Spaceframe joints with dissimilar tube sizes

and lastly the modulus of the Chinese steel you will be using :-)

Hmmm, tell me how to tell the difference between Chinese 1010, German 1010, USA 1010, etc. ..... and I'll let you know.

I'll let you know on the rest.

The wall where the pipe is welded to will quite quickly buckle then, depending on the pipe 's strength as to if it bends first or not, it will tear the wall area out whereas passing the tube through will just bend eventually.






I'll take that challenge.




Am I supposed to be surprised that having a 100mm long tube and putting a 40mm gusset up it's length produced a stiffer result? Show me a spaceframe with 40% long gussets all over it .... I believe it was you defined the wrong question.

So go back and use a 1 meter long 30mm tube, one through the box and one welded to the wall with a 30mm long gusset in FEA and see what results you get - this is the test I will do and we can compare those but make it 2mm wall.

Ok done that - can't post a picture of the results for technical reasons but the comparative stiffness' for the four still with the tube loaded axially and now 2mm wall thickness are;

with tube through the section 1.3x
with 30mm gusset 1.8x
with gusset and rib 2.3x



The 1m tube rather swamps the display so zoomed in with distortion exaggerated;

Edited by rachael, 04 January 2012 - 20:08.

I would think that when it was realized that gussets should be welded to the neutral axis (i.e. least loaded section of the tube), this became much less of a concern.

Ok, I can see that - I have always had the habit of welding to the 'neutral axis' anyway using 2 x thin gussets to the outside rather than a thicker singular gusset in the middle and that came about from simple experiments bending by hand way before computers of course*.

*Don't anybody get the wrong impression, FEA is brilliant and has helped tremendously towards better, reliable and cheaper products.

with tube through the section 1.3x
with 30mm gusset 1.8x
with gusset and rib 2.3x

I'll let you know, maybe before the end of the week.

Edited by cheapracer, 04 January 2012 - 13:24.

Odd design or not - the Birdcage was very successful in racing.
I recently saw on "Design for Victory" (one of my favourite shows) some fairly close-up views of the welding on the Birdcage - the welds appeared to be untidy verging on the diabolical.

Which is why John Barnard insisted on some parts being manufactured by Bob Sparshotts company, he knew the quality of the welding.

NeilR, my advice should be taken with a grain (to put it mildly) of salt, but that small triangular 'gusset' in that corner is less than optimal solution- the way I see it the most important welds will be placet in least favourable position in respects to tubes it's supposed to brace. Almost all sources nowdays recomment welds along neutral axis of the tube (putting the welds in areas with least stress).

I would consider using a modification of the 2nd method (sheet metal gusset) shown in the picture from Foale I enclose, but instead of welding it along the neutral axis, using the outermost edges of the structural tubes. That way, the open tube would be plugged, allowing easier attachment of the round tube...

Thanks for the thought. The gusset is there for two reasons. The first is structural, though it may not be required. The second function is to provide a greater bond area for the shear panel that will form the firewall at this point. The intended joint had been a but weld to a local thickener, with an internal diaphragm/web, which is 40mm into the tube. All steel is 350 mpa mild steel. A gusset as in you picture will make it very hard to bond the plate or honeycomb panel. The structural acrylic adhesive I have is good, but has a limited tolerance for gaps greater than 1.2mm
Rachel could you repeat your fea model with a internal shear web, full width and a doubling of the skin thickness via a plate? I would be interested to see if this works.

Thanks for the thought. The gusset is there for two reasons. The first is structural, though it may not be required. The second function is to provide a greater bond area for the shear panel that will form the firewall at this point. The intended joint had been a but weld to a local thickener, with an internal diaphragm/web, which is 40mm into the tube. All steel is 350 mpa mild steel. A gusset as in you picture will make it very hard to bond the plate or honeycomb panel. The structural acrylic adhesive I have is good, but has a limited tolerance for gaps greater than 1.2mm
Rachel could you repeat your fea model with a internal shear web, full width and a doubling of the skin thickness via a plate? I would be interested to see if this works.

Neil no problem - could you sketch what you mean exactly by internal shear web though. If you post some dimensions it's quick to setup similar or even more complicated models if there is a particular layout you want to look at.

Neil no problem - could you sketch what you mean exactly by internal shear web though. If you post some dimensions it's quick to setup similar or even more complicated models if there is a particular layout you want to look at.

