Power Supply designed for...

Power Supply Designed for Small-Scale Resistance Spot Welding

The power supply provides more flexibility, a smaller size,
and is portable for different application sites


Fig. 1 — SSRSW system diagram.

Resistance spot welding (RSW) is one of the welding processes that involves the joining of two or more metal parts together in a localized area by applying heat and pressure (Ref. 2). For applications such as medical devices and electronic components, the welded parts are thinner and smaller compared to common RSW applications. Therefore, the authors refer to this process as “Small Scale” RSW (SSRSW) (Refs. 1, 5).

Studies of the SSRSW have shown that 20% of the welding quality issues are welding schedule or power supply related (Refs. 3–5). Therefore, the study of different weld schedules and control schemes will contribute to weld quality improvement. This article presents the design and development of a novel power supply, which can provide a testing bench for these studies, and at the end implement the final results of the studies.

This power supply uses pulse width modulation technique, with low cost MOSFETs, to convert the power of a 12-V battery to the weld current up to 800 A. Microprocessor/controller technology is used in the novel design, which provides the flexibility for the application of different control schemes. The operation of the proposed power supply has been simulated, and the simulation results agreed with the experiments very well. Based on the experimental data, the weld quality and the stability of three different control modes have been analyzed, and the results indicated that the constant power control achieved the best weld quality.

System Description

Fig. 2 — System block diagram.
Figure 1 shows the system diagram of a SSRSW system, which consists of a micro welding machine, a DC power supply, and a 12-V battery.

The designed power supply consists of the following three major sections: the power section, the control electronics, and the microcontroller. Figure 2 illustrates the system block diagram.

Fig. 3 — The novel generic control scheme.
The power section is a DC/DC converter, which converts the input power from the 12-V battery to an appropriate output voltage to feed the welding machine. Power MOSFETs are used as the switching components, which switch at a frequency of 20 kHz. Filter capacitors are used to assist the ripple current required during the high-frequency switching. An output inductor is used to filter the output current.

The control electronics section includes the driver circuits for the MOSFETs, the sensing circuits for voltage and current, and other electronics circuit for the control purpose.

The microcontroller is the central control unit of the system, which includes both the hardware and software. Based on the given reference and feedback signals, the microcontroller implements the control schemes and provides the PWM signal to drive the MOSFETs in the power section. An 8-bit microcontroller, PIC16F73 from Microchip, Inc., was selected as the CPU of the system.

Control Schemes

Fig. 4 — Simulation model for constant current control.

The following four control modes are set in this power supply:
1) Voltage control mode;
2) Current control mode;
3) Power control mode; and
4) Generic control mode.

The voltage control mode is to control the output voltage as constant; the current and power applied at the welding machine tips change along with the load resistance. In the current control mode, the load current is kept as constant. The voltage needs to be adjusted according to load resistance, and the power changes accordingly. In the power control mode, the voltage is adjusted to keep the power constant. Consequently, the current changes as well.

A generic control scheme is discussed, which covers all three control modes above. Figure 3 presents the block diagram of the generic control scheme.

Fig. 5 — Typical waveform of the dynamic resistance.
A variable called equivalent power (Pe) is defined as Pe = Va x I(1-a), where a is a control index, which has the range between 0 and 1. The reference (Peref) and feedback (Peo) are all based on this definition. The PI controller takes the error signal (e) and calculates the duty ratio (D) of the PWM applied to the power supply.

The generic control scheme can represent all three control schemes discussed above by using a different control index a. When a = 0, the equivalent power (P
e) equals to the current (I), therefore the generic control scheme is equivalent to the current control. When a = 1, the equivalent power is the voltage, and the generic control scheme represents the voltage control. When a = 0.5, the equivalent power is defined as the square root of the output power, and the generic control scheme is the same as the power control.

