Reprint of Poster Presentation EML, May Welleman, E. Ramezani, J. Walmeyer, S. Gekenidis

Transcription

1 Reprint of Poster Presentation EML, May 1998 Welleman, E. Ramezani, J. Walmeyer, S. Gekenidis ABB Semiconductors AG Fabrikstrasse 3, CH-5600 Lenzburg / Switzerland Tel.: Fax: adriaan.welleman@chsem.mail.abb.com Abstract An overview is given about semiconductor components in a range up to V drm =12 kv and I p = 150kA, stackable up to more than 30 kv, for non-repetitive and repetitive applications in pulsed power technologies. This includes also the triggering units, the stack construction, as well as the demonstration of voltage and current waveforms at typical conditions for electric launchers and for laser technology. An estimation about reliability is presented, based on internal tests and field experience. This presentation will give the user of solid state components an overview over existing and newly developed devices available today, that are especially suited for pulsed power applications. Introduction Solid state switches are being actively used to replace spark gaps as well as thyratrons and ignitrons in various pulsed power applications. Each of the switch types has optimized features for specific applications. For single pulse and medium repetition rate applications in the range of a few hundred Hertz the semiconductor switches are getting more popular with respect to vibrations, orientation, maintenance, life expectancy and environmental issues. Over the last 4 5 years, ABB Semiconductors AG has been developing and producing a range of specific semiconductor switches for the pulsed power market. (1 of 9) :43:02

2 The range of these devices varies from designs with 6 wafers in one common housing up to reverse conducting or reverse blocking switches with integrated, very low induction and light-activated gate units. Since the power device itself can be combined with an integrated trigger unit, we have to take into account that for devices with integrated gate units the failure rate of the whole switch is not only given by the power semiconductor device, but also by the parts in the electronic control unit. Components ABB Semiconductors AG has a specific range of components designed for use in pulsed aplications and are being produced based on long term experience with production of high power semiconductors. This means that these devices are produced with the same technological processes and materials as those used for transportation or HVDC applications. The most important progress in this field is the development of devices which are able to handle repetitive switching. The table in Fig. 1 gives an overview of the available products (status March 1998): Device Type ABB p/n V rrm V drm I Pulse (50µ s) Principle Gate-Unit Discharge Switch Discharge Switch 5STH 20H V 18 V 80 ka Asymmetric Without 5STH 30J V 18 V 100 ka Asymmetric Without Discharge Switch 5STF 07Z kV 11.8 kv 60 ka Symmetric Multi-chip Separate Crowbar Diode 5SDA 27Z kv 12.0 kv 60 ka Multi-chip-diode Crowbar Diode 5SDA 27Z kV 12.0 kv 60 ka Multi-chip-diode with parallel resistors High di/dt thyristor 5STP 0365D V 6500 V 1.5 ka Thyristor Without Opening Switch 5SHY 35L V 18 V 4 ka* Asymmetric Repetitive Integrated Discharge Switch Discharge Switch 5SPY36L V 18 V 150 ka Asymmetric Non Rep. Integrated 5SPY36L V 18 V 140 ka* Asymmetric Repetitive Integrated (2 of 9) :43:02

3 Discharge Switch 5SPY36L V 4500 V 150 ka Reverse Blocking Non Rep. Integrated Discharge Switch 5SPY36L V 4500 V 140 ka* Reverse Blocking Repetitive Integrated Fig.1: Overview of Pulsed Power Components (*) Maximum values depend on repetition frequency Fig. 2: Thyristor based multi-chip design for single shot, high current, high voltage applications. Fig. 3: Thyristor based single-chip design for repetitive medium current applications. Fig. 4: Thyristor design with highly inter-digitated gate structure for high di/dt capability. Fig. 5: So-called "A-Z" switch, which is an inter-digitated thyristor design with concentric gate disc around the ceramic housing and integrated optically (3 of 9) :43:02

