MPS MPM54304 EVM RoadTest - Snippets

This page is a collection of snippets for the RoadTest of the MPM54304 <link to final review>

In this road test I aim to answer the following 2 questions:

(1) How is the MPM54304 programed the first time?

(2) How well do the software tools provided by MPS aid in the design, integration, and validation of an electronically programmable power solution?

For the purposes of RoadTesting the MPM54304, I will attempt to employ the programable DCDC converter into a multi-rail “Blinky” project.

Smoke Test

Before spending too much time documenting a review of the MPM54304 I wanted to give the converter a quick smoke test. I configured the EVM to parallel outputs 1 & 2 and programmed the output voltage to 5 V. I connected a WS2812B LED strip controlled by an ESP8266 running the WLED application to the 5V output of the EVM. This wasn't indented to be an accurate characterization measurement, just a quick test. A photo of the test setup is shown below:

After letting the test setup run for 10 minutes, there was no smoke! Additionally the DCDC converter didn't go into thermal shutdown or issue a thermal warning in the system status register.

All the status indicators in Virtual Bench Pro are green. This is a pretty good first test. Even if this isn't a state of the art DCDC converter, it's still impressive seeing a 7x7 mm DCDC converter output 25 W.

I am quite confident that I won't have any issues running the converter with a 30W load.

Reference Design and Layout

MPS provide Altium design files for the EVM which can be used as a sample schematic and layout. If you don't have an Altium license you can drag and drop the project files into the free online Altium viewer.

I wanted to know the drill hole spacing to mount a piece of perfboard to the EVM. We can simply take a distance measurement in the PCB viewer:

The EVM holes are 57.15mm center to center, too easy!

Solution Area

We can measure the bounding area on the layout of the EVM as,

Width: 11.14 mm, Height: 11.54 mm

Solution Area = (11.14 mm * 11.54 mm) = 128.6 mm^2

Power Density

We can determine the solution volume assuming a 1.6 mm thickness PCB and the secondary side having MLCC with a height of 1.5 mm, as

Solution Volume = (Area)*(2+1.6+1.5) = 656 mm^3  = 0.0400 in^3

Assuming the converter is thermally limited to outputting 30W, the power density is:

Power Density = (30W)/(656mm^3) = 46 mW/mm^3 = 750 W/in^3

When you contrast the power density of this solution to the size of a 10W USB wall charger, the figure is quite remarkable.

As an apples to flying flamingos comparison, the Google Little Box Inverter Challenge required challengers to design an inverter with a power density greater than 50W/in^3. The winning team achieved a power density greater than 143 W/in^3.

https://web.archive.org/web/20160310120416/https://www.littleboxchallenge.com/pdf/LBC-InverterRequirements.pdf

https://ai.googleblog.com/2016/02/and-winner-of-1-million-little-box.html

Solution Cost

The solution cost is based the minimum viable design using the same X5R MLCC capacitors used on the EVM module in 1k+ unit volumes.

Presumably, the MPS price includes programming the MPM54304 under a custom part number and shipped on tape and reel.

For a total solution cost of $7.28.

Rudimentary Power Specifications

VLED - One 150 LED WS2812B strip can draw up to 7.5A @ 5V when power is applied at both ends. So, we will software limit the strip brightness such that to stay within the maximum output current rating of the MPM54304. With outputs 1&2 operating in parallel the converter is rated for a maximum output current rating of 6A.

VDDIO – The RP2040 microcontroller will operate on any VDDIO from 1.8V to 3.6V. The external QSPI flash memory (W25Q16JVUXIQ) requires a 2.7V to 3.6V supply rail. For now, we will run VDDIO at the same voltage provided by the RPi Pico’s onboard DCDC converter of 3.3V (though we are not enabling this converter). The QSPI flash memory may consume 20 mA during a high-speed flash write. Regardless, we will be well within the 2A current rating of the MPM54304.

VCORE – All the RP2040 speciation’s are based for VCORE(DVDD) of 1.1V, so we will follow the datasheet specifications. The internal programable LDO on the RP2040 can be configured to output between 800 mV to 1.3V in 50 mV steps and is rated for 100 mA output load current. Again, here we are well within the 2A current rating of the MPM54304.

Rail Sequencing

The 5V VLED rail could come up at any time, but to make the graphs look “nice” we have the 5V rail come up first. The RP2040 does need VDDIO to come up before VCORE. Here is the sequencing diagram we will try to achieve with the MPM54304:

Modifying the PICO

We will connect to VDDIO directly on the Pico’s castellated pin headers.

The Pico was not designed with the intent of VCORE being provided externally. So, we will solder tack a mod-wire to the largest decoupling capacitor on the VCORE supply rail (C12 adjacent to pin 44). The RP2040’s internal LDO’s output (pin 44) is routed to VCORE underneath the chip. Isolating pin 44 from VCORE would not be an easy task.

So, we will have to work around the problem in software and leverage the design flexibility of the MPM54304 DCDC converter.

We can disable VCORE regulator by setting it to Hi-Z,

mem32[0x40064000] |= 2

First Power-Up

The MP54304 was not happy to see VCORE rise prematurely, due to the RP2040's internal LDO being enabled during its POR.

So, we will slow down the VDDIO rise time such that the MPM54304 can attempt to soft-start the rail in the its assigned time slot. In virtual bench pro, there are a number of start-up slew rates to choose from.

After slowing down the VDDIO slew-rate, we can see the MPM54304 starts up happily:

The MPM54304 is earning bonus points here for being highly flexible.

Here I programmed the MPM54304 to output 1V on VCORE so it is easy to see when the RP2040 disables the internal LDO. It took almost 170 ms to boot uPython and start executing main.py.

It blinks! With a 12 V input the MPM54304 draws 3.9 mA powering the Pico.

MPS Smart (Simplis)

I think MPS's simplis model might be a bit optimistic on conduction losses, nowhere in their datasheet do they suggest the converter will operate at 95+% efficiency. We will have to settle with the synchronous converters 90% full load efficiency.

References

[1] EVM Schematic: https://www.monolithicpower.com/en/documentview/productdocument/index/version/2/document_type/Schematic/lang/en/sku/EVM54304-MN-01A/document_id/9463/

[2] EVM PCB Layout: https://www.monolithicpower.com/en/documentview/productdocument/index/version/2/document_type/Layout/lang/en/sku/EVM54304-MN-01A/document_id/9464/

[3] EVM Datasheet: https://www.monolithicpower.com/en/documentview/productdocument/index/version/2/document_type/Datasheet/lang/en/sku/EVM54304-MN-01A/document_id/9460/

[4] Altium Viewer: https://www.altium.com/viewer/

[5] Simplis Model: https://www.monolithicpower.com/en/documentview/productdocument/index/version/2/document_type/MPSmart%20Model/lang/en/sk…

[6] Virtual Bench Pro: https://www.monolithicpower.com/en/virtual-bench-pro-4-0.html

[6] MPM54304 Datasheet: https://www.monolithicpower.com/en/documentview/productdocument/index/version/2/document_type/Datasheet/lang/en/sku/MPM5…