Now that we know they were shorted. Kris Verdeyen mentioned a closed FET as a sign of static electicity. Is it possible your wheels are generating static electicity, and the high voltage made it’s way to the wheel’s speed controller’s FETs? Also, what were the speed controlers connected to when they failed? Just the wheels or an arm or roller?
Could be. I never heard reports of static discharge from the drive teams until Nationals (when we didn’t have problems), but I’ll ask them. I never encountered static when the robot was driven on the carpet before shipping though. This is a theory that could easily be tested at home when the robot gets back.
The controllers that blew were on both the drive and winch motors, I don’t believe the arm (van door) controller was one of the victims.
Guys,
Static of sufficient energy to kill FETs would have been causing problems everywhere else, too. These FETs have internal protection to prevent damage (zener diodes) and are part of a circuit board where 12 FETs live. For only one quadrent to die from static is unlikely. It is still a big puzzlement though. For only one quadrant to fail (Not two opposite quads) then there had to be something wired incorrectly in the external circuitry. Something like one of the battery leads going to the motor and the other going through the controller. This could put full battery voltage across one quad set but not destroy anything else. It seems so unlikely though since the wiring error that would cause this would be easily diagnosed. It could happen, just not high on my list of possibles. It may remain a mystery for the rest of eternity.
If we discover anything new through testing during the off-season I’ll post the results. I think I’ll take one Victor (maybe an 883) and run controlled tortures. I might replace the FETs with lower amp models so they self destruct more easily (and are cheaper to replace).
I want to thank everyone for their thoughts and ideas!
FIRST people are just so VERY nice!
I am trying to replace the FETs on a Victor. First, they seem a bit tough to unsolder intact, so I clipped them off. Then, when I took the remaining pins out, the one with the square pad on the top side of the board lifted that pad off. Is this a problem? It didn’t seem to be connected to a trace, but it is hard to tell with that silicon goop on there. Are the holes plated through, that is, do the FETs need to be soldered on both top and bottom? Does anyone have any tricks or advice to share on replacing the FETs? Thanks!
Bob,
The best way to handle the pins is this…
- Use a solder sucker or solderwick to remove as much as solder as possible.
- Heat just the pin with your iron and wiggle the pin with a small screw driver. While the pin is moving, remove the iron and the pin will not solder itself back in place.
- Remove the device.
Any pad that lifts off is likely part of a plated through hole. When you install the new FET, just add solder to the bottom of the board and watch to see if the solder wicks up inside the hole. What is unknown is whether the board has more than two layers. Although some teams report that replacing FETs works, they can take out other devices. For the cost and reliability issues, I recommend a replacement. Anything that is stressed enough to take out FETs has other problems waiting to happen.
Edit…BTW a square pad is usually an indication of a particular pin (often pin 1) to assist with part insertion. Watch which way you put the FETs back in, they don’t like to be reversed.
You also may need to clean out the holes if the new one doesn’t go in easy. Forcing it may push out the plated through hole. If this is a multilayer board, pushing the hole out kills it in most cases.
Did you replace all of the FETs or just the blown ones? What part # did you use to replace them (was it the same as the ones IFI uses)? I’ve had some trouble finding fets with the same part # as the ones in one of our blown speedcontrollers.
I believe our electronics group typically replace only the blown ones, then bench test the rebuilt unit.
The FETs used are IRL3103 (International Rectifier) or an equivalent like FDB6035AL (Fairchild Semiconductor) for our competition victors, but close matches have been used for ones on the practice or older robots.
Digi-Key
http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=202764&Row=321579&Site=US
$1.63 individually or $1.16 ea. for 10
http://www.mouser.com/index.cfm?handler=displayproduct&lstdispproductid=199954&e_categoryid=277&e_pcodeid=51229 Mouser has them for $1.19 each if ordered individually or .90 each for 25
Al,
I’ve unsoldered parts from a lot of non-USF boards, including packages like these FETs, but the solder seems very difficult to melt out on these. My always reliable heated solder sucker can’t melt it. I put a larger 1/4" chisel point iron on all three pins and it took a while to melt. I guess the traces must be huge since these just pull the heat away. Parts of the board and the screw terminals get pretty hot. I seemed to lift a pad on one side or the other on just about all of them. Then I started snipping the FETs off and taking out pins one at a time. Still lifted pads. I guess you need a very hot iron. Since my sucker wouldn’t work, I ended up using a PC board drill to clean the holes out. Of course, if they are plated through, I may have partly or completely removed that. I tried soldering in a new FET, but there wasn’t anything for the solder to stick to on some pins.
