So I began thinking about Version 2 with the new goals: minimize per-module cost, and perhaps allow the user to provide his/her own AC DC converter.
If putting the voltage regulator on the module is too expensive, then voltage regulators must be somehow re-used for each module. The core idea of version 2 is to have several power rails at different voltage levels and each module will simply multiplex the power rails. Since I'm providing USB-PD, the finished product will need rails at the levels required by USB-PD: 5V, 9V, 15V, 20V (leaving 28V, 36V, 48V for a future "super" version). With an additional 2 variable voltage levels, most users should be able to get all the 20V and under voltages they need. Certainly I will. Though I will be still lacking the 29V @ 1.8A that my sit-stand desk requires.
In the interest of keeping prototype costs low, I am going to start with even fewer rails: 5V and 9V for USB-PD. One 3.3V rail for MCU power. And one variable voltage rail.
And so version 2, which I am working on now, looks like this:
I've learnt more about power electronics since I first selected the TPS55288 as my variable voltage regulator. I settled on the TPS55288 because there were no integrated-magnetics modules that could meet my current needs (highest available was 6A, I needed 9A). However, since then I have learnt about multiphase buck regulators. That is, you can connect two buck regulators to the same output with the clocks equally separated in phase and with a protective diode on the output. Then you double the current capacity on the output. This makes the PoL regulators viable again, since I know that I won't be painting myself into a corner with a hard cap on output current. The following TI ANs will probably be helpful if I should go that route: Synchronizing DC/DC Converters in a Power Tree, How to parallel two DC/DC converters with digital controllers, Multiphase Buck Design From Start to Finish (Part 1). Additionally Sequencing Power Supplies in Multiple Voltage Rail Environments may be relevant even before going multiphase.
Therefore I have selected the LMZM33606 PoL module for the variable voltage regulator and the APM81911 PoL module (with an eye on the TLVM1440 as well) for the 3.3V, 5V, and 9V regulators.
I can reduce cost in the future by replacing the power modules with custom-designed regulators using external-FET controllers and external inductors. I have my eye on the LM5145 for the variable voltage rails.
Features to be added in later versions are the following. More fixed voltage rails. More variable voltage rails. Increased maximum current on all rails.
As for the user-provided rectifier, I still tentatively plan on supporting it, though I will only step it down. And then if the user-provided rectifier voltage is less than the voltage of some of the fixed-voltage rails, I'll just disable those rails and have the USB-PD modules indicate to the connected device that those voltages are unavailable.
I plan on doing some power usage monitoring as well. And protection of the PSU, connected devices, and user in case of power faults, water spills, bad connected devices, or an imbecilic user. I'll look at some of the compliance that devices which go on market have to do (for me FCC Part 15 Unintentional Radiators, UL 1012 Certification for Power Supplies, RoHS compliance). And I'll have a look at a few adjacent ANs from TI on eFuses, voltage supervisors, and thermal design.
Finally, a couple of unrelated notes:
This article from Analog Devices describes another way of making a variable voltage buck regulator which manipulates the ratio on the feedback voltage divider using the DAC. It seems less elegant than controlling the reference voltage directly via the tracking pin, but perhaps there are some contexts when it's a better approach.
Also, I discovered a couple of products on Alibaba which one could use to relatively easily make a hacked-together imitation of what I want to build: the XY6020L, or for a more integrated option, the DP30V5A.
Please send comments to blogger-jack@pearson.onl.