Announcement

Collapse
No announcement yet.

Cycling rarely used electronics question

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • wombat2go
    replied
    Originally posted by PStechPaul View Post
    From the Mitsubishi document linked by wombat2go above:


    I didn't see a graph of Vce(sat) vs temperature, but there is a large effect of gate voltage. There is a chart showing Vce(sat) versus current, and at 125C it is a straight line, indicating mostly resistance, with a constant 1 volt minimum drop. The greatest difference is about 0.5 volts at 160 amps. The curves cross at 440amps, with Vce(sat)=2.8 volts, and at 800 amps it is 3.75 volts at 125C and 4.25 volts at 25C. But with 3400 watts dissipation, it won't stay 25C very long!
    [/SIZE]
    Hi PSTech
    From my experience with water cooling, , industrial silicon semiconductors paralleled will be sized for continuous duty ( 24/7)at a safety factor above the wet bulb temperature.
    So for duty at ambients on the planet, you could assume 38 ~ 42 Celsius on the heatsink below the case, subject to better information.
    In my experience the junction temperature ( silicon generally) should be around 80 to 95 C for active devices switching.
    Regards

    Leave a comment:


  • old mart
    replied
    I get the forum by googling it, www.badcaps.net

    Leave a comment:


  • ahidley
    replied
    Old mart. I used the search engine built into this forum and searched for Badcaps and did not find a forum with that word in the title. Perhaps you could post the address??

    Leave a comment:


  • old mart
    replied
    If anyone's interested, there is a forum called "badcaps forum".

    Leave a comment:


  • dian
    replied
    my old haidenhain dro started giving up the ghost a few years back. well, after reading about the caps somewhere i left it on for a week. has been working ever since. a vfd i bought about 20 years ago died after a few years. i was the caps. now, about 15 years later it still works, maybe because i switch it on from time to time when not in use?

    Leave a comment:


  • wombat2go
    replied
    Forest,
    Thanks, that is a very good resource for those of with interest in old gear.

    I read in it that the little pcb mount electros are better sealed than the upright can power filter types.

    In 2016 I restored this old blade chassis amplifier from a Magnavox stereo
    https://app.box.com/s/f8wk5i9uvpfw6r4037oaqd9yv92t4srw

    I think it from late 1950's the transistors are germanium and the power transistors, on pressed steel
    heatsink, were by Bendix labelled "68P1K" and "6624N" but could find no internet ref.

    In this amp, all the small electros were open circuit and the photo shows that I replaced them with Nichicon.
    But the DC ripple was good at full power as was the square wave test on the amplifier,
    so I left the nice looking power supply filter capacitor in place ! (at upper right).

    An interesting item about this amplifier was that instead of a 'loudness button" that we see in later stereos,
    this one has a passive circuit that varies the frequency response toward more treble boost as the volume pot is
    turned down.
    Also the dual treble pot had (I think was intentionally built with) different resistance values for each channel.

    Here is the frequency response test showing different levels between the channels.
    https://app.box.com/s/gwqz4xv9u87n48bmt5pjlkstu2i7tq86
    I have since replaced the dual pots with new ones with equal values. ( Those old dual pots are expensive)

    Leave a comment:


  • Forestgnome
    replied
    Found some solid information on electrolytic cap shelf life. Good info on pages 14-17.
    http://citeseerx.ist.psu.edu/viewdoc...=rep1&type=pdf

    Leave a comment:


  • J Tiers
    replied
    Originally posted by PStechPaul View Post
    From the Mitsubishi document linked by wombat2go above:


    I didn't see a graph of Vce(sat) vs temperature, but there is a large effect of gate voltage. There is a chart showing Vce(sat) versus current, and at 125C it is a straight line, indicating mostly resistance, with a constant 1 volt minimum drop. The greatest difference is about 0.5 volts at 160 amps. The curves cross at 440amps, with Vce(sat)=2.8 volts, and at 800 amps it is 3.75 volts at 125C and 4.25 volts at 25C. But with 3400 watts dissipation, it won't stay 25C very long!
    [/SIZE]
    pretty much any negative variation is less than good.... it depends on how well you can keep them at the same temperature, and also on how MUCH the variation is.

    No, they will not stay at 25C that long.... the issue is what happens when they are heating up. If one heats a little more, due to any sort of variation in parameter, it can hog current and have more dissipation than any of the others. depending on how much that is, it may decrease reliability, or cause that part to exceed limits when others are comfortably below the limit. The heating is inherently unstable, because more heating means better conductivity, and more current flow, which heats the part more...... it is a positive feedback that can rapidly go to the worst case.

    If it is a weak negative variation, it may not be so bad, the variation may not end up changing the current a lot. The "worst case" may be tolerable. That depends on the more constant resistive portion, which can provide an effect similar to putting in "ballast" resistors as was always done with paralleled standard bipolar parts in linear operation.

    Another way to help avoid problems is to mount all the paralleled parts as directly as possible to a "heat plate", close together, without insulation. You can do that because they are all acting as one big composite part. Then you insulate that plate electrically as a whole from the heatsink.

    What that does is to put the whole set on a relatively constant temperature base. The thermal resistance from each to the plate is as low as you can get it, and any heating from whatever part will heat the plate up. That provides thermal feedback to the other parts, so that they are closer to the same temperature as the hottest part.