I am sure if I play around longer I can get the knack, but I don’t think I will. You are right about the stress probably shortening the life of the rest of the board. I will either have them rebuilt (I see http://www.rotordesign.com/victor.html is doing them for $75 now plus shipping) or we will just get new ones. The rebuilts will be good for practice and we’ll save the new ones for competitions.
I guess we have a motor problem that is blowing them. We lost one Vic and the replacement also went. I wish the breakers ‘broke’ fast enough to protect the Vics. As someone mentioned, the Vics ‘sacrifice’ themselves to save the breakers. Well, for my sake, they don’t have to be so selfless. I wish they would let the breakers do their job.
Since I have trashed this board, I will probably rip it apart and see what I find in the way of layers.
BTW, I have now 24 brand new (and one slightly soldered) IRL3103 FETs from Mouser that I probably will not be needing any more. I don’t know if its legal to offer things for sale here so I won’t do that, but if anyone wants to discuss FETs, let me know.
Bob
Mark,
When one of our Vics went, it just smoked. It was running a match at the time. The crowd loved it. But when its replacement went in the pits, it was spectacular. We and another team saw a huge spark. I was about 50 feet away up in the stands and it sounded like a firecracker. All three FETs were cracked, of course, but their damage was incremental. One just cracked, another had a blob of white stuff on it, and the third had a larger blob of white. Must be vaporised innards. It could be that they went one at a time, but very close together. I am guessing the destruction got worse as the load on the remaining ones increased. As I mentioned earlier, we likely had a motor problem that caused it. Haven’t opened the motor yet.
Bob
Bob,
A good solder should go “though” the board and both sides will show a clean and shiny solder wick connecting pad and pin.
The higher the current, the wider and/or thicker the copper trace needs to be on the PCB. This larger copper will leach away any heat you are attempting to apply to the connection and frustrate your efforts.
About 3 years ago, I did a power electronics design PCB able to carry over 200 amps. The soldering process was to have two technicians simultaneously heat the pin/pad from both sides of the board and then to apply solder to one side and verify that it flowed through to the opposite side of the board and wicked to both pin and pad.
De-soldering was nearly impossible as a one person operation…
Based on my experience:
-
Cut all FET pins with a good pair of transverse cutters. You will be removing each pin individually…
-
Heat both sides of the PCB (a two person operation) and pull the pin out with a good pair of needle nose pliers or a small spring hook (recommended for low thermal mass). Note that most pad delamination happens at this step…
-
Note that there will probably be solder left in the hole… Heat from both sides with soldering irons with very thin tips.
-
Remove the soldering iron on one side and then place a vacuum solder sucker on the hole. Note that a professional desolder station is highly recommended as the solder sucker tip is also the heating element…
-
Remove the other soldering iron and engage the solder sucker.
This should clean the hole without lifting the PCB pad.
Note 1: The above is augmented if you have design information on the PCB (at what temperature will it delaminate?) and solder/desolder stations with adjustable temperatures (as design engineer, I had both).
Note 2: My experience is with 60/40 lead/tin solder. Newer “lead free” solders melt at an even higher temperatures…
At any rate, good luck…
Mike
Bob,
Mark has some good advice. Although I have never changed the FETs, what you describe sounds like a combination of a large thermal mass and holes that are too small for the leads. It is possible that the board was drilled for the right lead size but never compensated for the plated through hole.
As far as the fire and flame from the controller, let us know if you find a bad motor. Most often, it would have left evidence in some other way like driving problems. You may have a thermally induced short. This is where wiring breaks down under high temperature. You may have a wire in the motor that shorts when the motor gets hot but is fine under test in the pit. The phenom of one FET blowing after another is normal as the shared load gets dumped on the remaining FETs or the dead one produces a short on the other two. High junction temperatures inside the package cause the case to crack open. The light (arcing) you see is the FET trying to turn itself into an LED as the silicon junctions break down one after another. I love the smell of burning silicon in the morning.
For more info on replacing bad FETs on the Victors, you can go to www.delphiforums.com and go to the Battlebots forum. They have a few people in threads that fix a ton of them (even upgrade them).
Hey all!
