATION. INFINEON TECHNOLOGIES HEREBY DISCLAIMS ANY AND ALL WARRANTIES AND LIABILITIES OF ANY KIND (INCLUDING WITHOUT LIMITATION WARRANTIES OF NON-INFRINGEMENT OF INTELLECTUAL PROPERTY RIGHTS OF ANY THIRD PARTY) WITH RESPECT TO ANY AND ALL INFORMATION GIVEN IN THIS APPLICATION NOTE. Information For further information on technology, delivery terms and conditions and prices please contact your nearest Infineon Technologies Office (www.infineon.com). Warnings Due to technical requirements components may contain dangerous substances. For information on the types in question please contact your nearest Infineon Technologies Office. Infineon Technologies Components may only be used in life-support devices or systems with the express written approval of Infineon Technologies, if a failure of such components can reasonably be expected to cause the failure of that life-support device or system, or to affect the safety or effectiveness of that device or system. Life support devices or systems are intended to be implanted in the human body, or to support and/or maintain and sustain and/or protect human life. If they fail, it is reasonable to assume that the health of the user or other persons may be endangered. AN 2012-12 Revision History: (12-01-12) , V1.0 Previous Version: V1.0 Subjects: Buck Converter: Negative Spike at Phase Node Authors: Ejury Jens We Listen to Your Comments Any information within this document that you feel is wrong, unclear or missing at all? Your feedback will help us to continuously improve the quality of this document. Please send your proposal (including a reference to this document) to: [email protected] 2 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Table of contents 1 Introduction ................................................................................................................................................. 4 2 Circuit Analysis .......................................................................................................................................... 4 3 Switching Loss ............................................................................................................................................ 7 4 Detailed Spike Analyses ............................................................................................................................. 9 5 Driver .......................................................................................................................................................... 14 6 Summary .................................................................................................................................................... 19 3 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node 1 Introduction This application note is discussing the observable negative spike in modern hard-switching step-down converter applications. Ways to deal with the phenomenon and the root cause will be revealed. 2 Circuit Analysis The negative spike occurs during control-MOSFET (HS-FET) turn-off. The voltage at the phase node falls rapidly, crosses zero and reaches a rather significant negative voltage (perhaps of the same or even higher magnitude as the input voltage of the circuit) for a few nanoseconds before quickly declining. To understand this behavior the circuit of a step-down converter will be analyzed. Figure 1 Standard buck-converter with synchronous rectification X1 represents the HS-MOSFET with its output capacitance. The synchronous rectifier is represented by X2 – also with its inherent output capacitance. During the freewheeling period of the current in X2, the voltage at the phase node is clamped to GND via the body diode of the synchronous rectifier and should never fall below about -1 V. However, reality shows otherwise. Therefore it can be concluded that the circuit shown in fig. 1 is not representing the elements causing the negative voltage spike to occur. After introduction of parasitic inductances into the critical commutation loop the analysis of the turn-off event reveals the reason for the negative spike (fig. 2). 4 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node An assumption for further considerations is that the voltage drop across X2 is zero as the forward voltage of the body diode (~0.7 V) is small compared to the observable spike Figure 2 Highlighting the critical loop inductance, i.e. the current commutation path Lstray is the total value of the critical loop inductance consisting of L1 and L2. L2 is the part of Lstray that is located in the synchronous rectification path, i.e. between GND and phase node. L1 is the part of the L stray that is not part of L2. Another assumption is that the HS-MOSFET is switching so fast that the current transition is only limited by the stray inductance Lstray. This will provide the worst case for the negative spike. When the HS-MOSFET turns off the current through it will decrease. The resulting di/dt then induces a voltage on the parasitic inductance L1. Since the output current in the inductor L can be considered constant for the switching event the current ramp-up in stray inductance L2 has to occur at the same di/dt as in L1. The induced voltage across L1 and L2 can be that high that in a worst case scenario the HS-MOSFET reaches avalanche breakdown and the voltage is clamped by the MOSFET. In this worst case scenario the MOSFET is represented by a Zener-diode as shown on the right hand side in fig. 3. 