The drive roller, or largest roller, should have a curve in it so
that the center is larger than the edges.
Or pick two smaller rollers, spaced "far" apart in the loop.
For a quick and dirty test, add some adhesive-backed tape onto a
roller, centered and maybe a second, more narrow tape also centered (to
get that curve).
One common method is to flare the rollers slightly at the ends. That
way if the belt wants to ride up on the roller, there's a centering
force due to the stretching of the belt.
Failing that, something like what you propose could probably be made to
work--but I'd definitely avoid optical sensors for this if possible.
The problem is that they get dirty and stop working, which in an
automatic control situation like this would be dangerous.
Another simple method might be to have guide rollers at the edges of the
belt. Rather than encoders and motors and microcontrollers, you could
have the guide roller turn the adjustment screw via gears or a chain
drive. That way, any time the belt got far enough out of line to spin
the guide roller, it would be gradually adjusted until it didn't spin it
any more. That's sort of an integrating servo--if the gear ratio (i.e.
gain) is too high, the control loop will oscillate, and the poor thing
will bounce back and forth until it chews itself to bits.
Since the situation doesn't require too many smarts from the controller,
this would probably be easier and a good deal more reliable.
["Followup-To:" header set to sci.electronics.design.]
Phil Hobbs wrote:
How about a very simple electrical "middle of the road" thing? Mechanical
switches with rollers that activate a geared-down motor. One switch on
either side of the belt, each making the gear motor turn in opposite
direction from the other. Rugged, simple, and a lot cheaper than an
Others have suggested the crowned roller.
Take a look at how the web is kept centered in the presses. The same
methods should work for the belt. It involves adding an extra roller
or two to the path so that
You can also use a split roller to do the job. This means that you
have to add torque between the sections of the roller to push one part
of the belt forwards with respect to the other. This can be made to
work on a mechanically "automatic" manner much like the crowned
Crowned looks like the best option. Splitting the roller requires additional
support and bearings. We regularly have rollers reground by a specialty
service, and I will ask them about doing this to the (rubber) drive and idler
rollers. (Would having both rollers crowned be better? Or is one sufficient?)
Second, as someone suggested here, will consider the inclined rollers added
in the belt's path at the extreme ends of the belt's width to encourage
FYI, it's a sheet-fed operation, not continuous web printing (and,
specifically, this machine is a sheet-fed coater). Light weight product on
the belt (less than a few pounds over the entire length of the belt at low
speed). Belt is mesh carbon fiber. Seems somewhat springy, overall, but the
material, per se, isn't flexible.
Thanks to all for your suggestions.
You could put small rollers on the arms that do this. You really
don't want to add a drag to the one side. The added drag will work
against the effect of the arm. This method requires that the belt in
question be somewhat springy. I don't think the OP stated how springy
Have a roller tube made that is concave in shape. That tube goes at
one end, and the adjustment end is the other. Then, the taughtness of
the belt is what keeps it in the center of the "fall zone" of the
Great explanation and suggestion. Thanks.
Which brings up a question: does a concave or convex roller work best at the
drive (no adjustment) or idler (adjustment) end? Or would it matter where
either is placed?
Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here.
All logos and trade names are the property of their respective owners.