Instant response - tackling high voltage problems with high speed power supply control

TWI Bulletin, September/October 1994

Colin Ribton joined the Electron Beam Department of TWI in 1985 after gaining a degree in physics at the University of Nottingham. Since that time he has worked extensively on the research and development of high power electron beam equipment and has been substantially involved in several large equipment developments at TWI.

His flair for digital electronics and software has led to many important innovations in real-time seam tracking, control technology for high power switch mode power supplies and high frequency dot matrix electromagnetic deflection systems. Currently he leads four projects in TWI's Core Research Programme involving electron beam monitoring, real-time seam tracking, high voltage discharge control and computer aided field analysis,

Electron beam high voltage power supplies have come a long way over the past 30 years. Colin Ribton describes the benefits that can be obtained by using some of the latest technology.

Electron beam welding can be prone to defects, formed by process interruption, because of an electrical breakdown between the gun and the anode or the column wall. The breakdown reduces the accelerating potential to zero and the weld pool keyhole collapses as the beam power is removed. This can form void type defects in the fusion zone. Power supplies are required that can react quickly to avoid large surges of current or extended welding interruption when a discharge occurs.

The performance of the high voltage power supply, which provides the accelerating potential for the electron gun, plays a key role in consistent operation of electron beam welding equipment. The primary parameters which control electron beam welding are beam power, beam focus position, deflection amplitude and frequency, working distance and welding speed. The first three of these variables are affected by the accelerating potential, which in addition determines the fundamental power density of the beam as it emerges from the gun. The electron beam welding process generally is automated, and consequently consistent results rely upon the reproducibility of beam characteristics by preset electrical values.

Power supplies

The two fundamental requirements of an EBW power supply are high voltage regulation and rapid response to discharges. For welding in material thicknesses greater than a few millimetres the formation and stability of the weld pool keyhole are essential to consistent and successful performance of the process.

Typical figures quoted for load and line regulation are less than 1% although measurement of the high voltage cannot be achieved readily without use of specialised equipment and techniques. The response that the power supply should have to provide minimum interruption to welding is dependent upon how rapidly the vapour generated by the heating effect of the discharge and/or gas which may have initiated the breakdown, can be dispelled from the electrode gap.

Conventional designs

Some of the first electron beam welding power supplies used variable three-phase auto-transformers to provide control of the primary power to the high voltage transformer. As the powers of machines were increased, maintaining the accelerating potential accurately was recognised as being very important to obtain consistent performance for deep section welding. To improve high voltage regulation, motor-generator sets were used and the field winding excitation controlled by an electronic feedback loop. This configuration is used to the present day by some machine manufacturers and was used very widely until the introduction of power semiconductors, and the development of switch-mode power supplies.

Within the electron beam column the accelerating potential from the power supply is applied to gun electrodes typically separated by a few tens of millimetres. Although this region is always evacuated to a pressure of less than 10 ^-5 mbar, the ingress of dust, metal vapour or ions, or outgassing of gun electrode components inevitably results in occasional discharges (also called breakdowns or flashovers). The power supply response to the effective short circuit must be to decrease the transformer emf rapidly if damage to electrodes (and subsequent increased likelihood of further flashovers) is to be avoided. Conventional power supplies ( i.e. those that do not include an inverter driver) include a high current breaker contactor to protect the electrical components in the event of an overload. The breaker is fitted to the primary circuit of the transformer and opens when excessive current is detected. This causes the power supply to shut down until manually reset and the welding process is halted. Inevitably this leads to fusion defects.

During the past 10-20 years conventional power supply design has been improved significantly by using a high voltage, high power regulation valve in the secondary circuit. ^[1]   The valve carries out two important roles. First the high voltage can be regulated to reduce the DC ripple to a low level. Consequently, the stored energy in filters after the transformer-rectifier can be reduced without increasing the high voltage ripple. Secondly, during a gun electrode flashover the output impedance can be increased rapidly. However, at high voltages ( e.g. above l50kV) use of in-line valves cannot be considered to be economic or practical.

