|Product Performed to Expectations:||9|
|Specifications were sufficient to design with:||10|
|Demo Software was of good quality:||8|
|Product was easy to use:||9|
|Support materials were available:||10|
|The price to performance ratio was good:||10|
|TotalScore:||56 / 60|
Vishay SiC461EVB microBUCK Reference Design Evaluation Board
10A Synchronous Buck Regulator, 12V Output
By Gough Lui – August - October 2020
Power conversion is literally everywhere. From your smartphone, to your PC, to wall-wart power supply, switch-mode conversion has become the method of choice owing to its high electrical efficiency and mandates on energy efficiency. Switch-mode supplies used to be rather big and complex beasts, with many discrete components occupying not an insignificant amount of space. However, over the years, a number of semiconductor companies have been building highly-integrated controllers which simplify the task of creating a switch-mode DC-DC converter by reducing the necessary supporting component count and footprint through high-integration; and offering ever-increasing levels of safety protections and conversion efficiency.
Not being an electronics genius, the design of switch-mode supplies has always seemed to be to be a dark art. However, as someone who is very much interested in power conversion, possessing the equipment and inclination to explore, I wanted to see just how well Vishay’s SiC461 Synchronous Buck Converter performs given its claim of 4.5-60V input, 10A current output and over 98% peak efficiency – all of which are highly attractive specifications.
This review is with thanks to element14 and Vishay for selecting me as a RoadTester. I hope you find this review interesting, informative and useful. Please feel free to leave a like, comment, rating, bookmark or share this with others who may be interested – it’s been a lot of work (and a bit of fun) to deliver yet another one of my “trademark” reviews.
I will begin this review with a market survey, as I usually do, to try and understand how the Vishay SiC461 stacks up to some of the other competitors on the market. In order to do this, I have visited element14’s Australian website, searching for synchronous buck converter ICs that have a 10A current rating, single output and integrated MOSFETs.
These parameters were chosen to closely mirror what the SiC461 is capable of and to represent pseudo-equivalent products which designers may opt for. The reasoning behind this is that synchronous operation usually implies higher efficiency than free-wheel diode operation, so a designer looking for a synchronous buck converter is unlikely to want a non-synchronous converter and would rather pay a slight premium for a more sophisticated controller (although this is a possibility at the cost of efficiency). Likewise, the type of converter being a buck-converter implies reducing the input voltage and the current capability of up to 10A with a single output are key performance parameters which makes the other converters equivalent on a technical performance basis. Finally, the choice to opt for a converter with integrated MOSFETs purely lies in the reduction in complexity and footprint requirements on the board, as someone intending to use the SiC461 is likely to be drawn to the all-integrated nature of the IC which reduces component selection hassles and keeps things relatively compact.
With this in mind, a total of 17 contenders were found, from manufacturers including Microchip, Monolithic Power Systems (MPS), ON Semiconductor, Richtek, Texas Instruments and Vishay Siliconix tabulated below. Please keep in mind the disclaimer – the information is based on data extracted from manufacturer datasheets and does not consider or compare actual performance, all variables, operational differences and does not cover all products in the market. Prospective purchasers must do their own independent research to verify the suitability of these components prior to purchase. Information has been compiled in good faith and no responsibility is accepted for any errors or omissions. Use at your own risk!
On the whole, the Vishay Siliconix SiC461 is a rather unique offering. It is priced somewhere slightly above the middle of the market, the unit offers a wide-range voltage input unmatched by others, especially with regards to maximum voltage input. The minimum voltage input does not go as low as some others and the minimum output voltage is equal-highest among the bunch at 0.8V, but perhaps this is not unexpected as it targets higher voltage inputs. The peak efficiency is also the highest of all of the tabulated converters with the widest switching frequency range. A full suite of protections is provided and light-load efficiency boosting features are also competitive with the competition. Perhaps the only downside is the MLP55-27L package which is a little larger (1-2mm) than some of the QFN types but overall, the size is average as there are even larger ICs from the competition in the table.
It would seem that the Vishay Siliconix SiC461 is a good choice with very few downsides. However, it occurred to me that the Vishay Siliconix SiC466 appears to be virtually identical in specification in the tabulated parameters but has lower standby currents and is listed slightly cheaper, so may actually be a better choice compared to the SiC461 overall, especially where quiescent power is a concern. There is also the Vishay Siliconix SiC476 which seems to have very similar specifications and a slightly lower price except for a lower maximum input voltage of 55V instead of 60V. Perhaps that represents an older design or perhaps the dies that didn’t “win” the silicon lottery, but all of these alternatives appear to be a “drop-in” replacement. Also noteworthy is that the SiC461 is a part of a family of “drop-in” replacements, including the SiC461 (10A), SiC462 (6A), SiC463 (4A) and SiC464 (2A maximum current rating) which offer a scalable solution.
