Power supply for water electrolysis keeps blowing up. Please help.

Discussion in 'The Projects Forum' started by H2OWiz, Dec 12, 2007.

  1. H2OWiz

    Thread Starter New Member

    Dec 12, 2007
    8
    0
    I have been trying to get the following circuit to work but I keep burning out the MOSFET. I am using 2, 555 chips for frequency and Pulse width control which I am trying to use to drive the MOSFET as an on off switch. I just added the LM317 voltage regulator to keep the voltage on the gate of the MOSFET at 3 to 6 volts so I can control the amps to the electrolyzer. I am trying to get 15 to 20 amps out put. Now when I hit about 10 amps the MOSFET gate stops responding and the drain to source is full on which quickly kills the MOSFET. I had better run times before I added the LM317, about 1 minute, but still had the same problems with the MOSFET. One additional problem that I have now since adding the LM317 is the 555 that controls the pulse width is also burning out. How can such a simple circuit be so difficult. If any one knows what I am doing wrong please let me know, and thanks.
     
  2. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    Do you have an oscilloscope to check the waveform at the gate?
    You should be aiming for about 12volts after the lm317 to give the mosfet proper gate voltage so it can fully open.
    At what frequency are you working? (compute it if you don´t know).
    The 555 should do 200mA on the output so it´s hard too guess why it keeps burning out.
     
  3. H2OWiz

    Thread Starter New Member

    Dec 12, 2007
    8
    0
    I do have an oscilloscope but do not know how to use it. The problem with 12 volts on thye gate is it will allow 70 to 90 amps through the MOSFET depending on tempeture which is to many amps. See the attached gate voltage to drain current chart.
     
  4. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Here's a goofy idea ;)

    Use one MOSFET to charge up an electrolytic capacitor. When it's nearly charged, turn the MOSFET off.

    Use a 2nd MOSFET to drain the charge from the electrolytic cap through your electrolyzer. When the cap is nearly drained, turn off the 2nd MOSFET.

    Sort of a "bucket brigade". The size of the cap vs the resistance of the MOSFETs and associated wiring determines the RC time for the charge/discharge.

    I haven't gone through your 555 timer to see how fast you're going - but MOSFET gates have a considerable amount of capacitance, and it takes a while to charge them up and drain them down. If you're beyond, say, 20KHz or so, you'll have problems.

    MOSFETs do not like to run in a "partially conducting" state; it's either ON or OFF for them.

    You COULD run a bunch of them in parallel, if your load device can stand the current. MOSFETs are much easier to run in parallel than other semiconductor devices, as they have a positive temperature coefficient; as they heat up, their resistance increases, so their current goes down and neighboring MOSFET devices will conduct more. Of course, you need the capacity to drive the multitudes of MOSFET gates.
     
  5. Out To Lunch

    New Member

    Dec 7, 2007
    3
    0
    you don't understand MOSFETs or the circuit you are trying to build.

    the control circuitry basically creates an adjustable pwm (i am assuming - i never use the 555, but it really cannot be anything else). The pwm signal then turns the MOSFET on and off really fast. the MOSFET MUST be fully enhanced or you will not saturate it. Once saturated, the Drain-Source connection can be modeled as a small resistor - in the case of the IRFZ34N, it is 40milliohms.

    when the MOSFET is turned ON, the INSTANTANEOUS current through the MOSFET - and through the electrolyzer - will be based on the impedance of the MOSFET (40milliohms) and the impedance of the electrolyzer (i say impedance because there is likely to be some inductance in the electrolyzer).

    The AVERAGE current through the MOSFET - and through the electrolyzer - will be the instantaneous current when the FET is on multiplied by the ON time of the FET (commanly called Duty Cycle) divided by the period.

    I suggest that you get a much better understanding of how the circuit is supposed to operate and THEN you will know what to expect when you power it on.
     
  6. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    I'm not a MOSFET expert, but...

    Aren't you pulsing the mosfet, and controlling the pulse width as the means of controlling the average current? You still must quickly turn the mosfet FULLY on, and then quickly fully off, for each pulse, or else you can overheat the mosfet.

    If you don't get the MOSFET turned on fully, then Rds could be much larger than the Rds(on) spec, and, with a large current through it, the MOSFET might get way too hot, killing it. If you do it right, then even with 20 Amps DC through it, the Rds(on) spec of 0.04 Ohms would mean that the mosfet would only be dissipating about 0.8 Watt, and should barely get warm.

    With pulsing, you also want to make sure that the mosfet spends as LITTLE time as possible in the area BETWEEN fully on and fully off, since that condition will heat it the most. So, you probably want to bang the gate pretty hard and fast. For your IRFZ34N, the VGS(th) spec's MAX is 4v, and the VGS absolute max spec is 20V. So you'll probably want to hit it with the full 12 V 555 output, and use your pulse-width control to control the average current, instead of trying to limit the maximum current of each pulse.