4 activated trigger unit. Fig. 6: Crowbar diode, multi-chip design with sharing resistors. Assemblies All the described devices have disc-type housings, which makes it easy to stack them in order to realize high blocking voltages. In device stacks there is also the possibility to use air or water cooled heat sinks between the components which might be needed in case of high pulse repetition rates or high current amplitudes. ABB Semiconductors AG can supply complete switches for more than 40 kv, mounted in stacks with semiconductor devices, heat sinks and trigger units, optimized for most of today s requirements. Some examples of stacks for pulsed power are shown in Figs. 7 & 8. Especially for pulsed power applications, a reliable clamping system and homogeneous clamping with very high parallelism is most important to reach a high life-time. Reliability problems can easily occur if the importance of proper clamping is not understood. Fig. 7: Crowbar diode assembly with V rrm = 36 kv, I p = 60 ka, t p = 1 ms Fig. 8: Discharges switch assembly, Reverse Blocking, V rrm =20 kv, I p =100 (4 of 9) :43:02

5 ka, t p =50 µ s Trigger Units One of the most important parts of the switch system is the electronic unit which triggers the semiconductor at the right time, with the required energy, the required di/dt and with the typical waveform needed for the specific application. As most of today s semiconductor switches for pulsed power are supplied without trigger unit, the designer, or end user, is left with the problem how to trigger the device under conditions which are optimized for the application as well as for the devices themselves. ABB Semiconductors AG has decided to supply trigger units for most of the devices for pulsed power. These trigger units are optimized to safely trigger the component under general conditions, and they can also be optimized for special application conditions for example for series connection which allows only extremely short trigger time differences among the devices in series. Whereas the devices based on thyristor technology still have gate cables the separate trigger units, the new generation of switches have integrated trigger units, which results in almost no inductance (only a few nh) between trigger unit and device. Therefore, by utilizing a large-area parallel current path to gate and cathode, both the stray inductance and the series resistance are reduced to an absolute minimum, so that current rise rates to the gate of up to 5 ka/µ s can be achieved with a power supply voltage to the trigger circuit of about V. In case of non-repetitive applications, the supply voltage input terminal can be directly connected to the anode side of the device, and no separate voltage source is needed. Fig. 9: Trigger unit RAM with light transmitter (for multi-chip thyristor 5STF 7Z1201) Fig. 10: Monolitic integrated trigger unit with "A-Z" Switch as discharge switch (5 of 9) :43:02

6 (5SPY 36L4502) Reliability In contrast to quality (in its narrow definition), which includes fulfillment of the specification in all its aspects, reliability involves its fulfillment over the whole life of a product. Obviously, the reliability of the product is therefore only defined in combination with a defined application condition.a number of physical mechanisms are known to affect reliability in semiconductor devices. We can distinguish between effects of infant mortality, random failures and wear-out. Infant mortality failures are commonly known by the semiconductor manufacturer, and he prevents them by specific burn-in procedures where necessary. Random failures are characterized by a constant failure rate over most of the lifetime of a product. An example is destruction by cosmic radiation when a device is exposed to high blocking voltages over extended periods of time. ABB Semiconductors AG has established design rules which allow to specify the expected failure rate in fit.("failures in time", i.e. the number of failures per 10 9 device-hours) for a given design and operation condition. A typical value for a high-power semiconductor device is 100 fit. In pulsed power applications, this failure mode is not very relevant, however. Wear-out is a very important aspect for pulsed power, on the other hand. The dominant mechanism here is aging of the dry interfaces between the components inside a semi-conductor housing. This leads to an increase in electrical interface resistance and ultimately to device failure. In interdigitated structures, the cathode segment metallization can be deformed and pressed out into the gate region, which then causes a gate-to-cathode short circuit. The safe limit is determined by a number of times temperature can step up and down a certain amount. The smaller the temperature step, the more cycles can be allowed for. In case of a given application condition, ABB Semiconductors AG is capable to assess the estimated lifetime of a component. Fig.11 shows a scanning electron micrograph of two segments of an interdigitated thyristor after 30 million current pulses of 12 ka amplitude and 2 µ s duration. No wearout can be seen. For this case, wear-out has been expected to appear after 100 million pulses only. Another example is given in fig. 12. Here, a cathode segment is seen from the top, as it appeared after only a few thousand pulses of the same kind, but with unacceptably low and inhomogeneous mounting pressure. The width of the metallization has become strongly inhomogeneous (147 and 166 µ m), and a short circuit between cathode and gate is very likely now. The comparison of the two examples shows clearly that it is extremely important to pay high attention to the way semiconductor components are mounted. For best reliability it can therefore be advantageous to purchase full device stacks from the semiconductor manufacturer. (6 of 9) :43:02