    If they are each on their own electrical insulation, then they are much less "connected" thermally, and can have considerably different temperatures. That allows any negative tempco to more directly affect the imbalance between parts.
    Last edited by J Tiers; 03-30-2018, 09:07 PM.

    Leave a comment:


  • PStechPaul
    replied
    From the Mitsubishi document linked by wombat2go above:
    NOTE:

    It may be observed that Mitsubishi IGBTs have a negative temperature coefficient of saturation voltage over a wide range of collector currents. This is not a deterrent to parallel operation and, in fact, is an advantage as it yields lower conduction loss at high junction temperature. The homogeneous process characteristics of H-Series IGBTs produce VCE(sat) characteristics that track as a function of current and temperature such that, once a VCE(sat) rank is chosen, the parallel devices will share within the given derating factor.


    I didn't see a graph of Vce(sat) vs temperature, but there is a large effect of gate voltage. There is a chart showing Vce(sat) versus current, and at 125C it is a straight line, indicating mostly resistance, with a constant 1 volt minimum drop. The greatest difference is about 0.5 volts at 160 amps. The curves cross at 440amps, with Vce(sat)=2.8 volts, and at 800 amps it is 3.75 volts at 125C and 4.25 volts at 25C. But with 3400 watts dissipation, it won't stay 25C very long!

    Leave a comment:


  • J Tiers
    replied
    Power supply high voltage capacitors are much better than they used to be.

    Even so, over time, the dielectric layer degrades, and the capacitors lose voltage withstanding capability. They "leak" more and heat up as a result. If the leakage is enough, they may overheat, short, or just boil off their liquid. In the process they may literally explode.

    By periodically turning them on, and leaving on for a while, this degradation is reversed, the insulation film is built up by an electro-chemical process to a level that will withstand the applied voltage. But the process involves electrical current flow, and if the applied voltage is too high for the insulation film, a lot of current may flow.

    For devices that have been stored a long time, it is prudent to hold them at 25%, then 50%, 75% and finally full normal line voltage for a while to make sure not to exceed the possibly reduced voltage withstanding capability before it is restored by having voltage applied.

    Leave a comment:


  • darryl
    replied
    Low ESR caps are mandatory where switching speeds are high. This doesn't mean they will survive for decades though, and heat is not the only factor in their early demise. Consider that the ripple currents flowing in and out of the power supply caps are happening with much greater frequency, the caps are usually on the small side, and there will be a heat build-up- especially when the cap begins to fail. In any event, there's going to be a surge current at initial power-up, before any load is placed on the machine. It's going to happen every time you power up. If you want to power up the device from time to time just to keep the circuitry from dying through lack of use, it would be good to know how much steady state current the thing should be drawing at no load so you can do a quick test each time you power up. We've gone through this before- an inline meter and a safety resistor (light bulb in series, with a bypass switch) would be handy to have. Use it for testing all manner of plug-in electronics, and create a chart to show what the 'idle current' is for each device you'd like to maintain.

    Leave a comment:


  • BCRider
    replied
    Originally posted by Forestgnome View Post
    I've been repairing electronics for ages, and have never heard of or read an analysis of cycling to increase the life of capacitors. Heat is the most common cause of capacitor failure, which is why caps loaded next to CPU's on motherboards is such a common failure area. The other highest cause of failure is just poor quality in the construction of the cap. Following that, cutting corners on design parameters. Unfortunately you have most of that going for you in a Chinese designed and manufactured welder.
    I'm in the same boat. Stuff at home and at work that was used regularly or left on the shelf for a decade all seemed to work or fail with equal frequency.

    I've seen computers were the power caps on the motherboard went bad and it seemed to occur MORE often on computers left on than those used very infrequently and properly shut down for days at a time.

    Leave a comment:


  • J Tiers
    replied
    Originally posted by MattiJ View Post
    All good and sunshine until someone tries to parallel several mosfets for linear operation!
    I've done that in a number of designs. Worked well... when we got graded parts from the manufacturer, and used ballast resistors. Up to 1600W amplifiers. Did 2500W with 2 IGBT paralleled, 4 total, switching, half bridge audio output

    Leave a comment:


  • MattiJ
    replied
    Originally posted by wombat2go View Post

    Anyway it has been a long time since power electronic designers had to worry about the negative tempco by standing the switches on emitter resistors etc
    All good and sunshine until someone tries to parallel several mosfets for linear operation!

    Leave a comment:


  • J Tiers
    replied
    I just took a random sampling through Digikey.....

    I found that they are mixed in currently available parts. Some are negative, some are weakly positive, some are a bit more positive.

    The issue is that there are a number of parameters that affect on current. Matching can usually be accomplished only over one or two, unless they are linked. Not all are. So there will always be mismatches. A god strong positive tempco of resistance helps maintain good current sharing despite variations.

    I'd say that "a long time since designers had to worry about it" may be too strong.

    In any case, matching among discrete parts is not usually going to be nearly as good as matching between chips that were neighbors on the wafer. When you buy a high current device, the many internal chips are matched better than you will ever get consistently with discrete parts.

    When an equipment manufacturer chooses to parallel discrete parts, they are not getting as good a match, and not getting as good a reliability, as one "part" that may have a number of internal chips.

    You start out with the thermal linking on the baseplate, vs thermal linking through the individual part baseplates into the heatsink. Then you add in that the parts are selected to be wfer neighbors, and the result is very good. far better than you get any other way

    Leave a comment:

Working...
X