RoadTest: TI SWIFT™ Power Module EVM
Author: fmilburn
Creation date:
Evaluation Type: Power Supplies
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: None considered
What were the biggest problems encountered?: The product performed flawlessly. Limitations in the hobbyist grade test equipment arose and those are discussed in the review.
Detailed Review:
TPSM84A21 10A SWIFT™ Power Module Evaluation Module
Introduction:
First, I would like to thank Texas Instruments and Element 14 for the opportunity to test this product. As stated in my RoadTest proposal, the objectives were to:
The use of hobbyist grade equipment here is important. On the one hand, that is all many beginners and even advanced enthusiasts have available and there is valuable insight that can be gained with such equipment. On the other hand, the limitations of the equipment must be understood, compensated for when possible, and data rejected where the equipment is not up to the task. Where the testing was questionable I will identify it as such.
The evaluation board in the RoadTest is the Texas Instruments TPSM84A21 which features an 8V to 14V Input, 0.508V to 1.3 V Output, 10A buck converter power module. It comes in a 9 mm x 15 mm footprint and requires only one resistor to set the output voltage. The converter comes ready to use and installed on the TPSM84A21EVM-808 Evaluation Module (EVM).
In the course of the RoadTest many of the key items in the datasheet were addressed, including:
At the end of the report there is a list of links with additional information. I particularly recommend watching the following training series if you are new to testing power supplies or just need a refresher: TI Training: How to Test Power Supplies. It has great explanations and demonstrations on lab grade equipment with proper technique.
Not to spoil the ending but I learned a great deal, the module performed flawlessly, and it was easy to use.
Unboxing:
The product box arrived in good shaped packed within an Element 14 box with additional padding. The product box contained the TPSM84A21 10A SWIFT™ Power Module Evaluation Module in anti-static packaging along with a pamphlet directing users to the TI.COM website for additional information. The information on the website is comprehensive and complete.
The EVM board was nicely made and contained sufficient test points and access to the features of the module. Components and hardware were all high quality. The User Guide had schematics, PCB details, and a Bill of Materials sufficient to understand and test the board. Mounting holes are provided in the case the user desires to attach standoffs.
I must be a real geek but I love those Keystone Test Points with the loop for attaching mini grabbers.
The resistor for setting the output voltage are on the bottom of the board and are 0603 surface mount. If the user wanted to test at a different output voltage than what the board supplies, say the maximum of 1.35V, then they could remove one of the existing resistors with a SMD rework station and solder a new one but it might be nice to have an empty footprint or two. That is a pretty niggling point though.
Test Equipment:
Plan B: Originally I had planned to construct a simple constant current electronic load for the testing with the understanding that the power module would be the higher voltage TPSM84A22 Power Module. This turned out not to be the case and the parts on hand would not work with the TPSM84A21. Accordingly, Plan B was put in play which used a constant resistance load. The test equipment used is shown below.
The following table outlines the test equipment used and some of the limitations. Starting with the items in the box upper left and going clockwise in the photograph above….
Item | Quantity | Description | Comment |
TPSM84A21EVM-808 | 1 | SWIFT Power Module | The Device Under Test (DUT) |
Resistors | Assorted | Calibrated load resistors and shunt resistors for measuring current (0.1 Ohm to 2 Ohm) | Home made with what was on hand the load resistors would overheat if used for too long. |
Cheap Power Supply | 1 | Supposedly 15 V and 5 A with low output ripple | It was useable most of the time, kind of, sort of, well almost. |
100 MHz 2 Channel Digital Storage Oscilloscope | 1 | Siglent SDS 1102CML | The scope is entry level but has served me well. Not so long ago it would have been pretty hot stuff. During the RoadTest we will see the limitations of a 100 MHz 8 bit scope however. |
Alligator clips and jumpers | Assorted | Generic stuff | 18 AWG was used anywhere there was high current. Resistive loads were attached directly to the posts. |
Multimeters | 3 | Extech EX330 and two ANENG AN8008 | The AN8008 meters were bought after seeing them on the EEVBLOG. They actually aren’t too bad. Note the nifty tips on one of them from John Wiltrout. Mini grabbers were mostly used during the RoadTest though. |
Before starting the tests I did a comparison of all the multimeter readings at fixed voltages of 12V, 5V, 1V, 500mV, and 100mV. The differences were less than 0.5% and I considered that good enough.
Making Calibrated Resistors: Making and calibrating the resistive loads and resistors used for current measurement was more complicated. Resistance in the multimeter probe leads alone was in the 0.1 Ohm range and the meters only reported 1-2 significant figures in this range.
