RoadTest: Vishay Synchronous Buck Regulator EV Board
Author: hlipka
Creation date:
Evaluation Type: Development Boards & Tools
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?: There is no part from another manufacturer matching both the current capability and the voltage range - the Vishay 476 series comes close. Other notable part: Microchips MIC28517 (70V, 8A).
What were the biggest problems encountered?: Terminal blocks are barely large enough, and documentation for the particular EVB is hard to find.
Detailed Review:
A power supply is something every electronics design need. If you are lucky then you can just use a USB charger - 5V at 2A are plenty for many applications. When you need higher voltages or higher currents, its getting interesting. In the past I had the chance to review several switch mode converter eval boards, each with different capabilities. But as Gough Lui already pointed out in his market survey, there is currently nothing coming close to the SiC46x 8and SiC47x) series from Vishay - they can deliver up to 10A from up to 60V input, and do this over an output voltage range from less than 1V up to 55V. So when you have something needing high power, and you need to supply it from a higher voltage (e.g. 48V telecoms systems, or multiple LiPo battery cells in series) now you have something from the shelf instead of needing to design your own supply.
The board comes nicely packaged, with several pages of user manual for the board. That's nice since you don't need to go searching for it, then. The board itself is rather small for its power capability, especially when you compare it with the other ones I have seen in the past:
Here the SiC461EVB in all its glory:
The biggest part is not the IC itself, but the coil. The latter one is also responsible for the board being quite heavy - there must be quite some copper in it. Notice also the very heavy via stitching for all of the power planes - its needed so you can handle 10A at such a board (the other boards solve this by using very wide traces instead).
What you also might notice is that the terminal block looks quite small when compared to the other boards. Its in fact so small that I was wondering whether its rather for 10A at all. While the supplied user manual comes with a schematic, its missing a BOM. And all of the other versions of the user manual are for a "low power" version of the board which is not designed for 10A. After several tries I finally managed to dig up a file which had the complete design data for all version of the EVB, which included the BOM. So the terminal block is a Amphenol 20020327-D041B01LF (data sheet appended, as well as the schematic and the BOM), and its indeed rated for up to 10A maximum. It does just fit solid 1.5mm2 wires, which is the minimum you need to use for 10A (2.5mm2 would be better, but such wires don't fit). It would have liked to see better suited connectors here. (Since the terminal blocks are too small to hold the wires for my larger loads I needed to build banana-socket-to-solid-wire adapter).
Apart from that I like the board - its quite small, has mounting holes (though no standoffs provided) and the connectors are placed all on one side. When you want to use it in a project, there won't be any problems.
But for testing and evaluation purposes its missing some proper test points. There are just some vias spread out over the board. And only the one for the output voltage is actually suited to be used with a scope probe, the other ones are spaced to far apart. In the end found some wire wrap pins which did fit into the vias without soldering, and used these. (According to the BOM proper test points should be fitted there)
When you are testing power supplies (esp. switch mode ones), there a different characteristics you are interested in:
I originally planned to look at all of this. But when I read Gough Lui's review, with all its beautiful figures and the glory details, I decided that I neither can match that nor provide additional information. And concentrated on the rest.
Most difficult lesson I learned with my past power supply road tests: measuring signal quality is not easy, and probing really matters. When dealing with high frequency signal with short rise and fall times, do never ever you the ground lead of the probe. Always use the ground spring to have a connection as short as possible (that's why the test points do matter so much). Or, and that's what I did: solder the scope probe cable to the board. I have a BNC cable just for that purpose, and I soldered it to the two pins directly beneath the coil (sorry, forgot to take a picture
Its also best to use the 50 ohm termination of your scope (to match the cable impedance), or to use a separate 50 ohm terminator. After I nearly burned my fingers with mine (it can get quite hot - 50 ohm at 12V are nearly 3W) I decided to only take parts of the measurement that way and to just verify how much influence it has. All measurements where made with the cable soldered to the board in 1:1 mode (which limits the usable bandwidth, but that would be limited to 20MHz anyway).
