PWM Parallel Power Transistors

Discussion in 'The Projects Forum' started by damianhealey, Sep 29, 2008.

  1. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    I want to add a second power transistor to a PWM with the objective of reducing the power dissipation and hopefully heat generated. Can somebody tell me the best way to add a second transistor?

    The circuit can be found at http://alt-nrg.org/ click on the Plans & Schematics link and then My newest current sensing PWM-v2:

    Thanks in advance,

    Damian
     
  2. mik3

    Senior Member

    Feb 4, 2008
    4,846
    63
    As i can see you are using MOS so can easily parallel another one on it because MOS have a positive temperature coefficient thus the current will distribute equally on both of them. If one gets hot its resistance increases so less current pass through it and more through the other. This goes one until both have equal current flowing through them. Just take another one MOS and connect its source the the source of the existing one, its drain to the drain and its gate to the gate. Note that the turn on time will double because the input capacitance doubles. Reduce the 100ohm resistor to 50ohm to increase the turn on time. Another option is to drive them with an emitter follower to increase the turn on time.
     
  3. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    Just put the other HEXFET in parallel with the first one. Run the drive through another 100 ohm resistor.

    Your LM324 is not a particularly impressive op amp. I believe we have mentioned before that replacing the C section of the 324 with a comparator would improve operation markedly. You would probably be able to avoid that second HEXFET (it's not a power transistor, by the way).

    P.S. I'm going to chime back in after looking over the schematic a bit more thoroughly. The extra HEXFET will not help in the slightest. One of those monsters is rated at 110 amps continuous output, but only at a gate voltage of 10 volts or above. Your adapted design uses a 7810 regulator, so the LM324 op amp will not be able to raise the gate voltage much more than 8 volts, especially with that unnecessary 1K pull-down resistor. The design is inadequate for the purpose intended.

    As it stands, it's a waste of time to try to get it to work. Whoever "adapted" the design did not do you any favors.
     
    Last edited: Sep 29, 2008
  4. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    Thank you both for the input
    Response to Mik3
    My initial measurements suggest that the optimum frequency is below 10Khz, however their will be narrow pulse widths when the circuit is limiting – speed might not be an issue however I will certainly check it out on the scope.
    Response to beenthere
    The transistor does have a high sufficiently high current rating for the job however I want to split the power dissipation between 2 devices - could you explain what is wrong with this theory?
    With regard to the regulator this is more of a challenge – for my purposes the maximum peak pulse current will not exceed 60Amps. – would that be a problem if Vgs only reaches 8v? (another reason for 2 transistors ?)
    As a car environment is “noisey” there is a need for the regulator especially if the line is subject to high current pulses, my initial thoughts are can I put a diode or 2 in the earth lead of the regulator to up the voltage by about 1.2v -bearing in mind the operating battery voltage is over 12v, another solution would be a shunt regulator because of the limited headroom in the supply - a 12v Zener diode with suitable filtering?
    What solution can you suggest?
     
  5. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    You have an intractable problem using the LM324. The output can never swing to the supply rail - 10 volts in this case, and so can never fully turn the HEXFET on. The drive to the gate is limited by the 1K resistor to ground, so I am just guessing about the actual level the gate sees, but I do know it's not 10 volts. The data sheet for the HEXFET explicitly states that 10 volts is the minimum to fully turn it on. Less voltage than that, and the device can never conduct down to that stated 8 milliohms level.

    Substituting in a modern design that is designed for single polarity supply and a better output structure may help. The real need is to get the gate up above 10 volts. That is why I keep saying to use a comparator as the driver. An LM311 with a resistor to the battery supply will do so.

    If you are worried about alternator noise, use a diode to isolate the supply line and a nice big hash filter to keep the crud down. Keep the big filter cap, but place a couple of .1 uF ceramics across it for a lower impedance path to ground for high frequency noise. Your circuit won't worry one little bit about riding up and down a volt or two.
     
  6. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    The LM324 is not fast enough for operation above around 6kHz.

    I would replace that whole PWM section with a 555 timer PWM circuit.
     
  7. scubasteve_911

    Senior Member

    Dec 27, 2007
    1,202
    1
    You cannot just slap another power transistor (or HEXFET, yes, it is a transistor..) into your circuit. If these are mismatched, you may be doing more harm than good. Read the following for more information.

    http://www.irf.com/technical-info/appnotes/para.pdf

    Steve

    I know firsthand from trying to parallel FETs.. I had to build the most efficient switch mode supply possible from discrete components as a school assignment, we tried many things to be able to get less on resistance at the price of switching losses, but we were unsuccessful. These were from problems as outlined in the paper.
     
