How precise is Haydn diagonal rods when assembled

I got Haydn's magnetic rods some time before. I haven't built my delta printer yet, waiting for the PCB effector. But time is comming so I decided make some preparatory work. I measured ball stud diameter hoping sort out the pairs with best matches. Results are in the picture below,
My setup is as follows. 4 section plane through rotational symmetry axe AD give 4 circles which diameter are measured in 3 positions 13. That gives 12 measurements. It can be seen from the table that the diameters in horizontal plane are less then ones in oblique planes by 0.060.08 mm so sphere is not perfect somehow.
The most relevant for us are the zone of 23 measurements as that is working area for diagonal rod movements of delta printer. Fortunately the sizes in these zone are very cosistent for all the ball and have deviation of 0.010.02 mm and for nine of twelve balls is pretty near to nominal value for 3/8" balls. But three balls have diameter that are less by almost 0.1mm. The rods that I got have specified length on them that differ no more then 0.02 mm from each other. I suggest these were measured with etalon balls or just approximated from the length of rod alone assuming ideal ball size.If I would use existing ball studs I get the difference in diagonal rods on order of 0.1mm. Moreover I get one pair of rods with difference in length within it of the same value ( I have 3 balls with smaller diameter). Having no experience in 3D printer I am not sure how it may be relevant to precision of Delta. I am going to include another set of twelve ball stud into my order of PCB effector and electronics and hope that I can choose twelve identical balls (similar as David recomended choosing TRAX rods among double set).
Hope more experienced users can give an advice if it is worth to worry about that.

I'd be interested in the runout of the rod itself. Do you have any way to measure that? I put some 6mm carbon rods that I have into my lathe and was surprised how much there was.

Not tried yet but it is very interesting too. I think the rods have not much rotational movements when working in Delta.

and run out is not at all important as that wont affect the length

If you can use these measurements to get each rod pair exactly the same length, and then even get all 3 rod pairs exactly the same length and then couple that with perfection in the linear motion, the carriages, the effector and you will have the recipe for a very accurate printer.
However I did none of this, I do have linear rails, metal effector and carriages, and Haydn's magnet system. I can calibrate to as low as 0.003mm deviation so whilst I think you make a good point, I think we are talking about the tiniest of tiny practical effects when you actually assemble the printer, so many other things can affect the precision that .
That being said I am going to measure my arms and balls (I have a spare set of 12, so finding the right sizes, to match with the rods and swapping them all around might be worth the effort. It will either calibrate even closer to the holy grail of 0.000 or it won't).

DjDemonD, thank you for your reply. But my question was how crucial would be 0.1mm difference in the length of diagonal rod pairs for Delta precision. BTW I have 360mm Haydn's rods.

ah OK now that is a slightly different Question a 0.1mm difference is to much IMHO and would result in some Effector tilting as it moves around and this is what causes the biggest Issues when calibrating unless you are using the Nozzle as the probe.
the tilt can also cause a slight inaccuracy in the resultant print.I have just received a set of Custom Haydn rods which are matched to within 0.01 mm 4 are 450.57 and the other 2 are 450.56.

I would like to point out that ball studs are not ideal. They can have the difference in their diameters as much as 0.1mm that was in my case. Specified length of my rods are matched to within 0.02mm. But I suppose that length is calculated value for the nominal diameter of the balls. Combining the rods with the ball I get discrepancy in the length of final assemblies.

Yeah Doug is spot on to say 0.1mm is definitely not good. You can get away with that if its a difference between one arm pair and another. But it will show up as an obvious inaccuracy if its within one arm pair.
If you know the ball diameters (and they are actually spherical) and you know the rod lengths, I'd be very keen to see the effect of matching them precisely to get exactly the same arm lengths all around to 0.01mm precision.
Real world difference in your prints  zero! But it will be satisfying in the quest for perfection.
I no longer have the lengths marked on my arms, I foolishly removed the stickers. Does anyone know a clever way to measure them? I have 300mm calipers?

you will need 2 3/8 inch Diam ball bearings, you measure one of them and zero you callipers at that point (Ie 3/8 inches) then attach the balls one at each end of the rod and measure it you now have the true length of that rod repeat for all of them.
Doug.
I also have 300 mm callipers but could do with some 500 mm ones I could then verify my home assembled arms?

Thanks Doug that makes sense but I have 360mm arms and 300mm calipers, so what I wondered is whether there is a trick to measure something that is longer than the calipers? I could use the metal rod that protrudes from the caliper, but how to measure to high accuracy like this I don't know.

