PDA

View Full Version : Trouble with brown outs



billyzelsnack
12-27-2010, 01:22 PM
It's about time I learned how to fix this issue. It's been an issue on all my robots. The current situation is this..

I take the output of a 3S Lipo and split it into two power rails. One 12V and one 6V via an RC BEC. The RC BEC is rated for 3A continous. I then stick a diode on it to knock the voltage down to 5.35V.

On the 12V rail are the dynamixels.
On the 5V rail there is a Chumby (454MHz linux device), an AVR, and some 9G RC servos.

The Chumby boots up fine, but there is a 50% chance that the instant I command the RC servos that the Chumby will reboot. I am guessing this is either from a brown out or from some back EMF. I've actually not had any problem with the dynamixels (I am only using 4 at the moment), but I have had issues in the past.

There are products avaiable for RC servos to prevent the RC receiver from browning out.
http://www.amainhobbies.com/product_info.php/cPath/61_124/products_id/6570/n/Spektrum-Voltage-Protector
Also. The Robotis CM-5 controller has some huge capacitors in it. I'm assuming to handle the same issue.

I've tried sticking capacitors in parallel on the 5V rail. I've also stuck a series diode on the 9G servo side to help stop back EMF. I'm still having issues. I'm guessing there is something more subtle that I am missing and that I really should be doing some math rather than just throwing on random capacitors (470uF35V) or diodes.

Side note.. I actually destroyed my AVR trying to fix this. I'm pretty sure it is related to me being an idiot and accidentally soldering while I had the thing powered, but is it possible that throwing caps/diodes on the 5V rail powering the AVR could damage it? It's only possible that voltage could be lowered correct?

RobotAtlas
12-27-2010, 02:10 PM
Billy, have you tried using both large _and_ small capacitors on the line at the same time?
By large I mean as large as you can get, 450uF might not be enough when you are dealing with RC servos.
By small I mean about 1uF or less.
I'm a little rusty on the theory, but from what I remember you need to be protected from 2 different drops in voltage - low and high frequency.

Also this link looks like have some interesting info (I didn't read it, just glanced over):
http://www.maxim-ic.com/app-notes/index.mvp/id/4081

billyzelsnack
12-27-2010, 08:53 PM
Cool. I'll give that two capacitor size thing a try next time I tackle this issue.

darkback2
12-27-2010, 09:17 PM
Please keep us posted. My mech will sometimes have "brown outs" and I've tried putting diodes everywhere to no avail.

lnxfergy
12-27-2010, 09:19 PM
A couple of points -- yes, you definitely want various levels of capacitors on your power supply as they each filter at different frequencies. A google search for "decoupling capacitors" or "bypass capacitors" turns up plenty of explanations/theory about this. I typically have an array of 0.1, 10uF, and 47uF around my AVR circuits and it holds up pretty well even with quite a bit of other noisy circuitry on the same regulator.

That said, capacitors alone may not be enough. The Chumby is likely to be less resilient than an AVR to voltage changes (I've found AVRs to be tanks when it comes to EMI resistance). I'm assuming those "9g servos" are the very cheap micro servos, right? Such servos tend to emit terrible electrical noise as they move the servo motor. I often put them on their own regulator for that reason, as regulators may not be able react quick enough to handle the noise.

Another alternative would be to install an L-C filter on the regulator output. If the servos are drawing a lot of current quickly when they start to move, the added inductor can help as it will push some extra current out as the voltage drops and the inductor field collapses. Designing an L-C filter isn't quite as easy as tossing some capacitors on the circuit -- however Radio Shack carries some 100uH inductors which I've found to work fairly well in practice in the past (they are huge, but cheap and easy to procure for quick experimentation). Note that you'll need to provide diode protection for shutdown, as the field collapse creates a huge current rush, and you'll also have to check that the L-C doesn't rise too slowly for your circuitry on startup.

-Fergs

lnxfergy
12-27-2010, 09:26 PM
Please keep us posted. My mech will sometimes have "brown outs" and I've tried putting diodes everywhere to no avail.

Che -- I think billy's issues are mostly related to Voltage Regulator (in this RC BEC) characteristics, not the larger-scale battery circuit and wiring issues you have.

If I recall correctly, you have issues with servo LEDs actually blinking meaning the whole bus is falling, a lot, to at least <7.0V. There's two main causes I can think of here: 1) is your battery able to source enough current (I think you have quite a few servos on the bot, I can't remember if we discussed your battery), and 2) is your wiring capable of carrying enough current. The wiring on the Bioloid servos is pretty thin, and if you have very long runs of cabling you may need to hang extra capacitors out on the limbs tied directly to the AX wires to avoid voltage falling at the ends of the long runs.

-Fergs

gdubb2
12-27-2010, 09:56 PM
Personally, I've always been a proponent of separate supplies for critical things. Let one battery run all of the servos, but use a separate battery just for the electronics.

