To Roebel or Not To Roebel, That Is The Question.: Half A Roebel Bar
To Roebel or Not To Roebel, That Is The Question.: Half A Roebel Bar
To Roebel or Not To Roebel, That Is The Question.: Half A Roebel Bar
Ok, before we try to answer this question, we need to first understand how the Roebel bar is
manufactured.
Bending of
individual
strands
The next step is to add Nomex insulating sheets at the Roebel transpositions to lessen the strand
insulation degradation/damage due to relative movement. Note, we have the choice of either
Nomex paper or NMN laminate (Nomex Mylar Nomex). The choice of insulating sheet material and
thickness is determined by the transient electromagnetic force and a formula derived from years of
manufacturing Roebel bars. The use of these two materials will greatly affect the Tandelta result at
20% voltage (hmmm, not quite true, there is one type of Roebel ICP design which is not vastly
affected by the Nomex or conductive mastic transposition filler).
Now we must fill the Roebel transpositions with a conductive mastic (inner corona protection or ICP)
to prevent internal partial discharge due to voids. The mastic tape is a B stage tape which needs high
temperature to cure. The mastic tape is cut into small sections and is then manually pressed into the
Roebel transpositions. A final heat activated shrink tape is double wrapped around the bar to retain
its shape.
The next step is to fully consolidate the Roebel bar slot section (e.g. straight portion + a bit). The
consolidation is done in a heated hydraulic press which its temperature is monitored by a computer.
After the consolidation the worker removes the shrinkage tape and the insulation free Roebel bar
looks like the following. The grey mastic is now cured and is quite firm.
The Roebel bar then goes through a manual grinding process to remove excessive mastic and also to
remove any large indentation on the surface of the cured mastic. The final Robel bar goes through a
clean room where high powered clean dry air is used to remove mastic dust lodged between the
strands at the two ends.
Handling of the Roebel bars at this production stage requires extra care as the strands are only held
by the strand insulation. Two workers are required to lift bars longer than 1.5 meters to avoid the
slightest bend moving from one table to the other.
Once the Roebel bar is ready for taping, a series of electrical tests will be done. The most important
test amongst these tests is the strand to strand dielectric withstand test. This test is to detect strand
to strand insulation failure as a result of incorrect handling or false grinding. The strands will be
fanned out like what is shown below. Each strand is tested individually.
Once the Roebel bar is ready for taping, it will be put onto an automatic taping machine for further
processing.
Now coming back to the original question, when should a Roebel bar be used?
This is somewhat difficult to explain because there are many factors involved both from a magnetic
design and cost viewpoint. For me personally, I think there is not much difference in quality between
coils and bars nowadays if it is produced by the big players. Some customers might argue that the
bars have better Tandelta results compared to coils and also the coils have more voids in the end
turn compared to bars. The truth is that both the bar and coil will have voids in the endturns
because only the slot section is consolidated, not the ends (ends are far away from the ground plane
therefore the electrical stress kv/mm in the voids is unlikely to partial discharge, unless you have
used metallic brake pads and have conductive dust all over the endwindings).
Coils don’t normally have inner corona protection layer because there is no transposition in the slot
section. The transposition in the coils are done at the endturns and the voids are filled with a non-
conductive putty.
Bars can also have poor Tandelta results or delamination after thermal cycling, e.g. if the Nomex
paper is not fully dried before inserting into the transposition or the factory has high relative
humidity.
Figure 1 Coil dissection of a precious project (end turn VS. slot section)
So, from a quality point of view, there is not much difference between the two. But what about from
a manufacturing risk point of view?
I often see customers specifying 360 Roebel transposition for machines with a core height of less
than 800mm. Although this sounds impressive, but the risk associated with strand to strand
insulation failure has increased dramatically. Think about the following Roebel bar strand bending (tr
is the parameter which goes into the bending machine, subsequent strand bending is simply just
shifting the pitch, e.g. 2nd strand will be 3tr).
The secret number is something the Roebel bar manufacturer don’t publicly talk about. It ranges
from 2 to 3 for most OEMs. (a value of 2 is a very bad idea in my opinion)
Ok, there is an exception, with a special ICP design, the 20% Tandelta value is less correlated to the
mastic fillers (e.g. quantity and thickness). Debonding between ICP and main copper conductors is
less visible from the Tandelta results.
Copper strip