But it will show up as an obvious inaccuracy if its within one arm pair.
I really have 3 balls that deviate as much as 0.1mm from the rest of set so I will get one pair imperfect anyway. Hope another twelve ball studs will solve the problem.
But it will show up as an obvious inaccuracy if its within one arm pair.
As we are interested onley in rod length differences I think you can butt (sorry if it not right verb in English) one end of a rod into something hard and take measurement using the arm of calipers on another end of the rod and the end of the depth measuring stick on the other side. You get a virtual size of rods but their differences will be real.

Where are you based? I have 24 balls in total. If I get chance I measure them and work out which 12 I need, I will let you know what sizes the remaining 12 are and we can swap if you like? Depends on where you are based.
As for measuring technique it might work but be trusted to be accurate to 0.01mm? I'm not so sure.

Maybe 500mm calipers might have to be on the Christmas list.

Thanks Doug that makes sense but I have 360mm arms and 300mm calipers, so what I wondered is whether there is a trick to measure something that is longer than the calipers? I could use the metal rod that protrudes from the caliper, but how to measure to high accuracy like this I don't know.
You couldn't unfortunately. Do you know any friendly engineering companies that may loan you a 350375mm micrometer for a day
another way is that you could make a jig with a bit of timber (OR extrusion if your quick) set an end block up fixed at one end then a sliding but fixable block at the other locate an arm adjust the slider for a good fit. then using another block of say 100mm long in the gap measure with the inside jaws of your callipers add in the block length (Having first calibrated your callipers to allow for one ball) and you will have your rod length. 
I will have to get my 25mm micrometer out and start checking my balls now I have 18 extras.

Its always recommended to check your balls from time to time (sorry I couldn't resist).
I do have a friendly engineering company nearby so I might pursue that option.

Thank you for your kind offer but afraid it is impossible. I am in Russia.
Regarding to the measuring technique if you can make a simple jig I think it gives the same result as direct measuring with calipers assuming this is made accurately. You can't just get absolute value of the length but it is not relevant for the case. 
Hi,
I have these parts in three of my printers.
The bed calibration was spot on and and they print just fine.
You may have nothing to worry about.
Frederick

Hi,
I have these parts in three of my printers.
The bed calibration was spot on and and they print just fine.
You may have nothing to worry about.
Frederick
Thank you for your response. Maybe 0.1mm is not a big difference indeed for Delta. But I am apt to get another set of ball studs. Anyway they can be used for making other heads, laser graver for example.

Okay so I checked my balls (sorry)
Of the 6 spare ones I have not attached to a running printer
9.38mm x1
9.40mm x2
9.48mm x1
9.50mm x2I will check the other 12 that are currently deployed whenever I get chance, and then get the arms measured accurately, and see if a "perfect" combination can be found.

Hi,
Haydn had these parts made  given his attention to detail I would not think he would be selling parts that did not perform as intended.
Ask him if you have doubts.
Frederick

Okay so I checked my balls (sorry)
Of the 6 spare ones I have not attached to a running printer
9.38mm x1
9.40mm x2
9.48mm x1
9.50mm x2I will check the other 12 that are currently deployed whenever I get chance, and then get the arms measured accurately, and see if a "perfect" combination can be found.
I'll raise this with Haydn  looks like he may be making the arms to significantly closer tolerances than his supplier makes the ball studs.

Hi Frederick,
I know they work and I know Haydn puts in a lot of effort to make the arms very precise.
But the balls are not that precise.
Personally I consider this an opportunity to match balls to arms to get even more precision.
I know he bought 1000 or so of the ball studs in a job lot. All he has to do if he can supply 6 rods exactly the same length is supply 12 balls of the same diameter. But since the rods will vary by up to 0.05mm maybe he would be willing to tape two balls to each rod to equalise the total length.
Simon.
Hi,
Haydn had these parts made  given his attention to detail I would not think he would be selling parts that did not perform as intended.
Ask him if you have doubts.
Frederick

Okay so I checked my balls (sorry)
Of the 6 spare ones I have not attached to a running printer
9.38mm x1
9.40mm x2
9.48mm x1
9.50mm x2I will check the other 12 that are currently deployed whenever I get chance, and then get the arms measured accurately, and see if a "perfect" combination can be found.
I'll raise this with Haydn  looks like he may be making the arms to significantly closer tolerances than his supplier makes the ball studs.
Sure as I said earlier it's very marginal, and I suspect it will have almost no practical consequences, but then we go to extreme lengths for precision so measure your balls people!