    You might also be getting some high-frequency ringing (or even oscillation), somewhere, which could contribute more than you might think, to heating the mosfet. A good oscilloscope would be a big help. [Edit: I see that you do have a scope. Assuming you have a suitable probe, with a (very) short ground lead, see if you can look at the Gate-to-Source voltage. You should probably try to connect the scope while the circuit is still unpowered, as much as possible, since a probe slip that causes a short would not be a good thing. If you say what make and model your scope is, maybe you could get more-detailed help, with using it.]

    By the way: Can you detect any spikes in the power rails for the 555 ICs? (i.e. Vcc pin to local ground; probably difficult without a scope) At any rate, you probably want to bypass each of the 555 Vcc pins with 0.1 uF or so to ground, and maybe also with a larger cap, if there is much wire between the battery and the 555s.

    - Tom Gootee

    http://www.fullnet.com/~tomg/index.html
     
  7. H2OWiz

    Thread Starter New Member

    Dec 12, 2007
    8
    0
    On my first circuit with out the LM317 I tried to control the voltage to the MOSFET gate by controling the pulse width. This only worked for a minute then I lost the MOSFET. Someone told me that I could not do this because even though the the duty cycle was reduced I was still applying 12 volts to the MOSFET and this resulted in to much current through the MOSFET. See the gate voltage to current chart above. This is why I reduced the gate voltage to about 6 volts with the LM317. I am interested in trying the bypass caps. Do I simply connect one side to the power 555 vcc line and one side to ground? Also on the 317 adjust line where would I attach to this line. The frequencies I am using are all in the audio range as I have a small computer speaker hooked in so I can adjust by sound.
     
  8. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    Could you post a sketch or photo of the voltage at the gate of the transistor, with marked times and voltage?

    Limiting the supply voltage for the gate will only make it worse. Read again and carefully what gootee wrote and do what he says. Check the voltage at the gate, and the supply voltage on each 555 (at the supply pin!) with the oscilloscope,
    then post your results.
     
  9. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    And if you don´t know how to use an oscilloscope:
    1) set the input selector to ground and adjust the line to the middle
    2) connect the -lead to the ground and the +lead where you wan´t to measure
    3) set the input selector to DC
    4) set time base to 2ms and voltage to 5V
    5) power the device and watch the waveform, adjust timebase and voltage to view like 2 or 3 pulses at the highest possible size
     
  10. H2OWiz

    Thread Starter New Member

    Dec 12, 2007
    8
    0
    I will try to get the photos you need. Sounds like what i need to do is remove the LM317 and go back to supplying 12 volts to the gate of the MOSFET then try using the pulse width adjustments to lower the current to the working range of 15 to 20 amps.
     
  11. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    Yes that´s exactly what you should do.
    And by the way, you should have some heatsink attached to the mosfet, then you will have smaller chance of blowing it. The bigger the better
     
  12. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    I'm in a hurry. So double-check my math etc.

    It looks like it "should" work with 6-volt pulses, too, since VGS(th) MAX is 4 volts, for that mosfet.

    But it also looks like your LM317 was set up for a nominal voltage of about 3.8 volts, if I did the calculation correctly. You need to measure its output with a DC voltmeter, at least.

    BUT, even if the NE555's Vcc was actually made to be 6 volts, according to its datasheet it looks like its output would then probably only be swinging to anywhere from 3.75 volts to about 4.25 volts. So that might have been your problem, right there.

    It gets worse: Actually, even with your LM317's 1K adjustment pot set for max output voltage (i.e. 1K), it looks like you would barely have been getting just over 5V out of the 317, according to the output-voltage equation in the LM317 datasheet. So that would have put your NE555's output amplitude way too low for the mosfet to turn on fully.

    You might want to try it with at least 8 volts or so at the output of the LM317. You could add 1k in series with the LM317's adjustment pot, so that the adjustment range would be from about 6.5v to about 11.5v. (Of course, somewhere above about 10v (depends on the load current; see the dropout voltage vs load current plot, in the LM317 datasheet), the 317 would no longer stay in regulation.)

    You can measure the gate-driving NE555's output amplitude (and should also measure the mosfet's gate-to-source voltage's amplitude), with your oscilloscope. Note: When grounding the probe, make sure that you always use the CLOSEST relevant ground (or) reference point, and that you use a very short ground lead. An even shorter ground "clip" probe attachment would be better, since very fast edge times are involved. Otherwise, your probe will probably lie to you. Note that your scope should have a square-wave "probe calibration" output. Always use that, first, and turn the probe's adjustment trimmer to get the squarest square-wave display.