You’ve got mail! The SiC461EVB comes packaged in a cardboard fold-together mailer, of the sort you might find used to protect a DVD when mailed through the ordinary post, just a bit larger. The mailer opens by pulling on the tear-tabs. This particular board had a date code of 2017 (Week 17 of 2020?) and a SAP batch code of 201711US.
The board is packaged inside a static-shielding bag inside some pink bubble wrap. This is stapled to the datasheet, which is provided in full-colour print form.
The static shielding bag was sealed when received, containing a Revision 2 version of the evaluation board. This board is the 12V version, while a 3.3V and 5V version are also available. No choices were made during the RoadTest process, so it will be interesting to see if everyone received the 12V version or not.
The board itself is a high-quality heavy six-layer board, as necessary to optimise its heat dissipation performance with large copper planes with stitching throughout. Surprisingly, the silk-screen on the board claims to be Revision 3, in contrast to the outside label. The board is roughly credit-card sized but a little narrower, with a large amount of board area occupied by the inductor. There is evidence that the board was hand-assembled, including the appearance of some flux residue.
At this stage, a number of minor complications have already begun to arise – the first is the lack of any jumper shunts provided with the board. The mode selection is made using jumper shunts over the pins (J1-J4, J8-J9), so unless you have some handy, you might need to obtain some prior to being able to use the module. It would have been much better had at least two, possibly three shunt caps been provided, so that a user could select one of the four modes and provide a selection for the ultrasonic mode and enable line.
Another complication is the lack of fitted test-points. Plated-through-holes are provided are provided for test points VIN_S, GNDIN_S, VOUT_S, GND_OUT_S, PGD, SW and AGND, however according to the BOM, this should be fitted with Keystone 5002 test points. These points are especially useful for measuring the voltage without contribution of resistive losses, so omitting these test points means that an end user may have to solder their own connections to it. I did solder my own test pins to some of these test points, albeit with great difficulty as the 6-layer board very effectively wicked away all the heat my 60W iron could muster with its fine tip and assistance with the hot air gun was necessary to produce a serviceable but ugly-looking joint.
The board has an input electrolytic capacitor that sits somewhat proud of the board, but this picture should give you a sense of scale regarding the size of the shielded inductor and surrounding MLCCs.
Upon closer inspection of the SiC461 integrated circuit, it can be seen that it was not soldered down with much care. There were numerous solder balls, scrapes where the solder resist seems to have left the board, blobby-looking joints and chips on the edges of the package. It can be seen that the corner opposite the pin-one mark seems to be “sunk” further down compared to the other sides, suggesting uneven solder application and perhaps some force applied to the package during soldering. There are also some “toasty” looking MLCCs nearby where the metal end caps seem to have tarnished.
While the board may work under this condition, I was expecting the manufacturer would take more care in assembling evaluation boards, especially owing to the price they are usually sold for. There is always the slight possibility of a solder bridge from solder balls or a solder joint crack after some thermal cycling due to poor joints. A failure of the board would not make their product look good and shouldn’t be a risk evaluators have to be concerned about.
By contrast, the inductor doesn’t seem too badly soldered on – it is slightly askew but the joints look decent.
The input terminal blocks have been ordered such that the cables come in as Vin, GND, GND, Vout. The terminal blocks seemed a bit small – my test wires only just fit into the apertures.
To get up and running is a matter of plugging in a few jumper shunts to select a mode of operation, attaching the wires and powering up. However, in my case, I decided to go one step further …
… by first using some isopropyl alcohol and a cotton bud to try and clear all the flux residue from around the chip and then using a fibreglass brush to try and loosen and remove all the solder spatters to reduce the chance of faults developing during testing. One has to consider that some connections could potentially have 60V DC across them, so I’d want to minimise any chance of bridging.
Another thing that I did was to install some header pins into the VIN_S, GNDIN_S, VOUT_S, GND_OUT_S test-points to use with my power supply and DC-electronic load for accurate four-wire sensing. At first, my set-up instead routed the VIN_S and GNDIN_S pins to my Keithley Model 2110 5.5-digit digital multimeter just to see how much voltage drop I was getting over my test leads.
The board rested on top of a plastic chair, in the open air for ventilation.
Initial power-up at 24V input under a 10A load was successful. At this load, I’m losing about 600mV in my test leads, showing why using sense-leads is important.
It was possible to bring the input right up to the claimed 60V input with no problems. After this, I managed to connect the circuit with remote sensing on the power supply as well to conduct the experiments which follow in the next chapter.