5 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Figure 3 Analysis schematic for negative spike event The phase node voltage can be calculated by applying the Kirchhoff law for voltages with respect to the phase node: 6 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node It can be seen that undershoot of the phase node voltage depends only on the breakdown voltage of the HSMOSFET, the input voltage and the distribution of the partial loop inductance in the synchronous rectifier path to the total stray inductance. In case the MOSFET does not reach avalanche the term VBRDSS(X1) in the above equation can be replaced by the maximum voltage across the MOSFET. This maximum voltage across the MOSFET will always be higher than Vin during the turn off event as long as a parasitic inductance L stray is present. Therefore, even in the most common case when the MOSFET is not reaching avalanche (as switching speed is slow and energy stored in the stray inductance can be dissipated in the switching MOSFET while overshoot voltage will be limited by the output capacitance of X2) undershoot at the phase node depends on: 3 - ratio between L2 and Lstray - input voltage Switching Loss From the equations above one can see that the only design variables to reduce the negative spike lay within the board layout and package choice for the MOSFETs and capacitors as the choice of the package will influence Lstray and its distribution. To reduce the undershoot voltage L2/Lstray should be small. One way to accomplish that is to enlarge L1. This however would be a bad choice as it increases Lstray and will lead to more dynamic power loss. The impact of Lstray on dissipated energy can be calculated as: 1 EL 2 Lstray I Peak 2 This is valid for inductively limited switching, i.e. the current rising and falling slopes are only determined by Lstray and not by the MOSFET switching speed. Assuming a peak current of 35 A and a stray inductance of 1 nH the dissipated energy per switching period is: 7 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node ESwitch 1 2 1nH 35 A 612.5nWs 2 2 For a buck converter switching at 300 kHz each nanohenry stray inductance corresponds to an increase in switching loss of: PSwitch f SW ESW 300kHz 612.5nWs 183.75mW Therefore an increase of L1 by 1 nH would decrease the undershoot at phase node but at the same time increase the switching losses by about 180 mW. The only way to achieve the goal of reduced undershoot without sacrificing efficiency is by shifting the existing loop inductance so that L2 is minimized. In practice this is rather straight forward. Primarily a low inductive advanced package such as SuperSO-8, S3O8 or CanPAK should be chosen. The lowest value for L2 is achieved when: - the GND return path from the load is entering the source of the LS-MOSFET before reaching the closest input decoupling MLCC and - the inductor is connected as close as possible to the drain of the LS-MOSFET. These measures insure lowest undershoot without sacrificing efficiency. Connecting the driver should follow the thinking and take the reference potential also directly from the source of the LS-MOSFET. Unfortunately many times these considerations are not part of the design process and the undershoot issue will be found once the board is being tested. A typical countermeasure is the addition of an external gate resistor that provides a means to control the speed of the MOSFET. Another way to reduce undershoot is to use a slower older generation FET technology. Indeed, slowing down the switching speed of the MOSFET will effectively reduce the voltage spikes during transient events. However, reducing the switching speed of MOSFETs will increase switching losses and overall system efficiency will drop as can be seen in figure 4. 8 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Figure 4 Example of how switching speed impacts the efficiency, RDSON of all parts is about identical and its impact is minor as it can be observed by comparison of BSZ050N03MS (5.0 mΩ @ VGS=10 V) and BSZ058N03MS (5.8 mΩ @ VGS=10 V). BSZ060NE2LS has 6.0 mΩ @ V 4 Detailed Spike Analysis In fig. 5 the phase node voltage is shown for BSZ060NE2LS. The undershoot voltage reaches -10V as seen in the oscilloscope screenshot. For this particular measurement the following parameters are given: LStray = 1 nH IPeak = 35 A VIN = 12 V BSC060NE2LS QOSS(12 V) = 5.8 nC QOSS(25 V) = 8.9 nQ Question: W hen switching extremely fast, what limits the magnitude of the 9 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node spike? In order to find the answer one has to investigate the energy of the spike and its effects. Assuming the worst case, i.e. instantaneous switching, all energy stored in L stray is causing the voltage to rise across the MOSFET. Without its output capacitance any MOSFET would reach its breakdown