Great info, thanks.
One thing for sure - SOMETHING IS BETTER THAN NOTHING!
Planning on the type of rollover?
You could say a friend of mine had a simple slow speed roll. He was off angle on a side hill, going real slow in his TJ. Left side had to go over a rock, slightly too high, flopped onto it's right side almost in slow motion.
Then onto it's top, the fiberglass made that crunching $$$ sound.
Then onto the other side, still slow and easy.
Then onto it's wheels again - nice. But now the hill's too steep.
Ooops, over again, and again, and again - faster and faster - then he dissapeared into the canyon below!
It was too steep to climb down, had to make a trail down around the end.
Fearing the worst we finally got there.
He and his wife had climbed out. He was sitting on the side - As we drove up he said "anybody got a spare tire I can borrow?"
They were fine. We even drove it out.
ERW? DOM? PVC? Water pipe? No, the stock inadaquate TJ cage worked. Not to say the TJ cage will always save you though. But imagine if they didn't have it at all.
Are you only going on trails where you can flop it just once at low speed?
Proper design, proper welds, correct lenghts are far more important than the slight differences in material strength. I suppose even PVC with the proper design could be used - but???
I like things to be NOT WELD DEPENDANT. Find an old "store bought" or custom bar/cage. Cut through the welded areas - look for proper penetration. It's often scarey what you find. To think someone's life depended on that!
Look at cages after they crashed - see how and where they failed or bent. See what held fine. Ask WHY? Try picturing where a small gussett or brace would have helped. And where the ones that are there did their job. Each situation is different, what worked fine on one collapsed on another.
Good design - Meaning -- stresses pushing the bars together rather than pulling or twisting them apart, loads spread out over long distances, braces not positioned like spears to the occupants in case a weld breaks.
For the welding - I said I don't like things "weld dependant." My belief is that the cage should be mechanically sound - the welds only keep things in position till it's tested.
A technique I almost always use - internal stubbing. It's an internal gusset.
Say for example a "T" or butt joint.
Using 1.75 x .120 tube - cut a 1.5" hole in the top of the Tee (hole saw, not a torch.)
Slide a piece of 1.5 x .120 inside the fishmouthed connecting bar, poke it into the hole. Make sure there are no gaps, that it's a good solid connection. Make sure that internal piece goes in as far as possible - reverse fishmouthing gets it in deep. Extend it into the connecting bar at least twice the connector's diameter.
The inside tube supports that connection from the inside - mechanically it's already stout in most directions without even being welded.
Now when you weld it you can turn the heat up and get really good penetration without burn-through. Instead of welding 1/8 material with light amperage you are welding 2 pieces of 1/8 together - equiv to 1/4". Even it you never penetrate through the full 1/4, for sure you've penetrated the outer 1/8.
Mechanically it's strong, and the welding insures it's going to stay that way.
Sure, it's lots more work.
(And it sure makes it easier to construct the cage - you can usually put it all together without welding, making changes without scrapping welded pieces. Once it meets with your approval, tack it all and then weld away!
I started doing that technique right after I witnessed a rear brace's weld break, the brace slid around the main hoop and went through the seat and the driver's back and out his chest - all the way through and out the front! Ouch!!!!!!
Paramedic helicopter, fantastic life saving guys on it - They cut the brace off and took him in with it still in him.
Believe it or not, HE LIVED!!!!! I can't give enough credit to those flying paramedics!
So I guess my main point is "Thmimk" - and use the best design you can. But do SOMETHING!!!!