I have cut a 50mm square of the original material and have welded it into the tube 40mm in, bracing all four walls of the tube. The tube will then be capped by the same material (almost creating a steel 'cube') and then a 70mm x 50mm x 2.5mm plate added to where the round tube is but welded.

Wolf the planned bulkhead has this design:

Edited by NeilR, 04 January 2012 - 21:53.

[quote name='NeilR' date='Jan 4 2012, 21:48' post='5469010']
I have cut a 50mm square of the original material and have welded it into the tube 40mm in, bracing all four walls of the tube. The tube will then be capped by the same material (almost creating a steel 'cube') and then a 70mm x 50mm x 2.5mm plate added to where the round tube is but welded.

ok I think I understand. The joint you are talking about is at the frame corner rather than mid-span as I have modelled which will make a difference.

...and lastly the modulus of the Chinese steel you will be using :-)

Hmm. I'd been led to understand there aren't huge differences between steels. Measurable sure, but not huge.

Incidentally the old standard I have seen for testing joint stiffness is to cut all relevant mebers 1 foot to 500mm away from the nominal joint node, and then fix (123456) the cut ends except for the member under test.

The 1 foot thing is a bit arbitrary, and would be inappropriate for the larger joints we see nowadays such as B post to rocker. The 'logical' distance would be halfway to the next joint.

I'm not a huge fan of 123456 on plate elements , but it sort of replicates the test rig, which was usually a continuous weld to a 6mm plate.

As you might guess car companies have libraries of joint analysis, in the olden days our test labs spent a fair amount of enjoyable time building rigs to measure the joints.

You should be testing them in torsion and as cantilevers, not just axially.

Hmm. I'd been led to understand there aren't huge differences between steels. Measurable sure, but not huge.

Sorry that was a poor attempt at sarcasm on my part - 200GPa covers most steels.

Incidentally the old standard I have seen for testing joint stiffness is to cut all relevant mebers 1 foot to 500mm away from the nominal joint node, and then fix (123456) the cut ends except for the member under test.

The 1 foot thing is a bit arbitrary, and would be inappropriate for the larger joints we see nowadays such as B post to rocker. The 'logical' distance would be halfway to the next joint.

I'm not a huge fan of 123456 on plate elements , but it sort of replicates the test rig, which was usually a continuous weld to a 6mm plate.

As you might guess car companies have libraries of joint analysis, in the olden days our test labs spent a fair amount of enjoyable time building rigs to measure the joints.

You should be testing them in torsion and as cantilevers, not just axially.

I haven't used 456, just 2 on the cut plane and 13 on one edge of each of the ends. [123456 refers to the degrees of freedom at the nodes, 123 are displacements in x y z and 456 are the rotations about x y and z so 2 would be fixed in y and 6 would be no rotation about z]

No problem changing the way the load is applied - just trying to KISS (keep it simple stupid) ;-)

Sorry that was a poor attempt at sarcasm on my part - 200GPa covers most steels.

Darn, missed the little face thing. As you were then

The world of bicycle frames might be worth looking at...

Excellent recommendation. The best solution here is a lug fitting like bike frames use. One end of the lug can fit inside the open end of the square tube, with a perimeter weld. The other end of the lug should fit over the round tube, with a 3D or 4D overlap and a "fishmouth" weld joint. The lug can easily be machined/welded from pieces of plate and round bar.

The reason you want a lug here is due to the huge mismatch in local stiffness between the smaller round tube and the highly triangulated square tube chassis corner. The lug would help to diffuse stress in the small tube end.

slider

There were no 'points of view' in my post, simply details of how the model was set up and a statement on the relative stiffness' of the different arrangements. Clearly along with manners you need lessons in English as well.

Sorry if this came across as a bit rudish - that was not my intention. However it does seem these days that I endlessly get engineers quoting "FEA" at me to justify just about any position. If you have no idea how to do FEA (as I don't) it is annoyingly hard to argue against.

...... All steel is 350 mpa mild steel.....

NeilR,

Don't know how serious you are about your weld designs, but if you want to do a proper stress check, you need to apply the appropriate FoS, Kt and weld knock down factors to your analysis. With MIG welded mild steel tube (1020 or 1018) using ER70 wire, your max operating weld joint stresses should be 70MPa or less.

Good luck.

slider

Moderately serious. I am interested to see how my planned joint fares in Rachel's fea. I cannot continue the tube through the shs as the angle would mean it would go through two adjoining walls and not opposing walls of the shs. It would also be a pain in the arm to do!