The documentation provided includes a 28-page datasheet which comprehensively covers the features, operating principles, component selection methodology, protection features, performance and provides PCB layout guidance. Additionally, an 8-page reference board user’s manual is also provided that covers details specific to the design. Confusingly, documentation for an older design also seems to be available and comes up first if you try searching for it. Both documents appeared to be free of major errors and relatively comprehensive. Unfortunately, the reference board user’s manual appears to have some minor inconsistencies with the bill of materials versus the board that was received and the schematic within the guide is missing the distinction between thick-and-thin traces. Fortunately, alternate versions of these documents are available in the PCB files ZIP download, although even that is slightly confusing as there is an older version available, and neither version perfectly matches the Revision 3 board provided.
Design tools include an Excel component value calculator which was available at https://www.vishay.com/doc?75760 but appears to be since removed, now leading to a 404 error page. This tool consists of an Excel spreadsheet containing macros which allows users to enter number of parameters and compute the best supporting component values. This version was V1.59_MVP dated 19th July 2019, downloaded at the commencement of the RoadTest.
The tool itself is fairly simple to use and with the change of several input parameters in yellow, the values can be computed and bolded in case they are changed from the reference design. This is definitely a valuable offline tool for designing with a number of SiC-family converters. It does, however, illustrate just how many supporting components may need to change to optimally configure the converter for different conditions.
Perhaps the more comprehensive and preferred design tool is the Vishay Transim/PowerCAD service which is free to use and can be used to design and simulate designs including the SiC461. All you need is a compatible web browser and an internet connection.
While it is possible to get started without registering an account, you will very soon run into limitations with saving your design and running simulations if you don’t. As a result, I would recommend that you sign up for a “My Vishay” account from the outset, otherwise you may lose your “in-progress” designs.
The design process starts with picking your part and entering a number of operational parameters. Afterwards, the system will design your converter based on the reference design suggestions, which takes just under a minute.
Once configured, you have the ability to run simulations for steady-state, load-step and start-up behaviour.
The results are displayed as interactive graphs which provide quite a bit of detail on temporal behaviour of the converter. The timescale for the simulation can be changed, although the system doesn’t seem to like simulating durations much longer than a hundred or so milliseconds (not that it would be all that useful).
I did run into an interesting discovery with “edge cases”. In my initial simulation attempt, I decided to simulate a steady state load of 10A but the converter’s power good line only “blipped” and the output voltage fell short of the design. Repeating the simulation with an 8A load was just fine. As a result, I concluded that the simulation is particularly sensitive to the edge case of a full load – it’s designed to “fail” into protection, even though a real part may not be as decisive.
Perhaps the most impressive part was the Power and Efficiency tool which allows you to obtain efficiency versus load graphs, approximated thermals, operating modes and losses broken down into their types (e.g. low side, high side, PCB, capacitor, inductor, etc.). This seems to be extremely useful information for optimising a design before actually designing and building it.
While it is possible to export your findings into a report, I was not able to get the PDF export to work, while printing the report resulted in lost findings which were cut-off when using my preferred browser (Mozilla Firefox). Perhaps this is a shortcoming of the Transim/PowerCAD platform.
Performance testing was completed using a mixture of lab equipment including a Rohde & Schwarz HMP4040.04 Power Supply, RTM3004 Oscilloscope, Keithley Model 2110 Digital Multimeter and B&K Precision Model 8600 DC Electronic Load. While all of the units have lapsed their calibration interval, their accuracy has been tested to likely remain within their original limits based on cross-checking with the Keithley 2450 SMU’s readings (within its calibration interval). I would have used this latter unit for some experiments but due to a hardware fault, it is still being repaired at this time. Tests were performed with the board in open air, at a room temperature varying between 20 and 28°C.
This testing consumed the bulk of my RoadTest time and took several attempts, as initial naïve attempts resulted in unsatisfactory “noise” dominating the measured results and a limited test input voltage range up to 32V only. A more refined test method resulted from this experience and those initial results are not presented.
Efficiency was tested using a pyvisa script I wrote to automate testing via SCPI remote control over USB to the B&K Model 8600 and LAN to the R&S HMP4040. Testing was conducted at 0.2V steps between 13V and 60V input, with the load swept in 0.01A steps between 0A to 10A. After each change of current, 500ms wait was provided for the converter to settle and the readings from the load to be stable. After each change of input voltage, 500ms was provided for the converter to stabilise. The input voltage and current and output voltage and current pairs were measured eight times at 200ms intervals and the average of the values recorded for computation of efficiency and generation of plots. This helps to reduce measurement noise, improving measurement stability and likely also increasing accuracy of the computed efficiency figure. This results in a total of 235,000 test points, taking approximately a week of continuous operation to achieve. This was repeated for four combinations of modes – Mode 1 with Ultrasonic OFF, Mode 1 with Ultrasonic ON, Mode 2 with Ultrasonic OFF and Mode 2 with Ultrasonic ON. Modes 3 and 4 were untested, as I do not expect many applications to provide an external VDRV. Total test time exceeded five weeks continuous operations, with some additional time necessary to recover from transient communications faults.