The nominal resistor loads of 2, 1, 0.5, 0.25, and 0.125 Ohms were constructed from 2 Watt 5% tolerance axial through-hole 1 Ohm resistors. These are very inexpensive and I had a number on hand.
Please ignore the sloppy soldering as I was trying to get as much solder as I could into the connections to minimize resistance. Note the use of heavy wire “headers” on the 0.125 Ohm nominal resistor made from 8 resistors in parallel to reduce voltage drop at high current.
It might be thought that with 5% tolerance the nominal value would not be particularly close to the actual value. That turned out not to be the case. To determine the actual values, the power supply was set in constant current mode at three different current settings. The current was measured with one meter and a second meter used to record the voltage differential across the resistor.
In the following photo the test setup for determining resistance is shown.
The resistive load in this example is two 1 Ohm resistors in parallel, or 0.5 Ohm nominal. Multimeter clips are placed directly on the resistor. The source voltage is coming from the power supply which is out of the photograph. The voltage drop across the resistor is 197.7 mV and the current is 397.1 mA. Using Ohm’s Law, the measured value of the resistor can be calculated: 197.7mV / 397.1mA = 0.498 Ohms measured compared to 0.5 Ohms nominal. Not bad…
These measurements were repeated multiple times at different currents and recorded for all load resistors and current measuring resistors and resulted in the following table.
As a further check, a 0.1 Ohm 1% tolerance resistor was tested as well. The measured values from that resistor were less than 0.1% off. All of the resulting calibrated resistors are thought to be accurate to at least +/- 1% and probably +/- 0.5%.
Test Plans: A Test Plan was created that envisioned cycling through every output voltage on the EVM for each resistive load at a nominal 12 V input. The power dissipated across the load resistors was calculated as well as the rated power of the group assuming each resistor was evenly loaded (not necessarily a good assumption).
The calculated power being dissipated at each test point is less than the rated power, but… the resistors must be derated if the temperature rises above 70 degrees C and it was recognized that at heavier loads the resistors might overheat. During testing temperature measurements were taken at high currents and overheating would have occurred if a test was allowed to continue for an extended period. Accordingly, tests were kept short and power turned off in between. A proper test rig for extended use and testing of equilibrium temperature at high current would require higher rated resistors, heat sinks, or other modification.
At this point the preparations for testing were deemed complete and a spreadsheet prepared to record measurements.
Testing:
Test Setup: The basic testing setup is shown below as a simplified schematic and as a photo.
To minimize the error associated with voltage drop it is important to place the measurement probe / clip close to the component being measured. Measure the input power to the EVM at the EVM for example and not at the Power Supply.
Measuring Efficiency: Meters in the photo are arranged in the same order as the schematic. The multimeter at far left is measuring the voltage across a 0.239 Ohm resistor as 23.4 mV. By Ohm’s Law the input current is therefore 0.098 A.
The middle multimeter is measuring the output voltage as 599.2 mV (note that the jumper on the EVM is set at 0.6 V). Since the resistor being used as a load has a known resistance of 0.497 Ohms this measurement can also be used to calculate the output current: Iout = .599V / .497 Ohms = 1.21A.
The meter on the far right is measuring the input voltage of the EVM as 12.00V. All the information needed to calculate efficiency is now known.
Eff = Pout / Pin = (1.21) (.599) / (0.098) (12.00) = 0.616 or approximately 62%
As outlined in the Test Plan, these measurements and calculations were then done at a number of different points and recorded in the spreadsheet below.
Note the yellow cells where no data was recorded. These cases occurred at high currents using the 0.124 Ohm load resistor and are attributed to flaky behavior from the power supply. There was a repetitive clicking going on in these tests as the relay to the transformer taps of the power supply hunted back and forth for the correct setting. I seem to remember seeing something about this somewhere on the internet when doing a constant resistive load test and it can damage the power supply. I stopped testing immediately. To test that theory that the power supply was at fault a series of batteries was set up to supply approximately 10 V and the test rerun as shown in the bottom section of the spreadsheet. The results were normal with the batteries and the shortcoming of the power supply verified.
These raw results, plus additional testing will now be used to evaluate the information in the datasheet and use of the EVM.
Evaluation of the EVM:
Setting the Output Voltage (VADJ): Setting the output voltage is extremely easy. Power down the EVM and move the jumper on the VOUT SELECT header to the desired positon. Settings for 0.6V, 0.8V, 1.0V, 1.1 V, and 1.2 V are available. The resistors on the board are 0603 SMD 1%. I measured the resistance on all of them and all were within 1% of the values given in the User Guide. There is a handy table in the datasheet plus formulas for determining different output voltages.