Do you happen to know the measured yield strength of 350 MPa steel in the HAZ?

Thanks Charlie, the lower pic is very achievable for a hand made product.


side of chassis needs little help:



Corner node but weld is where I have concerns - the round tube should be higher than in pic to align with node, but you'll get the idea:




Rear of chassis:

I have to honestly say that there is (in my opinion of course) far too much agonising about perfect triangulation of spaceframes etc. Trying to get things perfectly triangulated while putting brackets etc. in the right spot is an almost impossible task. Trying to do this will just overcomplicate the design and drive you mad at the same time.
Some of the classic spaceframe designs have got what appear to be dreadful weak points and that didn't make them any less successful. Even that paragon of the perfectly triangulated chassis, the late C. Chapman took some dreadful liberties - for example on the Lotus 19/Monte Carlo two-seater sportsracer. The front suspension spring units mount directly into the centre of an unsupported tube. Costin and Phipps (pages 12 and 38) seem happy to accept this and justify it as being OK - presumably if Chapman did it - it must be alright.
And look at the chassis of the Atom in the neighbouring "Clever" thread - dear oh bloody dear. A lot of things untriangulated, unnecessarily curved tubes, pipes running in all directions. But there is little doubt that it "works" OK and they have sold a hell of a lot (and it seems to crash satisfactorily as well). If you handed that drawing of the Atom chassis in to an engineering lecturer, how many marks out of ten would you get? Maybe two (if you were lucky) - and maybe a suggestion that Media Studies is a good degree to do.
Presumably the photos are of the Godiva? It looks good - sturdy as buggery etc. No need to worry about tiny details about how tubes join onto other tubes etc. Just weld on a bit of plate and stick the tube to it (I think as Cheapy and others suggested) - it'll be right no worries. You really will never notice the difference.

Having said the frame looks sturdy (over sturdy even) - the seemingly separate rear spaceframe structure does not appear to be well-connected to the rest of the car - presumably there is going to be some lower frame members added?

Greg, not really. The steel is dual grade 350lo/450lo from one steel.
Kelpie, maybe. The thing is I am not comfortable using chapman as a model of current practice. Tyre loads and safety have changed substantially and I would prefer to look at mallock's work on the Saleen cars etc as evidence of good practice. Godiva is sturdy, but then for the obvious reason. Of course the other factor is that it is simply easier to make a chassis that will remain straight with bigger tubes during construction, well at least that was the professional advice from one bigger constructor. Larger tubes cope better with weld stresses etc. also at the end of the day an additional 25kg really does not matter.

Some of the classic spaceframe designs have got what appear to be dreadful weak points and that didn't make them any less successful. Even that paragon of the perfectly triangulated chassis, the late C. Chapman took some dreadful liberties - for example on the Lotus 19/Monte Carlo two-seater sportsracer.

Except for historical curiosity, I don't know why we should give a rat's ass what they were doing with the "classic spaceframe designs." The tire contact patch was smaller than a playing card and the engines produced less output than a current Hyundai Accent. Really, the purpose of the suspension was to prevent the four tires from touching each other while the function of the space frame was to keep the oil pan off the ground. Torsional rigidity was not required while impact safety was utterly irrelevant. Today we have totally different standards and requirements -- much higher ones.

Of course the other factor is that it is simply easier to make a chassis that will remain straight with bigger tubes during construction, well at least that was the professional advice from one bigger constructor. Larger tubes cope better with weld stresses etc. also at the end of the day an additional 25kg really does not matter.

That seems like an eminently sensible approach, especially when it's your butt in the seat.

[1] Cap, with 0.020" thicker material, that open square tube first (i.e. make sure it's welded to all 4 sides of that tube).

Wow, great thread seeing Wolf and Engineguy posting

I would solve the entire problem with a single cap as shown, ties the corner 3 tubes together, replaces the gusset and spreads the load from the smaller tube - something like 3 or 4mm plate.....

Edited by cheapracer, 05 January 2012 - 12:47.

Moderately serious. I am interested to see how my planned joint fares in Rachel's fea. I cannot continue the tube through the shs as the angle would mean it would go through two adjoining walls and not opposing walls of the shs. It would also be a pain in the arm to do!

any reason why you cant go to a larger diameter tube that will join much nicer into the box section ?