Been awhile…
Just my thoughts: I have seen a couple vics blown out in several ways. The most common I have seen (and yes, I am ABSOLUTELY SURE this is what I saw) is people pushing their robots when they are off. This creates juice from the motors, which goes where? the victor. (I really don’t recomend this, but if you backdrive the motors and look at the victors, sometimes the lights will turn on. Again, because of the previous point, I don’t recomend this.) Push it hard enough, they can zap themselves. This is what I think happened with the hand-cranking the winch, and also it makes sense that only one set of FETs would blow out if you’re only wheeling it one way, so the other direction should still work. Also, check the main capacitor in the middle. Another man I once met inspecting the Vics noticed some sticky stuff on one of his, and said it’s electrolyte. I don’t know how much truth there was to this, but if your capacitor’s top is bulging, then it’s gonna blow. Replace it before bad things happen.
Hope I am 1)helping, not hindering, and 2)not going over old territory.
Sparks
Oh, and Bob? That white stuff, I’d bet, is electrolyte. I could be wrong (you’re not supposed to see it, so I don’t know what color it is) but that seems to be the best guess I have.
Sparks,
I thought we lost you, glad to see you back. We have been back driving Victirs for years with out any ill effects. The motors do act as generators and there is enough leakage current that backs up through the controller to turn the LEDs on but unless you are pushing the robot at highway speeds in is unlikely you can genreate enough current to fry the controllers. The main cap in the middle of the FETs is rated pretty well for the voltage we use but where it is located subjects to very high heat when the robot is working hard, like during practice. The heat causes early failure and the case has a pressure relief that leaks the electrolyte out of the can. This is nasty stuff so don’t touch and then put your finger in your mouth. When it has dried for several days, it does turn into a white, crusty almost lime looking (as in hard water deposit) gunk and it eats copper and other metals. The white smudges are really the result of the magic smoke leaving the device at high speed. They are mostly the ashes of the silicon interior that vaporized during the event. If everything looks normal except the white residue, know that a closer look will show a crack in the case. In some of the more spectacular deaths, there has been loud cracks followed by microscopic plastic shrapnel as the cases of the FETs blow apart.
Al,
That sounds reasonable. My only point here is that some teams really gear their motors down, so it is sorta kinda maybe possible that by backdriving they are spinning the motors much faster than the wheels. I’m drawing at straws here, aren’t I. What I saw was a team pushing their robot in a parking lot, then complaining of a burning smell. On closer examination, the victors wouldn’t turn on. Since it was another team’s, I couldn’t get a close look at the vics, but a couple mentors blamed it on backdriving. I have avoided it ever since.
Thanks for setting me straight!
Sparks
Sparks, If you can remember the team I would like to find out exactly what they found.
Thanks,
On our robot, driving the arm at any decent speed, (its powered by a geared down Van Door Motor) will provide enough juice to turn on the RC, and our LEDs will continue to blink running off the 7.2v battery after the motor stops providing power. I’ve even seen the fans on the victors spin up from pushing the bot on the floor. Not the big fans though… I think the major source of problems is if there is somewhere for the current generated to go or not. Because if the current cant flow, the voltage generated increases. That is what could possibly cause problems.
If this were the problem, it would be very possible to provide something to sink the current and keep voltage spikes down. A 120mm fan or two would do the job, or maybe some sort of lighting. Just make sure that the load is wired so that it is on when the main breaker is on. Can I say for sure that this is the problem? No. Is it a possibility? yes. Look at the schematic symbol for some FETs and you see a reverse biased diode, with the H-bridge, that would act as a rectifier, and power up the bot.
I’m with Al here on the backdriving. I’ve seen some LED’s light up, and the fans twitch, but there can’t be anywhere near enough current flowing to mess up the fets.
Or is there?
Could it be that with dynamic braking on, the current on some victors just ends up as heat as opposed to getting sent onto the rest of the harness?
I’m just wondering that maybe if dynamic braking is not on, all that backdriven current gets disapatied through all the victors, RC, etc., causing little if any problems. If braking is on, that current would ‘stay’ in the victor, right? Or am I just missing something key about how braking works on a victor?
Just grasping at some more straws.
-Andy A.
Andy,
I am guessing on this, but I don’t think the jumper keeps the brake on when the controller is not powered. If that was the case, you wouldn’t be able to push the robot with the wheels on the floor. Also, as I remember, IFI started including a varistor across the output. This was something we previously had to add to the output terminals. The device is designed to eat up spikes from brush noise and back drive so I think it limits the voltage to 40 or 50 volts. I am going to have to look back at the specs on the devices we had to add in the past.