Table of contents

RoadTest: Enroll to Review the Rohm Buck Converter Eval Kit BD9G500EFJ-EVK-001

Author: hlipka

Creation date: 24 Dec 2023

Evaluation Type: Evaluation Boards

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?: Vishay SiC461 series, Microchip MIC28516/17

What were the biggest problems encountered?: The boards connectors are not optimal, and there are not test points available.

Detailed Review:

Supplying power to a might not be the sexy part of your project, but it will be the backbone of it. Doing it wrong can ruin everything. Maybe that's the reason I looking into these power supply board road tests so much. And I would like to send a huge Thank You! to both Rohm Semi and Element14 for giving me the chance to participate here.

A first look

Not only was the evaluation board nicely packaged - I also found another three bare ICs in my box. Its the first time this was done in one of my road tests, and it was a nice and welcome surprise (even though I did not find the time to build my own PCB to use them).

Upon looking through the material, I realized for the first time that Rohm Semiconductor is a Japanese company (I'm currently learning Japanese, and I have progressed far enough to at least be able to read the Hiragana and Katakana characters - so I understand maybe 10% of the Japanese leaflet...). This might also be an explanation for the one unusual thing I found with the board right away: its layed out with the inputs to the right, and the outputs to the left side. This is quite unusual - all of the other board I saw so far either have everything on one side (using just one big connector) or have the inputs on the left side. So be careful when working with this board - its easy to go with the style you are used to and connect the power to the wrong side...

That said, the board itself is very nice:

The layout flows very nicely through the board, so I think it should be quite easy to design the BD9G500 into your own PCB. It also does not use that much board space - the biggest parts are still the inductor and the diode (esp. when you don't need a higher voltage capacitor on the output). The evaluation board is much bigger than it needs to be (see the comparion to the other board above). But then it uses the space for a big ground plane, which together with quite a lot of via stitching serves as heat sink. There are some traces on the bottom side though which seem as if they should be routed differently (I think they split the ground plane in the wrong places).

Another issue I discovered when running my tests: there are no terminal blocks for the power connections, and the connector pins are too close to each other. Not having terminal blocks means you cannot screw in larger wires (for 5A I would use at least AWG 16, or 1.25mm2). And with the close spacing of the pin connections one cannot use proper crocodile clamps. So the only real option is to either accept higher connection resistance (and use small clamps) or solder the wires to the pins - which makes quick changes to the setup more difficult.

There are also no further test points to more easily connect a scope (e.g. to measure switch wave forms, or do ripple and noise measurements). So either one need to live with connecting the scope to the output pins, or build your own probe from a coax cable and solder it to the board (which I did for some measurements).

Transient response

The first test scenario was looking at the handling of load changes - after all one of the BD9G500 features is a fast transient response.

The startup behavior is rather unremarkable: the BD9G500 starts up as soon as the input reaches 7V, and then ramps up the output in about 20ms to its 5V output. There are no visible overshoots, no matter whether there is a load attached or not. One thigh I realized, though, is that its not possible to reach full 5A load when running with just 7V input. I guess the losses in the wires and the PCB (and the connection made) are large enough to drop below the voltage the BD9G500 needs for proper regulation.

For the transient response itself, I used the MOSFET-driven load simulator I build several years ago:

which can create load transients down to about 10ns rise time (probably a bit slower because of the test leads). I created a 30mA base load, and tested load steps up to either 2A or 5A, with varying input voltages (from 9V up to 30V). For these, I looked at the undershoot and time-to-recovery when the load was turned on, and the overshoot when it was turned on.

For the measurements, I connected the scope to the sense outputs of the board. That way, they do not get to see the voltage drops from the PCB or the connectors.

For a 2A load step at 12V input, , the measured drop was about 300mV for about 20µs, and then it recovered to a drop of about 50mV, after about 1ms. With higher input voltage the recovery was a bit faster - at 30V the initial drop was done after 10µs, and it was only 200mV. With 9V input, the recovery was a bit slower.

For a 5A load step, the timing was about the same. But the initial drop was 500mV now, and the recovery ended up with a drop of about 100mV now.

The turn-off behavior was about the same in both scenarios. There is a very short overshoot (less than 1µs) followed by some ringing - parts of this are most likely due to the non-ideal measurements connections here. So the overshoot measured (somewhere between 0.6V and 1V) is probably not as high - during the ringing itself its about 200 to 400mV. Time to actual recovery is again about 1ms.