    Last edited: Sep 30, 2008
  8. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    If you want your circuit to be more efficient, consider building a "buck" topology converter, configured to regulate current.

    If you're using a PWM circuit to drive a hydrolizer cell, then you're basically turning the cell off for a period of time (stopping gas production), and then feeding it too much current the rest of the time, which wastes power and heats the electrolyte. Unless, of course, you've found an elusive frequency that provides for magic multiples of gas production.

    With a "buck" topology converter, an inductor keeps the current flow relatively constant, which will keep your gas production relatively constant. It will certainly be more complex than a simple PWM circuit, but it will also be far more efficient.

    Here's a good starting place for a 20A buck DC-DC converter:
    http://www.national.com/pf//LM/LM5642.html
     
  9. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    Thanks for the input.

    I have reviewed the spec sheets for the LM324 and LM311 – I should have done this before building the circuit – I have not built up any discrete circuits for over 25 years (I moved into microelectronics)
    The LM324 will only go up to about 8.7Volts (2*Vbe + Vce of the current source) and as the LM311 has an open collector output it can go up to Vcc – so clearly I will take your advice for the next iteration and use the LM311.
    I am running at 15A at time 0 and the RDS is approximately 0.012ohms and when the current has risen to about 40A it is around 0.016ohms – my application works OK – of course I would like to reduce the power dissipation for reliablilty.
    Because I have the PCB and components for the current design I need a quick fix - I will increase the value of the pull down resistor and add a second power transistor to reduce the dissipation.
    My final question – my idea about raising the regulator voltage with a diode in the common terminal – would this work?
     
  10. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    Just isolate the line with that diode, plus the filtering I suggested.
     
  11. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    With regard to the buck topology - that is certainly the way to go! My measurements confirmed the inefficiency of the current schematic - however it is an improvement on a standard PWM without current sensing.
     
  12. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    Great reference on parallel transistors !
     
  13. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    Please can you explain why the LM324 is not fast enough for operation above 6khz?
     
  14. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    Take a look at the spec sheet. The slew rate is so poor it can't significantly amplify a waveform above that frequency. Slew rate the maximum rate that the output can change. If you are trying to amplify a waveform that calls for a rate of change greater than the op amp's slew rate, then the output becomes distorted and loses amplitude.
     
  15. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Your regulator has a fixed dropout voltage. 78xx series regulators have about 2v dropout. This means with 14.5v in, the most you could hope to get out of it is 12.5v.

    You could use diodes between the ground terminal and ground to raise the output voltage, within reasonable limits.

    However, by adding a 2nd MOSFET, you will be doubling (minimum) the capacitive load on the LM324's output, which will further degrade the gate rise/fall times, thus increasing the time the MOSFETS will spend in the linear operating region.

    Before you add a 2nd MOSFET, try raising the voltage regulator output, and decrease the frequency of the oscillator portion of the circuit. Decreasing the frequency/PRF will porportionately decrease the time the MOSFET will spend in the linear region.

    But really, the "C" section of the schematic as it exists will be abysmal in performance until that section of opamp is replaced with something more suitable.

    As I mentioned before, even an inexpensive 555 timer would make for a huge performance enhancement.

    LM324's are terribly slow. They're OK for low-audio-spectrum triangle wave generators, which is basically what the first portion of the circuit is doing. They are not good at all for square wave generators.

    An ordinary bjt 555 timer is capable of running more than 70x as fast as an LM324 (500kHz vs 6kHz), with many times the current source/sink capability (100mA vs 30mA). The circuit is even much less complicated than the LM324 version - fewer components to break.

    [​IMG]

    The circuit is variable from 0%-100% PWM duty cycle. Note that bypass caps are not shown.
     
    Last edited: Sep 30, 2008
  16. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    OK, just threw together a little demo for you.

    The first circuit is basically what you have, except with different opamps and a few other things - nothing major really. The important item to note is where the voltage (green trace) and current (yellow trace) are being taken from, and what they look like. I used a small capacitor and a low-value resistor (R17) to simulate your cell.

    Note how the current ranges from 0A to around 80A. During the time there is 0A current flow, nothing good is happening in the cell. During the time there is 80A current flow, there are some good things going on (gas production) and some bad things (heating of electrolyte, wasted power).

    The 2nd circuit is nearly identical. However, there's an added part; L1, a 230uH inductor. Look at the traces now; current is actually fairly steady between 27A and 38A.

    What's happening is that the inductor basically turns the current-regulated PWM circuit into a buck regulator. D1 is now acting as a "flywheel" diode; it provides a return current path for the inductor through the cell. When the MOSFET is turned on, current flows through the cell and the inductor; when the MOSFET turns off, L1 keeps the current flowing via D1 back around through the cell.

