Help: Power drop on Servo Controller

constraints).

I don't know what an LM1117 is [LM117 ???], but most voltage regulators
require at least 2v of overhead minimum. IE, that's 5+2 = 7v "minimum".
So, things to try:

a. use LDO v.reg.
b. use LDO v.reg = 4v to drive the PIC Vcc buss at 4vdc.
c. For not too much extra space, add a 6th AA or AAA cell in series
with the other 5 to boost the power to the v.reg up to 7.2v, and also
use an LDO v.reg or else a 4v v.reg for the PIC.

Item c. is similar to what I did on one of my servo-driven bots, and it
worked fine.

- dan michaels
www.oricomtech.com
=======================

The LM1117 is a low dropout voltage regulator.
http://www.national.com/pf/LM/LM1117.html

Good point though that regulating 6V to 5V is somewhta critical. Since the
PIC runs down to 3.2V, I used the same regulator in the 3.2V version, and
while the servos are happy with 3.2V on the PWM line (digital servos), I
still get PIC resets when all 16 servos move at once.


The sixth cell is a good idea, however the servos are 4.8 to 6 Volts. I'll
give it a shot, but in the long run, I'd have to change the chassis. I don't
like dangeling batteries ;-)

Matt

Matt:

In general, you want to keep your power draw equipment (e.g.
servos) off of the power supply that is driving your logic
circuits (e.g. microcontrollers.)  Two battery packs are
a good way to do that.  However, you have said that you
do not have the room for the second battery pack.

It turns out that servos can be directly connected to your battery
pack.  So hook the servos up directly and run the microcontroller
off of a low drop out voltage regulator like the LM2940-5.  It can
live with no more than approximately 5.5 volts of battery power.
Be sure to plant a .1 uF decoupling capacitor right next to your
PIC to save it from transients.  Also, do not scrimp on the voltage
regulator capacitors; use some nice tantalums.

Since I am paranoid, I drive the servo pulse signal through
some open collector circuits (a PN2222 and a pull up resistor,
with a base current limit resistor).  That way the servo pulse
signal always swings from 0.2V up to the full battery voltage.
Some people do not bother with this step, since many servos
are perfectly happy to eat signals that swing from 0 to 5V
even though the servo is running at a higher voltage.  It is
your choice.

I know that you don't have room for two battery packs, but
for other people reading the list -- if you have the room
for two battery packs, use two; you will tend to have fewer
problems.  If you are really careful, you can get it to work
all off one battery pack; however, lots of people are beginners
and the cost of an extra battery pack is well worth the hassles
that will be avoided.

My $.02,

-Wayne

P.S. To contact me directly, send mail to Wayne .at. Gramlich @dot@ Net.

Yes, I am afraid it'll end up beeing just that. Metal work coming up ;-)


I added a .1uF and a 10uF electrolyte right on the power lines of the PIC,
and a 10uF right in front and right after the voltage regulator, to no
avail. No tantalums though.


Yes, that is a pretty good practice. I do use a pullup to remove the
power-on-jerk and another resitor in series to limit current, but in this
particular robot, I use digital servos which draw a few uA on the PIC's.

Thanks for the informaton

 Matt

You already got some good replies, but I wanted to add just one thing:
is this really a *six* volt nicad pack, or is it just 4.8 volts?
Remember that nicads put out 1.2 volts nominal per cell, not 1.5 volts,
so a pack of four cells will be only 4.8 volts. Even with a low dropout
regulatoryou're getting pretty close to brownout level voltages for an
MCU, and any surge to the servos may dip the terminal voltage below a
magic threshhold.

I know Brian Dean, who designed the MAVRIC controller, has some tricks
he does to help assure proper voltage levels on his AVR-based boards.
He's a regular contributor here, so you might drop him a line.

-- Gordon

Yes, I have five NiCad cells, giving me 6.2V when they are fully charged.


Yes, clearing the Brownout detection bit helped a little. The PIC crashes
less frequently, but if it does, it usually needs a power cycle to awake
again. Maybe I should simple rewrite the PIC software to deal gracefully
with brownouts... .

