Though the nitty gritty of all his assumptions isn't necessarily completely valid (namely his assumptions regarding pressure curve effects on the bolt thrust), the
Orion's Hammer blowback explanation and calculation page has a very thorough rundown of the concept.
In short;
-Chamber pressures pushes on the bullet and bolt equally, but the bolt head usually has more area and receives more force (barrel receives the difference plus some friction with the bullet, minus some friction with the case)
In long;
-Momentum of the bullet and bolt are essentially equal (there's some minor variance since the barrel is blown forward ever so slightly at first, and if a hammer is used, its momentum must be added to the bolt's)
-Bolt thrusts are so much higher than return spring/buffer forces, that they have an infinitesimal impact on delaying the breech (though the mass of increasingly heavy springs plays a small part), but play a huge roll in determining the rate of cycle and bolt travel length
-A too-light bolt will allow brass to bulge, tear case heads or extractor grooves, and even allow case rupture under pressure. Most designs have the ejection port large and very near the breech so a case rupture (most often due to out of battery detonation in an open-bolt design) is not contained and causes little damage to the gun or shooter
-Nearly all designs ever built to be blowbacks have far heavier bolts than mathematically needed, both for margin, and to keep cycle speeds down to make the guns controllable
Delayed blowback uses mechanical leverage to magnify bolt mass displacement due to case head displacement, with the effect of increasing the momentum needed at the bolt face to open the action. Since the leverage can run the gamut from 1:1 all the way to infinity:1 based on geometry, a vast range of cartridges can be accommodated (7.5 Swiss to 9mm Luger, unless H&K ever made a 50cal or larger version of their CETME clone).
When the ratio is
at infinity:1, you have a locked breech; the case head cannot budge the bolt mass no matter how it tries, and instead the entire (slide+barrel) assembly receives momentum as the bullet is fired. This has the benefits of negating the problem of case rupture from a too-light bolt, as well as slowing down the cycle speed of the recoil operation by adding the mass of the slide and barrel to that of the bolt, and finally by "trapping" and nullifying any extra force delivered to the moving part of the action because of a wider case head (it instead sees only the bore area, and its momentum is therefore equal to the bullet's recoil --thus "recoil" operated)
TCB