    Bypassing: Yes. Attach the 0.1 uF caps as close as possible to, or better yet right at, each Vcc pin, and then to the nearest matching or related ground. If there is more than several inches of wire or pcb trace length before the 555s' Vcc pins, also add a larger cap in parallel with each 0.1 uF cap. Typically, 10 uF is a reasonable guess. But more or less is OK (hint: use your scope). Adding such a cap probably can't hurt, in any case. If you want to "go all the way", you could also try putting a small series resistor in the Vcc line, just upstream from each Vcc pin. You could start with something like 33 Ohms, and check the Vcc pins with your scope. (Actually, with your fixed frequency, you could probably do even better, if needed, by also using small chokes (i.e. inductors).)

    For the LM317, you should probably add a small cap from the ADJ pin to gnd, too, to greatly-improve the output noise and ripple, and maybe also try increasing the output cap to a max of about 100 uF. I would also add a 0.1 uF cap in parallel with the LM317's input cap. The datasheet probably also shows how to add the two standard regulator protection diodes around the LM317, which would probably be a good idea. If not, look at some other three-terminal regulators' datasheets.

    - Tom Gootee

    http://www.fullnet.com/~tomg/index.html
     
  13. JoeJester

    AAC Fanatic!

    Apr 26, 2005
    3,373
    1,157
    If you want to limit the current through the electrolyzer to 20 amps, then insert a 0.6 ohm resistor between the electrolyzer and ammeter.

    Then you can put your circuits back to 12 volts knowing the max current through the MOSFET will be 20 amperes.

    Be sure to place the bypass caps on Vcc and Ground on both 555s.
     
  14. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    Wasting 240W on a resistor is a bit too much I think. Where would he find such resistor?
     
  15. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    Hi H2OWiz,

    About that "Ground":

    At the least, the conductor from the mosfet's source pin to the battery's negative terminal should have nothing else connected to it. i.e. The "ground" rail from the rest of your circuit should be a completely separate conductor, all the way to the negative battery terminal.

    At best, every single thing that connects to that ground rail should have its own separate ground conductor, all the way back to the battery's negative terminal.

    It would probably also be a good idea to run separate conductors from the LM317 output pin to each thing that is connected to that rail, in your drawing, bypassing to ground at the consumption end, wherever needed (check with your scope).

    And it would probably also be a good idea to tightly twist together each Vcc wire with its corresponding ground wire. And also, try to put some distance between conductors with highly-dynamic or large currents and everything else.

    -----

    The basic idea behind most of the suggestions above is often called a "star ground", since, topologically, it resembles a star shape. The theory is that all grounds should only connect at one point, to avoid having any ground-return conductor sharing any current with any other ground-return conductor.

    "In general", as you probably already know, having ground-return currents sharing conductors can be "a bad thing". That is one basic reason for using a "star" grounding scheme, and keeping all ground returns separate.

    "But why?", you ask. I'll try to explain, a little, because, in many circuits, it can be critically important to do it right.

    Every conductor is just another non-ideal component. Any current in any conductor will induce a "distributed" voltage, incrementally, all along that conductor, because of the conductor's parasitic resistance and inductance (and sometimes capacitance), exactly like when current flows through a regular component. (There are calculators and tables on the web that will give you the values of the parasitics, for specific PCB traces' and wires' geometries and materials, etc.)

    In the case of a ground-return current, we can assume that the total induced voltage appears back at the NON-ground end of the ground-return conductor. That means that the "ground" for that device or circuit will not be at zero volts (or, actually, not equal to whatever the voltage is, at the "ground" end of the conductor).

    And if the ground-return current is dynamic (i.e. changing), then the "ground" point from which it came will have a voltage that is changing. That is often called "ground bounce".

    If, for example, the ground point that is bouncing happens to be the ground reference for the input of an amplifier, the amplifier will see only that the difference between its input pin and its input's ground reference is changing, which means that the ground-bounce voltage will be arithmetically added to the amplifier's input signal. Not good.

    And having a ground-return conductor share the return currents from two or more different devices or circuits or sub-systems allows a ground-bounce voltage to be induced, in each place, by the ground-return currents from each of the other places. Not good.

    It should be easy to see, now, that if the ground reference point for a device or circuit shares a ground return conductor with any large, dynamic ground-return current, the result could be "less than desirable", at best.

    Note, too, that while the induced voltage from a current flowing through a resistance is proportional only to the current's amplitude (Ohm's Law, V = I x R), the induced voltage from a current flowing through an inductance is proportional to the rate-of-change of the current's amplitude (V = L * dI/dt). So, even a small current, if its amplitude changes quickly, can induce a relatively large voltage, when flowing through an inductance. The point is that dynamic return currents can make ground-bounce voltages much worse than some people might imagine.

