DC/DC efficiency measurement with the E36312A supply

or how to save your multimeters

The Incident

At some point, Keysight announced a contest called "Power up your bench!". The rules were simple:
Submit an idea or story how this new E36312A Bench Power Supply would enhance you lab, or how you would use it for some task and which benefits you get from it. Then, with some luck, get a free E36312A power supply. Guess what happened to me?

box from Keysight

And what's inside this box?

Keysight E36312A

No doubt, I'm a lucky winner of this contest!

The story

So this is the story I submitted to the contest. You may notice I've read the E36312A datasheet before writing this up.

Go for it

Now, after getting familiar with all the menus and controls of this power supply, I started to evaluate one of the DC/DC converter prototypes for its efficiency. As I wrote in my story, the goal is to use many of the features of this power supply to replace other lab instruments. I managed to measure the Efficiency over Output Power Diagram using an external load resistor and the bench power supply. Nothing else.

No multimeters

So the test setup is like this:
  • Connect the input of the DUT to Channel 2 of the supply, using 4W sensing
  • Connect the output of the DUT to Channel 1 of the supply, using 4W sensing
  • Connect the external load resistor to Channel 1 output, effectively in parallel to the DUT's output

  • The Device Under Test is a small prototype of a step-down (buck) regulator. I've used some screw terminals matching the connectors at back side of the power supply to connect it in 4W sensing mode:
    > prototype buck regulator   prototype buck regulator

    The load resistor is a somewhat larger 4R7 resistor:
    external load resistor

    The whole test setup looks like this: I've connected the load resistor using the front panel binding posts, as one would expect, these just wired in parallel to the back panel connectors. The load resistor is mounted to the heat sink right to the supply. The back panel connectors are used to connect to the DUT, using 4 wire sensing mode.
    Test Setup with E36312A and DUT

    How is this intended to work then?

    The power supply channel 2 is used to supply the DUT with a suitable voltage. Remote sensing is used to get most accurate results.

    Channel 1 is used as a variable load for the DUT. This works by connecting the load resistor in parallel to the DUT's output. At a constant voltage (this is assumed for the regulator's output), the resistor sinks a constant current. To measure the output voltage of the DUT under load, the channel 1 output is set to the lowest possible current and a slightly higher voltage than the DUT's nominal output voltage.

    Turn on both channels of the bench power supply. channel 2 gives the reading of the DUT's input voltage / current / power, while channel 1 gives the reading of the DUT's output voltage. A small current (about 2.5mA here) flows from the power supply into the load resistor, slightly reducing the output current of the DUT. At about 0.9A output current, 2.5mA doesn't make a big difference. In this configuration one can get a direct readout of the DUT's input power at the load determined by the load resistor.

    By slowly increasing channel 1's current setting, the power supply brings more current into the load resistor, reducing the DUT's load current by the same amount. At some point, the bench power supply has taken over the full load, the DUT's output current is zero. Using this technique, one can get an input power reading from the bench power supply at any load between zero to full load of the DUT.

    If the current setting is above the full load current, the output voltage starts to rise. To prevent it rising too high, the voltage setting is just above the DUT's nominal output voltage. Be careful with that, some kind of regulators do not like this "feeding from the back". Especially regulators with synchronous rectifiers can start pushing power from their output to the input now, this is absolutely undesireable here. So be careful to set the output current of channel 1 not higher than the calculated maximum load current.

    Semi automatic measurement

    And now for the second goal: Not to have to write down all the values and manually adjust all the current limiting settings.

    First, determine the minimum load current set point:
    Slowly increase the channel 1 current limit setting, until the bench power supply takes over the full load. The DUT's output current is then near zero, the input current is low.
    Minimum load at DUT
    Remember the load current (0.925A).

    Now, use the "Output LIST" feature of the E36312A to set up a list of current values, starting at this current down to smallest possible value.
    List of current steps   List of current steps   Manual trigger for LIST
    I've made smaller steps at the beginning, and larger steps towards the end. This gives some finer resolution at small output power. Note the trigger "BOST" at the first step.

    Now, set up the digital inputs and outputs to route the trigger from the Output LIST to the Data Logger:
    Trigger input and output setting

    Setup the data logger:
    Data logger setup
    Using this setup and the 3 seconds dwell time, we get 3 samples per current step. Typically, the first or the last of these steps is invalid due to the LIST stepping, so we should use the middle sample of each step for the later analysis.

    Use some wire to connect Pin 1 and Pin 2 of the Digital Port, located at the back side of the power supply. This routes the trigger from the LIST execution to the data logger to provide a means of simultaneous starting both.

    Put a USB thumb drive into the front panel USB connector. This is where the logged data gets saved to.

    Assuming your DUT is still hooked up, start the LIST execution and watch the readings while the power supply runs through all the programmed steps. When finished, the data logger writes a file containing binary data to the thumb drive. Use "Export file" from the data logger's File selection menu to export the data as a .csv (comma separated values) file. This should be readable by your favourite data analysis tool. This is how the exported file looks like: default.csv.

    Create a nice chart

    For this example, I've used LibreOffice Calc to import the data, do some calculations and finally get a nice chart.

    Do some spreadsheet magic

    First, we need to calculate the DUT's input and output power from the saved data. The data has Voltage and Current of each channel we used here. Input power is straightforward: Multiply Channel 2's voltage and current reading per row to get the power reading.

    To calculate the output power, we first need the actual DUT's output current, not the bench power supply output current. As the current into the load resistor is constant here, the DUT's output current is the difference of load resistor current and bench power supply output current. We've started our output LIST with this value, so the Sample "1" shows this current for Channel 1.

    For each sample, subtract the channel 1 current reading from the sample "1" ch 1 current reading and multiply by its output voltage to get the DUT current.

    Efficiency is now calculated by dividing output power by input power

    Charting

    Now use the Chart tool to create a simple X/Y chart of the calculated "Output Power" and "Efficiency" columns. The result looks like this:

    Chart

    There's some spiking in the diagram, these are the invalid samples I mentioned above, but yout get the figure. Since I don't know a simple way to remove all unused samples from the data using LibreOffice, I left them for now. For sure there's someone in this world who knows a simple trick to do this with LibreOffice.

    Here's the spreadsheet for those playing along at home: Efficiency Chart Spreadsheet.

    Don't treat this as a precision measurement, especially at low DUT output currents, there might be a large error due to the difference current calculation method. For larger currents, the result looks quite plausible but still to be verified by the more conventional method using a bunch of multimeters.

    Table of Contents ... und ein Zaehlpixel hab ich auch :-)