7 Fig. 11. Scanning electron micro-graph of two segments of an inter-digitated thyristor after 30 million current pulses of 12 ka amplitude and 2 µ s duration. Regular and homogenous mounting pressure. Fig. 12. Scanning electron micro-graph of one segment of an inter-digitated thyristor after a few thousand current pulses of 12 ka amplitude and 2 µ s duration. Low and inhomogeneous mounting pressure. Conclusions It has been shown that ABB Semiconductors AG produces a range of special products suited for pulsed power applications. These products include standard and special silicon designs with one or several chips in one housing. In this way, extremely high current and voltage ratings can be achieved, which are otherwise not accessible to a device in a single housing. The designs are only slight variations of the standard product range and therefore benefit from our large experience in manufacturing processes, device quality and reliability. Circuit requirements and reliability considerations make it favorable to realize a very close interaction between power semiconductor device, gate trigger unit and mounting stack. ABB Semiconductors AG therefore additionally supplies components with an integrated, optimized trigger unit and is able to mount the devices in a stack with well defined and homogeneous mounting pressure for utmost reliability. We are able to supply reliability predictions under customer application conditions, and we have experimental evidence of the validity of our prediction rules. ABB Semiconductors, Inc. 575 Epsilon Drive (7 of 9) :43:02

8 Pittsburgh, PA USA Phone: (412) Fax: (412) Solid State Switch Issues For High Current, Short Pulse Applications Date: June 29, 1998 Total # Pages: 7 (including this one) To review solid state switches for pulsed discharge applications in terms performance, reliability, and cost, we should consider the following issues: 1. Bipolar vs. MOS-Bipolar 2. Circuit Issues: tq, Voltage, stack size, etc. 3. Dynamic Performance: peak current; di/dt limitations, dynamic loss and heating to estimate TJ,?TJ per shot and device life; 4. Operating Reliability 5. System Costs for the configuration required to meet life requirements. Bipolar vs. MOS-Bipolar Bipolar: Low On-State Loss (Latching Device) Fast Turn On Can Be Turned Off Low Parts Count: Low Failure Rates (8 of 9) :43:02

9 Packaging for Double Side Cooling, High Thermal Cycling Available to Large Are, 6kV Gate Units and Power Supplies More Expensive Cannot Limit and Commutate Short Circuits MOS-Bipolar: 4x On-State Loss 2x Switching Loss Limits and Commutates Short Circuit Current Isolated Mounting Available Random Failure Rates ~ 10x Bipolar Device Life ~ ¼ - ½ of Bipolar Possibility of Explosion Solid State Switching for Pulsed Discharge Applications June (9 of 9) :43:02