    Were the frequency (PRF) of the PWM circuit significantly higher, L1 could be much smaller, thus requiring less copper and be lower in resistance. However, it needs to be a pretty large value due to the slow PRF. Speeding up this particular design isn't in the cards at the moment.

    If you want to make an inductor with characteristics similar to what is in the schematic, you'll need about 36 feet of AWG 10 magnet wire, and a spool that has a 2" diameter center, and 4" diameter flanges about 0.51" apart. Wrap 45 turns of the magnet wire on the spool, as neatly as you can, 5 turns per layer, 9 layers deep. The spool will be nearly filled when you're done. Your inductor will be close to 235uH.

    You could get by using less wire if you used a ferrite toroid, but even that would need to be quite large at the low frequency you're operating at.
     
    Last edited: Oct 1, 2008
  17. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    You second circuit is certainly the way to go and I will consider after I have done some initial tests on the cells we are using. We anticipating to finalize a cell that runs at about 15Amps - the frequency is going to be around 3kHz as the efficiency is best when the tubes are resonating (mechanical not electrical resonance). From what you described we should be able to use a smaller inductor as your circuit looks like it was running at approximately 190hz.

    Once again thank for the great input - and I need to ask Santa for a copy of circuit maker :)
     
  18. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    OK, a few corrections/additions:
    1) I realized earlier today that in my simulation, I completely disregarded the 1.5v-2v (ballpark) Vf (forward voltage) across your cell (I'm assuming a single-stage cell where you have alternating plates connected to +v and -v). I added a few diodes in series to simulate the middle of that voltage drop (about 1.8V). This reduced the average current considerably; down to roughly 17A. Adding more diodes would result in the current being reduced even further; unfortunately with this "student edition" (aka free) it is limited to 50 components, and I'm up against that limit.

    2) I've changed the scale and markers; you can see that the actual frequency of the simulation is 1.822kHz. I didn't have the markers set in the previous simulation.

    3) As far as the inductor size; my suggestion is a trade-off between resistance of the wire in the inductor, the frequency the circuit, the capabilities of the op amp in use, minimizing thermal rise while keeping the current flow through the cell relatively constant, and keeping the size of the inductor reasonable. The larger the inductor is, the more constant the current flow through your cell will be, but you'll wind up with a higher resistance unless you go to a larger conductor, which will make the inductor larger in diameter... etc. Can you say "spiraling out of control?" :) With that LM324, you need to keep the frequency below a couple kHz, or you'll be running the MOSFET in the linear region too much of the time.

    4) With the addition of the coil, there is no need to add a 2nd MOSFET. The load on your single MOSFET will significantly decrease due to the regulating effect of the inductor.

    5) You might be tempted to run down to your local big blue or big orange hardware store and pick up a roll of AWG 10 wire for the inductor. Don't do that. The insulation is far too thick and will keep the heat in until it melts; you need magnet wire.

    6) As far as Circuitmaker Student; it is obsolete and no longer supported. For simulating linear circuits, check out Linear Technology's LTSpice available as a free download from their website. I really should've used LTSpice for this, but I already had a simulation of the circuit from a year or so ago, and didn't feel like going through the effort to re-create it in LTSpice.
     
    Last edited: Oct 1, 2008
  19. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Here's a better option for that inductor.

    Go to this page:
    https://www.amidoncorp.com/items/21
    and order one FT-114-J. This is a J-material ferrite toroid that is under 1.2" in diameter, and is an excellent candidate for this project. It has an AL of 3170; L(mH) = AL x #turns x #turns / 1,000,000

    Wind on at least 9 turns of AWG 10 magnet wire, as evenly as you can. That will make a 257uH inductor, +-20%. More=better, but don't get the wire bunched up.

    Here's a tutorial on winding toroids:
    http://www.kitsandparts.com/howtowindtoroidswithoutpain.php

    You'll only need a couple of feet of wire.
     
    Last edited: Oct 3, 2008
  20. damianhealey

    Thread Starter Active Member

    Aug 27, 2008
    30
    0
    Thanks I have made a note of the toroid details, I have installed the LTSpice and found models for the op-amp and power transistor; the next step will be to capture the schematic of the full circuit and simulate with a typical range of currents.
    For bench testing I have removed the regulator from my PWM - it is running cooler with the higher Vgs.
    I might see if the magnetic field from the toroid has any impact on HHO production – I am always skeptical about what I see on the internet as there is a lot of BS .
     
Loading...