Last but not least one needs to have a power source which does not introduce any unwanted noise or ripple into the measurements. So when you use a lab power supply, make sure its noise and ripple specs are better than the device you are examining! Fortunately I had a transformer around which can supply 24V at 10A, so I mounted it on a small wooden plate, added an active rectifier (a 10A rectifier bride usually needs cooling as it will dissipate around 14W then) and 10mF of capacitance.
{gallery:width=800,height=480} 24V power supply |
---|
24V transformer with active rectifier |
Active rectifier |
That gives a quite clean input voltage, with a 100Hz ripple with about 2.5V peak-to-peak at full load.
{gallery:width=800,height=480} 24V input voltage under load |
---|
Input voltage |
Input voltage, AC-coupled |
For all tests involving a range of different input voltages I could not use the transformer, but used my Peaktech 6181 lab supply (which deliver 60V at 6A in series mode or 30M at 12A in parallel mode). I assumes that the input signal is a bit more unclean than from the transformer.Do not forget to turn of any kind of switch mode supplies (and LED lamps) nearby, as they can easily disturb your measurements. As a reference, I took a scope shoot of the scope connected to the EVB, but without input voltage. Looks quite clean:
When I turn on my Anker switch mode charger on my desk (without anything connected), you can see how the signal gets disturbed (note the changed timescale):
Its not much, but usually I would also have my computer running, my lighting would be LED lights (used an old incandescent desk lamp instead), there are phones nearby, and so on. When you are interested in the details of the output voltage, you do want to see only the noise generated by the switch mode converter you are looking at, not the other ones in your room.
When looking at the signal quality, there are three interesting modes the SiC461 can run in:
Under significant load these should behave the same, but they differ in how they work in a low-load scenario. I will explain this in more detail below. I skipped modes 3 and 4 - they behave the same as 2 and 1, but disable the internal regulator to reduce power consumption.
Generally I tested the different modes at different loads:
(I planned to use my new electronic load for most of the tests, but I did not finish its software in time, and the temporary reference voltage I used is has some large ripple on it, so its not really usable for now).
The I looked with the scope to check if there is any low-frequency ripple (e.g. 100Hz from the input), and from there zoom in until we can see the individual switching cycles. This allows to check the switching frequency, and the amplitude of the switching artifacts at the different frequencies.
Mode 1 is a power-saving mode. This means the SiC461 can go into 'discontinuous conduction mode' - which in turn means the current through the inductor can fall to zero. This happens only at low load currents, and since then the frequency can be lowered further, and no current is flowing when its not needed, this mode can be more efficient than in 'regular' mode.
{gallery:width=800,height=480} Mode 1 output signal |
---|
1k load |
140ohm load |
140ohm load, zoomed in |
10ohm load |
10ohm, zoomed out |
5A load |
5A load zoomed in |
5A load, further zoomed in |
10A load |
10A load, zoomed in |
10A load, switch artifact |
So what can we see? First, the output ripple itself does not change (at least not significantly) with rising load - its about 70 to 80mV peak-to-peak. We also can see that the output frequency for higher loads stays constant (its configured to about 300kHz on the EVB). But it falls with lower loads, which can be seen nicely on the 1k load: the output rises sharply whenever some switching occurs, and then falls slowly down. On the 1A load I added a zoomed-out scope shot, where the two control loops of the SiC461 can be seen. One is for the steady state and is slower - this is the super-imposed wave. And then there is a feed forward loop which handles fast transient responses.
I also looked at how the output ripple changes when the input voltage changes. I varied the input voltage from 15 up to 60V.
{gallery:width=800,height=480} Output ripple vs. Input voltage |
---|
1A load, 15V input |
1A load, 20V input |
1A load, 25V input |
1A load, 30V input |
1A load, 40V input |
1A load, 60V input |
10A load, 15V input |
10A load, 20V input |
10A load, 25V input |
10A load, 30V input |
10A load, 40V input |
10A load, 60V input |
Whats notable is that the ripple depends quite a lot on the input voltage - higher input voltage means higher ripple. And also the switching artifacts can be quite big with higher input voltages.
In DCM the frequency can be very low - in the single kHz range or even lower. Depending on your load and the chosen inductor, this can result in audible noise from the coil. When this happens, and is unwanted (because its e.g. a consumer device) the SiC461 can be set to 'ultrasonic mode'. The SiC461 is then forced to use a minimum switching frequency greater than 20kHz to avoid audible noise.
