Introduction
My third and final test on the Tektronix AFG31052 Arbitrary Waveform Generator is to investigate the use of the Double Pulse software for testing the performance of MOSFET and IGBT devices. Naturally, I have to try and build in some aspect of an insulation tester into as many of the blogs I write on element14 and the main voltage generation circuit within a lot of insulation testers is a small pulse transformer switched by a MOSFET.
It therefore seemed a good idea to modify the test circuit within the Tektronix application note to see how MOSFETs perform inside the voltage generation circuit of an insulation tester.
A pulse transformer from an old insulation tester was made available, with the rest of the components either purchased or obtained from my stock pile. I identified several MOSFETs from the insulation testers I have carried out teardowns on.
Some of the devices seemed to be a little obscure and I had to scout around a bit to minimise the number of orders to obtain samples. I hope to test a few of the devices to obtain a comparison of the output circuit efficiency from different insulation testers. To be honest, I was not expecting too much from this. The output circuits are low power and an insulation tester is operated intermittently by its nature, so I did not imagine that efficiency would be a high ranking factor with the insulation tester designers.
As a secondary aspect to this, I wanted to look at the pwm output capability of the AFG31052 at testing the MOSFET / transformer arrangement to see how the different voltage output levels were generated by the insulation tester.
Double Pulse Software
The double pulse software does not come installed on the AFG31052 and needs to be downloaded from the Tektronix website. A login for the site is needed to perform this.
The downloaded file is then unzipped onto a USB stick and loaded into the AFG31052 using the 'Add Apps' function on the home page of the waveform generator. This only took a few seconds and loaded without any issues. An icon for the software then appears on the home page next to the ArbBuilder software as shown in the following picture taken from the Tektronix Application Note.
The double pulse software has only one screen for the settings. One half of the screen is for general setup of the pulse parameters and the other half is set aside for specifying individual pulse widths. A number of options are available for running the double pulse test, from continuous running based on a timer setting to manual triggering of the test, for a single shot function.
I tried running tests with both a manual trigger and a continuous running setting.
Test Setup
The test setup is quite a busy one requiring two power supplies, an oscilloscope and the AFG31052. The differential probes are powered from the USB port on one of the instruments.
For the majority of the tests, the current was measured via a 1 Ohm shunt resistor. I did do some tests with a current probe, but found there was a lot of noise, as the signal was at the bottom end of the capability of the probe. Averaging the data collections, does help, but then I am potentially affecting the spikes in the signal. There is also an issue with the time synchronisation and the rise time of the current probe.
To test this I made up a crude 50 Ohm load with a BNC and 4 mm sockets to connect the oscilloscope probes and differential probes to. I then used the double pulse signal to provide a fast square wave across the 50 Ohm load to allow me to synchronise the four channels. Below is the arrangement with one of the oscilloscope probes removed for better clarity. This test setup is something I will look to better in the future.
The initial screenshot of the oscilloscope shows all four channels enabled and differences can be clearly seen.
The following image gallery shows the measurements and adjustments made between the channels.
| {gallery} Channel Synchronisation |
|---|
Channel 1 and 2 delay measurement - no adjustment required |
Channel 1 and 4 (Voltage) delay measurement - 3.9 ns adjustment required. |
Deskew applied to channel 4 |
Channel 1 and 3 (current probe) delay measurement - 179.5 ns adjustment required |
Maximum 125 ns desk applied to Channel 3 |
Extra deskew applied from Channel 1 |
Aligning the differential voltage probe (Channel 4) was easily achieved. The probe had a rise time of 8.7 ns compared to the rise time of 4.2 ns of a probe connected directly to the Channel. The issue is seen in the last three screenshots. The total deskew between Channel 1 and Channel 3, the current probe, is 179.5 ns. The MDO34 has a maximum deskew of 125 ns. This isn't a big problem as the remaining deskew required can be achieved by using a negative deskew on the other channels.
However, even when aligned, the signal on the current probe is measuring a 98.88 ns rise time. This is considerably more than the rise time on the voltage probe and the direct connected channels. Given that the experiment is measuring on and off times of the MOSFETs which will be less than 50 ns and 200 ns respectively, I do not feel that the use of the current probe is appropriate for this kind of testing. Hence the use of a shunt resistor, for the actual timing measurements made.
Test Results
Whilst a number of MOSFETs were tested, the screen shots from the oscilloscope are predominantly from the tests on the LR3410, as this appears to to be the MOSFET utilised the most in the insulation testers I have reviewed.
