Servo Mirror SAGA Continues

That's a pretty neat project.  It must be art, since I don't see a
purpose, but it's still pretty neat art.  :)


Just get a 5V power supply from Radio Shack (or your preferred
electronics dealer), I would think.  Actually, when you're powering 900
servos, you might need something able to handle a higher current than
the typical wall warts can do -- something more like a lab bench supply,
which is quite a bit more expensive.  A good electrician can often adapt
a power supply from an old computer, and get the same effect for
considerably less money.

On the other hand, when you're prepared to drop a grand in servo
controllers and close to ten grand on servos, a few hundred bucks for a
lab bench power supply is chump change.


I love Pololu products, including the micro serial servo controller -- I
have one myself, which I'm using on my 4-servo walker:
  <http://www.strout.net/info/robotics/projects/4s1.html>

...but, in all honesty, when you're dealing with this many servos you
should by using the Lynxmotion SSC-32 controller.  It's by far the
cheapest controller, at a little over $1/servo.  (She'll still need 29
of them, so she might contact them directly and try to negotiate a
quantity discount.)

Lynxmotion engineers also hang out on their forums (at lynxmotion.com)
and I bet would be happy to provide advise on this project, especially
on the best way to power 900 servos and 29 SSC-32s.

Best,
- Joe

Your project sounds like one of Daniel Rozin's mechanical
mirrors. Perhaps you could contact the artist and ask for
some tips?

http://www.smoothware.com/danny/index.html

TK

Richard Greenberg wrote:

Hi Richard,

Richard Greenberg wrote:

Kind of like the wooden mirror project:
http://www.smoothware.com/danny/woodenmirror.html
which is pretty cool to see.


Basically, you'll need a bunch of power supplies. To keep things
efficient, you'll want the output of the power supply to be around 6v.

The servos are going to consume most of the current. According to the
spec, the HS-81 is the least expensive and draws 220 mA (no load
operating).

So you're looking at 1/4 amp per servo. That's about 225A assuming all
of the servos can move at one time. If you can stage the movement, then
you can reduce the current consumption.

The cheapest power supplies are going to be something like 400w PC
power supplies, which are essentially commodity items.

You can pick up surplus stuff sometimes, like these guys have a 5v 100A
supply
http://www3.telus.net/public/dt0116/transfor.html

Tigerdirect.com has a 400 watt supply that's rated to give 5v @ 40A for
$35.

You can probably run 3-4 of these from a single AC circuit.

I'd recommend getting something like the LynxMotion SSC-32. It's only
$39 and can control 32 servos. This will be cheaper per servo than the
Pololu servo controllers.

Then you have to figure out how to get 30 of these connected together.
Instead of using full blown RS-232, you could connect all 30 RxD lines
together to a single TxD line from the PC, and broadcast to all of the
nodes. You could modify the protocol (the SSC-32 has full source code
available) and precede each command by a node id.

If you use the TTL version of the serial port, you could even connect
all of the Tx lines together providing you disable the TX portion of
the UART when not transmittings and configure the pin as a GPIO input.

Anyways - that should give you some points to ponder....

--
Dave Hylands
Vancouver, BC, Canada
http://www.DaveHylands.com/

Yes, you guys that suggested this connection must have missed the early
talk about this in this group.  The wooden mirror and the other work of
that artist was the motivating idea for the project.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

I guess that you've been through all this, but I do wonder about all those
servos - its going to be noisy for sure. Since what you want is 90 degree
movement and the load is negligable, I'd go for the cheapest motor I could
buy, mount the disc directly on the shaft and place a return spring and a
couple of end-stops. Then the deflection is proportional to (average)
current. The current only flows in one direction, so a simple transistor
switch driven directly from a micro port would be about as minimalist as you
could get.

Power supplies and distribution for a project like this are things not to be
taken lightly - large currents can easily create common mode 'noise' that
will cause chaos.

Its a fun project, but don't underestimate some of the engineering
challenges ...

Dave

Wow!