System Simulation

Fig. 6 — Simulation results. Upper: blue — load current, red — battery current; bottom: blue — capacitor voltage, red — load voltage.
The operation of the DC power supply is studied through simulation. Figure 4 shows the simulation model for the closed-loop constant current control mode of the SSRSW system.

The load-dynamic resistance is the key component in this model. The load is the welding machine, and its impedance consists of the following components:
1) Resistance of the electrodes;
2) Bulk resistance of the workpieces;
3) Contact resistance between the electrode and workpiece;
4) Contact resistance between pieces; and
5) Resistance of the cables.

Fig. 7 — Block diagram of experimental setup.
An equivalent RL component should be an adequate representation for the load. However, to simplify the simulation, components 1 and 5 are considered in the miscellaneous loss resistance, and the cable inductance is lumped in with the battery inductance. Therefore, in the simulation model, the load only represents the components 2, 3, and 4, which will be changing during the welding process. That’s why it is also called a dynamic resistance.

In reality, the dynamic resistance is a complicated function of different variables like current, voltage, and temperature. To simplify the study, the following assumptions are made:
1) The welding process takes 20 ms; and
2) A single resistance curve can be used regardless of power supplied.

Fig. 8 — A — Output waveforms under open-loop voltage control; B — constant current control; and C — constant power control modes.
According to these assumptions, the dynamic resistance is described as a function of time. Figure 5 gives an example waveform of this function taken from Ref. 7. Based on the experimental data from the prototype, an analytical expression for the dynamic resistance is obtained as Equation 1

Rwelder = (1.1433t-0.8867).(u(t)-u(t-1))
+ (0.7937(t+5.4253) + 0.03). 0.025
.(u(t-1)-u(t-20))                                 (1)

where R
welder is the dynamic resistance in mΩ, t is the time in ms, and u(t) is the unit step function.

Figure 6 shows the simulated load voltage and current, battery current, and capacitor voltage during switching. The simulation results are verified by the later experimental results.

Experimental Results

Experimental Setup

Fig. 9 — Nugget diameters under different control modes.
The experimental setup includes a SSRSW machine, the developed DC power supply, current and voltage sensors, a data acquisition system, and a personal computer (PC). Figure 7 shows the block diagram of the setup.

Fig. 10 — Comparison of variances of nugget size for three control modes
The SSRSW machine is a Model 80 Series from Unitek Peco. The welding force can be adjusted by selecting a trigger point within a range of 2 to 20 lb. When the force reaches the trigger point, a trigger signal is sent to the DC power supply, which controls the welding current flowing through the electrodes and the workpieces. A voltage sensor and a current meter are used to measure the electrodes’ output tip voltage and welding current, which are transmitted to the PC through the data acquisition system.

The data acquisition system is a Single-Board Simulator from DSpace, Inc., which includes a connector panel, PPC controller board, and the software. The connector panel samples the measured signals, and the PPC controller board is inserted into the computer and communicates between the PC and the connector panel. The software does the job of signal processing. A Tektronix TDS 2014 oscilloscope is used to monitor some critical signals during the welding process.

Experimental Results

Fig. 11 — Tip voltage waveforms under voltage control mode.
Parameters obtained from the lobe tests were implemented in welding tests for the three control modes. The welding test results provide the data for further analysis and comparison of the three control modes.

Figure 8A–C shows the output waveforms under different control modes. The experimental results show that the novel power supply can provide sufficient power and control the welding process successfully.

Experimental Data Analysis

Fig. 12 — Weld current waveforms under current control mode.
Based on the experimental data from the welding tests, further statistic analysis has been performed. The effects of the three control modes are compared according to the statistic analysis results.

The variance of the nugget size is used to evaluate the general effect of the control mode on the welding quality. The smaller the variance, the less the dispersion of the samples, and the more consistent the weld quality. The repeatability of the system is used to check that the change in consistency of the weld nuggets is due to the mode of the power system and not a result of the implementation of the control modes.

The measured nugget sizes under the three control modes are presented in Fig. 9. Note that the open loop or voltage control has least repeatability and, hence, requires loose set points.