Operation in Mode 1 with ultrasonic mode off should be the most efficient mode that the SiC461 is capable of when operating standalone (that is, without external VDRV input). This is because Mode 1 allows for the converter to go into discontinuous switching when in light load with no lower-bound limit to the switching frequency (hence ultrasonic mode off). The downside to this is the possibility of audible noise generation as inductor coils may sometimes whine, buzz or click in time with the current due to the magnetostrictive effect.
The converter claims a peak efficiency of about 98% and the test results seem to support this assertion. As expected, the best efficiency occurs where there is the least difference between input and output voltage – in my case, exceeding 98.5% in the region below 15V and around 2.5A. The peak efficiency begins around 2.5A load at 13V input voltage and increases to about 6.7A when at 60V input voltage. Efficiency remained above 90% for all loads above 1.5A regardless of input voltage. All of this is rather impressive.
Unfortunately, it seems that the operation of discontinuous mode was not consistent. The graph shows sections where this mode seems to have kicked in (between 31-43V, 47-48V and 52-56V) and provided a significant boost to light load efficiency, but outside those ranges, it seemed the converter did not exhibit power-saving characteristics. As the testing provided at least three seconds of dwell time every time when returning to zero-current load, this should have been enough to allow the converter to engage low-power mode. Instead, I suspect a problem with the soldering could be to blame for the inconsistent operation.
Looking at the lower current sections, where the power save mode was in force, an efficiency of above 90% could be maintained for loads of 200mA or greater, which is absolutely amazing. Unfortunately, where this mode was not operating, efficiencies declined noticeably with a 20-30% efficiency gap in 200mA light-load efficiencies opening up at higher voltages.
The experiment was repeated for ultrasonic mode switched on, which causes the switching frequency to be limited to no less than 20kHz. This is expected to reduce light-load efficiency slightly.
The efficiency curve is arguably indistinguishable for the most part from the previous, although there is a bit of a perturbation around 42V likely due to temperature changes in the test-room. The problem with inconsistent power save mode also rears its head in this test, but engages more frequently – in this case between 19-26V, 32V, 35-38V, 40-46V and 48-60V. This errant behaviour is unexpected and is perhaps due to faulty soldering on the evaluation board due to its intermittent nature.
As there are more power-save mode sections, the impact of the power save mode is quite impressive – for lower voltage inputs, efficiencies above 90% could be maintained at loads of below 100mA.
Zooming in further, the effect of the power saving mode even with ultrasonic mode enabled is clear to see. For a 24V input, 90% efficiency is exceeded as soon as a load of 55mA or above is placed on the converter. Without it, it is estimated that the converter would only have an efficiency of about 50% at this point. However, the low load efficiency accuracy is not so good for the test set-up which is explained in a later part of this RoadTest.
Mode 2 operation removes the ability for the converter to fall into power-saving mode. This mode is perhaps useful if you are concerned about EMC problems with more widely varying switching frequencies or want more consistent load regulation performance. This forces the converter to be in continuous PWM mode.
Unfortunately, this test seems to be marred by some unusual values right about 48V. This results in the “vertical” section seen in the graph. I don’t believe it to be a problem with the test equipment (see next section), but perhaps to be an issue with how the converter’s switching may be “resonating” either with the load or power supply under a very particular condition. Regardless, one can easily guess what the correct values should be, but in the name of scientific integrity, I did not interpolate over those questionable data points.
Because there is no power saving mode, the “blocks” near the light load region at the bottom have disappeared, while the converter’s efficiency characteristic at high loads remains practically unchanged. The results are again impressive, especially as peak efficiencies above 95% are maintained even at the extreme high end of input voltage.
Efficiency at low load across the range of input voltages becomes highly predictable, following a pattern due to the relatively fixed nature of the quiescent losses.
This combination would perhaps be the least-efficient configuration for the SiC461 reference design board, as it constrains the converter to continuous PWM and with an enforced minimum switching frequency.
However, this mode also proved to be the most stable – producing an absolutely perfect graph of efficiency with no unusual spikes or dips. This is why I am confident that the issues seen in the prior parts are not an artifact of the test equipment or the test protocol. The efficiency is virtually indistinguishable from the above example, however – more information in a later section that follows.
Again, the results are very similar to the results above, which are still impressive overall, but perhaps not as impressive as if the power-save mode was operating correctly.
Efficiency in all modes is summarised with these stacked plots which look at the efficiency combined for all input voltages from 13-60V. It is seen that where the power save mode is active, a bump in efficiency can be seen for low currents. There are some glitches with these graphs mainly due to the glitches with the converter during testing.
