OK, I got the picture with the mosfet reverse conduction. I was a little confused because of my experience with vlsi design where mosfets are symmetric. Looks like power mosfets are not symmetric, and include a reverse conduction "body diode", so they can't block any current in reverse. So he places the mosfet in in reverse so that it can block in reverse and then turns it on at the same time he turns on the other mosfet. This obviously forces a double voltage drop. In his case, both drops together are probably under 0.6 volts.
RDS for power mosfets from the IR datasheet:
small transistors size of Tim nolan's (to-220)
v RDS MaxI
60 .028 30
100 .077 17
200 .18 10.5
250 .28 7.9
Larger TO-3P
V RDS MaxI
60 .014 62
100 .055 41
200 .085 23
250 .14 18
the IRFIP150 (100 volt .055 41 amps) is around $3 a piece
You can see that as soon as we jump out of the 60 volt transistor range, the voltage loss rises rapidly. For the IRFIP150, it is quadruple the loss of the 60 volt version. Just passing 40 amps through it will cause an over 2 volt drop and incur the transistors maximum power dissipation of close to 100 watts. To get this loss under control we will have to have 4 of the things. ($12) That will only drop the voltage drop to 0.5 volts which doesn't beat a diode. And that is only a 40 amp charger ... (whewsh) To get the same efficiency for 60 amps (a reasonable size charger) ... 6 transistors..... $18 + heatsink + silicone padding/screws (probably reducible for large quantity orders). A 40 amp relay is $5. A 60 amp relay is $5 (I got them for $4). A 120 amp relay is $12 at aliexpress.com .. .. and the relay solves the reverse blocking problem for free, and without voltage loss. So you can see that if your panels match your batteries, relays have the clear advantage for higher voltage charging. Their problem is that they don't pwm well.
Panels have open circuit voltages considerably higher (like 20% higher) than their maximum power voltage, so charging a 48 volt bank will require an open circuit voltage of 75 volts to do pwm and something closer to 150 or 200 to do mppt. You need some spare voltage on the transistors for protection against inductance, etc. On the other hand, the battery helps to give you that spare voltage. To do mppt for a system that can accept panels during the day and rectified 120vac at night, you need the 250 volt transistors. Just 10 amps per transistor will incur 1.4 volts of loss .. 80 amps ... 8 transistors, probably $25 + heatsink bla bla bla, and if you want to block, use a diode or a relay cause your transistors are imposing a lot of loss. To get your transistor loss down to diode level, you have to triple your transistor count ... (24 transistors) .... ouch. Oh woops ... I forgot something... to get the 80 amps out, you may only need 40 amps in for mppt, That sounds like it could halve the transistor count, but it really doesn't because the 40 amps will be bunched up into half of the duty cycle, making it 80 amps and causing the full voltage drop half the time. As the voltage drops down, the duty cycle rises, causing the same voltage drop, but for more time. So there is some efficiency advantage at higher voltages because the voltage drop is smaller compared to the total voltage. It shows up in less loss because of a shorter duty cycle.
Obviously, to move a lot of power at 12 volts, we are talking about really high currents, like 100s of amps (400 for a 5 kw system). So you can see how the 12 volt solution works great for a small system, but the low RDS doesn't save you if you want to move much power. To move 200 amps for 2.4 kw, you have to have 10 60 volt transistors at 20 amps each, incurring a 0.28 volt loss which doubles to 20 and 0.56 for a reverse voltage blocking system. So that's about 5% power loss. To do it with relays, we are talking 4 relays at $5 each, free blocking and no appreciable loss. Now if we put the four relays into a 3 module system. first module does 100 amps, second and third do 50 each. The controller detects the current flow of each module, and for multi-stage charging, it simply turns off the modules it doesn't want to run in stead of pwm-ing it. Only thing remaining is to match the panels to the batteries, and you'll still be ahead if your panels have an extra 0.5 to 0.7 volts.
A 24 volt system may still be able to use 60 volt transistors. This creates a competency window for transistors. We can basically cut the percentage power losses of the 12 volt system in four, or only cut them in two and halve our transistor count. Unfortunately, we can't mppt up to very high voltages this way, but simple pwm power losses will be relatively low.
So it appears that transistor/mppt charging is only competent
a:with small, low power systems
b:systems where we want to raise the voltage to avoid distance IR losses
c:when panels of the proper voltage are more expensive, demanding use of higher voltage, cheaper panels
d:when we are too clueless to match our panels to our batteries
e:maybe 24 volt systems that use low resistance transistors and don't mppt up to over 50 volts.
and it seems to me that d: is probably the most common of these conditions