To compare ultrasonic mode with non-ultrasonic mode, I just looked at the low-load scenario (so no load, 1k and 140 ohm). All tests are run with 24V input.
{gallery:width=800,height=480} Compare Ultrasonic mode on and off |
---|
Ultrasonic mode off, no load |
Ultrasonic mode on, no load |
Ultrasonic mode off, 1k load |
Ultrasonic mode on, 1k load |
Ultrasonic mode off, 140ohm load |
Ultrasonic mode on, 140ohm load |
We can very nicely see how low the switching frequency drops in power-save mode with low loads, and how it stays above 23kHz with ultrasonic mode enabled. We can also see that at 80mA (the 140ohm load) the switching frequency is already 15kHz. So it seems that at 100 to 150mA it does not make any difference (but this will depend on the input voltage - something that could be investigated further).
In mode 2 the power-save mode is disabled and the SiC461 operates in continuous conduction mode. This is less efficient at low loads 8since there is always some current flowing through the inductor, and there is always switching), but should should result in faster transient response (since there is always some switching the output is much tighter controlled) and also less ripple.
{gallery:width=800,height=480} Mode 2 measurements |
---|
No load |
No load zoomed in |
1k load |
1k load zoomed in |
140ohm load |
1A load |
1A load zoomed in |
10A load |
10 load zoomed in |
We can clearly see the constant switching frequency over all load ranges (although the scope at some shots got confused, the zoomed in versions are better in that regard).
Transient response means how fast and clean the power supply can react to changes in the output current. Modern devices, such as FPGAs or CPUs can switch their power modes with microseconds (or even faster) and its important that the output voltage is stable under such conditions. My electronic load could create current ramps down to about 5 microseconds, but to really stress-test the SiC461 I used by load step generator. Its basically just a PWM generator with an attached power FET (a PSMN1R5-40) which can create load transients down to 10ns rise time (maybe a little bit slower given the length of my test leads). I tested load steps from 10mA to 1A, 10mA to 5A, and 1A to 5A (all with 20V and 60V input). Note that the images vary quite a bit in their vertical and their time scale. Input 2 (green) is the trigger from the load generator.
{gallery:width=800,height=480} Transient load response (rising) |
---|
load step 10mA-1A, 20V input |
load step 10mA-1A, 60V input |
load step 10mA-1A, 20V input, zoomed in |
load step 10mA-1A, 20V input, further zoomed in |
load step 10mA-5A, 20V input |
load step 10mA-5A, 60V input |
load step 10mA-5A, 60V input, zoomed in |
load step 10mA-5A, 60V input, further zoomed in |
load step 1A-6A, 20V input |
load step 1A-6A, 20V input, zoomed in |
load step 1A-6A, 60V input |
load step 1A-6A, 60V input, zoomed in |
Especially on the zoomed in scope shots one can see the very fast response - basically after the first switching cycle (which is about 3µs) the output voltage rises again. But then on the zoomed out shots the second control loop kicks in a little bit later (in these shots the first transient response can only be see as a very small blip downwards). So there is a very short (less than 50µs) undershoot of about 100mV, maybe 150mV when the load kicks in at the wrong moment.
I also looked at the falling transient response:
{gallery:width=800,height=480} Transient response (falling) |
---|
load step 6A-1A |
load step 5A-10mA |
Interestingly the response to a falling load current is worse - it can easily reach 150mV and also seem to take longer to be regulated (which, to a certain extend, is due to the output capacitance - it takes a while until the caps can be discharged with a very low load).
But all in all it behaves quite nice here (although the comparable TI SWIFT™ Power Module EVM - Review I tested earlier handles this even better). I would expect it to be even faster with a higher switching frequency (it could go up to 2MHz which would be 7 times faster). Its still not bad for a single-phase buck converter.
Voltage ramp-up means to look at the output voltage when input voltage is applied. There should be a clean curve for the output, and no (or only minimal) voltage overshoot - otherwise your precious electronics might be destroyed. Since the SiC461 also has a 'power good' signal, you want to know how it behaves. Its usually used to enable downstream regulators so they turn on only when their input voltage is stable, to avoid any fluctuations.