Easy fix to a complicated problem..

Gary

elaughlin
12-27-2010, 10:48 PM
Che -- I think billy's issues are mostly related to Voltage Regulator (in this RC BEC) characteristics, not the larger-scale battery circuit and wiring issues you have.

If I recall correctly, you have issues with servo LEDs actually blinking meaning the whole bus is falling, a lot, to at least <7.0V. There's two main causes I can think of here: 1) is your battery able to source enough current (I think you have quite a few servos on the bot, I can't remember if we discussed your battery), and 2) is your wiring capable of carrying enough current. The wiring on the Bioloid servos is pretty thin, and if you have very long runs of cabling you may need to hang extra capacitors out on the limbs tied directly to the AX wires to avoid voltage falling at the ends of the long runs.

-Fergs

Fergs,

Do you think I would encounter this type of problem on my mech using the bioloid wires?

I have 15 servos, arbotix, 2 tank guns, camera, wifi encoder, lipo audio alarm, target setup, going to be hooked up through the bioloid cables. This will be run from the 3S 11.1V 2100mAh Pro Lite V2 20C LiPo Battery.

The last servo in the leg is the farthest run on the daisy chains.

I got a little worried from looking at this thread and seeing what you said.

billyzelsnack
12-27-2010, 11:52 PM
A couple of points -- yes, you definitely want various levels of capacitors on your power supply as they each filter at different frequencies. A google search for "decoupling capacitors" or "bypass capacitors" turns up plenty of explanations/theory about this. I typically have an array of 0.1, 10uF, and 47uF around my AVR circuits and it holds up pretty well even with quite a bit of other noisy circuitry on the same regulator.

I've read a bunch of stuff, but I just don't have a lot of holes in my knowledge and I feel like I am missing something.


I'm assuming those "9g servos" are the very cheap micro servos, right? Such servos tend to emit terrible electrical noise as they move the servo motor. I often put them on their own regulator for that reason, as regulators may not be able react quick enough to handle the noise.

Hey! HXT900's are sweet servos. :)




Another alternative would be to install an L-C filter on the regulator output. If the servos are drawing a lot of current quickly when they start to move, the added inductor can help as it will push some extra current out as the voltage drops and the inductor field collapses. Designing an L-C filter isn't quite as easy as tossing some capacitors on the circuit -- however Radio Shack carries some 100uH inductors which I've found to work fairly well in practice in the past (they are huge, but cheap and easy to procure for quick experimentation). Note that you'll need to provide diode protection for shutdown, as the field collapse creates a huge current rush, and you'll also have to check that the L-C doesn't rise too slowly for your circuitry on startup.

I think this type of thing is just too advanced for me. What I need to do is get some sort of scope and watch what is going on while I am playing around.

billyzelsnack
12-27-2010, 11:53 PM
Personally, I've always been a proponent of separate supplies for critical things. Let one battery run all of the servos, but use a separate battery just for the electronics.

Easy fix to a complicated problem..

Gary

I could do that, but this is eventually going to end up in my biped and weight is a big concern for me.

lnxfergy
12-28-2010, 07:24 AM
Fergs,

Do you think I would encounter this type of problem on my mech using the bioloid wires?

I have 15 servos, arbotix, 2 tank guns, camera, wifi encoder, lipo audio alarm, target setup, going to be hooked up through the bioloid cables. This will be run from the 3S 11.1V 2100mAh Pro Lite V2 20C LiPo Battery.

The last servo in the leg is the farthest run on the daisy chains.

I got a little worried from looking at this thread and seeing what you said.

This isn't much of a problem for quads -- because your cables aren't as long of runs. In a biped, the cables go all the way down the leg in a single, long, continuous run (typically). In a quad, you typically only have 2-3 servos (a leg) on a run. I did have some issues with a hexapod at one point -- but only because I was trying to go 0.6m/s in a tripod gait with a cruddy old tenergy battery (replacing the battery fixed the issue). The Pro Lite batteries are very good at sourcing large amounts of current quickly, which further reduces the chance of issues.

-Fergs

lnxfergy
12-28-2010, 07:28 AM
I think this type of thing is just too advanced for me. What I need to do is get some sort of scope and watch what is going on while I am playing around.

This is a pretty good idea.


I could do that, but this is eventually going to end up in my biped and weight is a big concern for me.

If you can't get capacitors alone to fix the problem, I'd recommend splitting the HXT-900s off onto their own regulator (less weight than extra batteries, but lots of noise protection over sharing a regulator). The 1 and 3W switchers (http://www.trossenrobotics.com/store/p/5987-25W-Step-down-adjustable-switching-regulator.aspx) that Trossen sells are really good at keeping noisy servos from interfering with other electronics. On Nelson's face I used a 25W step down switcher to power 6 micro and 2 hobby servos, and it did a pretty good job of isolating the noisy servos from the rest of the circuit.

-Fergs