I don't see the problem. If you cap the open tube and join the round tube at/near the intersection of the four square tube segments, then weld the round tube 360, you have the cat's ass right there. If you add a doubler (Cheapracer) or two 90-degree gussets (Engineguy) you have medieval war engine capabilities -- ideal for ramming embattlements and so forth. Good to go.

In thinwall construction the problem with a small tube intersecting a large one is that the load is focused on a small area of material in a tearing mode. I don't see that here.

Except for historical curiosity, I don't know why we should give a rat's ass what they were doing with the "classic spaceframe designs." The tire contact patch was smaller than a playing card and the engines produced less output than a current Hyundai Accent. Really, the purpose of the suspension was to prevent the four tires from touching each other while the function of the space frame was to keep the oil pan off the ground. Torsional rigidity was not required while impact safety was utterly irrelevant. Today we have totally different standards and requirements -- much higher ones.

Yeah, screw all those fussy little tubes and critical welds... bah!

119 lb sheet box frame from 1400 lb Porsche 904...

I don't see the problem. If you cap the open tube and join the round tube at/near the intersection of the four square tube segments, then weld the round tube 360, you have the cat's ass right there. If you add a doubler (Cheapracer) or two 90-degree gussets (Engineguy) you have medieval war engine capabilities -- ideal for ramming embattlements and so forth. Good to go.

In thinwall construction the problem with a small tube intersecting a large one is that the load is focused on a small area of material in a tearing mode. I don't see that here.

100% agree with this

100% agree with this

Thanks, that's very generous. I wasn't sure about the portability of the term cat's ass as a structural benchmark. Superior to a frog's, certainly, but not waterproof.

The world of bicycle frames might be worth looking at...

Excellent recommendation. The best solution here is a lug fitting like bike frames use. One end of the lug can fit inside the open end of the square tube, with a perimeter weld. The other end of the lug should fit over the round tube, with a 3D or 4D overlap and a "fishmouth" weld joint. The lug can easily be machined/welded from pieces of plate and round bar.

I mentioned bike frames not just for the use of lugs, but also double- and triple-butted tubes. I don't know if these have ever been used in automobile chassis.

I'm not sure as well whether butted tubes have been used in automotive applications, or for that matter in motorcycles or aircraft where for instance Reynolds 531, the quintessential bicycle tubing, was also reasonably commonly used at one time. It makes mechanical sense to use butted tubes in a space frame whether from a pure weight/stiffness standpoint or even short butts in the HAZ allowing thinner wall in the centers but there are probably usually cheaper ways to shave a few kgs.

I mentioned bike frames not just for the use of lugs, but also double- and triple-butted tubes. I don't know if these have ever been used in automobile chassis.

Look for posts by McGuire, Cheapracer, and myself in this thread:
Mr. Gasket SS frame lugs...
.

I have cut a 50mm square of the original material and have welded it into the tube 40mm in, bracing all four walls of the tube. The tube will then be capped by the same material (almost creating a steel 'cube') and then a 70mm x 50mm x 2.5mm plate added to where the round tube is but welded.

I think this is what you described minus the extra plate (still a shell model but with thickness to aid visibility);



This is 1.8x stiffer than yesterdays baseline;



However I don't think this is a good design because it has a severe stress concentration here that is likely to crack the weld;



I understand that cutting an oblique hole is not easy but Cheapy's design is particularly good in that it doesn't have this stress concentration feature and the joint is effective however the tube is loaded.

any reason why you cant go to a larger diameter tube that will join much nicer into the box section ?

Yes. A larger tube will be too big to attach to the damper tube due to the angle and fish mouth required. Yes I could preen the ends over to shorten them, bt the tubes are already generously sized and good enough for the job. In actual fact the 38mm tube at the chassis end when cut makes for a roughly 44mm oval shape.
Kelpie there will be a fairly substantial lower subframe running under the engine and transaxle, which are transverse.
Rachel thanks for that answer. I had considered the weld stresses...what effect would the added plate have. Also what effect would the oblique angle of attachment have?

Edited by NeilR, 05 January 2012 - 21:17.

Given that these are quite large diameter tubes by some standards is it possible to cut into the smaller tube and add fillets to incease it to the bigger size?

OR , maybe easier, waist down the bigger tube with welding to the smaller size through the joint area.

