RoadTest: MPS Four-Channel Output Power Module EVM
Evaluation Type: Semiconductors
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?:
What were the biggest problems encountered?: The main issue was locating the software and operating the clunky Gui
MPM54304 4 Channel Buck regulator Module Review
What's supplied in the EVK?
Having used literally every DC-DC vendors single IC, it makes a change to evaluate a module that houses a the Die along with 4 custom designed power inductors in a very compact module measuring a staggering 7 x 7 mm. All that’s required are few R’s and C’s and you have an ultra compact 4 channel step down (Buck) converter in less than 12 x 12mm. Applications are generally going to be FPGA or mixed signal SOC that require multiple rails or any system where multiple DC-DC converters or Linear regulators can be replaced with a single module and not forgetting the hobbyist or educational establishment that can quickly assemble a fully integrated custom PSU that can easily be controlled via MCU or PC with a handful of components! The internal die within the MPM54304 is the MP5470 and this is a 4 channel fully integrated COT buck regulator that can be used stand alone, supplied in 5 preprogrammed configurations, programmed on the fly over I2C interface from a host MCU or can even be flashed twice with a fixed configuration set by the user during manufacture of the product. One more really nice feature is that individual rails can be set with external feedback resistors enabling legacy analog AVS schemes seen on many of the modern SOC’s and to improve transient response and improving stability with the addition of a feedforward capacitor in the feedback loop. External feedback also provides a mechanism to break to loop to allow measurements of Gain/Phase to prove the device stability.
Why would you use this module and what’s the Pros and Cons?
For the bulk of high volume designs a module may seem expensive/excessive but these modules are becoming cheaper and can be sourced from multiple vendors, they also do have some distinct advantages over individual converters.
Getting it all fired up!
Not as easy as you may think, there is no paperwork supplied in the package so you need to get online and grab the datasheets and software
Note for MPS: There is no link on the EVK datasheet that even points to the software so they are listed here
MPM54304 module Page: https://www.monolithicpower.com/en/mpm54304.html
One other issue i found was that there are only 3 pins on the EVK so you have use the supplied 3 way ribbon and need to be careful getting the correct pinout. The following picture outlines the correct pinout
EVK MPM54304 comes preconfigured in option 0000 (Shown here) so without squirting any data the EVK is operational. With an input voltage of 10V ICQ is around 33mA with no load on any output. Following the warning on the I2C unit and connecting the I2C on the EVK First then applying power to the EVK you can plug in the adapter and the drivers self install installed. If you cant find them there is a link on the software page that pings you to Silicon Labs website for a driver.
Once Visual Bench is installed don’t opt for the Auto Find Board as it doesn’t work! Start a fresh project and manually add the MPM54304. There is also no mention of the default I2C address in the datasheet so without looking at the schematic its a bot of a random selection and is frustrating that its nots mentioned. it’s 0x68… If its connects the marker next to the device tab goes green and you are ready to rumble!
For this evaluation I opted for voltages that the “elcheapo” active loads could support as these are limited to 1.5V minimum. Rail voltages selected are: V1 = 1.8V, V2 = 2.0V, V3 = 2.5V, V4 = 3.0V, Input voltage range for testing was 5V to 12V, its capable of 16V but i limited the range for this evaluation. The way in which each supply voltage is adjusted is by altering the feedback reference voltage from 550mV to 1820mV with an option of direct or Ref x 3 thus the 4 rails can be anywhere from 0.55V to 5.4V with resolution as low as 10mV on the direct setting. The datasheet states that the device can cover 4 to 16V but obviously there is a dropout limitation if the supply voltage is lower than the requested rail output so be careful! If a much higher current is required an option exists where you can connect a pair of Bucks in parallel but for this testing I left each rail separate using 4 loads and 4 separate voltages. The user can adjust almost every parameter and I really like the fact the register structure is the same for all 4 Bucks easing the code overhead. The operation of each Buck is pretty self explanatory with a Global clock and phase adjustment of each regulator then the independent controls for each rail. There isn't much you cant tweak in this device so it has a huge use case.
Visual Bench Pro Gui
I personally found the Gui a little clunky and it wasn’t that easy to see how you actually send the data to the IC and then noticed the Volatile and Non Volatile Memory arrows on the opposite side of the window that read and write the registers (Ringed in red). Select the Volatile Memory option is you just want to play around with the settings or the Volatile if you want to flash the values to ROM. One other small issue was once the device is power cycled you cant just send the same settings, I had to physically change each setting to a different option, send it and then change back before the IC would accept the changes? Maybe a Gui bug but its really annoying!
Leaving the frequency at 800kHz and the rest on the defaults I set the voltages, increased the current limits and loaded each Buck with its maximum current and took a look under a thermal imaging camera to see what was going on in the module. From the next set of images its clear to see how the internals are laid out.
The MPS PCB design does indicate that each PSU is close to the corner of the package but we can verify by loading individual rails and can pinpoint the exact placement of the Inductors and the die. With a little digging and matching up parameters its clear this module houses the MP5470 Die.
Rather than testing each rail separately and since I had all of the equipment on hand, I opted to load all the rails with some real world scenarios to see just how far this tiny module could go. Looking at the Max Power loss Vs Temperature chart in the datasheet its clear to see that there are limitations on the thermal performance and to test this out I placed the Module under some fairly heavy loading or at least as far as my bench PSU would go.
You can see from the Max Power Loss chart that the maximum losses need to be de-rated versus ambient temperature to keep the die within the specified 125C limits. The closest I could get was the 3A, 3A, 1A, 1A loading case without my bench PSU current limiting at lower input voltages. The efficiencies measured were very close to what's published in the datasheets and I also correlated each individual Buck and these too were almost identical to the datasheet curves. Also included are some other lighter and split loading cases.
For a commercial product running at 60C ambient the power loss curve states the recommended maximum loss at 60C is around 2.7W and I assumed that either the die or the internal inductors were close to the 125C maximum/Curie temperature as the curve seems to hit 0W at 125C. Since I had a thermal imaging camera I could at least prove this theory and also look how well balanced the die and Inductor temperatures are within the module.
Ripple and Spurious
Ripple was measured under the 3A, 3A, 1A, 1A loading case and the results are very impressive. All supplies were measured and as the individual converters are likely to be identical it was no surprise that the ripple and spurious measured was the same on all rails. The below images show the output ripple and also filtered to 7MHz to show any issues with switching transients from the other converters, Output ripple under load was around 10mV which is staggering considering there is only 1 output capacitor per converter. I also captured the input ripple raw and filtered as many designs are very susceptible to noise emanating from the input of the switcher and its often overlooked during the design phases. FFT of both the input and output noise are shown as in many design that include sensitive or low noise devices knowing the frequency and magnitude can help you identify any potential issues and with this particular module you also have the option to change the global clock and tweak the phase of each converter if needed.
Unfiltered Capture of the Output Ripple
Filtered Capture of the Output Ripple
FFT Output Ripple
Unfiltered Capture of the Input Ripple
Filtered Capture of the Input Ripple
FFT input ripple
From this brief evaluation the MPM54304 is a beast of a module with excellent efficiency and low noise performance. Based on the results taken I will definitely be using the module in some space constrained and EMI sensitive projects and possibly develop a small form factor, digitally controlled MCU driven PSU as a test bench for evaluating other SOC's.