As I wasn’t sure whether the issues with the power-saving discontinuous switching mode not always kicking in was caused by a transient fault that is reproducible or a transient fault that causes random mode switching, I re-ran the “Mode 1 – Ultrasonic OFF” test at a cost of an extra week just to find out.
As it turns out, I was able to measure the effect of the power-saving mode across a wider range of voltages, but the mode did not consistently remain engaged at low loads. It appears to support the assertion that the fault is with the evaluation board, as the failure is not consistent with voltage and is not due to the test set-up.
Looking at the test results at low currents, it seems that efficiencies of 90% and above can be achieved as low as 10-50mA depending on voltage input, although the accuracy of the test rig at these low currents is not likely to be more accurate than 5%.
It’s all good to come up with a number of plots which clearly show efficiency across the space of input and output conditions, but it’s also important to understand the accuracy of the measurement figures. As a result, I decided to take the results from the Mode 2 - Ultrasonic ON tests and determine given datasheet error margins for both the power supply and electronic load, what the worst-case error in terms of efficiency percentage points would be. This resulted in a rather complex array of calculations …
Once computed, the plot of efficiency errors is as follows:
On the whole, the measured efficiency should deviate less than 1.1% for the majority of the plot, except for the region below 1.5A. That means if the plots measured 98% efficiency, then the real value should be guaranteed to be in the interval from 96.9% through to 99.1%. The deviation around the 3A line is due to the range-switch in the B&K Model 8600 DC Electronic Load.
At lower currents, the error increases, concentrating in the bottom-left corner of the graph. Values below 0.2A have efficiency deviation up to 2.3% due to the limited measurement resolution of the power supply’s inbuilt metering, although in reality, I have confidence that it is less based on my review testing of the HMP4040’s metering accuracy. Regardless, this is something to be aware of and puts the measurement results into context.
Generally speaking, the use of forced PWM (Mode 2) as compared to allowing the converter to go into discontinuous conduction modes (Mode 1) is associated with a loss of efficiency at light loads as the converter will be switching at all times. Likewise, enforcing the use of Ultrasonic Mode is also likely to cause a loss of efficiency at light loads by forcing the converter’s switching frequency to be above 20kHz regardless of load. However, the use of these modes is perhaps advantageous for tighter load regulation, elimination of potential audible noise issues and (maybe even) maintaining switching frequency for EMC compliance reasons.
I wanted to understand whether we could observe the impacts of each of these modes on converter efficiency, so using the data obtained above, I decided to take the difference in efficiency between the modes to see how each of them may impact performance.
Due to the inconsistency of Mode 1 operation, I decided to use the data from the second Mode 1 Ultrasonic Mode OFF run compared with the Mode 2 Ultrasonic Mode OFF run that has a "blip" in it. This would give us a glimpse into the efficiency gain as a result, although it would not be able to illustrate it for the full range of voltages due to the inconsistent operation of my evaluation board.
The majority of the diagram is in the navy blue colour that indicates less than 0.5 percentage points of efficiency shift either direction, with only a few outcrops exceeding this just slightly. This is expected due to the accuracy limitations explored earlier. The glitch at about 48V does result in a "blip" where the efficiency difference goes out of range and should be ignored. The benefit of Mode 1 power saving can be seen in the low current regime, especially below about 1.7A.
The efficiency gains generally increase at lower currents (although the accuracy of the test decreases somewhat as well). This is unsurprising, however, it means that a big gain in efficiency is experienced using Mode 1 over Mode 2, especially where current is below 0.5A where greater than 5% in efficiency percentage points are gained.
Unfortunately, as Mode 1 operation with or without ultrasonic operation was not consistently achieved, taking the difference in the efficiency is not possible to understand the impact of Ultrasonic Mode. I did previously take the difference between Mode 2 Ultrasonic On versus Ultrasonic Off, but later on, I realised that the Ultrasonic selection has no effect in Mode 2 as the converter switching frequency is already ultrasonic at all times thus the result is meaningless.
Output voltage stability was assessed as a by-product of the above efficiency testing by plotting the voltage as measured by the sense terminals of the B&K Model 8600 DC Electronic Load which were connected to the output voltage test points on the board.
In all, output voltage stability was similar in all modes, varying by about 30mV more as a function of input voltage than as of load. The exception is at the low input voltages below 15V where there were severe voltage drops at increasing loads, especially at the maximum load value of 10A. The cause of the output voltage variations is likely a result of the IC and perhaps variations its internal linear regulated supply. While thermal changes in the feedback resistances could have played a role, I would have expected a load-dependent voltage stability effect as the temperature of the board is very much modulated by the load current.