Switch mode inverters

An alternative power supply configuration which has been implemented by a number of machine manufacturers during the past decade uses switch-mode inverters. In the early 1980s this work was pioneered by TWI to powers of l00kW. ^[2]   Inverters for electron beam welding typically operate at up to 30kHZ and consist of a three-phase rectifier, high power semiconductor inverter and high frequency transformer, see Fig.l. High frequency operation leads to compact transformers (normally only a minor advantage for most potential equipment users), but more importantly gives high ripple frequencies (of twice the inverter frequency) in the usual case of single phase transformers. The higher ripple frequency after the transformer rectifier allows lower energy DC filters to be used which reduces the overall stored energy of the power supply. Low stored energy, high frequency supplies can be controlled better to prevent damage during gun discharge and, as will be described can be configured so that the discharges do not interrupt welding.

Avoiding high voltage discharges

There are a number of measures which may be taken to reduce the number of gun electrode flashovers. However, although these will reduce the probability of a gun discharge, the risk can never be entirely eliminated. Gun design has been improved significantly with the development of high power diode guns at TWI which have 'fail-safe' characteristics under discharge conditions. The cost of repairing damage caused or discarding components because of uncontrolled discharges is often high, as electron beam welding is a process generally applied to pre-machined workpieces, which can be of thick section for high power machines. Even a very low risk of flashovers during welding may not be economically tolerable for manufacturers. Therefore, there is a requirement for supplies which may respond to discharges in such a way as to minimise interruption to welding and reduce or eliminate consequent defect formation.

TWI developments

A number of improvements to the switch mode power supplies used at TWI were required to allow the control system to change the high voltage and beam current levels rapidly to permit control during discharge events. Those areas of the power supply control system which needed to be addressed were:

• Voltage and beam current measurement techniques;
• Intelligent selection of response to discharges;
• Frequency response of beam current feedback control circuits for diode guns.

Further reduction of discharge probability was addressed by design of a control circuit especially for the gun conditioning period, aiming to operate the power supply in a low power mode. Finally, a control circuit was required to produce a pre-defined high voltage recovery curve following a gun flashover, this curve being of an optimised form to avoid weld defects and reduce the risk of further re-striking of the discharge.

To control discharges it is necessary for the power supply to respond within a few tens of microseconds of detection of the flashover. During this time both the high voltage and beam current levels will be changed rapidly. As this must occur within feedback control loops it is a requirement that the output values can be monitored accurately during these transitions, and consequently that the speed of response of the monitoring circuits matches that required for the power supply response.

High voltage measurement

Accurate measurement of the accelerating potential is not readily achieved. The problems encountered with high voltage monitoring circuits have been a result of the parasitic capacitance of the high voltage divider resistor and the response of the circuit under transient conditions ( e.g. when a discharge occurs). These led to generally low frequency responses inadequate for the discharge control circuits. As direct measurement of the circuit response time was not possible, computer analysis techniques were used to simulate the divider and subsequent improved circuit designs under transient conditions.

Beam current measurement

In practice the earth return current does not provide a good signal-to-noise ratio. If the earth return current is monitored directly across the resistor there is generally more noise than signal and filtering is necessary to remove the high frequency elements. This lowers the frequency response of the beam current controller and can lead to instabilities when the beam is switched on from zero current or following discharges in the gun. The higher frequencies are a consequence of monitoring a small voltage signal (up to 0.7V across a 1 Ω resistor for a maximum of 666mA beam current) in an electrical environment where high current switching is occurring at dI/dt rates of greater than 35 x 10 ⁶ A/sec ( e.g. 500A →0A in 15 µsec). Even when non-contact probes are used ( e.g. a DCCT device) current spikes are detected which do not appear on the real beam current.

These are caused by the stray capacitance of the transformer to its enclosure (ground). The stray capacitance current amplitudes are further increased by the high frequencies produced by rectifier diode recoveries and square edged input current waveforms from the inverter.

The DCCT device used has a 3dB band-width of l0kHz. In the earth return current monitoring circuit the frequency response is limited to up to 200Hz. To use the frequency bandwidth, and to maintain beam feedback stability an alternative method of measuring beam current has been found. This uses the DCCT mounted within the gun column and allows measurement of the beam current as it passes through the core of the device. Column mounted DCCTs are successfully used within two TWI gun columns. Measurement of the high beam current can be made accurately (±0.2%) at up to 10kHz bandwidth with very favourable signal-to-noise ratios (typically better than 1000:1).