Efficiency: Efficiencies are recorded in the table above. The data overlaps that of a graph in the datasheet for 1.2 V, 1.0 V, and 0.8 V. The chart below plots measured data over datasheet curves. It can be seen in all cases that the measured efficiency exceeds that plotted in the datasheet.
This means that either TI is conservative in their datasheet or my measurements are a bit high. I believe my measurements to be reasonably accurate and it will be interesting to see if other RoadTesters get similar results. In any event, efficiency gets a thumbs up from me.
Power Dissipation: The datasheet has a plot which shows the safe operating area for various output currents and ambient temperatures with and without airflow. The safe area is quite large and at roughly 20 deg. C ambient in the testing area no overheating was expected. The maximum temperature seen on the module was 33 deg. C. It must be remembered however that the test apparatus was only operated for short periods and turned off in between tests so it probably never reached equilibrium operating temperature.
Output Voltage Ripple: The datasheet gives the Output Voltage Ripple with 20-MHz Bandwidth peak-to-peak as 8mV. Before going any further it should be noted that my oscilloscope probes do not have ground springs and I used the normal (long) ground wire with clips. The way the test points on the board are set up and the unorthodox way the load resistors were attached made me uncomfortable using the “tip and barrel” approach. The TI Training video admonishes against using the measurement technique I used but we work with what we have.
Switching ripple noise occurs at the same frequency as the switching frequency of the module, or 4MHz in the case of the TPSM84A21. We expect to see a 4 MHz triangular wave form. Here is a screen shot from the oscilloscope:
The time scale is 250ns per division and the wave peaks are superimposed upon them which allows easy confirmation that we are seeing the switching ripple at 4 MHz. The cursors are set apart by 9.6 mV and the ripple is inside these limits. The 8 mV ripple given by the datasheet is within the accuracy of the measurement.
At this point I wondered what the power supply voltage at the input looked like. Pretty poor and well over the specification given by the power supply manufacturer. However, given that the output ripple looked OK, additional bulk input capacitance (as provided for by the empty pads on the EVM) was not pursued.
Start-up Waveforms: This is a thing of beauty and a joy to behold.
The time scale is 5 ms per division and Vout is set to 1 V, Vin is 12 V. The curves follow the datasheet well. The module does not start coming up until the input voltage reaches approximately 7.5 V. There is a pre-programmed soft start where the output ramps up in 4.1 ms to reduce the initial surge in current. Worked perfectly.
Transient Response: The transient response as outlined in the datasheet depends on setting but as an example the electrical characteristics section of the datasheet gives +/- 10 mV for a 1 A/us load step. Curves are also plotted.
Not having a programmable load I switched a resistive load in and out with a transistor using a microcontroller (TI MSP430F5529LP) toggling at 20 Hz. Settings were as follows:
My best guess is that about 40 0mA was being switched in and out of 4 A.
The setup looked like this:
The top yellow trace on the oscilloscope is the measured transient load voltage being switched in and out. The bottom blue trace is the output voltage which ideally is stable. All that can be seen is noise on the output so it is “good” but this is not a particularly meaningful test and cannot be compared to the datasheet.
Enable (EN) and Shut-down: The TPSM84A21 has an enable pin which can be used to control on and off. The enable pin is pulled high to turn on. When pulled low, the module stops switching and enters a low power state. It has an internal pull-up so that it can be left floating and the EVM will operate. During my tests it was reliable and easy to use. Note from the efficiency table above that the power losses when there is no load were in the 35 mA range so being able to disable is useful. I did not take a screen shot with the oscilloscope or make more detailed measurements of power.
Gate Driver (VG): An internal linear regulator is used to supply internal power to the controller. The measured voltage on my sample was 4.78V (typical from the datasheet is 4.8V). It is also connected to the Gate Driver (VG). To improve efficiency an external power supply between 5.0V and 5.5V can be attached to VG. This was not tested.
Power Good (PGOOD): The PGOOD pin is pulled low when power falls outside 95% to 105% of the internal voltage reference, during a soft-start, or when a fault is detected. It needs to be pulled high when used. During testing this was done by connecting it to the Gate Driver (VG) through a 22.6 k resistor. It was tested by lowering the input voltage via the power supply until the pin went low. This occurred at approximately 7.5 V and testing was deemed successful.
Thermal Shutdown: As mentioned in the section on Power Dissipation, thermal shutdown was not tested and the module never overheated.
Overcurrent Protection: This was not intentionally tested. However, when the power supply was oscillating around the transformer tap there was overvoltage, and thus overcurrent would have occurred momentarily. No magic smoke or subsequent damage was observed.
