Post by thread:%Industrial Electronics Problem [OT]%

To the forum: The problem I have does not involve a PIC; however, there appears to be an abundance of knowledge in this forum regarding electronics based on some of the messages I've read so far. Is there anyone out there in the forum that has experience with Design of Experiments (Taguchi-type) in addition to their electronics troubleshooting skills. I have been seeing repetitive problems with a MM54C14J Hex Inverter (an old ceramic-style part) where the Vcc is getting creamed. Because the explanation of the failure will be a bit too long to adequately put into a message, the best place to get further details of the failure type is to go to http://web.gmtcom.com/~k3jsch/sixsigma/access.htm The problem occurs regardless of the following: 1) Whether it has been in service for only a few hours or at least 8000 hours. 2) Whether it is operating at a temperature of 55 degrees C or slightly past 100 degrees C 3) Whether it is operating in a 60Hz domestic unit or a 50Hz overseas unit. 4) date code of IC Any suggestions, etc., are welcome. What I am trying to do is set up an experiment at our production facility to inject noise, etc. into the design as it is running. Factors I am looking at are : 1) Levels of noise voltage to inject into the cable between the microprocessor unit that this IC resides on and the gate firing board of the SCR bridge that is directly mounted over the 3-phase, 6SCR bridge that ultimately runs either the left track or right track motor. 2) Brand name of chip. 3) Whether or not modification to the shielded cable and/or its placement in the circuit would make a difference. Regards, Kelly Schauf spam_OUTk3jschTakeThisOuTmail.gmtcom.com

Not that I'm any great expert on RF, but this looks familiar. One of those "This one bites everyone" things that is good for "Breadth of experience" for newbie & old-timer alike. It's probably not a ground loop, I'd guess you've looked at that! Here's MY best guess; Is the capacitance of that shielded cable, what's killing the 54C14J's? That's in common between all the failing parts. And I know I've almost been BIT by this one before, driving MOSFET's gate capacitances would have killed my 4000 series CMOS circuits dead if I hadn't thought it through in advance. (Current limiting resistance... Could be FET oscillation, I know I'm not an expert on that just now <G> My "quick dirty" solution: Split the 1k resistor at the output of that shielded cable, into 2 resistors - one, say 100R, between the 54C14J and the cable, to limit 54C14J max output current, and the other, say 900R at the input to the FET. (You want to prevent both FET oscillation and drawing too much current out of that FET.) There's probably a better optimized resistor value splitting to be done there, there are some past posts in the PICList archives that covered this IIRC. (http://www.iversoft.com/piclist/ - I'd go look there.) I don't usually memorize solutions, just "Gotchas", and look up the solutions when needed. And check datasheets <G> Mark Kelly Schauf wrote: {Quote hidden}> > To the forum: > > The problem I have does not involve a PIC; however, there appears to be > an abundance of knowledge in this forum regarding electronics based on > some of the messages I've read so far. > > Is there anyone out there in the forum that has experience with Design of > Experiments (Taguchi-type) in addition to their electronics > troubleshooting skills. I have been seeing repetitive problems with a > MM54C14J Hex Inverter (an old ceramic-style part) where the Vcc is > getting creamed. > > Because the explanation of the failure will be a bit too long to > adequately put into a message, the best place to get further details of > the failure type is to go to > http://web.gmtcom.com/~k3jsch/sixsigma/access.htm > > The problem occurs regardless of the following: > > 1) Whether it has been in service for only a few hours or at least 8000 > hours. > 2) Whether it is operating at a temperature of 55 degrees C or slightly > past 100 degrees C > > 3) Whether it is operating in a 60Hz domestic unit or a 50Hz overseas unit. > > 4) date code of IC > > Any suggestions, etc., are welcome. What I am trying to do is set up an > experiment at our production facility to inject noise, etc. into the > design as it is running. Factors I am looking at are : > > 1) Levels of noise voltage to inject into the cable between the > microprocessor unit that this IC resides on and the gate firing board of > the SCR bridge that is directly mounted over the 3-phase, 6SCR bridge > that ultimately runs either the left track or right track motor. > > 2) Brand name of chip. > > 3) Whether or not modification to the shielded cable and/or its placement > in the circuit would make a difference. > > Regards, > > Kelly Schauf > .....k3jschKILLspam@spam@mail.gmtcom.com

