Oh, didn't see the attachment
Yes, in place of R5, except it's usually placed in the drain instead of the source circuit. Same idea.
Eww, LM358, I hope that's not what's actually there. No way in hell a 358 is fast enough for a switching converter current sense, marginal at best at unity gain, no way in hell at a gain of 30!
Bizarre, they used a semi-discrete 10.83V regulator with current limiting instead of a 7812 (or similar). It probably works worse than a 7812, and I don't see how it could be cheaper. Why?!
Example:
http://seventransistorlabs.com/Images/Mag_Amp_PSU.pngI haven't built this, but it should work. Main thing lacking is slope compensation, which I think for the values shown is actually a problem (it'll operate in the chaotic region -- nasty).
By the way, this was sized for 3A 30V output I think. The snubber is already recommended, and for 70A capacity, you'll *need* not just one, but several, spread amongst the switching transistors (plural) and diodes.
If you really insist on building this thing... you'll need to clarify that "70A peak" demand. For how long? If microseconds, it doesn't matter -- it draws from the caps and doesn't affect anything else. If miliseconds, enough cap can be added to accommodate it. But more than that, and you must have the converter sized for
at least that much current. The only thing you can skimp on then is heatsinks, which isn't saying much.
Doing just 20A capacity, continuous, in just a few transistors and diodes (one or two each), is possible, but you'll have an easier time using 2-4. You'll need some heatsinking, possibly a fan, but it won't be obnoxious. A snubber on every other transistor is probably a good idea, so 1-2 of those. And as many caps as you can stand, since skimping on those *will* release the magic smoke\\\\\steam.
Tim