The first screenshot depicts the overall waveform captured during a test on the LR3410. For the tests on this MOSFET I decided to use an averaging measurement and run the test for a few seconds instead of the single shot methodology discussed in the application note from Tektronix.
Over the whole switching cycle, the average power dissipation was 87.48 nW, with a maximum current flow of 668.4 mA and peak reverse current of -684.2 mA. These measurements are much less than those detailed in the application note. The predominant reading for this will be the lower voltage I am feeding to the MOSFETs. As I wanted these tests to be representative of the operation of the MOSFET in the insulation tester circuit, I applied 9 Volts against the 100 Volts used by Tektronix.
I have also replaced the recommended 1 mH inductor, with a transformer from an insulation tester that has a much different characteristic. This has undoubtedly reduced the power dissipation within the MOSFET during these tests.
| {gallery} Inductance Measurements |
|---|
1mH Inductor measured |
1mH impedance measurement at 1kHz |
Transformer Inductance Measurement |
Transformer impedance measurement at 1kHz |
Timing characteristics were successfully measured for the MOSFETs. The 'on time' for the MOSFET is extracted from a measurement from the rising slope of the second pulse. The time differential is measured between 10% of the switching pulse ramp, which is 1.2 Volts in my case, to 90% of the Vds ramp, which is 8.1 Volts for my test. This is referred to as 'td(on)' within the standards and is identified in most of the data sheets I have for the MOSFETs I am testing.
In the two screen shots below, the traces are all measurements across the bottom MOSFET. Channel 1 is the drain-source voltage, Channel 2 is the gate pulse applied, Channel 3 is the current through the MOSFET, measured off a 1 Ohm resistor in series with the drain connection and the Maths Channel is the Voltage multiplied by the current.
The 'off time' is taken from the descending slope of the first pulse. The measurement point is taken between 90% of the switching pulse, 10.2 Volts, and 10% of the Vds voltage, which is 0.9 Volts for my test setup. This is referred to as td(off).
These measurements were made manually on the oscilloscope using the cursor function.
There is a nice little graphic in the Tektronix Application note that summarises these two measurements along with the rise and fall times.
Based upon the data obtained the energy consumed during the switching process can be calculated using either the integration function on the oscilloscope or a manual calculation detailed in the Tektronix application note.
The other measurement made was the reverse current through the body diode of the first MOSFET as the electric field in the pulse transformer collapsed after the bottom MOSFET is switched off. This reading is captured with the voltage across the first MOSFET alongside the drain current measurement. The Tektronix application note carries this test out at a supply voltage of 20 Volts, but as may intention is to test the MOSFET performance in the insulation test output circuit, I kept to the 9 Volts supply I was using. As a result, the leakage currents are quite low and not so well defined in the oscilloscope traces.
In these traces Channel 3 remains as the current measurement and Channel 4 is the drain source voltage across the top MOSFET.Channels 1 and 2 have been turned of for better clarity. The Maths Channel is again the voltage multiplied by the current.
The relatively small reverse current can be seen in the measurements tab on the right of the screenshot. For a comparison, I did raise the MOSFET supply voltage up to 20 Volts and captured the screenshot.
You can see that the reverse current was jus over double that when the supply was only 9 Volts, showing a relatively linear characteristic.
MOSFET Comparison
The results of the MOSFET testing are tabulated below with a few accompanying graphs.
The switching times measured are compared to the nominal times from the manufacturer's data sheets. A mixture of results are seen across the devices. The 02N120 IGBT device seemed to perform better than its specifications for both on and off times, as did the 16N10L MOSFET. Both the 050N10 and K2231 MOSFETs had slower switch off times than expected, but faster switch on times. Both the 320N20N and LR3410 exceeded its parameters for all the tests.
The reason for the different timings obtained is likely to lie in the test methodology. Looking at the data sheet for the 320N20N MOSFET, that is seemingly one of the worst performing devices, the test conditions for measuring the switching times are considerably different to the operating conditions within the insulation tester output circuit.
In an insulation tester, the supply voltage is generally 9 V or 6 V, but the test conditions required 100 V, as per the Tektronix Application Note. The gate voltage was 10 V, I utilised a 12 V supply, as unfortunately my variable supply was feeding the MOSFET circuit. The gate resistor is specified as 1.6 Ohms for the test, where as the gate resistor I fitted was 22 Ohms, which is closer to the general specification of 25 Ohms for operating conditions.
It would seem likely that the manufacturer has set up the test conditions for the switching parameters to favour the device, so the ability to test the MOSFETs at the actual operating conditions would seem to be beneficial for more critical applications.