"mirror"

On Fri, 01 Dec 2006 03:23:00 GMT, "Richard Greenberg"


Power is probably the easiest part. I'd use a large 12v lead acid
battery with a ~10 amp auto battery charger connected to it all
the time as the base supply. To power the servos (which should
actually have little load on them), I'd use a 7805 voltage
regulator chip (rated for 1 amp) with a diode on the ground lead
to supply ~5.7v output for the servo. As the servos will be using
little actual current, you probably could have several servos
powered via each regulator chip. The below page has a simple
schematic showing how I do this with my servo setup. As the
student design is some what of a clone of the bottom circle
mirror, my achedemic expectation would be for the students to be
more inovative and significantly reduce the number of servos and
servo controllers used to have a significant reduction in the
cost of the project (cost reduction will be the name of the game
when they graduate).


http://www.geocities.com/zoomkat/ezservo.htm
http://www.smoothware.com/danny/circlesmirror.html
    

Thank you all for your responses.  The prototype mirror is now responding to
video input.

http://richgreenberg.com/downloads/
mirror.avi
http://richgreenberg.com/downloads/
mirrorVid1.avi

From what I've been reading of your responses, it sounds like a given power
supply with about 6 volts out can be used to power both the controllers and
the servos?

Also, I'm thinking that from the power supply it would look something like
this?
http://richgreenberg.com/downloads/
diagram1.doc

With 900 motors turning could a single outlet power the whole thing?  I
would hope so.
Thanks again,
Rich




"mirror"

Looks good!


Yes, that's correct.  But, if you use a single supply at 162 amps (from
calculation below) you will need very large wires running from the supply
to your distribution panel.  Doing some research on the net, it looks like
you something like 4 gauge wire to do 150 amps.  I don't even know where
you buy that size wire.


Good question.  I just looked up the spec of a standard servo (Hitec
HS-311) and it says it draws 180 ma at 6V with no load (your disks are
probably the same as no load).

900 servos times 180 ma is 162 amps.  That's very high and that doesn't
include whatever your control boards will add to that.  You should try to
measure the current draw on the servos you are actually using to see what
it is.  You should consider using the lowest power servos you can if you
aren't already using low power servos.  The lowest current draw servo I
could find was 100 ma.  That would be much better at 90 amps total but
those servos cost twice as much as the standard servos which draw 180 ma.

A high wattage PC power supply (420 watts) can deliver 42 Amps at 5 volts.
So you would need 4 or 5 of those if you could run your system at 5 volts.
PC power supplies are at least cheap.

162 amps at 6V is 972 Watts.  A single 15 Amp outlet can go up to near 1500
watts so it's possible.  But it depends on the efficiency of your power
supply (or supplies).  I have no idea how efficient they are so I don't
know if it will work or not (my guess is they are fairly efficient).  I
think it will probably work but it's close.  If you servos use more power,
or the control boards us a lot of power, you might not be able to run it on
a single outlet.

Ah, I just found a spec on a PC power supply that said it was 70%
efficient.  Which means if you need 972 watts for your load, the power
supply will be sucking 1388 watts from the wall - which is near the point
that you will trip a 15 A home circuit.  So it could be close.

If you take care to limit servo power, you should be able to make it work
(you might need to write the software to limit the speed of the servos or
limit how fast they are all turning at once to limit the max power.

If you intend to run it in commercial buildings (like a school) it will
probably have 20 amp circuits which will give you a bit more room to work
with.

You should hook up an amp meter to your prototype and see how much current
it's drawing and then do the math to see what happens for 900 servos.

Basically, power supplies in that range are special beasts that will cost
you a fortune (I couldn't find any searching the net - the highest I could
find were 100 Amp and those were very expensive).  You are probably going
to have to split up the load and use multiple power supplies because I
doubt you will find anything big enough to run your entire system at a
reasonable price.

A previous post suggested you use a 12V battery and a "high current" 10 amp
charger.  If you need 162 amps, an "high current" 10 amp charger won't even
be close.  A car battery can deliver many 100's of amps but not for very
long.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

Sometimes, but I don't recommend it.  Servos are electrically noisy, and
in my experience often cause the controller to reset.  Life is less
painful when you power the controller from a separate power source.

HTH,
- Joe