Diagramming one of the possible use cases for my "RPC" (Regulating Power Converter).
One key difference between my RPC and any off-the-shelf switching power supply is it's ability to operate smoothly in conjunction with another active load such as a Grid-Tie Solar Inverter
Listing out some of the features I'd want to include. "Arduino" is more of a place-holder here, since it's something I've used and *could* implement.
Realistically, I'd use a version of my SmoothPower & PIC18 (developed years after this session) in conjunction with a LoRa module.
The Physical Layout of this design was not defined beforehand.
Defaulting to Left-to-Right and Top-to-Bottom
Using the LM339 is not so convenient being open-collector
Pull-down rate is fast because of transistor,
Pull-up is slower because of pull-up resistor
This can complicate the feedback loop design
Suggest using more standard Opamp rather than "comparator"
Still using the LM339 open-collector comparators, did not necessarily test ok. Asymmetry caused by open-collector implies need to effectively compensate for 2 disparate frequencies in feedback loop - an added and unnecessary challenge: use Opamp instead
Using the same basic topology, but now using an Opamp (with symmetrical pullup and pulldown) will make tuning the feedback loop much simpler
A small amount of power is needed to run the PWM system itself, power the ICs, drive the gates.
Here is a no-nonsense, high-voltage-capable startup regulator.
Typically, you want to use this only for startup and fill-in scenarios, but it's not efficient, so add a "bootstrap" to supplement, a source that's much more impedance-matched (ie from a pickup coil on the main transformer or output filter) than this brute-strength linear one.
When current is flowing in the freewheeling diode in your buck converter, there would be the I-squared-Vf losses associated with that diode. If you turn on a transistor, ie a MOSFET, in parallel with that diode, you can reduce those losses considerably
Detailing the part of the Buck Converter that detects freewheeling current and sends the signal to turn on a MOSFET switch to turn it on while that freewheeling current is flowing.
This was a first pass.
This toroid has a single large wire through the center where the high current flows. The many turns of small wire go to an amplifier.
This amplifier would make an audiophile cringe. This is like a class AB amp turned inside out. Its purpose is to turn on hard when there's a small current, and turn off hard when the current peters out.
It's an intentionally highly non-linear amp.
For very fast response, ie turning on the freewheeling MOSFET very soon after the series-pass MOSFET has been turned off (and freewheeling current starts), I added differentiating components with the 2 orange capacitors.
Here's one way you might make that amp. It's more of a logical diagram, without the differentiation.
Instead of this method with the inverter and NPN, I experimented with a PNP in the high-side of the totem-pole. That has better inherent blow-thru protection as opposed to this design that can see blow-thru if turn-on and turn-off times of the buffers aren't carefully considered.
This one I've subjected to some testing. This instance should handle about 100V assuming certain loading.
I'm working my way up to solar-string-inverter peak voltage ie 600V.
The main series-pass NMOS and highside driver UC27714 and the startup linear Vreg should be OK but stresses on the switching elements will depend on the main "transformer" coil inductance
With one probe measuring the freewheeling-pickup current-transformer, and the other the drive to the would-be freewheeling MOSFET.....
I built 3 of these, using various values and types of inductors, just what I had on hand
Not enough current is flowing yet to even trigger a sync-rect turn-on. At this low power level, your SMPS's freewheeling current will exclusively flow in a diode. That diode may be the intrinsic body-drain diode of your freewheeling MOS transistor, or it may be a parallel one, depending on design parameters.
As soon as the main high-side transistor turns off the Freewheeling Pickup jumps to a negative value. That triggers the pickup amp which then delivers the Freewheeling Gate drive
The Freewheeling Gate Drive does turn on nicely, but it turns off too soon. Could adjust effective DC bias point/hysteresis trigger thresholds of saturating amp to lengthen on-time.
Don't want to go too long. Absolutely MUST turn off before Series-Pass element turns on.
Editing the pickup circuit can improve sync rect. Here adjusted the bias of the saturating amp, and we stay on during more of the freewheeling phase.
Tried editing another component value that slowed the turnon/turnoff rates of the amp
Slowing the Rise-Time of the gate drive.
This could be an advantage for certain reasons, but in general we want faster rise- and fall-time
in the gate drive