10 A SOLID STATE CROWBAR FOR RF TUBE PROTECTION J.F.Orrett, D.M Dykes, J.E.Theed, A.J.Moss, P.A.Corlett, J.H.P.Rogers, S. Buckley, S.A.Griffiths, R.J.Smith, E. Wooldridge Abstract Daresbury Laboratory is the home of the worlds first dedicated Synchrotron Radiation Source (SRS); which has been in constant operation since In 2002 the original High Voltage Power Supply (HVPS) was replaced by a modern low stored energy Pulse Step Modulator (PSM) based power supply. However, the original power supply is still a capable unit and is being recommissioned as the HVPS for Daresbury Laboratory s Energy Recovery Linac Prototype (ERLP). During the recommissioning the original ignitron based crowbar will be replaced by a solid-state unit. This paper will discuss the criteria on which the choice of crowbar switch was based. In addition to our evaluation methods, acceptance test data will be presented. Other modifications to the HVPS will be a new PLC control system and crowbar trigger circuits. These improvements will produce a versatile HVPS which will be used to provide HV DC for klystrons and inductive output tubes (IOTs). INTRODUCTION The original SRS HVPS is to be refurbished and recommissioned to provide HV DC for the ERLP and RF Test Area. The known problems with the HVPS were; Unreliable Crowbar, prone to spurious trips. Control System (24 V relay) was prone to failure and difficult to fault find Environmental issues, mercury filled ignitrons in the crowbar, asbestos resistor mats and PCB filled capacitors All the above issues have been addressed; however this paper will mainly deal with the Crowbar Switch replacement and Crowbar Trigger System CCLRC Daresbury Laboratory UK or IOT the capacitor bank will attempt to discharge through the lowered resistance of the tube, this would result in permanent damage. In order to protect the RF tube a fast operating switch is placed in parallel to it. When activated this switch closes and diverts the rapidly rising capacitor current away from the RF device and so protects it. This system of diverting energy away from the device that is being protected is crowbarring. 11 kv Transformer 8 Ohm Resistor 2 H Smoothing Choke 20 Ohm Resistor VCB RMU Rectifier Crowbar Circuit Roller Regulator Klystron or IOT 24 µf Capacitor Bank Figure 1: Block diagram of power converter The two factors which determine the amount of energy dissipated within the RF tube are; crowbar operating time and HT voltage. During an arc incident the arc current will rise at a rate dependant on the inductance and resistance of the capacitor discharge loop. This current will continue to rise until the crowbar operates, it will then decay at a rate set by the inductance and resistance of crowbar-rf tube loop. CROWBAR SWITCH REPLACEMENT. In order to procure an effective replacement, it was necessary to assess what was required of the switch in the event of an RF tube arc. The HVPS is of conventional design and is capable of delivering between 7 and 52 kv DC at up to 16 A. Due to its conventional design, the HVPS has a bank of smoothing capacitors with a total value of 24 µf. These capacitors ensure that we have smooth DC output, with a ripple of ~ 0.7%. These capacitors store a large amount of energy, if the HVPS was delivering its maximum output of 52 kv the stored energy would be E= CV 2 /2 or 32 kj. In the event of an arc in a klystron Figure 2: Arc and Crowbar current paths Rising current (pre crowbar) = V 0 1 e R' R' t L'

11 Decaying current (post crowbar) = R = 30.2 Ω L = 70 µh R = 20 Ω L = 50µH V 0 = HT t 1 = Crowbar operating time The inductive output tubes (IOT) to be used on the ERLP are operated at ~ 25 kv. The calculated arc current for 25 kv HT with the crowbar operating at 6 µs from arc initiation is shown below. Current (amps) 8.00E E E E E E E E+02 Arc Current (crowbar at 6 us) V0 (1 e R' 0.00E Time (seconds) Figure 3: Arc current rise and decay R' t1 L' ) e The sooner the crowbar operates the less energy will be dissipated in the RF tube. The relationship between crowbar operating time and arc energy is shown below. R'' ( t t1 ) L'' Energy (J) Energy in Arc at 50 kv Time of crowbar (s) Figure 5: Arc energy against crowbar operating time (50 kv) The HT has been increased by a factor of 2 and the energy dissipated has increased by 2 2 or 4x. The HVPS will be used to power two types of RF tubes a 500 MHz klystron which operates at ~ 50 kv and 1.3 GHz IOTs which operate at ~ 25 kv. To give a good level of protection to these devices the crowbar switch would need to operate within ~ 8 µs. Other factors taken into account during procurement were; price, lead time, integration in to control system and sub assembly life time. It was decided to purchase a thyristor based solid state switch. SOLID STATE CROWBAR The ABB crowbar switch is based on an optimised version of the 5STP 0365D002 high voltage, high di/dt thyristor. Energy in Arc Energy (J) Time of crowbar (s) Figure 4: Arc energy against crowbar operating time (25 kv) The other factor determining arc energy is the level of HT. It seems reasonable to expect arc current to be proportional to the HT voltage and arc energy to be proportional to the square of the voltage. Calculations support this and the graph below shows arc energy when HT is 50 kv. Figure 6: Thyristor Fourteen of these thyristors are placed in series, and each has its own parallel resistor. This arrangement ensures that each thyristor experiences only a fraction of the HT hold off voltage, and that this voltage is shared equally between them. The thyristors are triggered using a pulse from the trigger generator. This pulse travels through a loop to which all the thyristor gates are inductively coupled, ensuring simultaneous triggering.