High voltage discharge detection

Operation of the power supply depends upon measurement of the high voltage and the beam current and detection of gun flashovers. The high voltage and beam current measurements are used within electronic feedback loops to control and regulate the output of the gun. Detection of discharges is important for power supply protection and for arresting flashovers before irrecoverable damage is caused to the workpiece or the gun electrodes. Following discharge detection both the beam current and the high voltage regulating circuits respond in a manner designed to reduce workpiece damage to a minimum.

Moreover, the survival of the power supply during a discharge would be at risk if prompt detection and output current limiting were not carried out. Discharges may be detected by measuring one or more of the following:

• The transformer input current;
• The transformer input current rate of increase;
• The transformer output current;
• The transformer output current rate of increase;
• Electromagnetic radiation from the flashover ( e.g. RF, light, X-ray).

At present a combination of the first two is used to detect gun flashover although other techniques are under consideration.

Gun flashovers may occur during welding or during initial application of the HT to the gun (gun conditioning). During welding the priority is to interrupt the process for the minimum time to cause the least possible disturbance to the welding operation. During conditioning many discharges will occur as particles of dust are 'cleaned' off the gun. To reduce the conditioning time to a minimum and to clean the gun most effectively the priority is to ensure that minimum energy is dissipated in the discharges.

Gun conditioning control system

Switch-mode inverters offer a high degree of flexibility, allowing control in a number of operational modes. This feature has been fully exploited in development of electronics which allow a 'conditioning mode' of power supply operation where the maximum inverter output is limited to a small fraction of its normal potential. The inverter is enabled for up to one half-cycle (100µsec) ten times a second which gives an approximate output power of 0.1% of the maximum. For a l50kW inverter this corresponds to 150W. During the conditioning mode very little power is drawn as the beam is switched off. The loading on the output is due to the high voltage divider (where some 40W is dissipated at 250kV) and insulation leakage to ground. When the inverter is first switched on it is necessary to charge the output capacitance to the operating potential. In normal modes of operation this would take less than three milliseconds but in conditioning mode it takes at least five seconds and longer if there is any corona discharge in the gun. The high voltage output from the transformer during conditioning is shown in Fig.2.

If a discharge occurs when conditioning, the voltage reduces to zero and there is no energy input from the inverter for up to l00msec. Following this the high voltage increases in small steps, recovering finally after some five seconds. The delay between the discharge and the next pulse from the inverter gives sufficient time for the flashover to collapse the voltage and self extinguish. Regulation of the high voltage is poor compared with other inverter modes, e.g. a typical voltage ripple of 3% is experienced. When the conditioning mode is switched off the regulation returns to being better than 1%, as is required during welding.

High voltage discharge control

In the normal mode of operation of the inverter it is preferable to ignore some of the minor discharge events. These small flashovers extinguish without causing a complete collapse of the high voltage but may cause half-cycles of the inverter to give outputs which exceed the first current limit. During these events the inverter protects itself by switching off one of the conducting transistors but does not carry out any special action to stop the discharge.

At the other extreme, severe flashovers may cause the output of the transformer to be effectively shorted to ground for a period exceeding the half-cycle of the inverter (100µsec). When this occurs a very high rate of increase of current occurs through the primary winding. The strategy developed at TWI to clear the discharge is to hold the accelerating potential demand signal at or near zero for a period (termed the 'dead time') and then to recover the potential linearly to its normal level in a second period (called the 'ramp time'), see Fig.3. The duration of the dead time and the ramp rate are fully programmable. It is conjectured that during the dead time the initial discharge self-quenches. As the high voltage is ramped up, micro-discharges occur at intermediate levels but the recovery of the voltage is unaffected. However, the exact physical response within the gun column is not yet clear; this method of re-establishing the high voltage has been developed empirically.

It has also been found that at low beam powers, when the volume of vapour produced during welding is relatively low, the ramp rate may be increased by a factor of 3 or more. At higher beam powers, when more vapour is produced, the lower ramp rate is more effective and avoids restriking of the discharge during the voltage recovery. To take full advantage of this effect, a circuit was implemented which adjusted the rate of recovery dependent upon the beam current demand.