Another solution could be to insert a sleeve of about 2- 3 mm material inside the bigger tube so it will slide down to the joint area and fix it with plug welds. Then you have internal reinforcement to prevent wall "panting" at mismatched joint. If the insert was about 30mm longer than teh stressed area the plug welds would be away from any connection welding to minimise distortion

Yes. A larger tube will be too big to attach to the damper tube due to the angle and fish mouth required. Yes I could preen the ends over to shorten them, bt the tubes are already generously sized and good enough for the job. In actual fact the 38mm tube at the chassis end when cut makes for a roughly 44mm oval shape.
Kelpie there will be a fairly substantial lower subframe running under the engine and transaxle, which are transverse.
Rachel thanks for that answer. I had considered the weld stresses...what effect would the added plate have. Also what effect would the oblique angle of attachment have?

I have a feeling that adding the plate would make it worse - sounds counter-intuitive but the plate will 'break' the stiff load path where the stress concentration is. The joint behaviour will also be non-linear both with load and with load direction as the plate will separate from the box section - difficult to describe but imagine a cardboard tube glued to a piece of card and then the card taped down round the edges to a table. As you push and pull on the tube the card will lean on or pull away from the table depending on the load direction. This is a much more complicated fe model beyond 'free'!

RachAEl

I have a feeling that adding the plate would make it worse - sounds counter-intuitive but the plate will 'break' the stiff load path where the stress concentration is. The joint behaviour will also be non-linear both with load and with load direction as the plate will separate from the box section - difficult to describe but imagine a cardboard tube glued to a piece of card and then the card taped down round the edges to a table. As you push and pull on the tube the card will lean on or pull away from the table depending on the load direction.
RachAEl

That's the reason I suggested capping the (2"x2" sq?) tube first, and no subsequent larger plate... but you explained my concern clearly where I didn't.
.

Look for posts by McGuire, Cheapracer, and myself in this thread:
Mr. Gasket SS frame lugs...
.

Well, I've read it a couple of times and I can't see anything about double-butted tubing...

http://www.equusbicy...ebrochure03.pdf

Edited by Tony Matthews, 05 January 2012 - 23:24.

In actual fact the 38mm tube at the chassis end when cut makes for a roughly 44mm oval shape.

I would not plate the end , run the end of the tube into the box section by a couple of mm and weld it up , no need to introduce flat plates and stress points

Wow, great thread seeing Wolf and Engineguy posting

I would solve the entire problem with a single cap as shown, ties the corner 3 tubes together, replaces the gusset and spreads the load from the smaller tube - something like 3 or 4mm plate.....

Exactly Cheapy - that would do the trick. Looks neat, is simple and will never fail (unless you have an Atom-style crash).

I do not want to seem to jump on the bandwagon, but I did a quick sketch last night but opted not to post it, because I was under the impression NeilR wanted that slanted part of his triangulasr gusset (or should say brace) to bond the firewall to. And that he'd have trouble bonding it to the single sheet gusset that was welded to the side of the main tube (firewall, if tunning over the tube accross the frame would, in theory actually be bonded to it quite nicely, on account of adhesive being loaded in pure shear), but I presumed there was an issue of thicknesses of the parts that would lead to the gap he mentioned...

But I'd post my version, which is quite similar to Cheapie's, except for having more pronounced gusset function, and would welcom criticism if warranted...



EDIT now that I've put up this purdy, if not quite useful model, and assuming I read NeilR's posts correctly... I'm thinking it would not be terribly hard to blend those two in single 'solution'. If one was to get rid of the curvature in the middle section of the gusset and add the mating surface for the firewall and perhaps even help some gusset effect and at almost no cost, save some small nuisance. I'm not saying it would be pretty (or even sensible), but it would be quite doable- all it would need would be some fancy cutting of the plate and making 2 90° bends, which I'd think quite feasible and on very low budget... It's quite late around here, so if I don't come up with a model within 10-15 mins... don't take offence if I do it tommorow.

Edited by Wolf, 06 January 2012 - 02:32.

Sorry for not bothering with proper bending radii, and being generally a bit untidy and slow, but here's something that hopefully might work for NeilR...

I have a feeling that adding the plate would make it worse - sounds counter-intuitive but the plate will 'break' the stiff load path where the stress concentration is. The joint behaviour will also be non-linear both with load and with load direction as the plate will separate from the box section

That may be true for a sign base standing in the wind but in this application within the total sum it's never going to happen.

Wolf, I like your curved gusset, it is a stronger way and looks purdy - I think that's enough though and no need for the inside corner box gusset, again looking from a total sum POV as the frame is considerably overbuilt anyway.