Kelly, I couldn't guarantee that I know what is happening here but I suspect that I could STOP it happening for very little effort, BIMBW * i Doesn't seem likely but - High dv/dt on FET switching can reflect into gate. Oscillation can occur on gate under "correct" conditions. The voltage here would need to be high enough to drive IC into latch up via 1K drive resistor. ii I don't know how long the cable is but significant energy *may* be able to be stored in the cable. What is it's capacitance. You are driving it directly (NO resistor from gate output. Is this within spec? (ie - depending on connection method there may effectively be capacitance to ground bypassing the FET gate resistor). iii Regardless of mechanism, the *cause* would *seem* to be energy from the cable getting into the gate's output. Faced with this problem I would. a) Strongly consider placing the 1K series resistor at the gate output rather than at the FET (isolates gate from direct capacitance) or split the resistor to eg 470r at each end of cable. At present the IC output is directly exposed to anything that *may* appear on the cable. b) Place Schottky catch diodes to ground and supply at the IC output. With 1K resistor and say 30ma capability (a very small Schottky (BAT85 etc?)) you can withstand 30 volt continuous above or below power supply rails without sinking ANY current in the ICs substrate diodes. As long as the problem is not from somewhere else (eg spikes via the Vcc rail) these steps would be very very (very ...) likely to prevent damage, regardless of the cause. With the amount of energy you have in the general vicinity I would want to be extremely certain that my power supply was not the source of problems BUT I would pick the above as most probable. regards Russell McMahon * BIMBW = But, I May Be Wrong :-) PS - if I had to "test" this situation rather than fixing it I would consider applying, to the IC output, capacitors of the value equivalent to that of the cable, charged to increasingly greater voltage until I replicated the fault (or couldn't). From: Kelly Schauf <k3jschKILLspamMAIL.GMTCOM.COM> >To the forum: > >The problem I have does not involve a PIC; however, there appears to be >an abundance of knowledge in this forum regarding electronics based on >some of the messages I've read so far. > >Is there anyone out there in the forum that has experience with Design of >Experiments (Taguchi-type) in addition to their electronics >troubleshooting skills. I have been seeing repetitive problems with a >MM54C14J Hex Inverter (an old ceramic-style part) where the Vcc is >getting creamed. > >Because the explanation of the failure will be a bit too long to >adequately put into a message, the best place to get further details of >the failure type is to go to >http://web.gmtcom.com/~k3jsch/sixsigma/access.htm > >The problem occurs regardless of the following: > >1) Whether it has been in service for only a few hours or at least 8000 >hours. >2) Whether it is operating at a temperature of 55 degrees C or slightly >past 100 degrees C > >3) Whether it is operating in a 60Hz domestic unit or a 50Hz overseas unit. > >4) date code of IC > >Any suggestions, etc., are welcome. What I am trying to do is set up an >experiment at our production facility to inject noise, etc. into the >design as it is running. Factors I am looking at are : > >1) Levels of noise voltage to inject into the cable between the >microprocessor unit that this IC resides on and the gate firing board of >the SCR bridge that is directly mounted over the 3-phase, 6SCR bridge >that ultimately runs either the left track or right track motor. > >2) Brand name of chip. > >3) Whether or not modification to the shielded cable and/or its placement >in the circuit would make a difference. > > >Regards, > >Kelly Schauf >.....k3jschKILLspam.....mail.gmtcom.com >