voltage and clamp the voltage until the energy has been dissipated. In reality the output capacitance absorbs energy and a non- infinite value for the spike voltage can be obtained. Figure 5 Measured spike 10 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node For the example discussed before the energy dissipated in a switching period (equals the energy EL stored in Lstray) was 612.5 nWs. The output capacitance of the MOSFET holds the energy related to the input voltage: ECIN 1 QOSS VIN VIN 0.5 5.8nC 12V 34.8nWs 2 At breakdown voltage the output capacitance contains the energy of: ECVBRDSS 1 2 QOSS VNBRDSS VIVBRDSS 0.5 9nC 25V 112.5nWs That leaves a fixed amount of energy that can be absorbed by the output capacitance: EAbsorb ECVBRDSS ECIN 112.5nWs 34.8nWs 77.7nWs Here is the dilemma: 612.5 nWs have to be absorbed by the MOSFET’s output capacitance but it is only capable of absorbing 77.7 nWs. The excess energy of 612.5nWs 77.7nWs 534.8nWs will be dissipated in avalanche. However, the waveform in fig. 5 does not show any indication of avalanche. That means the energy stored in Lstray will be dissipated in another way. In the following it will be shown that the additional power dissipation emanates from increased turn – off loss due to source feedback on the gate drive voltage of the HSFET. So far we assumed instantaneous inductive switching during turn-off. In this assumption it is not considered that the source feedback voltage caused by the source inductance of the HS-MOSFET during the turn-off process will effectively slow down the switching speed of the device. To calculate the impact of this feedback mechanism the current slew rate is derived from: di VBRDSS VIN 25V 12V 13A / ns dt 1nH Lstray Therefore, the turn-off of 35 A peak current without source feedback would take tOFF I Peak 35 A 2.7n di 13 A ns dt 11 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node However there is the need to account for the voltage feedback of the source inductance of the HS-MOSFET. As an example 15 % of Lstray will be assumed as HS-MOSFET source inductance LHS-source which leads to an introduced gate to source feedback voltage of V fb LHS so u rce 13A / ns 0.15 1nH 13A / ns 1.95V With 1.95 V feedback the turn-off becomes delayed and energy is dissipated in the longer transition. Figure 6 Transfer characteristic of the BSZ060NE2LS (Tj as parameter) 12 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Transitioning from VGS = 3 V (35 A) to VGS = 2.2 V (~0) in case of the BSZ060NE2LS (see fig. 6) yields the following transition times: RGtot = 1.5 Ω No feedback: 1.95 V feedback: RGtot Ciss 1.5 670 pF 1ns Ciss = 670 pF t ln v t ln v v V0 et / 2.2V V 1ns ln 3V 0.31ns 0 2.2V 1.95V V 1ns ln 3V 1.95V 1.43ns 0 The commutation starts when the gate-source voltage has reached 3 V. The maximum energy that will be dissipated in the turn-off process is: E tra n s V DS I D (t)dt 0.5 V DS I Pea k t tra n s = 0.5 ∙ 25V ∙ 35A ∙ 1.43ns = 625.625nWs 13 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node That is more than the amount that was not accounted for (534.8 nWs) in the capacitive storage. Hence the MOSFET will not reach its breakdown voltage. For the case shown here it means: - The device is working in the optimal range. No avalanche is observed (note that a 25 V rated MOSFET has in fact a higher breakdown voltage which also increases with operating temperature further). - The small loop inductance (1 nH) enables fast transitions (high di/dt) and leads to small energy dissipation. - The main portion of the inductive energy is dissipated in the channel of the MOSFET as switching loss. Only a fraction is causing the charge of the output capacitance. - The small output capacitance causes voltage spikes that are significantly observable but have low energy associated. - Reliability is not impacted by the spike occurrence because the higher efficiency leads to lower operating temperatures. This will more than compensate for the brief voltage events. The calculations have been done under the assumption of a SuperSO8 package that has some source feedback. Given the use of a CanPAK with proper routing and driver decoupling the source inductance can be much smaller almost eliminating source feedback voltage which results in higher efficiency. Specifically with CanPAK a symmetric decoupling can be done reducing the stray inductance well below 1 nH which attributes to very benign voltage spikes overall. 5 Driver The negative voltage spike at the phase node is connected to the driver IC. Usually driver ICs have sufficiently large ratings for the phase node voltage allowing for negative spikes. However, especially when having 5 V driver ICs, the rating is usually smaller and reason for concerns. Fig. 7 shows part of the driver output stage connected to the power circuit. 