View Full Version : Potential DSC Damage
As I can't find this specification on the USFIRST-provided spec sheet (http://www.usfirst.org/sites/default/files/uploadedFiles/Robotics_Programs/FRC/Game_and_Season__Info/2012_Assets/Digital%20Sidecar.pdf) (or a confirmation through forum searches), I'm turning to the CD community for a definitive answer to this question: Is the ground on the GPIO pins of the DSC 12V tolerant? Alternatively, will plugging in a 12V signal to the power or ground pins on the GPIO damage the DSC?
The reason I ask is because we've been having issues with an apparent short to our frame, which induces sudden motion upon providing power to the robot, but the short point is unidentifiable, and we suspect the damage lies in the DSC.
The current situation is as follows: Closing the circuit breaker (giving power to the PDB) will, on occasion, incur forward motion in a number of motors that persists for a few seconds. I know for sure that the motors on PWMs 1-3 and 6-7 are moving forwards; 4 and 8-10 have no connections, and 5 may be doing so (but is unconfirmed). Later testing only saw motion in a subset of these motors (which is undetermined).
The backstory: The sensors (details later in the post) were wired with +12V and GND to a 20A PDB, a 1K pullup across the 5V and signal, and the signal wire was mistakenly plugged into both the +5V and GND (at two different points).
The one common factor that we have confirmed is that this phenomenon only occurs when the signal from our conveyor sensors is plugged into the DSC (on the signal pins), in particular when one of them is activated. We are using Banner QS18VN6D sensors (http://info.bannerengineering.com/xpedio/groups/public/documents/literature/63908.pdf) (the white signal line) for this purpose.
The camera and cRIO are isolated from the frame and we cannot identify any other potential shorting point.
Edit: Omitted detail: All 3 lights on the DSC are lit. The RSL has yet to be connected, so I can't describe its behavior.
We didn't get a chance today, but we plan to confirm tomorrow whether the point of failure is the DSC (if we can reproduce the issue) by cutting its power and unplugging the DB37. Until then, it would be great if someone could confirm our hypothesis.
Kevin Sevcik
19-02-2012, 23:41
I won't hazard a guess. I suggest you PM EricVanWyk (http://www.chiefdelphi.com/forums/member.php?u=17696) and direct him to this thread. (Assuming he hasn't found it already.) He has some deep insider knowledge about the DSC and should be able to answer you quickly and authoritatively.
MAldridge
19-02-2012, 23:41
it is definitely not a good idea to be dumping 12v into the sidecar. I would suggest you sink your power elsewhere, and use a transistor arrangement to connect the two pins (sig,gnd) so that there is no extraneous power on the sidecar.
I won't hazard a guess. I suggest you PM EricVanWyk (http://www.chiefdelphi.com/forums/member.php?u=17696) and direct him to this thread. (Assuming he hasn't found it already.) He has some deep insider knowledge about the DSC and should be able to answer you quickly and authoritatively.
I've already done so :o
Nonetheless, I still wanted to put this out there for anyone else who's familiar with and knowledgeable enough about the DSC to respond.
BitTwiddler
19-02-2012, 23:44
As I can't find this specification on the USFIRST-provided spec sheet (http://www.usfirst.org/sites/default/files/uploadedFiles/Robotics_Programs/FRC/Game_and_Season__Info/2012_Assets/Digital%20Sidecar.pdf) (or a confirmation through forum searches), I'm turning to the CD community for a definitive answer to this question: Is the ground on the GPIO pins of the DSC 12V tolerant? Alternatively, will plugging in a 12V signal to the power or ground pins on the GPIO damage the DSC?
The reason I ask is because we've been having issues with an apparent short to our frame, which induces sudden motion upon providing power to the robot, but the short point is unidentifiable, and we suspect the damage lies in the DSC.
The current situation is as follows: Closing the circuit breaker (giving power to the PDB) will, on occasion, incur forward motion in a number of motors that persists for a few seconds. I know for sure that the motors on PWMs 1-3 and 6-7 are moving forwards; 4 and 8-10 have no connections, and 5 may be doing so (but is unconfirmed). Later testing only saw motion in a subset of these motors (which is undetermined).
The backstory: The sensors (details later in the post) were wired with +12V and GND to a 20A PDB, a 1K pullup across the 5V and signal, and the signal wire was mistakenly plugged into both the +5V and GND (at two different points).
The one common factor that we have confirmed is that this phenomenon only occurs when the signal from our conveyor sensors is plugged into the DSC (on the signal pins). We are using Banner QS18VN6D sensors (http://info.bannerengineering.com/xpedio/groups/public/documents/literature/63908.pdf) (the white signal line) for this purpose.
The camera and cRIO are isolated from the frame and we cannot identify any other potential shorting point.