(Hmmm, Russell, your post just got here. Time lag?) One minor suggestion/question for the list, would it be better to put the Zener or catch Schottky's on the output pin of the 54C14J, or at the outside (past the output resistor), at the cable connection? That's where I'd be tempted to put them, then any HV noise put into the the cable is solved right there - you cannot get up *to* even 30 volts on the other end of that 54C14J output resistor. Also, Kelly, one question: Your web page doesn't give relative quantities made of the various units, some sort of "The failure rate IS (or IS NOT) fairly uniform across models made" information might be good to provide people in some cases, you could provide that without disclosing sales data directly (in case that's sensitive information.) In this case, I suspect it's that cable/54C14J connection. (I learned something good that I keep losing, from Dave VanHorn, reminder of Triboelectricity. Good to be reminded of that. I need to go write a note to optoisolate one thing <G>) Mark Russell McMahon wrote: {Quote hidden}> > Kelly, > > I couldn't guarantee that I know what is happening here but I suspect > that I could STOP it happening for very little effort, BIMBW * > > i Doesn't seem likely but - High dv/dt on FET switching can > reflect into gate. Oscillation can occur on gate under "correct" > conditions. The voltage here would need to be high enough to drive IC > into latch up via 1K drive resistor. > > ii I don't know how long the cable is but significant energy *may* > be able to be stored in the cable. What is it's capacitance. You are > driving it directly (NO resistor from gate output. Is this within > spec? (ie - depending on connection method there may effectively be > capacitance to ground bypassing the FET gate resistor). > > iii Regardless of mechanism, the *cause* would *seem* to be energy > from the cable getting into the gate's output. > Faced with this problem I would. > > a) Strongly consider placing the 1K series resistor at the gate > output rather than at the FET (isolates gate from direct capacitance) > or split the resistor to eg 470r at each end of cable. At present the > IC output is directly exposed to anything that *may* appear on the > cable. > > b) Place Schottky catch diodes to ground and supply at the IC output. > > With 1K resistor and say 30ma capability (a very small Schottky > (BAT85 etc?)) you can withstand 30 volt continuous above or below > power supply rails without sinking ANY current in the ICs substrate > diodes. > > As long as the problem is not from somewhere else (eg spikes via the > Vcc rail) these steps would be very very (very ...) likely to prevent > damage, regardless of the cause. > > With the amount of energy you have in the general vicinity I would > want to be extremely certain that my power supply was not the source > of problems BUT I would pick the above as most probable. > > regards > > Russell McMahon > > * BIMBW = But, I May Be Wrong :-) > > PS - if I had to "test" this situation rather than fixing it I would > consider applying, to the IC output, capacitors of the value > equivalent to that of the cable, charged to increasingly greater > voltage until I replicated the fault (or couldn't). > > From: Kelly Schauf <EraseMEk3jschspam_OUTTakeThisOuTMAIL.GMTCOM.COM> > <snipped>

Hope I'm not repeating what someone else has already suggested but- How well bonded are the earths between the control board with the failing IC and the FET driving the pulse transformer. Any earth potential rise at the FET end could cause the output of the IC to exceed its max (or min) limits causing latchup and failure. If the cable is longish or there is any sort of inductance or signal coupling in the earth loop then this could create a damaging situation - particularly with the power levels and risetimes likely to be occuring in the environment. Note that the earth return inductance will not need to be very high to generate substantial voltage from the high di/dt values you're likely to find in this area. (V=L * di/dt) If L=1uH I=500A in 1uS then V=500V & more significantly perhaps E=1/2 * L * I^2 = 125mJ = more than enough to do damage! My reccommendation would be to move the driver IC down to the same PCB as the FET and drive it from a differential signal, rather than the single ended one shown. Standard RS485 driver / receiver ICs might be an option. The differential receiver would need to have a good common mode performance but this is not too hard to arrange with even just a "long tail pair" of transistors and a constant current source. Alternatively, isolate the FET driver from the logic using an