The quiescent current of the converter was measured using a Keithley Model 2110 5.5-digit Digital Multimeter at 10PLC for greatest accuracy. Power being supplied by the Rohde & Schwarz HMP4040 is sent through the meter in series for measurement, however, the sense terminals are connected such that any voltage burden from the multimeter is compensated for. The quiescent current was measured at 200mV intervals from 13V to 60V for the combination of all-loads-disconnected while the converter is Enabled, in Mode 1/ 2 and with Ultrasonic Mode Enabled/Disabled or with the converter Disabled. Recordings were automated by a pyvisa script and averaged over 100 samples.
Under a no-load condition, the current consumption ranged from about 10mA to 29mA depending on the input voltage. The choice of modes did not seem to have much of an effect, although measurements below 15V input showed some variation and instability as the converter may not have been operating smoothly. The current plateaued for voltages above 30V, however, as quiescent power is the product of the current and voltage, the quiescent power consumption generally increases with voltage, ranging from about 150mW through to 1680mW.
The quiescent current for the case of the converter being disabled was so low by comparison that I plotted a separate set of graphs for it.
With the converter disabled, the quiescent current drops dramatically, registering about 30µA up to 115µA, corresponding to a quiescent power of about 0.4mW up to 7mW depending on voltage. This sort of measurement is starting to approach the limits of the capability of the Keithley 2110 which would have an error of about 2.05µA for measurements of this magnitude. A clearer result may have been obtained with my Keithley 2450 SMU, however, it is still on its return journey from repair.
It may be noted that these results seem to be in contrast with the datasheet claims of “250µA operating current not switching, 5µA supply current at shutdown.” The reason for this is due to the datasheet’s claims being with reference to IVCIN which is the current input to the converter only, excluding the power into the MOSFET stages (VIN). In my case, my measurements made on the reference design are as per their design, measuring the supply current which is connected to the board which is connected to both. This means that any MOSFET-related losses are also taken into effect. Furthermore, because of the no-load condition being the baseline, this is not a comparable measurement to the converter not-switching, instead representing a more realistic application scenario.
Power Good output was assessed by watching the converter power up under zero load and 10A load. Overload behaviour was assessed by increasing the load on the converter until the output ceased to be stable at an input voltage of 24V from a Rohde & Schwarz HMP4040. Output behaviour is assessed using a Rohde & Schwarz RTM3004 Oscilloscope for both the output and power good signal. The converter was operated in Mode 1, Ultrasonic Mode OFF for maximum flexibility.
Whether the converter was unloaded or loaded at 10A did not make much difference to its power-up time. It took about 5.5ms from applying power to the Power Good signal being asserted as the voltage crossed 11.8V.
The power-down behaviour is much more load dependent, as without a load, the capacitors must naturally discharge. This process took about 28ms to reach the point where Power Good signal was deasserted, crossing 11V. Under load, this fall was much quicker, taking under 100us to see the Power Good signal deasserting while crossing through 10V. In this case, it seems the converter may have been challenged by the rapidly collapsing voltage and did not quite drop the Power Good signal as quickly as it should have. The fall was not monotonic, however, with the converter trying its very best to maintain the load even though the voltage is collapsing (as evidenced by the switching noise and the platau in the voltage trend).
When the SiC461 is overloaded, the output voltage tends to waver a bit. In the above, at a load of 11.4A, the converter finally tripped its Power Good output low, even though it was still continuing to (try) and provide a steady output. Increasing the current load resulted in the converter shutting down and instead attempting to restart on a periodic basis. The right graph shows what happens when the load is reduced to 11A - the converter attempts and successfully restarts on its own, maintaining a steady output and asserted Power Good signal. This means the SiC461 is suitable for applications where transient faults may occur as it is robust to overload and capable of automatically restarting once a fault clears.
Holding the converter at an overload of 11.248A and letting it warm up, the instability in the output was able to cause the Power Good signal to oscillate, indicating it is reflective of the state of the output although its behaviour does not seem all that well defined for a noisy output such as the above. However, since the converter is being operated outside of its intended 10A current rating, this is not a problem in practice.
Transient behaviour was assessed using the load-step feature of the B&K Precision Model 8600. A number of load steps between 1A and 9A were generated using a range of slew rates at a fixed input voltage of 24V. The resulting transient response behaviour is captured using a Rohde and Schwarz RTM3004 Oscilloscope. The converter was operated in Mode 1, Ultrasonic Mode OFF for maximum flexibility.
With the load steps occurring with a rise/fall time in the order of 62ms, the deviation in voltage was well controlled, about 120mV peak-to-peak (1%) when discounting the switching noise.
At a more aggressive rise/fall time of 62us, the deviation has increased, but the overall deviation still remains well controlled at about 200mV peak-to-peak (1.67%) when disregarding the switching noise.