We didn't get a chance tiday, but we plan to confirm tomorrow whether the point of failure is the DSC (if we can reproduce the issue) by cutting its power and unplugging the DB37. Until then, it would be great if someone could confirm our hypothesis.
Is the 5V indicator on the DSC still lit? I'm not sure you did any damage to the DSC by routing the signal line to either ground or the 5V pins. Worse case I see is if the output was at ground and you plugged it into the +5V pin in which case the on-board regulator would go into current limiting mode. Now if you had routed the 12V to the DIO or +5V pins then damage might occur. Be advised, the DIO input pin connects straight to the cRIO DIO module through the 37 pin cable.
Is the 5V indicator on the DSC still lit? I'm not sure you did any damage to the DSC by routing the signal line to either ground or the 5V pins. Worse case I see is if the output was at ground and you plugged it into the +5V pin in which case the on-board regulator would go into current limiting mode. Now if you had routed the 12V to the DIO or +5V pins then damage might occur. Be advised, the DIO input pin connects straight to the cRIO DIO module through the 37 pin cable.
All 3 lights on the DSC are still lit.
The 12V and GND lines of the sensors were never connected to the DSC.
BitTwiddler
20-02-2012, 00:19
All 3 lights on the DSC are still lit.
The 12V and GND lines of the sensors were never connected to the DSC.
As long as you were using the NPN version of the sensor (the model number you gave says it is) then I don't think you got any 12V into the DIO input so you should be OK there. You may need to look at the wiring to the sensor to see if there is a frame short there. I don't think the DSC is the source of the short but I could be wrong.
As long as you were using the NPN version of the sensor (the model number you gave says it is) then I don't think you got any 12V into the DIO input so you should be OK there. You may need to look at the wiring to the sensor to see if there is a frame short there. I don't think the DSC is the source of the short but I could be wrong.
The terminal block we're using for the sensor connections has been checked for a frame short, so we know the sensors themselves aren't the issue.
BitTwiddler
20-02-2012, 00:45
The terminal block we're using for the sensor connections has been checked for a frame short, so we know the sensors themselves aren't the issue.
Looks like you'll need to pull out the ohmmeter at it's lowest scale and probe various points to ground. Pay attention to the values you read as they should get smaller as you get electrically closer to the short.
BTW, the specs for the cRIO digital IO module (http://www.ni.com/pdf/manuals/374069e.pdf) indicate that the inputs can tolerate up to 30 Volts. (Not happily for long but it will tolerate it).
Looks like you'll need to pull out the ohmmeter at it's lowest scale and probe various points to ground. Pay attention to the values you read as they should get smaller as you get electrically closer to the short.
The intermittence of the problem prevents us from effectively isolating the issue, as does a lack of multimeters with that level of precision.
Thanks for your help! The 30V tolerance is good to know.
EricVanWyk
20-02-2012, 01:34
Please don't do it again, it wasn't designed for this sort of situation! :-)
The short, conservative, answer is that this fault condition is not within spec, and further operation is not guaranteed. The GPIO signals are not intended for 12V operation in any configuration, the GPIO power and ground pins are intended as outputs only. Unfortunately, that isn't made clear in the documentation you linked to. I will try to include this information in the next revision.
The issue is that the GPIO pull-up resistors and ESD protection diodes provide paths for this extra voltage sneak across the rest of the board, so it could have damaged any system within the DSC. It is very likely that the board needs to be replaced, most of the circuitry is now suspect especially since an unrelated section is faulting.
The longer answer is that "a few seconds of forward motion from multiple motors" has to be the weirdest symptom I've ever heard for a DSC fault. My best guess is that it is somehow generating a single PWM pulse. Here are the problems with that theory:
1) It should generate less than a second of torque before the motor controllers shut down again. Maybe inertia makes it look like several seconds?
2) The pulse length window is pretty short. It feels unlikely that a fault could accidentally create a valid signal.
Are you using Victors or Jaguars?
Do you have control of the motors after their initial jump?
Have you tried swapping in a different DSC?
The NI9403's datasheet states that it can withstand up to +/-30V (with no impedance) on up to 8 channels. You likely applied +/-12V with 0-20 kOhms impedance to many of its channels. I can't guarantee that you did not damage this unit, but I suspect the DSC much more than the 9403.
Thanks for your response.
The short, conservative, answer is that this fault condition is not within spec, and further operation is not guaranteed. The GPIO signals are not intended for 12V operation in any configuration, the GPIO power and ground pins are intended as outputs only. Unfortunately, that isn't made clear in the documentation you linked to. I will try to include this information in the next revision.
Good to know.
The longer answer is that "a few seconds of forward motion from multiple motors" has to be the weirdest symptom I've ever heard for a DSC fault. My best guess is that it is somehow generating a single PWM pulse. Here are the problems with that theory:
1) It should generate less than a second of torque before the motor controllers shut down again. Maybe inertia makes it look like several seconds?