12 Secondary Trigger Cct. Auxiliary triggers Trigger Cct. VCB Trigger Primary Trigger Cct. Crowbar 24 uf 2.2 ohms 2 H 8 ohms 20 ohms -7 to -25 KV Figure 8: Crowbar Trigger System Figure 7: Crowbar switch diagram. The crowbar switch is a self contained unit requiring only 40 V AC (via isolating transformer) and an optical input to trigger the system. It has two optical outputs crowbar ready and crowbar fired. The compact nature of the device, and its use of optical inputs and outputs simplifies commissioning and interfacing with the control system. CROWBAR TRIGGER SYSTEM In order for the crowbar switch to operate it must receive an optical input. It has been the responsibility of Daresbury Laboratory (DL) to design and build a suitable fault detection and trigger system for the crowbar switch. This system must fulfil a number of roles; Detect the initiation of an arc in an RF device. Provide an optical trigger to the crowbar switch. Provide a suitable level of redundancy; failure to crowbar during an arc incident is likely seriously damage the RF tube. Provide a means of operating the crowbar by the PLC Control System, or for auxiliary triggers, such as focus fail. There are three current transformers (CT) used in the system, two on the earth side of the capacitor bank are used to detect the rising current associated with an arc in an RF tube. The third CT, in the crowbar switch detects the operation of the crowbar. Each CT has an associated detection circuit. In order to simplify fault finding and spare holding, the three detection circuits are identical. Internal jumpers enable these to be configured for the different roles. The Primary Trigger Circuit detects the onset of an arc and produces the optical output to fire the crowbar. This circuit is configured for speed; minimising the amount of electronics between the input of the voltage pulse from the CT and the output of the optical pulse to the crowbar. This circuit also uses the most sensitive CT. The Secondary Trigger Circuit fulfils two roles, it acts as a back up to the primary circuit and gates together auxiliary triggers from the PLC or other systems. It then triggers the crowbar via the final stage of the primary circuit. The Vacuum Circuit Breaker (VCB) Trigger Circuit has a CT in the crowbar switch line. This CT will detect the current going through the crowbar switch when it closes. The trigger circuit will then produce an output to open the VCB at the input to the HVPS, effectively removing the mains and ensuring that the crowbar is not left conducting the short circuit current of the HVPS. TEST RESULTS To fulfil its required role the crowbar switch has to achieve a number of things; Handle a hold off voltage of up to 52 kv. Carry a conduction current of up to ~ 5 ka. Have a sufficiently high di/dt. Have a fast operation time. Results of MICROCAP simulations of the HVPS for an arc incident followed by crowbar operation are given below.

13 4.64K KLYST3F.CIR Temperature = 27 Case= K 2.72K 1.76K 0.80K -0.16K 99.98m 99.99m m m m m Left Right Delta Slope -i(r11) 1.361K 0.733K K e08 i(l2) 0.000K 1.982K 1.982K 6.936e08 T m m 0.003m 1.000e00 Figure 9: Crowbar and arc currents The blue curve represents the arc current and the red curve the crowbar current. The highest rate of di/dt experienced by the ideal simulated crowbar switch is indicated by the two boxes and is 1982 A in 3µs or 660 A per µs. The highest current the switch experiences is ~ 4.6 ka. To offer the maximum protection to the RF tubes the real crowbar should exceed these figures. Figure 11: Delay from arc initiation to crowbar conduction Figure 10 shows that the switch can easily achieve the required levels of di/dt and current carrying. Figure 11 shows the delay between arc initiation and crowbar conduction. This was measured using the DL trigger system and gives a good indication of how it should operate in situ. During acceptance checks at ABB the system was successfully tested for HT hold off and then given a test using a simulated arc. The simulated arc test was carried out using the DL built detection circuits to trigger the crowbar. The results are shown below. Figure 12: Crowbar Switch Assembly CONCLUSIONS Figure 10: Peak Current and di/dt Tests and simulations show that the Solid State Crowbar is a system which offers a high level of protection for RF tubes. The system will be tested more extensively as the HVPS is recommissioned. In September the system should be operational and will be used to test the 1.3 GHz IOT s for ERLP. The combination of the refurbished SRS HVPS and a modern solid state crowbar has produced a capable, versatile and flexible system which will be used for the foreseeable future, as a HVPS for the ERLP and a DL RF Test Area.