I don't have a basic stamp here to check, but I will look at more
schematics... . The problem seems to be the number of servos sucking at the
battery. 16 seems to be too much.

Thanks!

Matt

. The problem seems to be the number of servos sucking at the

Oh man, 16 servos, that's a lot of juice. If this is a walker, and
regular 44 oz-in servos, then prolly > 2.5 Amps average, to > 4.5 amps
peak. If bigger torque-ier servos or digital, then lots more Amps than
that.

Given this humungus load, a couple of other things to try besides the
drop-out ckt on the v.reg are ....

1. some large filter caps on the servo power buss right where the
battery connects. All of those wires [including those from the battery
to the controller] have a lot of inductance and you're prolly getting a
lot of gnd and power buss bounce.

2. secondly, don't pulse all the servos simultaneously, rather pulse
them in sequence, like the r/c receivers do. If you look at the recent
thread on servo internal pulses/etc, you'll see that Gordon McComb
discovered that power is only applied to the internal motors in the
servos for a short period [2-5 msec or so] after they receive the
control pulses. The rest of the time, there is no voltage applies
across the motors. If you pulse all the servos at the same time, then
the power surges will all superimpose.



- dan michaels
www.oricomtech.com
======================

Forgot to mention you might also look to see what products like the
Basic Stamp 2 use. Seems to me they use an ultra-low drop out regulator
(<0.25 volts?), because they allow the use of nicads on the BOE Bot,
which is designed to support up to four servos.

Just looked up their schematic...note the honking 15 uF cap across the
voltage regulator output. That can't hurt the situation. The regulator
is a National LM2936.

-- Gordon

Depending on the power constraints of the PIC and if the voltage drop is
momentary and not constant, you can use a diode and a capacitor:

Connect a diode between the power supply and the PIC, you'll get about 0.6
volts drop over the forward biased diode. Make sure you put a big enough
capacitor on the PIC.

When the voltage drops from the motor, it will drop below the voltage stored
in the capacitor (which will continue to power the PIC circuit), once the
diode is turned off, your PIC power is effectively isolated from the
battery drop. When the battery voltage rebounds, it will again forward bias
the diode and charge the cap.

The voltage drop under load is based on the internal resistance of the
supply. NiCad batteries have a pretty low internal resistance and typically
behave well in this mode. I would say that your batteries need to be
recharged, replaced, or that there is some source of resistance between the
batteries and the circuit.

The servos use too much power in starting, the voltage drops, the PIC
resets.


I *always* use a separate battery.


You can use smaller batteries.

If you can't, use bigger caps on both the motor side and the PIC side
to help reduce the transients.
--
D. Jay Newman
http://enerd.ws/robots/

It's a good idea in theory, but it makes the power system more complex.

You still need to manage the charging differently. Using two batteries is
like partitioning your hard disk (on Windows) to C and D. You will
eventually run into the problem when C+D would be big enough, but neither C
not D has enough space.


Pr, as I said, add diode isolation and capacitor decoupling to the PIC.

This really should do the trick for most hobby-type servos. I would
definitely go this route prior to using a separate supply.

I'll frequently use a separate battery for larger motors because I'm
lazy, but for servos you probably don't have to do this -- and really,
you can frequently get away with not doing it for bigger motors.

If you DO go with separate supplies, keep in mind that the servo control
line and servo power supply will need to share a common ground.

--
(Replies: cleanse my address of the Mark of the Beast!)

Teleoperate a roving mobile robot from the web:
http://www.swampgas.com/robotics/rover.html

Coauthor with Dennis Clark of "Building Robot Drive Trains".
Buy several copies today!

Unless you optically isolate them!!! :-)

The servo control line (AKA "signal" ) is the line actually going into
the servo -- a low on this line is taken in reference to the black gnd
(usually labeled "negative") wire.

Obviously, whatever is pulsing the signal line itself has to do it using
some kind of isolation scheme if separate supplies are used. Personally,
I'd just go with the cap/diode method, although maybe there's an upper
limit on the number of servos this works reliably with. Dunno.