Output Undervoltage / Overvoltage Protection: Not tested.
Undervoltage Lockout(ULVO): This was observed after testing PGOOD by cranking the voltage on the power supply back up after triggering a shutdown. As described in the datasheet the switching started up again at something greater than 7.5V with hysteresis around 200mV. The test was successful. It is possible to set a higher threshold but this was not tested.
External Clock Synchronization (SYNC): There is a SYNC pin for connecting an external clock. The datasheet states the external clock should be within +/-10% of 4 MHz (internal oscillator frequency). The switch from the internal oscillator to the external clock can be made while converting power. When the external clock is removed, the transition back to the internal oscillator occurs after 4 internal clock cycles. The external clock must be a square wave with a duty cycle between 20% and 80%. The amplitude transitions must be below 0.8V low side and above 2.0V on the high side.
The tests were done by generating a square wave using the PWM feature on one of the timers of a Texas Instruments MSP430F5529 microcontroller. It was arbitrarily set to 3.66 MHz and 50% duty cycle for the first test. The photo below was taken before connecting to the SYNC pin to understand the shape of the resulting square wave.
There is ringing but the square wave stays inside the parameters set in the datasheet. The second test was done at 4.27 MHz and resulted square wave barely qualified as such but I ran it through anyway.
The external clock was attached both before start-up and while running (i.e. hot connect) in several trials without problems. The results are shown in the table below.
I was surprised that the efficiency improved when running at the slower clock speed (86% efficient for the external clock at 3.66 MHz and 85.6% efficient for the internal oscillator running at 4 MHz). It was somewhat worse at the higher external clock speed of 4.28 MHz. These tests are indicative but somewhat suspect in my mind and I would be interested in what other RoadTesters observe.
EMI: Emissions were not tested.
Layout and Design: There is a complete section in the datasheet on design considerations and layout which the evaluation board appears to follow closely. A four layer board with a large internal ground plane is required to match the thermal values in the datasheet. Large copper areas (not traces) are used to minimize resistance and thermal stress for VIN, VOUT, and PGND.
Conclusions:
As stated in the introduction, a primary goal of this RoadTest was to satisfy a personal desire to become better acquainted with power modules. This goal was decisively met. I learned a lot about power modules, how they are tested, and the limitations of the hobbyist grade test equipment and methods I have been using.
The following summarizes my feelings on the equipment and methods used:
What Worked | What Could Be Improved |
Efficiency measurements match the datasheet well | A better power supply and a programmable electronic load would improve measurement efficiency and probably accuracy |
The multimeters I have worked reasonably well and gave acceptable results with the shunts and resistive loads. | The high conductivity resistor loads were prone to overheating at high currents. |
Start-up waveforms looked very good and ripple reasonable although ripple did not have high resolution. | Transient waveform measurement was not useable and would require a programmable electronic load. Higher resolution oscilloscope would also be helpful. Ground springs for the oscilloscope probe tips could improve accuracy. I tried making some myself but they didn’t help. |
The module functions were easy to control with the MSP430F5529 | A better clock source or digital signal generator would be good. I have a low quality one I built myself but did not use it since the microcontroller seemed to work OK in the end. |
The TPSM84A21 performed flawlessly throughout and the EVM was easy to use. The Texas Instruments Datasheets and User Guide were complete and comprehensive. I also found the videos in the TI: Training section on How to Test Power Supplies succinct, informative, and very approachable. A point was knocked off the price to performance since the EVM would be expensive for a hobbyist. But for the intended market it would be cheap compared to having one of your own engineers design and make up an evaluation board.
If you made it this far I hope you found the RoadTest informative and useful. As stated in the introduction the equipment used, methods, and write-up are aimed at enthusiasts. I am interested in your feedback on errors, omissions, and areas for improvement. Please don’t hesitate to comment below.
Links:
TPSM84A21 SWIFT Power Module Product Page
TI Training: How to Test Power Supplies
Texas Instruments Webinars on Element14: Do More with Power Modules
Additions and Revisions:
18 Feb 2018: Replaced Efficiency Measurement Schematic with easier to read version, replaced external clock photograph, added sub-headings
Top Comments
Frank,
Great job on the roadtest/review of this product. You made a plan, collected data and presented results in a clear manner.
Any plans on what to do next? Is this module going to make it into a project…
Hi Frank,
Great, detailed review!! I'm not sure what we'd do without TI EVMs.. they really are excellent.
Excellent review and detailed description of the tests you carried out. Thanks for posting.
Kind regards