14 REFERENCES Daresbury Laboratory internal publication SRS/TDN/83/01 THE CROWBAR SYSTEM FOR THE SRS STORAGE RING KLYSTRON C.R.Dunbar and F.B.Pemberton

15 A SOLID STATE CROWBAR FOR RF TUBE PROTECTION J.F.Orrett, D.M.Dykes, J.E.Theed, A.J.Moss, P.A.Corlett, J.H.P.Rogers, S.Buckley, S.A.Griffiths, R.J.Smith, E.Wooldridge CCLRC Daresbury Laboratory UK ERLP HIGH VOLTAGE POWER SUPPLY SOLID STATE CROWBAR ACCEPTANCE TESTS The HVPS is of conventional design and is capable of delivering between 7 and 52 kv DC at up to 16 A. The HVPS has a bank of smoothing capacitors with a total value of 24 µf. These capacitors ensure that we have smooth DC output, with a ripple of ~ 0.7%. VCB 11 kv Transformer 2 H Smoothing Choke 8 Ohm Resistor 20 Ohm Resistor The ABB crowbar switch is based on an optimised version of the 5STP 0365D002 high voltage, high di/dt thyristor. To fulfil its required role the crowbar switch has to achieve a number of things; Handle a hold off voltage of up to 52 kv Carry a conduction current of up to ~ 5 ka Have a sufficiently high di/dt Have a fast operation time Results of MICROCAP simulations of the HVPS for an arc incident followed by crowbar operation are given below. 4.64K KLYST3F.CIR Temperature = 27 Case= 1 RMU Rectifier Roller Regulator Crowbar Circuit Klystron or IOT Figure 4: Thyristor 3.68K 24 µf Capacitor Bank Figure 1: Block diagram of power converter These capacitors store a large amount of energy, if the HVPS was delivering its maximum output of 52 kv the stored energy would be E= CV 2 /2 or 32 kj. In the event of an arc in a klystron or IOT the capacitor bank will attempt to discharge through the lowered resistance of the tube, this would result in permanent damage. Two factors which determine the amount of energy dissipated within the RF tube are; crowbar operating time and HT voltage. During an arc incident the arc current will rise at a rate dependant on the inductance and resistance of the capacitor discharge loop. This current will continue to rise until the crowbar operates, it will then decay at a rate set by the inductance and resistance of crowbar-rf tube loop. Arc Current (crowbar at 6 us) Fourteen of these thyristors are placed in series, and each has its own parallel resistor. This arrangement ensures that each thyristor experiences only a fraction of the HT hold off voltage, and that this voltage is shared equally between them. The thyristors are triggered using a pulse from the trigger generator. This pulse travels through a loop to which all the thyristor gates are inductively coupled, ensuring simultaneous triggering. 2.72K 1.76K 0.80K -0.16K 99.98m 99.99m m m m m Left Right Delta Slope -i(r11) 1.361K 0.733K K e08 i(l2) 0.000K 1.982K 1.982K 6.936e08 T m m 0.003m 1.000e00 Figure 6: Crowbar and Arc Currents Figure 6 shows that in order to give the required level of protection, the crowbar switch should be able to handle a current > 4.6 ka and have a di/dt better than 660A/µs. During testing this level of performance was achieved, as can be seen in figure E E E+02 Current (amps) 5.00E E E E E E Time (seconds) Figure 2: Arc current rise and decay (25kV) V 0 Rising current (pre crowbar) = L 1 e R' V Decaying current (post crowbar) = 0 (1 R' R = 30.2 Ω L = 70 µh R = 20 Ω L = 50µH V0 = HT t1 = Crowbar operating time R' t ' e R' t1 L' ) e R'' ( t L'' t1 ) Figure 7: Current Handling and di/dt Figure 8 below, shows the total delay from simulated arc to crowbar conduction. Energy in Arc 7 Energy (J) Time of crowbar (s) Figure 5: Crowbar Switch The crowbar switch is a self contained unit requiring only 40 V AC (via isolating transformer) and an optical input to trigger the system. It has two optical outputs crowbar ready and crowbar fired. The compact nature of the device, and its use of optical inputs and outputs simplifies commissioning and interfacing with the control system. Figure 3: Arc energy against crowbar operating time (25 kv) Accelerator Science and Technology Centre