Pushing it right to the limit of what the B&K Precision Model 8600 DC Electronic Load can generate, transitions with a rise/fall time in the order of 9us were generated. This speed of transient has bought even some lab power supplies to their knees, collapsing their output voltage to zero for non-trivial amounts of time. General-purpose converters rarely stand a chance (or so I thought). The SiC461 put up a formidable performance, with a deviation about 600mV peak-to-peak (5%) disregarding the switching noise, in all cases recovering within about 50us! It seems that the SiC461 is not only efficient - it's especially adept at handling transients as well.
Measurement of ripple and noise was attempted with a 10x passive probe and a Rohde and Schwarz RTM3004 Oscilloscope with 20MHz bandwidth limiter engaged (as is common with ripple/noise measurements). The converter is loaded at a number of fixed current values (0A, 1A, 2A, 5A, 10A) using a B&K Precision Model 8600 DC Electronic Load while being powered at 24V from a Rohde & Schwarz HMP4040. The converter was operated in Mode 1, Ultrasonic Mode OFF for maximum flexibility.
When unloaded, it seems the power saving measures are clearly evident, resulting in a ripple with a low-frequency characteristic of about 35Hz. The peak-to-peak ripple measured about 58mV inclusive of a sharp transient at switching.
However, once loaded at 1A and 2A, the switching frequency can be seen to increase, changing the nature of the ripple. The ripple amplitude remained controlled at about 44-46mV peak-to-peak, although some of this could be induced switching noise into the measurement.
Increasing the load does increase the ripple, for example at 5A the ripple measured 61mV peak-to-peak while at 10A, it measured 102mV peak-to-peak. On the whole, however, the ripple voltage is very acceptable and remains low-enough for many sensitive computer-type electronics applications - e.g. the ATX specification mandates a peak-to-peak ripple of 120mV on the 12V rail.
Switching frequency measurements were made with a Rohde & Schwarz RTM3004 oscilloscope connected to the SW test point, observing the switching frequency given a range of load conditions when supplied with a fixed input voltage of 24V when configured in combinations of Mode 1/2 with/without Ultrasonic Mode.
The thermal performance of the chip was assessed by running the board at an input voltage of 60V for five minutes at each load current from 0A to 10A in 1A steps. The temperature of the IC was measured using a non-contact IR thermometer, which was scanned across the surface to obtain the highest reading at each current level. The results are plotted below:
The results correspond rather closely to the graph in the datasheet (Fig. 16 – SiC461 Load Current vs. Case Temperature, Vin = 48V, Vout = 12V) despite the more strenuous conditions due to the input of 60V, with an ambient temperature of 24°C (the deviation representing error in the IR thermometer used). The result shows a roughly linear relationship of temperature corresponding to the load current, but with the provided 6-layer reference design, it seems there is sufficient operational margin at the load of 10A to the OTP point of 150°C trip / 115°C recover depending on the ambient. This is quite impressive considering the design is passively cooled.
Optimal configuration of the converter for a different voltage range requires the changing of many components, as the design tools had alluded to. However, if we are not entirely concerned with the absolute best performance, it should still be able to change the output voltage up to about 25V (due to the voltage rating of the onboard capacitors) by changing the value of R10 and R12 which form the feedback pair. The value of R10 is 10kΩ (the maximum recommended), while R12 is used in the reference designs to set the voltage. As a result, I decided to compute the actual value of R12 necessary to achieve some common voltages below 25V and see how they might be best achieved using a single E24-series resistor or a pair of E24-resistors in series (as that’s what I have to hand).
As a rule, better results were had with pairs of E24-resistors as it allows us to tune closer to the target value. In the case of 5V, 18V and 24V, a perfect match can be attained, while the biggest error of 0.16% is seen at 2.5V. The practicality of the tolerance of the resistors themselves (1%) would probably become the dominating factor. Using a single E24-series resistor, the error ranges from about 2.2-3.3%, which is somewhat less acceptable. Of course, higher-precision E96 resistors can be purchased to obtain the necessary values for a design and that seems to be what is done in the case of the reference design board and it’s odd 31.2/52.3/140kΩ ± 0.1% values.
In theory, I could use a precision potentiometer of about 300kΩ to provide the range of output voltages from 0.8V to 25V, however I didn’t have any to hand and was cognisant of the possibility that heat from the converter would cause enough resistance value drift to affect the converter’s output voltage stability.
While I did intend to reconfigure the output voltage, I was initially stopped by the fact that R12 is not in a particularly accessible position, right next to a number of other components near the SiC461 IC. The resistor is also an SMD type with a relatively small 0603 footprint, while all the resistors I have in stock are through-hole units. But as I had already finished my intended "battery" of tests, I decided to give it a go by modifying the converter to output 5V through using a 51k resistor in series with a 1.5k resistor to replace R12.