The variance of nugget size is expressed as the following:

where Di is the diameter of the nugget sample i, D is the average of the nugget
diameters, and s2 is the variance of the nugget size.

Fig. 13 — Output power waveforms under power control mode.
The variances of nugget size under the three control modes have been calculated and the results are illustrated in Fig. 10. It is obvious that the variance for the power control is the smallest, and then the current control mode, followed by the voltage control mode. The analysis results indicate that the power control mode achieves the best welding quality for the material of workpieces used in the study.

Table 1 — Variables tor Testing Repeatability

Control Mode Repeatability Variable
Open-loop voltage control Tip voltage
Constant current control Welding current
Constant power control Tip power

fig. 14 — Variances of the repeatability variables for three control modes.
The variables used to measure repeatability for the three control modes are listed in Table 1. For the open-loop mode, the control variable is the duty cycle, which is not an appropriate variable for testing repeatability of the controller implementation. However, control of the duty cycle, theoretically, is a form of voltage control. Hence, we have used tip voltage to measure repeatability for this mode. Figures 11 through 13 show the measured waveforms of the repeatability variables from the repeated experiments under the three control modes, respectively.

In order to compare the three control modes, the normalized values are used in the definition of the variance of the repeatability variable. The variance of the repeatability variable at time moment t is defined as:

where s2(t) is the variance of the repeatability variable at time moment t, n is the number of samples, Ci(t) is the repeatability variable at time moment t, and C(t) is the average of the repeatability variable at time moment t.

Figure 14 shows the variances of repeatability variables as functions of time for three control modes.

The variance of the repeatability variable over the whole welding process is defined as:

Figure 15 shows the variances of repeatability variables over the whole welding process under the three control modes. From small to large variances, the control modes are the current control, the power control, and the open-loop voltage control. The analysis results indicate that among the three control modes under study, the current control mode is the most stable.

The figure also shows that the variances of all three-control modes are getting larger at the end of a welding period, and it indicates that the preset welding length 20 ms may be too long for the tested materials.


Fig. 15 — Comparison of the variances of repeatability variables over the whole welding process.
A low cost, highly flexible welding power supply has been built. This power supply is ideal for testing new control parameters. It has been determined that constant power control mode produces more consistent sized nuggets. The implemented three control modes, voltage control, current control, and power control modes can be seen as three special points in the generic power control mode. The V-shaped nature of our existing results suggests that a for the generic control has not yet been determined.


    1. Steimier, D. W. 1998. “Downsizing” in the world of resistance welding. Welding Journal 77: 39–47.
    2. Welding Handbook, eighth edition, 1991. Miami, Fla.: American Welding Society.
    3. Zhou, Y., Gorman, P., Tan, W., and Ely, K. J. 2000. Weldability of thin sheet metals during small-scale resistance spot welding using an alternating-current power supply. Journal of Electronic Materials 29(9): 1090–1099.
    4. Zhou, Y., Dong, S. J., and Ely, K. J. 2001. Weldability of thin sheet metal during small-scale resistance spot welding using high-frequency inverter and capacitor-discharge power supplies. Journal of Electronic Materials 30(8): 1012–1020.
    5. Chang, B. H., Li, M. V., and Zhou, Y. 2001. A comparative study of small-scale and ‘large-scale’ resistance spot-welding. Science and Technology of Welding and Joining 6(5): 273–280.
    6. Fundamentals of Small Parts Resistance Welding, UNITEK PECO Company.
    7. Tan, Wen. 2004. Small-scale resistance spot welding of thin nickel sheets. Ph.D. thesis, Doctor of Philosophy in Mechanical Engineering, Waterloo, Ont., Canada.

L. J. BROWN and J. LIN (jlin33@uwo.ca) are with the University of Western Ontario, London, Ont., Canada.

Based on a paper presented at the AWS Welding Show and Annual Convention held April 25-­28, 2005, in Dallas, Tex.