Beam current control during discharges

The electron beam welding systems designed by TWI over the past 10 years have used indirectly heated diode guns as the beam source. These have considerable advantages over the conventional triode gun systems, in particular the configuration of the gun is such that a flashover can lead only to a reduction in the beam current. However, indirectly heated diode guns have posed particular problems for the design of beam current control circuits where closed loop feedback is used to regulate the gun power output.

Control of indirectly heated diode guns

• The response curve of beam current against BBV has a very high gradient so that small changes in the BBV can cause very large changes in beam current;
• The response curve of the gun shifts considerably as the whole gun warms up;
• The time delay between applying BBV and obtaining beam current can be up to three seconds. This is associated with the heat capacity of the cathode and its thermal dissipation.

As with all types of electron beam welding gun; the beam current should be accurate to, say, 0.2% of full scale and excessive current settling times are undesirable. When discharges occur the beam current is reduced to zero as the high voltage diminishes. The normal operation of the feedback circuit would be to increase cathode heating in an attempt to stabilise beam current. The most important characteristic during a flash event is that the cathode heating is reduced and to achieve this the control circuit reduces the beam current demand to zero in a similar fashion to the high voltage. Actual beam current settling times will be cathode and gun design dependent.

Power supply operation

Experience of using power supplies with little or no flashover control systems has shown that when a microdischarge occurs during welding it frequently builds up to being a more severe discharge leading to either a power supply shut down, leaving a hole in the fused metal, or an extended interruption to the welding process which produces defects. Those defects formed are normally voids within the fusion zone caused by weld pool instability.

Discharge response circuits have been developed progressively to improve performance of the power supply on a wide range of welding applications. One particular requirement, which led to some of the control circuit development, was to ensure that a high power (75kW) beam could be used without interruption for 2 ¹/ ₂ hours. This test was successfully completed several times and further development has concentrated on reducing defect formation in welds, as well as continuously maintaining the beam power.

Trials were executed with an IDH 100kW diode gun operating at 150kV accelerating potential. Conditions were chosen from recent welding application work at travel speeds of 100 and 150 mm/min. All melt runs were made with the beam axis horizontal. Two fully penetrating conditions were examined in a 150mm thick carbon-manganese steel commonly used in offshore applications and in a 30mm thick 3% nickel steel supplied to a naval specification. These materials were chosen as there was a need within an industrial application to make similar welds for long periods without interruption or defect formation from discharges.

In both cases, melt runs were performed without flash events and then repeated with simulated flash events. These were generated by producing a dead time (flash event) signal which was applied simultaneously to both the beam current and high voltage control circuits. Radiographs of the welds were made using X-rays transverse to the weld plane. Comparison of these results showed that this interruption to welding did not cause defect formation.

Fair testing of the discharge control circuits under experimental conditions cannot be achieved without large workpieces and long melt run times. Flash events are quite rare when carrying out welds on the materials used in these trials, but over a period of 2 ¹/ ₂ hours several events may be expected depending on machine characteristics, particularly the pumping system and gun column configuration. As a consequence long welds of up to six metres or more are required for trials to investigate effects of real flashovers.

An Industrial Member company is using this system on large component fabrication. Although full welding parameters have not been made available, 30mm deep FP (beam axis vertical) welds have been made without fusion zone defects because of the flash events which occurred, although some disturbance of the weld top bead was observed. Further experience with the discharge control system will inevitably lead to progressive improvements but initial tests are encouraging.

Main conclusions

This work (which has been more fully reported ^[3] ) has led to improvements to electron beam power supplies such that gun discharges during welding do not cause defects under trial conditions. The main conclusions of this work are:

• A control circuit has been designed and built which successfully prevents formation of defects when discharges are simulated;
• Techniques for measuring the high voltage and beam current have been developed to allow rapid control of the output;
• Gun conditioning has been achieved in a shorter time with less electrode damage by use of a novel control circuit;
• The output power from indirectly heated diode guns has been accurately controlled using specially designed high speed closed loop beam current control circuits.

References