You can see the through-hole resistors standing proud of the board, after I had used hot-air to remove the existing R12. As the resistors are carbon-film, 5% type, the voltage accuracy is not expected to be very high and temperature-related voltage drift is a possibility. The hot air managed to affect my input voltage test point, causing it to fall out of place leaving a filled-in hole, so I ended up mounting the pin on the opposite side of the board for the final test - a test of the conversion efficiency and voltage stability now that it has been modified to output 5V. In order to reduce the test time, I decided to test at 0.5V input voltage steps between 5.5V and 60V and at current-load steps of 25mA between 0A and 10A which should take about 36 hours to complete.
Lower efficiencies over most of the test area is expected as the difference between input and output voltages are greater on the whole. However, it seems that the choice of component values (especially the oversized inductor) may have contributed to the unstable efficiency characteristic below about 16V where the efficiency is suddenly about 2% worse at a given load. In spite of this, above 16V, the peak efficiency is around 95%, while below, there are pockets which reach 97-98% but the actual voltage difference is so small as not to be of great use. However, one interesting outcome is that it seems that the power saving mode is operating correctly and consistently after the modification. Perhaps the extended heat-gun treatment to change out R12 managed to reconnect a broken joint which may have been responsible for mode selection.
The output voltage stability may have been affected by my choice of wide 5% tolerance carbon film resistors, but overall, it seems the voltage stability was not bad, varying about 50mV through most of the operating region. The unstable efficiency region also shows a positive deviation with regards to output voltage, while the remainder shows mostly a smooth characteristic with a slight voltage reduction as the load increases.
As this modification to the board disregards optimisation of the other component values to ensure proper operation - for example, compared to the 5V reference design, the inductor is oversized and the switching frequency and ripple injection resistor values are different, I did not feel the need to further characterise the board in its modified state as better performance is almost certain with the correct optimal design values. The converter was, however, still perfectly capable of operation in this sub-optimal state, producing some acoustic noise especially at low loads and having an unstable efficiency characteristic.
It seems that the Vishay SiC46x series of microBUCK synchronous buck regulators have a lot to offer. They’re not the cheapest on the market, but they do offer leading specifications across the entire board when it comes to voltage range, switching frequency range and peak efficiency. The size of the IC and the completed solution is quite reasonable, especially when considering the current/power handling capabilities. It comes with sufficient documentation and software tools to design and simulate designs down to even estimating the thermal performance.
I evaluated the SiC461 on a Revision 3 reference design board and on the whole, performance testing showed that the efficiency met the stated claims while also demonstrating excellent output voltage stability, transient handling and ripple/noise levels. I was very impressed to see that it handled the torturous 9uS 1A to 9A load-step without collapsing entirely as many power supplies often do. The power saving mode behaviour also showed some significant efficiency gains at low load, providing a very broad band of high efficiencies. Continuous PWM and ultrasonic modes both performed as expected and the converter was not damaged by overloading, cutting back its output and automatically restarting.
The quiescent current and power consumption when tested in Mode 1 and Mode 2 showed a dependence on voltage, consuming between about 10mA to 29mA (or 150mW to 1680mW) when operating with no load. When disabled, the quiescent current dropped to 30µA to 110µA (or 0.4 to 7mW). This figure was derived under realistic conditions, however, the datasheet claimed quiescent power is for only VCIN and excludes the power to the MOSFETs (VIN) under non-switching condition only thus being less relevant to the real-life condition. This level of consumption is reasonable for a converter of this power-level, but perhaps not the best.
The main downsides were unfortunately related to the evaluation board itself. Shipped without any jumper shunt caps, I had to find my own to make the necessary mode selections. Soldering on the board, especially around the SiC461 was poor, with lots of solder spatter remaining on the board and the chip being mounted unevenly as if the solder was not properly applied. I suspect construction issues were the reason behind the inconsistent power-save mode behaviour experienced in the RoadTest. The test points on the board were not populated with Keystone test points as per the bill of materials, which made testing slightly more challenging because its six-layer design makes soldering difficult. It was also discovered that the supplied BOM and board files were not consistent with this particular version of the board, leading to some uncertainties. Finally, the design of the board does not make it easy to change components either – for example, to alter the behaviour or output voltage. For the price of the evaluation board, I would have expected somewhat better.
While the Vishay SiC461 microBUCK synchronous buck converter is an excellent product in almost every way, the evaluation board could have been better designed and manufactured. However, if you are considering the Vishay SiC461, you might also want to consider the footprint-compatible SiC466 that has a lower quiescent current on the datasheet or the SiC476 that seems also to be footprint-compatible but with a mild-reduction in the maximum input voltage to 55V. These could be worthwhile alternatives, especially in case of supply issues.
Thanks to element14 and Vishay SIliconix for the opportunity to test this rather amazing product. Let me know what you think in the comments below - and if you found this interesting, entertaining, useful or educational, feel free to leave a rating, like, bookmark or share with your friends and colleagues.