14 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Figure 7 Clamped voltage protection for phase node by (red) clamp structure, voltage drops across R Clamp The internal phase node voltage is clamped by a dedicated clamp structure consisting of a resistor RClamp and a Zener-Diode. The capability of how much energy can be absorbed by this clamping structure depends on the permissible power dissipation of RClamp and the Zener structure. The driver datasheet usually specifies a permissible negative (and positive) voltage that can be applied for a limited time at the phase pin and exceeds the max DC rating for that pin. This indicates the presence of an internal clamp structure. In practice the spike magnitude is often bigger but the time of exposure much shorter than the stated datasheet conditions. In such a case the dissipated energy per cycle has to be considered and compared with the maximum allowable per cycle energy dissipation of the internal clamp structure. As an example here is an excerpt of a driver datasheet specifying the following: - VVCC = -0.3…+6.5 V - VPVCC = -0.3…+6.5 V - VPhase = -1.0…+30 V - VPhase-pulsed fsw = -10….+35 V for tpulse < 30 ns = 0.1…2 MHz 15 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Based on the DC vs. pulse rating for the positive spike one can conclude that the clamp structure can sustain additional +5 V for 30 ns on a cycle by cycle base for up to 2 MHz switching frequency across its clamp resistor compared to the DC rating. The energy across the clamp resistor can be calculated as: That implies that a spike of double magnitude but a quarter of the pulse length compared to the maximum datasheet rating will result in the same dissipated energy across the clamp resistor. ECla mp VS p ike The total power dissipation of the clamping structure is: VS p ike VCla mp RCla mp t Pu lse For shortened pulse duration the spike voltage of identical energy can be obtained as: VS p ike On the concrete example the maximum clamping energy is: VCla mp 2 2 VCla ECla mp RCla mp mp 4 t Pu lse ECla mp 35V 35V 30V 5250nV 2 s 30ns RCla mp RCla mp Assuming a spike duration of 5 ns instead will permit a maximum spike voltage of: 30V 2 30V Vo versh o o t VS p ike VDCra tin g 2 15V 35.7V 30V 20.7V 4 5250nV 2 s 30V 5ns Using the same energy in the claming resistor the spike voltage for the negative spike can be calculated as: 2 S p ike 2VS p ike VCla mp V VS p ike VCla mp 16 2 Cla mp ECla mpmax RCla mp V ECla mpmax RCla mp tCla mp tCla mp 0 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Usually these negative spikes are in the 1 ns ... 4 ns range. Therefore it can be concluded that the driver operates very reliably although the driver specification does not account for that condition. Still it should be checked with the driver manufacturer whether there are other reasons that might prevent the driver from operating under high negative spike conditions. Another important aspect for the driver is the way it is connected to the power MOSFETs. 17 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node Figure 8 Ground and phase node connections impacting negative spike observable by the driver The ground connection between driver and power stage should be done in the way depicted in fig. 8, i.e. the driver ground should be connected to the synchronous rectifier MOSFET directly without having a current path in common with the power circuit. This reduces the L2 contribution to the package of X2. For the phase connection of the driver the following has to be considered: Making connection “b” instead of “a” (see fig. 8) reduces the L2 contribution to the package of X2 on the phase node side and therefore will reduce the negative undershoot as explained in section 1. However, all parasitic phase node inductance will be source feedback for the control FET X1 and hence slow down the switch. That in return will increase switching losses 18 Application Note AN 2011-02 V1.0 February 2011 Buck Converter: Negative Spike at Phase Node and efficiency will suffer significantly. The only solution to reduce the negative spike level and keep efficiency high is to choose connection “a” while keeping the switching node interconnect between both MOSFETs as small as possible. 6 Summary As shown the phenomenon of strong negative phase node spikes is part of fast switching modern voltage regulators. It is mainly caused by the fast commutation of the switched current in the critical loop combined with a specific distribution of parasitic inductances. For integrated devices such as DrMOS, the current commutation speed (di/dt) can be significantly higher than for discrete MOSFETs. At the same time the MOSFETs in these integrated devices are almost ideally coupled to the driver with very little source feedback. That makes the negative spike issue for DrMOS (and other integrated devices) virtually nonexistent for the driver. For discrete designs it is important to follow these basic design rules: - Keep the commutation loop as small as possible to allow for lowest inductive energy in the circuit - Minimize the partial loop inductance in the synchronous rectifier path to achieve the lowest negative spike - Couple the driver ground as closely to the synchronous rectifier ground as possible, connect it in Kelvin style, i.e. avoid common current between the driver ground connection and the power ground current flow - Do not compromise switching speed for the sake of lower spike unless there are no other options. 19