Many high current switchmode DC to DC converters, particularly on computer motherboards, use two or more power FETs in parallel to increase efficiency or, in other words, reducing power loss to make products more eco-friendly or green.  Keeping the amount of heat dissipation as low as possible also reduces the possibility of damaging the board, particularly if the airflow from the fan is restricted.  Putting two similar devices in parallel theoretically reduces the power loss in each device by a factor of four.  This means that the temperature rise in each device is a quarter of what it would be with one device, so the heatsink size can be reduced or the manufacturer can use surface mount devices instead to save space on the board.

For the devices to share the current equally, they must be matched for on-resistance (R_DS(ON)) and switching time.  The latter is determined by the FET gate capacitance and intrinsic series resistance (Rg), which is typically 10 ohms or less. The normal technique of measuring these parameters is with a precision LCR bridge. Tests with these instruments take between 200msec and one second. This is far too slow for production line testing of these devices as the full datasheet set of tests needs to be done in less than 200ms. The bridge instruments are also very expensive as only a small part of their capability is used for this particular test.

Our customer makes discrete device test equipment for end of production line testing, at 12-14,000 units per hour or more than three per second.  This rate includes a full datasheet test on a device and mechanically indexing the device to the next test station.  Their tester did the R_DS(ON) measurement and needed the Rg test, so they commissioned us to design and develop a board to do this test.

Our design (called the Rg board) uses a high speed analogue interface to a digital signal processor (DSP) which forces a low voltage sinewave onto the gate of the device under test. It then measures the the gate-source voltage and the current into the gate. The resulting magnitude and phase of these measurements is used to compute the series resistance and gate capacitance in under 5ms. The result is then sent to the tester through a serial data link.  The board we designed is shown below with the auxiliary power supply, communications and digital circuits to the left and top, and the analogue circuits to the bottom right.

The challenge is that the phase angle between the sensed voltage and current is very small with a few ohms in series with a gate capacitance ranging from over 10nF to under 100pF.  A 1MHz sinewave is generated using a digital to analogue converter and sent to the device.  The voltage and current are measured synchronously and stored over a fixed number of cycles to be analysed by the DSP.  The number of bits and the correlation technique enable us do the measurements with high resolution.  Careful calibration is needed to get the accuracy required.  This nulls out the parallel capacitance for small devices, the series inductance for large devices and the effect of intrinsic phase lag in the analogue interface circuits.  A second current range is used for devices under 200pF to improve the resolution of the measured current.  For devices under 100pF, the tests can be repeated in a fast software loop and the results averaged to improve the resolution.

The analogue interface must be close to the device under test (DUT) to reduce the effects of stray inductance and capacitance in the leads between the measuring head and the DUT.  The Rg board can be combined with the high voltage test head or installed in a stand-alone test head as pictured below.

There are three variants of stand-alone Rg test heads which incorporate the main Rg board with a relay matrix and a Kelvin test circuit. The Kelvin test is to check the continuity of the force and sense contacts onto the pins of the device under test. The first variant had a relay matrix to allow the Rg board to switch between two devices.  The second variant was designed to test only one device but did incorporate the Kelvin test circuit.

One semiconductor manufacturer wanted to do Rg tests with the gate under forward bias while a second manufacturer wanted the tests done with the drain biased. This had to be achieved without compromising the accuracy of the test system and adding as little to the total test time as was reasonably feasible. The board pictured will bias n-channel or p-channel devices with up to 10V on the gate or 30V on the drain.

This third variant of the auxiliary board was designed to work from a 12V supply and has on-board switchmode converters to supply +5V for the Rg board and ±35V for the voltage bias circuits.  It includes a relay matrix for the Kelvin test circuit and to connect the device source to drain.  It also has the capability to drive up to eight external relays on the handler interface.  This board is primarily designed to test only a single device which allows us to minimise any extra stray capacitance and inductance in the measurement path.

The bias voltage can be switched to the gate or the drain of the DUT or it can be switched to an external application board on the handler.  The eight external relay drives can be used with a suitable application board to test multiple devices and has been used to make additional capacitance measurements on the DUT.