2) The pulse length window is pretty short. It feels unlikely that a fault could accidentally create a valid signal.
When it first surfaced*, it actually just spun around (http://www.youtube.com/watch?v=airDtge-MN8&feature=youtu.be). Unfortunately, I wasn't present to observe this occurrence, so I can't provide more details than are present in the video. Later on, the issue was indeed momentary twitching upon powering the robot that would terminate after a few seconds.
*When the issue first surfaced, we had two unresolved issues. 1) There was a case short through the camera (Axis 206). This issue has since been rectified. 2) When the circuit breaker was open (no power applied), but the battery connected, we could read 12V across the +12V from the battery and the chassis. However, I'm afraid that this wasn't the issue at fault, as a)we measured 12V across the chassis, not 5V, which is the nominal spec on the camera's power line and b) the issue persisted following removal of the camera.
Are you using Victors or Jaguars?
We are using 6 Victors, with 2 Jaguars on CAN.
Do you have control of the motors after their initial jump?
Yes, we do.
Have you tried swapping in a different DSC?
That will be the second thing we'll be doing when we go in today (the first will be, if the issue remains reproduceable, disconnecting the DSC from all sourced power).
The NI9403's datasheet states that it can withstand up to +/-30V (with no impedance) on up to 8 channels. You likely applied +/-12V with 0-20 kOhms impedance to many of its channels.
The chassis short was +12V, with 7 kOhms of impedance.
I can't guarantee that you did not damage this unit, but I suspect the DSC much more than the 9403. Hopefully we'll be able to provide more details after the DSC is switched out.
We've swapped out both the DSC and 9403 (given time constraints, we're not taking any risks - we'll assess failure points post-ship) and the issue seems to have desisted (although whether the issue had been persisting with the DSC in question is in itself debatable).
We've also opened the DSC in question, but there is no noticeable damage to it. Here it is: http://dl.dropbox.com/u/11542623/IMAG0071.jpg
If you'd like, we could send it to you so you could take a look at it.
We've swapped out both the DSC and 9403 (given time constraints, we're not taking any risks - we'll assess failure points post-ship) and the issue seems to have desisted (although whether the issue had been persisting with the DSC in question is in itself debatable).
We've also opened the DSC in question, but there is no noticeable damage to it. Here it is: http://dl.dropbox.com/u/11542623/IMAG0071.jpg
If you'd like, we could send it to you so you could take a look at it.
Looking closer at the left edge of the board it appears that the resin has melted. We occurred similar issues last year and damaged 3 DSCs before locating the issue. By having the power and signal wires reversed on our pressure sensor we were sending 5V "the wrong way". In one instance the voltage was enough to desolder some of the contacts on the board. If the resin looks wavy and non uniform you have a definite wiring problem.
Looking closer at the left edge of the board it appears that the resin has melted. We occurred similar issues last year and damaged 3 DSCs before locating the issue. By having the power and signal wires reversed on our pressure sensor we were sending 5V "the wrong way". In one instance the voltage was enough to desolder some of the contacts on the board. If the resin looks wavy and non uniform you have a definite wiring problem.
I don't recall seeing any resin damage (what you're seeing may be just the poor quality of photo resizing done by mspaint), but I didn't take that close a look at the board - making sure the robot got rewired correctly was the more pressing issue at the time :o I'll make sure to take another look at it tomorrow.
But three sidecars... :ahh:
Thanks for your input :)
EricVanWyk
22-02-2012, 14:12
Fe_Will -
Could you send me further information on how you had mis-wired the pressure sensor? A link to the sensor's datasheet and a quick schematic of the wiring would be great. I'd like to take a look to see if it possible to prevent this sort of damage in the future.
slijin -
I'm glad to hear that you have resolved the issue, at least enough to compete.
From looking at the video, the good news is that it happens very shortly after power on. This leads me to believe the fault is within the DSC and not the 9403. It might be interesting to try again with the cable between the DSC and the 9403 disconnected.
Fe_Will -
Could you send me further information on how you had mis-wired the pressure sensor? A link to the sensor's datasheet and a quick schematic of the wiring would be great. I'd like to take a look to see if it possible to prevent this sort of damage in the future.
I really don't have either... It was a KoP pressure sensor and the 5V wire was switched with the signal on the sensor side. The initial wiring was done Thursday at the Autodesk Oregon Regional, a hasty fix for sure. At first we thought we had a chassis short. During testing we discovered that the DSC was bad. We replaced it and continued looking for our perceived short. After burning out DSC #2 we started looking at what was going on with that part of the system. A few changes later we mounted #3 and turned it on. Everything was fine until the pneumatic system started charging and then we melted that one.