So I tested to ramp-up into a 1A an 5A load, with 20V and 60V input (and into no load as comparison). The green signal is the power-good signal. I also tried to zoom in to see the overshoot at the output.
{gallery:width=800,height=480} Output on startup |
---|
Startup into no load, 20V |
Startup into 1A, 20V |
Startup into 1A, 20V, zoomed in |
Startup into 1A, 60V |
Startup into 1A, 60V, zoomed in |
Startup into 5A, 20V |
Startup into 5A, 60V |
So we can see that the overshoot at the output is very low, I'd say maybe 100mV or so. This would need some more detailed testing to capture exactly this signal.
What I found interesting is that there seems to be no startup control or slow-start. The higher the input voltage the faster the output voltage ramps up, and the load does not seem to matter at all (compare the startup, into no load vs., into 5A - both take about 40ms for the full ramp-up).
I just fixed some thermocouples to the SiC461 itself and to the inductor, and let it run for 10 minutes at 10A load. Interestingly I did not measure temperatures as high as the datasheets suggest (and as Lui did see during his tests). For 24V input I measured the SiC461 at 73°C and the coil at 65°C. For 60V input the SiC461 ran at 85°C, and the coil at 74°C (with an ambient of 25°C). Maybe my desk is a little bit colder...
The last part to look at are how the emissions (and ripple and noise) spread out over the frequency. So first I did a re-run of my mode 1 tests, but this time I turned on the FFT function to get to see some spectrum information:
{gallery:width=800,height=480} FFT - spectrum evaluation |
---|
FFT: 140ohm load, ultrasonic mode on |
FFT: 140ohm load, ultrasonic mode on |
FFT: 140ohm load, ultrasonic mode off |
FFT: 140ohm load, ultrasonic mode off |
FFT: 1A load |
FFT: 1A load |
In hindsight using a 1k load for the ultrasonic mode on-off comparison would have been better, with a 80mA load the difference is just not big enough. But we can clearly see that higher loads move the noise to higher frequencies, and also to higher levels.
For EMC tests I used my DIY probe (build after Dave Jones' explanations):
This is the smallest probe I have, and it fits just around the coil. I probed around the board to see what happens. And while I cannot actually quantify whether the emissions are good or bad levels (I would need to compare with other supplies and boards), it still looks quite interesting:
{gallery:width=800,height=480} EMC probe results |
---|
EMC results, 5A load, at the inductor |
EMC results, 5A load, at the SiC461 |
EMC results, 5A load, at the input caps |
EMC results, 5A load, at the output caps |
EMC results, 5A load, at the terminal blocks |
EMC results, no load, at the inductor |
EMC results, no load, at the Sic461 |
(All tests where run at 24V input, so there should not be any noise coming from the power input).
Its interesting to see that some of the noise gets coupled backwards to the input capacitors - I had seen already when looking at the quality of the power input. When this voltage is used for something else, there should be additional filtering of the inputs.
So, all in all I like this board, and I like the SiC461. When you are in the need for its capabilities it can save you a lot of headaches (and time). It performed quite well, although not spectacularly (in nearly all aspects there are better switch mode converters available, but will have the same power delivery capacity). But since I think the Sic461 is more intended to power a complete sub system where there are additional downstream regulators and additional filtering, its par for the course.
The evaluation board itself leaves mixed feelings. Its rather small when compared to other ones, and still performs quite good. The lack of proper test points makes testing more difficult than it should - and then such an evaluation is its main purpose, isn't it? The use of a just properly rated terminal block does not increase confidence - for a board which is 'high power' by designation this should not happen. It should also not be complicated to find the proper documentation for the board.
So for the board its a 'better than average, but nothing to write home about', while for the SiC461 its a 'when you need it its really great'.
Top Comments
Hi colporteur - seeing as I was addressed, I think I should probably chime in even though this is hlipka's review to address your concerns.
On the whole, I'd say the board does have sufficient test…
I did link your name only because I didn't want to reference a comment you made without you being aware of it. I may attribute a comment to you that is not correct and feel it is only right you have a…
I'm confused. Not with your review but rather the issue about test points both you and Gough Lui made. What is a reasonable expectation for a development board? Poor soldering, minimal components,…