The BD9G500EFJ Eval Kit is configured with 200kH switching frequency - so basically regulation kicks in with the first switching cycle after the load step. This is quite fast, and what I had expected from a buck regulators claiming a fast response (to be faster, it would need higher switching frequencies, which can create other kinds of issues to deal with). But the voltage drops I see are higher than my expectations, and also higher than the claims in the data sheet. It would need further investigations (and maybe even my own board) to find out why this happens. OTOH - these drops are not large enough to cause problems - such big load steps are uncommon, and can be rectified by local bulk capacitors in the circuit.

Ripple and noise

To properly test ripple and noise, I connected the scope to the output cap of the board. Due to the lack of dedicated test pins, I soldered a coax cable directly to the cap, and terminated the cable with 50 ohm at the scope:

But first I did some reference measurements with the scope probe connected as before (with no load connected, and with a 2A load, all with a 12V input):

As you can see, the signal looks quite noisy here. Compare this to the output waveform with the probe directly connected:

Although the ripple voltage is nearly the same, the noise is nearly gone. To a big part this is because the scope creates a big inductance with its ground connector, and with it longs test leads, and to another part because these long leads also pick up other noise in my work area.

Here the full test series, also over a range of input voltages (they were created as part of the efficiency test in the next section, see there for more details):

It was interesting to see how the ripple waveform changes with higher inputs. My guess is that the higher input voltage also means higher current flowing into the inductor, so the switching behavior changes and the behavior of the inductor change.

Efficiency

For efficiency measurements, I did a step-trough over a a wide range of input voltages and output currents. By hooking up two DMMs to input and output as well I could then observe the real input and out values, in addition to the actual currents.All of the instruments are driven via SCPI, and I did write a Python program to run the test scenario and collect the data.

From these I can get the input and output power, and calculate the real efficiency. In addition, I hooked up the scope as well and measured the output ripple over these load scenarios.

The result quite well with what Rohm states in the data sheet: the efficiency is usually above 80%, with the exception of lower loads (below 1A) and high input voltages (above 40V or so). But esp. for high loads the efficiency stays above 80%. This is not the best in class, but still reasonable.

In this test I did drive the BD9G500 over its stated current limit - up to 5.6A. For higher input voltages this was OK, but for lower ones it isn't. This probably has two different reasons: one is that the voltage losses on the input side are to high (driving the BD9G500 into shutdown), and that the switching current get high enough to reach the current limiter (which has a similar effect).

The ripple voltage is mostly constant through the parameters, but it is bit higher in the areas where the efficiency is lower.

I did also measure the temperatures while applying a 5A load over different input voltages. The BD9G500 stays quite cool in all scenarios (the most I did measure was around 60°C), but the diode reached around 74°C with an input voltage of 30V. This is because the BD9G500 is not a synchronous buck converter - when it would replace the diode with an additional FET these losses could be reduces significantly (which would also improve the efficiency). For the BD9G500, the cooling is also supported by the large ground plane to which it is connected via its thermal pad. SO in a denser layout higher temperatures should be expected.

Basic EMC tests

I do not own specific EMC test gear (ones which will give you absolute values), but just a DIY EMC probe which can be used for comparative tests (see Dave Jones' video about building one). I did use the same setup in a previous road test (the one for the SiC461), so I had at least something to compare against.

When sweeping the probe over the board, the only place where any radiated emission could be seen was directly on the inductor.

The important difference is the spike right at the left side, when measuring over the inductor

Especially it seems nothing gets coupled into the input or output capacitors and wires - that was an issue with the SiC461 board. The data sheet does not make any particular claims about reducing EMC with the BD9G500, so it seems that the PCB layout is responsible for most of that (and at least it proves that a good EMC result can be achieved).

Conclusion

So where does this leave us? As a chip itself, I very much like the BD9G500. It seems easy to be used (and esp. its SOIC package make PCB design easier for hobbyists like me), and provides excellent results. Thanks to Rohm sponsoring some additional chips I surely will design them into my projects soon. But for the evaluation board itself disappoints a bit. Its fine when you just want to use it in a project, where you just add it to the rest of the circuitry and check how it behaves. But doing more detailed experiments and measurements are harder than they should be, due to the lack of dedicated test points, and its sub-optimal connectors.