Greg, not really. The steel is dual grade 350lo/450lo from one steel.
Kelpie, maybe. The thing is I am not comfortable using chapman as a model of current practice. Tyre loads and safety have changed substantially and I would prefer to look at mallock's work on the Saleen cars etc as evidence of good practice. Godiva is sturdy, but then for the obvious reason. Of course the other factor is that it is simply easier to make a chassis that will remain straight with bigger tubes during construction, well at least that was the professional advice from one bigger constructor. Larger tubes cope better with weld stresses etc. also at the end of the day an additional 25kg really does not matter.

Do you have a picture of a Mallock/Saleen chassis you can post?

I can post some in a couple oF days. I'm away from my computer at the moment.

I can post some in a couple oF days. I'm away from my computer at the moment.

Logic and deduction tell me that you must be near "A" computer of some kind.

iPad and pictures and passwords are on office computer

Do you happen to know the measured yield strength of 350 MPa steel in the HAZ?

Greg Locock,

The wisenheimer answer would be 350 MPa, right?.

But kidding aside, the only sure way to answer your question would be to perform a metallurgical analysis of the welded joint. But as I noted in a previous post, when analyzing something like a welded joint it is standard practice to apply knock-down factors for things like weld defects, stress concentrations, thermal strains, etc. While your base material might have 350 MPa tensile strength before welding, the weld filler material and the HAZ might have porosity, oxidation, inclusions, etc. For analysis purposes, you should assume that the material in the HAZ has a strength equivalent to the lesser value of base material or filler. Then you would apply the appropriate Kt, R factor, knock-down factor, and FoS to your calculations.

With welds in heat treatable steel alloys, the strength in the HAZ should be considered as annealed, unless a post-weld heat treatment is performed.

slider

Logic and deduction tell me that you must be near "A" computer of some kind.

Mallock Saleen




Mosler:


Artega
http://www.artega.de...gt/details.html

This is a great thread because so much basic "engineering intelligence" is being discussed and analysed with modern tools.

Could I just make a comment from a purely amateur viewpoint, lacking FEA etc.

The pictures show a chassis with mostly square tubes, if I understand it correctly square tubes are less efficient then round tubes of the same diameter in terms of weight versus stiffness. The first spaceframes ( aircraft) and early racing car spaceframes were all round tube. Even the US builders of Oval track and drag cars used, mostly, round tubes however bad the frame design was.

Square tube frames were introduced by the UK racecar builders to simplify production ( no notching/fishmouths ) and to make attaching panels etc. easier.


So, with no disrespect to anybody if the chassis is not designed for maximum efficiency, but maximum stiffness is vital ( hence the joint flexing question) then why not

1) Reduce the size of the bigger square tube but increase its wall thickness so as to have the joint properly matching.

Or

2) switch the critical joints to round tubes with proper fishmouthing so the differences in width are dealt with via the oblique angles at which the two tube walls meet. I don't have the FEA skills to validate it but I think that the round to round joint would be much stiffer and a bit of old fashioned hammering would blend two sides of the joint almost perfectly.

One last thought, all of this is subject to the skill of a the welder to get full penetration without distortion. Stress releiving looks unlikely given the frame size so I suspect the real world results are likely to be as much welder-skill driven than mismatched wall alignment driven, hence the earlier suggestion of brazing.

The pictures show a chassis with mostly square tubes, if I understand it correctly square tubes are less efficient then round tubes of the same diameter in terms of weight versus stiffness. The first spaceframes ( aircraft) and early racing car spaceframes were all round tube. Even the US builders of Oval track and drag cars used, mostly, round tubes however bad the frame design was.

I believe this is one of the more misunderstood and overhyped aspects of hobbyist chassis design. Yes, round is slightly stiffer than square per unit weight. However, square is stiffer for a given tubing size. There's the grand morsel of theory for everyone to beat to death, anyway.

But the real issue is not in theory but in practice. In order to use the chassis frame for anything, you will need to mount suspension, drivetrain, body, fuel tank, etc. and there's the advantage of square/rectangular tubing.

Let's be honest here: the real reason enthusiast builders use round instead of square is that in their minds, round looks racy and square looks agricultural. If they backed up a few steps and took a more balanced and objective approach, they'd be more likely to choose square tubing. I like the car pictured here. It represents sound, practical, intelligent design.