--
(Replies: cleanse my address of the Mark of the Beast!)

Teleoperate a roving mobile robot from the web:
http://www.swampgas.com/robotics/rover.html

Coauthor with Dennis Clark of "Building Robot Drive Trains".
Buy several copies today!

: I have built a servo controller around a PIC Microcontroller. The servos are
: pwered by a 6V NiCad battery pack that also powers the PIC through an
: LM1117.

: Unfortunatly, as soon as I get some load on the servos, the PIC resets due
: to the poer drop.
: How can I make sure that the PIC will always get sufficient DC, so it won't
: reset?
: Adding a separate battery for the PIC is not an option (space constraints).
: Any help is greatly appreciated.
:  Matt

  Guard the power line into the PIC circuit.  The easiest way is to put
a Schotkey diode between the battery back and the input to the regulator.
Now, put a big electrolytic cap on the line that goes from the cathode of
the diode to the input of the regulator, the size depends upon the current
used on the PIC board, my guess is that 100uf to 470uf should be enough
to survide the dv/dt power needs of the controller.  Now use an LDO
regulator, one that only needs .5V or so above your Vcc needs.  This will
survive a couple ms voltage sag on the battery side.  Now, you need to
make sure that you don't have EMI seeping into your board and messing up
your oscilator circuit.  To help with that "guard" your crystal pads with
a copper flood on both sides of the board and heavily bypass your Vcc
line.  Another good idea, if extreme, is to put a choke on the power lines
to the processor and make sure power enters that processor "side" of the
choke at exactly one point, same with grounds to the processor.

These measures aren't too extreme and will create a more robust circuit.

have fun,
DLC
--
============================================================================
* Dennis Clark         dlc@frii.com                www.techtoystoday.com   *
* "Programming and Customizing the OOPic Microcontroller" Mcgraw-Hill 2003 *    
============================================================================

Wow, thanks for all the help from the NG. So putting it all together, I got
it checked in the skope and working! Yeah!

So here's a summay:

Problems:
- regulating 6V down to 5V is a bad idea, because of voltage drop.
- running up to 24 servos simultanously on a walker robot puts huge strain
on the battery, each servo sucking up to 1Amp!
- noise!

Limitations:
- due to space and simplicity in the charging unit, it is not possible to
have a second battery

Solutions:
- the voltage regulator runs through a filter, then a low drop diaode, then
a cap, then the regulator, then two more caps
- Connect the regulator as close to the battery as possible, don't run the
lines by a few motors first (stop thinking digital! Wire is always a
resistor and a cpacitor!). Clean up the layout!
- Worst case, use a switching regulator that even pumps 3V up to 5V if need.
- a 300Ohm resistor on every servo data line in series keeps the noise from
the servos to the PIC low (not neede for digital servos)
- 3k pullup on the servo data line keeps the servos from jerking all at
onece when power is switched on
- I change the software to send pulses to the servos in groups of 4 with 2ms
delay from group to group, so that there is only on high load (leg) servo
per group. This spreads battery load over the 20ms cycle.
- with digital servos, the load on the battery is even greater. On the other
hand, they are happy with 3.2V pulses. So in that case, I use a 3.2V
regulator which runs more stable on 6V battery anyways.

Problem solved! Thanks for all the help!!

       20uH           +---+
Bat ---UUUU---|>|--o--|reg|--o---o--> pic
                   |  +---+  |   |
                  ===   |   === ===
              470uF|    | 10u|   |.1uF
GND ---------------o----o----o---o--> GND

On Wed, 4 May 2005 04:14:33 +0200, "Matthias Melcher"


I use a diode and capacitor to isolate my servo controller from
momentary voltage drops. There is a basic schematic at the bottom
of the below link. The diode acts somewhat like a check valve
preventing the capacitor from discharging back to the servo
motors when the voltage drops low. The diode I used has .7v
voltage drop, so you might want to use a type with a lower
voltage drop and a really big capacitor. You might be able to use
an NPN transistor as a low voltage drop diode if needed.

http://www.geocities.com/zoomkat/ezservo.htm