[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: 48kW DRSSTC and RELIABILITY



Original poster: Terry Fritz <vardin@xxxxxxxxxxxxxxxxxxxxxxx>

Hi Dan,

There are lots of systems out there with 10,000s of parts. If any "one" fails, the systems goes down. I am typing on one right now ;-))

In general, the probability of a single part failing is minor compared to true design flaws, unintended operation, accidents, external causes.... Many times I have "wished" that a system's failures were just a matter of random component failures... But most real system failures don't follow that simple of logic.

Military deratings and calculated MTBFs tend to be low by a factor of 5. Often they do not take "modern" components or non-military environments into account. Thus, if one goes "just by the numbers" many perfectly happy commercial products will look terrible to some military specs. To make a long story short, stay within the rated specs, and the parts won't die for no good reason... There are millions of complex SMPS systems out there that do just fine... Our little DRSSTCs are basically the same from a plain Jane electronics standpoint.

However, we have two big problems.

Operating environment. - I don't mean aircraft or sea launched missile things... I mean Joe in has back yard in the rain, Sally at the Halloween party, Gus at the school show, ....... Goodness only knows what odd things they will do wrong to cause a failure. Running without the secondary, running without ground, running off one or "three" phases when they should be running off two... Someone really WILL try to cook food on the primary... Those are the things that will kill a commercial DRSSTC. Having a chip randomly fail is just not going to be a problem. It's all those goofballs out there that are going to run it all wrong...

The second thing is the high voltage and pure "newness" of such systems. The basic electronics is plain, but these power supplies have big sparks and big voltages that are "abnormal" and really not well understood. The effects of the sparks on the electronics and all the sneaky ways that they can do damage is pretty unknown and the learning curve could be pretty messy. Suppose an arc to the case induces a -2 volt spike somewhere that blows an IC that blows.... "Finding" that hidden problem could be a nightmare!!!

So, IMHO, trying to derate the components to improve reliability might be a waste of time other than checking to be sure they are not run outside their specs (often MTBF studies are really "just" to check for that only). I would worry more about FMEA (Failure Mode Cause/Effects Analysis) where you brain storm everything that can go wrong or be done wrong to an otherwise happy system to make it fail. Then work to protect against those situations.

Second, LOTs of good hard testing!!! The usual verification tests to make sure it runs in hot and cold weather, low and high line voltage, vibration, etc. to catch true design flaws.... And then some field testing where "normal user" types get to try and break it by doing normal and stupid things to it... They can't run it in driving rain or anything obviously stupid, but what happens if the control cable gets jerked out will it is running (that will happen a lot) sort of stuff.

Unfortunately, getting all that done is a pretty tough task!!! Then there might be EMC, CSA, FCC and other compliance issues... A lot of that can be worked around since DRSSTCs are pretty specialized "not really" commercial and "sort of experimental" things... But even then, you really need to check to be sure that they will not attract any "special attention" from the rules folks...

Then there is sales, marketing, and all that...  But that is not "special".

At least that is MHO on it all... If you want more humble opinions, I am free ;-)))

Cheers,

        Terry


At 07:20 AM 12/4/2005, you wrote:
Reliability is a very complex issue.  In the design and test of our
commercial DRSSTC system, MiniBrute system, we are finding is
exceptionally difficult to get a unit to have enough performane vs. cost
(bang for the buck), yet at the same time be reliable enough to sell as
a commercial unit without it becoming a warranty nightmare.

We had originally designed our system to follow standard NAVMAT
deratings on all components,  and found that this does provide a good
factor of reliability, but at the same time, reduces performance by more
than 50%.

The trade-off is do we sell a coil that will outperform any other coil
in the world for the size, power input, but will likely be plagued by
failures / reliability problems, or design one very conservatively, that
doesn't put out much arc (would be considered non-spectacular by most
amateurs)

We have almost come to the point to abandon the entire project
altogether as a good performance vs. cost vs. reliability is difficult
to obtain.

The biggest factor in the reliability front of the DRSSTC which makes it
much different than other systems, is that there are MANY sources of
single point failures which degrades reliaiblity considerably.  With a
conventional tesla coil, you might only considered the capacitor as the
single-point failure mechanism or the weakpoint in such a system, and
such a system can easily be made bombproof.  Not so with a DRSSTC.  And
this is ESPECIALLY true for high power DRSSTCs which have many complex
subcircuits including AC/DC frontends, some with active PFC, high power
gate drive circuits, complex fault detection, etc...

Dan



>Anyway, commercial product (not 833C
>compliant) is allowed to fail and will someday. Electron tunneling will

>put small holes in silicon no mater what you do (run at spec), and
>someday a path to metal will happen.

If you ever run your DRSSTC long enough for this to happen, I'll eat
my hat, and any other items of clothing I might have on at the time.
Assuming I haven't died of old age that is. I'm no reliability
expert, but as far as I understand, if the device is run within its
ratings, systematic failures like this should not happen for >100,000
hours.

There isn't much work done on reliability of IGBTs that are pushed
outside their ratings (well not in the public domain anyway- maybe
you know some "trade secrets" that I don't) but I'm willing to bet
that that 100,000 hour figure drops like a stone as soon as you go
outside the datasheet peak current. I don't know why it would, but I
just have that feeling. After all, if the devices were capable of
taking higher peak currents with good reliability, then the
manufacturer would print that bigger number on his spec sheet to make
the device look better.

So I reckon it all comes down to the same tradeoff as Terry Fritz
explored with MMCs. He discovered that you can run the capacitors
outside of their AC voltage rating. It shortens their life
drastically, often to less than 100 hours. But that is still plenty
enough for hobby Tesla coil use. We discovered that you can run IGBTs
outside their peak current ratings, but the life implications are
nothing more than a "Bad feeling" just now. That is in opposition to
the case with MMC caps, where there is even an equation to tell you
how long they will last at a given overvoltage. If someone could
figure out a similar equation for overcurrented IGBTs, it would be
worth a NoPig Prize ;-)

Steve Conner
http://www.scopeboy.com/