opto isolator and keep the earths and supplies separate. The other replies with respect to bypass diodes and series resistors should also help but I'd look carefully at the return path impedance and coupling. Richard (rprosserspam_OUTswichtec.co.nz) From: Kelly Schauf <@spam@k3jschKILLspamMAIL.GMTCOM.COM> >To the forum: > >The problem I have does not involve a PIC; however, there appears to be >an abundance of knowledge in this forum regarding electronics based on >some of the messages I've read so far. > >Is there anyone out there in the forum that has experience with Design of >Experiments (Taguchi-type) in addition to their electronics >troubleshooting skills. I have been seeing repetitive problems with a >MM54C14J Hex Inverter (an old ceramic-style part) where the Vcc is >getting creamed. > >Because the explanation of the failure will be a bit too long to >adequately put into a message, the best place to get further details of >the failure type is to go to >http://web.gmtcom.com/~k3jsch/sixsigma/access.htm > >The problem occurs regardless of the following: > >1) Whether it has been in service for only a few hours or at least 8000 >hours. >2) Whether it is operating at a temperature of 55 degrees C or slightly >past 100 degrees C > >3) Whether it is operating in a 60Hz domestic unit or a 50Hz overseas unit. > >4) date code of IC > >Any suggestions, etc., are welcome. What I am trying to do is set up an >experiment at our production facility to inject noise, etc. into the >design as it is running. Factors I am looking at are : > >1) Levels of noise voltage to inject into the cable between the >microprocessor unit that this IC resides on and the gate firing board of >the SCR bridge that is directly mounted over the 3-phase, 6SCR bridge >that ultimately runs either the left track or right track motor. > >2) Brand name of chip. > >3) Whether or not modification to the shielded cable and/or its placement >in the circuit would make a difference. > > >Regards, > >Kelly Schauf >KILLspamk3jschKILLspammail.gmtcom.com >

> One minor suggestion/question for the list, would it be better to put >the Zener or catch Schottky's on the output pin of the 54C14J, or at the >outside (past the output resistor), at the cable connection? That's >where I'd be tempted to put them, then any HV noise put into the the >cable is solved right there - you cannot get up *to* even 30 volts on >the other end of that 54C14J output resistor. With the diodes at the IC output AND a resistor on the PCB before the cable the diode sinks current when the "spike" is dissipated across the resistor. The resistor takes the brunt of any voltage spike. Diodes should be Schottky to keep drop below 1 silicon junction drop and also are "fast". Diode power is low as it is 1 Schottky junction drop x the available fault current. If a zener is placed outside an on board resistor it has to dissipate all the energy itself. Gruntier diode needed and runs the risk of being driven quite high if spike is fast. > Also, Kelly, one question: Your web page doesn't give relative >quantities made of the various units, some sort of "The failure rate IS >(or IS NOT) fairly uniform across models made" information might be good >to provide people in some cases, you could provide that without >disclosing sales data directly (in case that's sensitive information.) >In this case, I suspect it's that cable/54C14J connection. He implied that there were probably about 1000 all up - he has not been given the exact figure I think. regards Russell McMahon

Russell McMahon wrote: {Quote hidden}> > > One minor suggestion/question for the list, would it be better to > put > >the Zener or catch Schottky's on the output pin of the 54C14J, or at > the > >outside (past the output resistor), at the cable connection? That's > >where I'd be tempted to put them, then any HV noise put into the the > >cable is solved right there - you cannot get up *to* even 30 volts > on > >the other end of that 54C14J output resistor. > > With the diodes at the IC output AND a resistor on the PCB before the > cable the diode sinks current when the "spike" is dissipated across > the resistor. The resistor takes the brunt of any voltage spike. > Diodes should be Schottky to keep drop below 1 silicon junction drop > and also are "fast". Diode power is low as it is 1 Schottky junction > drop x the available fault current. If a zener is placed outside an > on board resistor it has to dissipate