The energy consumed during the switching process is very low, due to the relatively low supply voltage.
Overall the 02N120 IGBT consumed the most energy over the switching cycles. The 16N10L consumed the least energy. The 320N30N is the most balanced over the on and off switching cycles.
You can see that the average powered measured across the complete double pulse cycle is very low. The IGBT, at just under 2 mW, consumed the most power over the complete cycle. The 320N20N MOSFET consumed the least power, with the LR3410 and 16N10L having similar average power consumption.
The final plot shows a comparison of the reverse current measured for all of the devices.
This moves onto a couple of interesting screenshots of the 02N120 IGBT. You will notice in the results table, I have noted that a Schottky diode was fitted to measure the switch on time. The device does not come with a body diode installed and when initially tested, a large spike was seen between the two gate pulses, as there was no diode for the energy stored within the pulse transformer to be dissipated through. As seen below, for a 9 V source, this was a considerable spike, albeit only for a short duration.
With a diode installed across the drain and source connections of the top MOSFET, this spike no longer appeared and the waveform followed the same pattern as that seen for the MOSFETs. The IGBT is only used by Gossen Metrawatt in their instruments. I have not fully reviewed the output circuit of the instrument, but I assume that the manufacturer has installed an external diode to the IGBT to prevent any spiking during circuit operation.
The reverse current can clearly be seen through the device in the screenshot below, depicted in Channel 3.
PWM Signal Tests
With this aspect it is my intention get a better understanding of how the high voltage output is generated within an insulation tester with the use of the MOSFET switching the transformer using a PWM signal. This signal can be generated by the AFG31052 using a pulse output and enabling the modulation function. This brings up a set of parameters for generating PWM pulses, with the option of using the internal PWM reference, or an externally generated one. At the moment, I have only investigated using the internal reference.
To help with the settings, I did capture the MOSFET gate pulse from one of the other insulation testers I have at my disposal. From the two examples below, the frequency seems to be around the 10 kHz mark and the pulse width is varied dependent upon the voltage output being generated.
The setup for the tests is similar to the MOSFET characterisation tests. The main difference is that the output is measured on the secondary side of the transformer. A rectifier diode is added to the output, I also experimented with a capacitor and a load resistor as I tried to get the system to work.
The following table an plot shows the problems I encountered in trying to get this to function.
As the duty cycle is increased on the signal, both the current through the MOSFET and the output voltage on the transformer are increased. The pulse width captured on the oscilloscope seemed to be a little erratic.
The current however, is too high. From the units tested, the supply current is between 130 mA and 380 mA for an insulation test, which I reach at 30% duty cycle setting. The corresponding output voltage is only 7.31V, which is nowhere near where it should be. I can drive the MOSFET harder, and get an increase in current, but the output voltage still remains much to low.
Example of output from transformer with 10% duty cycle applied;
The corresponding settings on the AFG31052 are;
I have tried various different setups and frequency settings, but they all produce a similar response.
This is disappointing, but is as far as I can take it for the purposes of the RoadTest, otherwise I will struggle to complete it. I am pretty sure it is something I am doing wrong. I think my next step will be to solder the transformer back into the specific insulation tester and test its operation in there. This will prove to me that the transformer is okay and I haven't inadvertently damaged it, and will also furnish me with some extra data on the PWM signal.
Conclusions
The Double Pulse software is a useful application for testing the performance of the MOSFETs. Although not provided with the instrument, the download and install is seamless and quick. As with the other functions on the home menu, I could only seem to access it via the touch screen.
The software seems to work well in conjunction with the evaluation board. Care does need to be taken with regard to probing and the setup of the test, to ensure that the results are representative. There is plenty of data within manufacturer's data sheets for the MOSFETs on test setups for this kind of application. As I decided to test the MOSFETs based around their operation within an insulation tester, my test results differed quite a lot from the manufacturer's specifications.
Whilst the AFG31052 was more than capable of producing a pulse for synchronising the channels across the oscilloscope, the test would benefit from a neater setup for making all of the connections and is something I will look to improve.
I think I will return to this aspect when I can source a 100 V power supply with the capability to supply enough current into the MOSFET, so that I can test them directly against the manufacturer's specifications.
I did not manage to master the PWM output capability and successfully drive the pulse transformer to produce an insulation test value. I will need to take this away and carry out some more research and testing to see if I can resolve the issue. In principle the AFG31052 seems to be full capable of producing the signals required, whilst there is an application example within the user manual, it did not seem that helpful to my specific application.
Top Comments