An alum from another team was able to solder the contacts back onto the best board of the three, we ran with it for two events after rewiring the sensor. The other boards were damaged beyond repair and disposed of. Good thing we had spares and someone who was able to fix the one we had. Unfortunately, our main concern was competing and not further testing the DSC.
Kevin Sevcik
22-02-2012, 17:23
Eric,
The pressure sensor is a Nason SM-2B-115R/443 (http://www.nasonptc.com/pdfs/SM_specs.pdf). It's just a pressure actuated switch, so it's nothing more complicated than a NC dry contact.
I'm not sure if I understand the mistake Fe_Will made in wiring the sensor. Standard wiring is signal to one side and common to the other, as with a limit switch. Reversing the connector where it plugs into the DSC wouldn't cause any problems. I think he's saying they wired it by chopping the end off a standard PWM cable, and they accidentally wired red and black to the switch instead of white and black. I assume a short from 5V to common would damage the DSC in some fashion, yes?
EricVanWyk
22-02-2012, 17:45
I assume a short from 5V to common would damage the DSC in some fashion, yes?
Thanks for the clarification.
Usually, it would be fine. The 5V supply will detect the short and shut itself down. Periodically it will wake up and try again. I don't remember the exact numbers, but with a dead short I think it averages out to 100-200mA (don't quote me on that). I've left the supply shorted out for hours at a time on the bench without issue.
There is a "danger zone" where it doesn't draw quite enough current to detect the fault. In a worst case scenario, a specially tuned resistance could draw up to 5 Amps continuous. Maybe the sensor + wire path has exactly the wrong amount of resistance?
I'm still mildly surprised it caused permanent damage. 5 Amps is a lot, but there is a solid connection to the power and ground planes to wick away the heat. I'm wondering if bonus current is being added to the circuit - maybe from the compressor?
I think he's saying they wired it by chopping the end off a standard PWM cable, and they accidentally wired red and black to the switch instead of white and black. I assume a short from 5V to common would damage the DSC in some fashion, yes?
Yes it was a PWM cable that was used, is this not common practice?
I can't speak for the DSC, as far as I am concerned it is plug and play. All of the boards had melted resin on some or all of the board. It pooled in some areas and was dry in others in the severe board. In the least damaged case it had a wavy, non-uniform appearance.
Matt Krass
22-02-2012, 19:48
I'm still mildly surprised it caused permanent damage. 5 Amps is a lot, but there is a solid connection to the power and ground planes to wick away the heat. I'm wondering if bonus current is being added to the circuit - maybe from the compressor?
Don't you mean Wyk away the heat? ;)
Joking aside, 5A is a fair amount of current, if it was pulling that much for a long time (in the last few days of build our poor robot was pretty much on nonstop for 15 hour stretches), it seems feasible that the board would eventually start suffering damage. The impression I got from this post:
During testing we discovered that the DSC was bad. We replaced it and continued looking for our perceived short. After burning out DSC #2 we started looking at what was going on with that part of the system. A few changes later we mounted #3 and turned it on.
was that there was a fair amount of time between power up and the actual damage, which may have also been done incrementally?
The 5V regulator for the DSC is the TPS5431DDA, correct? As I see here:
http://www.ti.com/lit/ds/symlink/tps5431.pdf
That is a nifty buck converter, nice high speed switching frequency too :)
It also has a thermal shutdown, but that is at a (typical) value of 162C, which is pretty high, I don't have the figures on the PCB/resin material to estimate how hot it is getting, nor what temperatures it will tolerate, so could you tell me if that kind of failure is feasible with long term exposure to those temperatures?
If so, perhaps using the (currently disconnected) enable line on the converter with a temperature sensor on the board near the regulator and a comparator to provide a much lower thermal shutdown? I can't imagine our typical operations heat the system that much (but, again, I don't have all the numbers here). Just throwing ideas out there.
Matt
EricVanWyk
22-02-2012, 20:22
Yes it was a PWM cable that was used, is this not common practice?
I can't speak for the DSC, as far as I am concerned it is plug and play. All of the boards had melted resin on some or all of the board. It pooled in some areas and was dry in others in the severe board. In the least damaged case it had a wavy, non-uniform appearance.
Can you send me some high res pictures? I'll PM you my email address.
Don't you mean Wyk away the heat? ;)
Groan... :)
It rhymes as "I like VanWyk", if I can steal the campaign slogan. Re-rhyming that to wick seems a little too... personal.
Al Skierkiewicz
24-02-2012, 08:16
Fe,
We need some pictures but I suspect there is more to this story. The PWM cable alone cannot support enough current to melt anything on the DSC. It would burn and smoke long before anything on the DSC let alone melt the solder on the pins. You are using a Spike to drive the compressor and not trying to drive it direct from the DSC correct?
vBulletin® v3.6.4, Copyright ©2000-2017, Jelsoft Enterprises Ltd.