    That's basically it. And the same sort of ideas also apply to shared supply conductors.

    If you want to include these types of effects in spice simulations, I have some downloadable LTspice circuits that already have the basic setup for it, at http://www.fullnet.com/~tomg/gooteesp.htm .

    Basically, I have replaced every ground symbol with an inductor and series resistance, which are tied to a star ground point. (You can right-click on each inductor, to set the series resistance, in LTspice. Or you could use a separate resistor.) I have used node labels to simplify/clarify the wiring, and so all of the ground-returns' parasitic components can be conveniently clustered together, near the star ground point, making it easy to bridge connections between different return paths (i.e. to share and un-share them), for easier experimentation. And it would be relatively easy to copy and paste or otherwise reproduce the star ground section on your own schematics, and then add node labels for each of your ground points, replacing your ground symbols.

    Sorry to have blathered-on, for so long, about all of that. It might not even be a big problem in your current circuit. But it's good stuff for people to be aware of, in any case.

    - Tom Gootee

    http://www.fullnet.com/~tomg/index.html
     
  16. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    I don't know if he needs or wants to limit the mosfet current with a 0.6-Ohm resistor or not.

    BUT, regarding FINDING such a high-power rated resistor, they are actually fairly common. (And if you're ever looking for large capacitors, General Electric makes some that are the size of railroad boxcars!)

    I'm glad that you brought this up, because I spent a bit of time looking at commercially-available power resistors, earlier this year.

    First, note that ready-made commercially-available high-power resistors don't have to be used. Even a relatively low-power rated resistor can dissipate _enormous_ amounts of power, without being damaged, IF it is properly heatsinked. One practical way to heatsink almost any resistor is to immerse it in a one-gallon paint can filled with oil (somebody please remind me what type of oil is best, for that (i.e. mineral oil, motor oil, or what?)). Google, for example, "Cantenna", and you'll probably find specs and DIY construction details.

    But it turns out that there are some pretty good ready-made high-power resistors available that are also not overly expensive:

    I just flipped open the latest 1,982-page free catalog from mouser.com, and immediately found the Vishay/Dale RH-series heatsink-encased 1% wirewound power resistors, which come with ratings of up to 250 Watts. The 250-Watt models don't come any lower than 1.0 Ohms, and cost $74.99 for qty 1. Ouch. But I guess he could also parallel five to ten of the 50-Watt models, which only cost $3.67 each for qty 10, and come in values from 0.01 Ohm to 100K Ohms. (Or maybe just get ONE and put it in a gallon can of oil.)

    Even cheaper would be the Vishay/Dale HL-series silicone 5% wirewound resistors, which are available with 25, 50, 100, and 225 Watt ratings. Minimum value is 1.0 Ohm. But paralleling would be necessary, anyway. The low values of the 50 Watt versions are only $3.04 each for quantity 10. And the 100 Watt ones are $6.64 each for qty 1. The 225-Watt version's low values are $11.16 each, for qty 1.

    Most wirewound resistor types are not significantly inductive, if they have very low resistance values. But, in case it would matter (and I guess it just might matter, for 20-amp pulses), there are also non-inductive power resistors, such as the 100-Watt Vishay/Dale LTO100 series 5% thick film resistors, which are in heatsinkable TO-247 cases. They're a bit expensive, at around $12.50 each, and don't have many values available. But, the similar non-inductive LTO50-series 50-Watt TO-220-cased models are only $3.49 each for qty 10.

    The Ohmite 600 Watt and 1000 Watt heatsinkable Planar models are over $100 each. And their flat (i.e. heatsinkable) aluminum-cased wirewound WFH series' 160, 230, and 330 Watt models are quite expensive, too. But their 270-series vitreous enamel 5% wirewound models include 50-Watt ones that are $7.63 qty 1, and some 100-Watt ones that are about $10 each qty 1.

    Sorry to be so long-winded, again! But I think it's good for people to get a feel for what's actually available.

    - Tom Gootee

    http://www.fullnet.com/~tomg/index.html
     
  17. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    Hmmmmmm, but that still doesn´t justify wasting 240W in heat when he can just reduce the pulse width to reduce the current. Also the inductance of the circuit will create a time constant which will lower the current.
     
  18. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    Hi kubeek,

    You are absolutely right. And I wasn't trying to justify it.

    - Tom
     
  19. recca02

    Senior Member

    Apr 2, 2007
    1,211
    0
    Hi Mr. Gootey,
    how about the ones used in transformers?
     
  20. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    Hello again, recca02. Thanks. I assume that that would work, but can't say that I know what type of oil is used in them, either.
     
Loading...