all the energy itself. Gruntier > diode needed and runs the risk of being driven quite high if spike is > fast. Sure, good explanation; I'd thought that even with a FAST dv/dt the zener/Transzorb on the cable end would catch the spike sooner than if the zener was at the gate end, due to higher voltage across it (Transzorb is what I'd use - I think of them as the "Same device", transzorb's a "Rough hewn" Zener from a black box level, really. And sure takes higher wattage!) With my headache yesterday, I wasn't getting how the Zener can be driven higher with a fast spike on the outside of the resistor, finally with less headache today it dawns on me: Duh!: That gate's an ACTIVE output, Mark. Not tri-stated, which I do a lot. Trying to drive that active output 'past the stops', should do a GOOD job of pulse width stretching, your "the resistor takes the brunt" finally clued me in! <G> For production stuff, the gate end is the way to go - I do custom few-at-a-time stuff, some of my stuff HAS to have huge "herking" protection to prevent any failures if I can; I'd guess Schottky's on the gate end with a Transzorb ("cheap P6KE6.8", 600W, 88 cents or so, beats a service call any day!) on probably EACH end of the cable, and perhaps a Zener at the FET's gate as a backup? I'd rather have some cheap redundancy, than drive 300 miles to replace a $1 part, or trust "Joe Ten-Thumbs" there who's more likely to use Acid core solder than not, to solder in replacement parts, and definitely more likely to melt the parts soldered in & still end up with cold solder joints, than not. Motivates me to put everything in connectorized modules, it's hard to plug the Red DB-25M into a Green DB-09F socket! He'll probably use a hammer <G> ("Joe" does what he does really well, it's nice and technical and incomprehensible to me - he just doesn't "Grok" electronics at all; I wouldn't ask my Doctor to solder for me, he doesn't ask me to diagnose medical problems for him, same with "Joe" here, the world's a safer place if he doesn't solder...) Different markets, different techniques! Consumer appliances are different than industrial embedded stuff <G> > > Also, Kelly, one question: Your web page doesn't give relative > >quantities made of the various units, some sort of "The failure rate > <snipped> > He implied that there were probably about 1000 all up - he has not > been given the exact figure I think. > > regards > > Russell McMahon Missed that number - Thanks, Russell! Still would be good to know (in other circumstances) if failure rate was "uniformish" or not, a minor design change is the solution here, though, though the more I think about it the more uncomfortable I am with my ASSUMPTION that ground loops weren't the culprit; I'd be tempted (in addition to definitely a series resistor at that gate!) to consider going Optoisolated at the end of the driven cable, and have the optoisolator drive that FET, just because any ground loop that changes the Voltage at the ground end of the FET could change whether the FET's driven hard or not, potentially. I'm probably being paranoid, though; I definitely don't think that's the problem we're against here <G> Mark

Mark Willis wrote: > Motivates me to put everything in connectorized modules, it's hard to > plug the Red DB-25M into a Green DB-09F socket! I still can't figure out what these "DB-09F" and similar fittings are that people say they use! DB-25s I understand, and DE-9s and DE-15s and similar, but DB-09 must be an odd species indeed. Does this refer to a partially-populated shell with insertable crimp pins? Do you blank out the unused positions? You can certainly produce "coded" connectors, but they're not particularly goon-proof! > I wouldn't ask my Doctor to solder for me, he doesn't ask me to > diagnose medical problems for him, Well, not being *your* doctor, I suppose I couldn't object to that but otherwise I probably would! ;-) > same with "Joe" here, the world's a safer place if he doesn't > solder... > Different markets, different techniques! Consumer appliances are > different than industrial embedded stuff <G> I don't follow. You reference to "acid-core" solder denotes a plumber. I should certainly not expect a *plumber* to be working on appliances (except I have seen one or two replace a heater element). An appliance serviceman or electrician however - have you *really* seen one using plumber's solder? That *is* more than a bit slanderous! -- Cheers, Paul B.