Maximum Power Point Tracker (MPPT) charge controller - what drives the weight?

I'm having trouble finding any literature paper that studies this problem. What drives the weight of an MPPT charge controller? How would I go about reducing that weight if I had to design a custom design? The only thing that even alludes to the power-to-weight of an MPPT is this paper, behind a firewall so I can't read it: https://ieeexplore.ieee.org/document/7990724

The clue here is that it's the power transistors that are massive? If I had to design an MPPT with an expected range of input/output voltages and currents, how could I go about determining the heavier objects? Maybe this schematic was potentially the most helpful thing I found, which details the inside of an MPPT: https://www.ti.com/lit/ug/tiduej8a/tiduej8a.pdf?ts=1593845160703&ref_url=https%253A%252F%252Fwww.google.com%252F

But there's no mention of weight of anything. Of course, I could buy an OTS MPPT and strip off the housing, etc. to reduce mass, but I'm trying to go for a more sophisticated design approach. I guess weight has never mattered to MPPTs, being installed near large solar panels on houses, etc., but it is an interesting problem (I think?). What are the components by weight? How can they be reduced (e.g. GaN power transistors apparently?)? And what's the driving factors (e.g. is it current that pushes the need for heavier everything?)

Apologies if this is basic electronics - I am new to the electronics world.

• You can minimise the mass of the unit by designing for one battery voltage, eliminating the display etc but then it becomes targeted to a particular use and not so sellable. – Solar Mike Jul 4 '20 at 7:34
• MPPT charge controller includes tiny PCB's that can charge on 12V battery at 1 or 2 Amps, and it also includes giant inverter/charger/controllers that can charge a 10kWh battery bank or even bigger. But in general, I think thermal components such as heat-sinks will end up being the heaviest. Current probably is the single biggest driver toward larger heatsinks. – mkeith Jul 4 '20 at 8:03
• One approach, if component cost is not a major constraint, would be to use very low resistance switching elements (possibly MOSFET's) and try to get very high efficiency. If you can achieve 99% efficiency with 1000W you only have to dump 10 Watts. But if it is 95% efficient you need to dump 50 Watts. So you can maniacally pursue high efficiency at every point. Forced air cooling will probably be a win for weight, but not necessarily for reliability. If some other type of cooling is available due to the particular use case for this unit, maybe you can use that. Liquid cooling for example. – mkeith Jul 4 '20 at 8:08
• Another approach is to increase the switching frequency. MPPT charge controllers are (usually) basically buck converters, which store energy in an inductor when the switch is ON, and release it when OFF. The size (and weight) of the inductor depends on the energy storage needed, so cycling it faster reduces the energy per cycle and thus the inductor size. But switching faster implies more switching losses, and more heat, and bigger heatsinks, so ... balance this with mkeith's efficiency drive. – user_1818839 Jul 4 '20 at 12:21
• USUALLY buck converters are most efficient when Vout is close to Vin. This is because the duty cycle is highest, and ripple current is lowest. In a buck converter with a large step-down ratio, the inductor will actually be a transformer. This means the duty cycle can be 100 percent when Vin/Vout = k, where k is some constant number. If both Vout and Vin are highly constrained, k can be chosen so that the duty cycle is close to 100% all the time, and thus more efficient. – mkeith Jul 4 '20 at 18:24