• Author Author: Jan Cumps
  • Date Created: Date Created
  • Views 2950 views
  • Likes 7 likes
  • Comments 20 comments
Related
Recommended

TI SWIFT™︎ Power Module EVM RoadTest: Module Review and Test Setup

Jan Cumps

I'm reviewing a single component DC converter for the TI SWIFTTm Power Module EVM RoadTest.

This article is a review of the switching device and the test setup.

 

Feel free to comment with the tests you'd like to see done with the module.

image

 

 

The Switching Device TPSM84A21

 

The TPSM84A21TPSM84A21 is a hybrid DC to DC Buck converter. All electronics to make it switch are in the package. It doesn't require external inductors or capacitors to work.

In the simplest setup, no other component is needed. The unit will then switch 8 - 14 V DC to 0.508 V. Maximum current of the output is 10 A.

image

 

When you want to set the output to another value between 0.508 V to 1.35 V,  the only external component is a voltage selector resistor (RSET).

This component is not involved in power conversion and can be a low power small one.

image

For ripple reduction, the module has input and output capacitors on board.

If your application is OK with the ripple figures of the device (documented in the datasheet), you don't need external capacitors.

If you need better filtering, you can add input and output capacitors. Guidelines are in that datasheet.

image

 

The device has additional functionality such as Power Good, under-voltage lockout and a possibility to externally synchronise the 4 MHz switching frequency.

Check this functional block diagram for the possibilities.

image

 

The Evaluation Module

 

As expected with hybrid componenst, there's not much more on the evaluation board than the  TPSM84A21TPSM84A21.

The only passives are a set of resistors to select different outputs, and to test the device enable, current limit, sync and power good.

 

image

The board gives you good access to all the interesting measure points. It also has footprints for additional capacitors if you want to test such a setup.

For the input and output rails, there are test points very close to the device. They are intended to be probed with an oscilloscope probe with ground spring.

This doesn't work that well with the probes of the Tektronix and Rigol scopes I have.

Their pins are thinker and the distance between center pin and ground spring wider than what the measure point on the PCB takes.

All other test points are solder-on loops that allow easy access with meter test leads or oscilloscope probes.

 

image

There's an empty footprint for a coax connector to sync the switcher's clock externally.

 

Test Setup

 

The EVM is connected to an electronic load that draw a current of up to 7.5 A. It can log the current and voltage measured.

The load is connected to the EVM with short thick wires. With these low voltages and high currents, resistance of the wire matters.

I'm using a linear power supply that can deliver more than enough power to feed the EVM at maximum load.

image

For measuring, I have a few DMMs and a 4 channel oscilloscope.

 

The PSU, load and oscilloscope are programmable. I can measure and log info in scenarios where the input voltage and output loads change over time.

Things I can automatically log:

  • Current at EVM input and output
  • Voltage at the EVM output
  • Any 4 points that I probe with the scope.

 

Let me know in the comments below if you're interested in particular load operating conditions o the EVM.

I will try the ones that I can perform with the above setup without damaging one of the components.

 

Edit: requests received:

 

 

Related blog
TI SWIFTTm Power Module EVM RoadTest: Module Review and Test Setup
TI SWIFTTm Power Module EVM RoadTest: Noise Measurements
TI SWIFTTm Power Module EVM RoadTest: Efficiency
  • in reply to jc2048

    It takes a while to do this, because I put 2 second pauses between setting and measuring - for stability and because there's a gap of 1 second between sample and availability of data in the load.

    But everything is automatic, yes, from initialising the two instruments to the ramp for input voltage and output loads, and the logging into a spreadsheet. 900-ish measurements.

    The calculation of the efficiency and the graphs were done by me..

    I've also noticed the irregularities around 2.3 A load and the behaviour at the end of the curve. I have to spend some quality time to work on the calibrations. Nowhere near a reasonable precision at this point.
    Stability is good, but set and measure aren't anywhere near the precision I want them to be.

     

    For the higher currents, the resulting dissipation may be giving your load some measurement problems. One way to find out would be to give it a 'cool off' period between readings and see if the resulting curves diverged from these. (Just a thought, if you wanted to experiment further and had the time.)

     

    No, it's consistent. It always shows this poor behaviour at that range.

     

    I'm getting to like the idea of SCPI. Perhaps we should have a SCPI Project 14 contest and invite people to SCPI-fy an item of non-automated test gear.

     

    The library I'm using also works on a BeagleBoard (I tested that) and I'm sure it 'll work on a Raspberry Pi too.

    I've also made some devices with that lib on the Hercules family of microcontrollers.

    With some optimisations it even runs on an Arduino UNO.

  • in reply to Jan Cumps

    Wow! Presumably you just set it going and it did the rest for you.

     

    How long does it take to work through them?

     

    Where did the odd wrong readings come from (they are curiously similar in form)? Is that something not settling quickly enough?

     

    For the higher currents, the resulting dissipation may be giving your load some measurement problems. One way to find out would be to give it a 'cool off' period between readings and see if the resulting curves diverged from these. (Just a thought, if you wanted to experiment further and had the time.)

     

    I'm getting to like the idea of SCPI. Perhaps we should have a SCPI Project 14 contest and invite people to SCPI-fy an item of non-automated test gear.

  • in reply to Jan Cumps

    Graph of efficiency with output set to 0.6 V, input between 8 and 13.5 v

     

    image

  • in reply to Jan Cumps

    Graph of efficiency with 8 V input, as logged by LabVIEW automated run.

    image

    Not calibrated, but close to usable.

  • in reply to Jan Cumps

    image

    I've made a test run with spreadsheet write. Not all info is in there. I'm missing enough info to calculate efficiency but want to give it a go.

    Here are the doing a run from 8 to 12 V, stopped at 10 V:

     

    Here's a snippet (yellow is manual manipulation done by me):

    image