PWM frequency problem between MOSFET and Driver

Discussion in 'Analog & Mixed-Signal Design' started by Yin Min, Jun 24, 2019.

  1. Yin Min

    Thread Starter New Member

    Jun 24, 2019
    7
    0
    Hi,

    Now, I am facing some issues with PWM frequency. I generated one frequency program with C. Frequency is time varied frequency( Hz to KHz) using adjustable PWM. My problem is that when I adjust PWM minimum and the load current is high (>2.77 A), MOSFET is not working properly. When I check with oscilloscope, not getting Frequency continuously. When I check with LED, LED is blinking. Then when I check with another time varied frequency without adjustable PWM, it is OK and can get high current (>5A). Is there any solution, please??

    Thanks for your help,
    Yin Min
     
  2. MaxHeadRoom

    Expert

    Jul 18, 2013
    18,248
    5,604
    Show circuit, also what are you using to drive the Mosfet, micro etc?
    Max.
     
  3. Yin Min

    Thread Starter New Member

    Jun 24, 2019
    7
    0
    Hi Max,

    I used Driver IC 2109s and PIC is 12f675. The circuit is as below.
    Thanks.


    upload_2019-6-25_12-3-13.png
     
  4. JohnInTX

    Moderator

    Jun 26, 2012
    3,686
    1,913
    If you are using a PIC to generate the PWM, post your PIC code and schematic showing how the PIC is hooked up.
     
    Joël Huser likes this.
  5. Alec_t

    AAC Fanatic!

    Sep 17, 2013
    9,743
    2,352
    What is the bridge supply voltage?
    What is the voltage on the VB pins of the ICs?
     
  6. Yin Min

    Thread Starter New Member

    Jun 24, 2019
    7
    0
    Hi,

    VB pins of the driver ICs is 16V. The supply voltage of MOSFETS is 48V. The circuit is working well with low current load (< 2.77 A). Only when the load current is higher than 2.77A, MOSFET Q2 gate become abnormal because the signal gate voltage of Q2 that I measured is not stable. Then, I noticed 10 ohm resistor between Q2 gate and HO pin of driver IC. So, I changed 20 ohms resistor in place of 10 ohms. Now, it can work higher load current with minimum PWM of frequency and there is no more output LED blinking. But the temperature of MOSFETS Q2 and Q1 is increasing obviously than Q3 and Q4. After 10 minutes, I stopped.
    Now, I confused:(
    Is changing higher value resistor making MOSFETs spoiled?
    Does these MOSFETs cannot stand PWM changes of frequency? (because there is no issue with same frequency with fixed PWM)
    Any advise, please..

    Thanks for your help.
    Yin Min
     
  7. MaxHeadRoom

    Expert

    Jul 18, 2013
    18,248
    5,604
    Why are you using a Pic that does not have the PWM module?
    There are other 12 pin versions with PWM, also for low power applications there is the LMD18200 H bridge IC that can be picked up cheap already mounted on PCT board that just requires a single PWM signal.
    Max.
     
    Joël Huser likes this.
  8. Yin Min

    Thread Starter New Member

    Jun 24, 2019
    7
    0
    Hi Max,

    Thanks for your suggestion.
    Because I am currently using this PIC with the other frequencies. Now, I start want to adjust PWM of frequency. So, I modified the existing frequency and used potentiometer to adjust the pulse width.

    Thanks,
    Yin Min
     
  9. sagor

    Member

    Mar 10, 2019
    82
    20
    Just a suggestion, the 8 pin 12f683 has a PWM module in it, and is pin for pin compatible with the 12F675. It may make coding of the PWM easier...
     
    Yin Min and MaxHeadRoom like this.
  10. MaxHeadRoom

    Expert

    Jul 18, 2013
    18,248
    5,604
  11. JohnInTX

    Moderator

    Jun 26, 2012
    3,686
    1,913
    Increasing the gate resistors may cause the MOSFETs to run hotter at high loads because it slows down the turn-on and turn-off and increases the power dissipation during switching.

    I don't know why the MOSFETs would have a problem with changing frequency and duty cycle within reasonable limits. In general though, you pick a frequency and vary the duty cycle of the PWM to vary the output.

    Since you are generating the PWM in firmware, I wonder if you are running into read-modify-write problems on the PORTs of the PIC. If you are setting and clearing the PWM output using bit-wise instructions i.e. bcf/bsf (or in C, PORTx.BIT = 1 etc.) the port IO can be corrupted due to noise, pin loading and other factors. Increasing the load current will increase the switching noise too. Making the R's bigger slows things down a bit and maybe that's why it works a little better with the 20 0hm gate Rs. The way to beat the problem is to
    1) shadow the outputs - maintain a copy of the output image in a byte variable, modify the variable (not the PORT pins) then write the variable as a byte to the PORT.
    2) use a PIC with a hardware PWM. That will help the PWM but not the other outputs.
    3) use a PIC with LATx registers (10F322 has them. 16F1840 is a nice one).

    You also may just have coding problems, noise into the ADC etc. Post the code if you want and we can take a look.

    Good luck!
     
    Yin Min likes this.
  12. Yin Min

    Thread Starter New Member

    Jun 24, 2019
    7
    0
    Hi JohnlnTX,

    Thanks for your valuable help. You are right. I am using bit wise instruction in my coding. Maybe because of this. I will try again and will come back.

    Thanks,
    Yin Min
     
  13. jeffl_2

    New Member

    Sep 17, 2013
    4
    1
    I could be way off because I don't have experience with the 2109 driver, one thing that appears to be kind of nice is it has a built-in 540 ns deadtime. I'm not trying to criticize the circuit design choice of a non-complementary bridge (all N-channel devices) because that will obviously result in the lowest total RDSon (and you only have "Miller effect" considerations for the low side devices). However these circuits can create problems that are tricky to resolve. You appear to be attempting to get at least 240 watts and to reverse polarity which makes me think this is a driver for something like an electric bike or street surfboard, at least it appears the load is some kind of motor. With something like that not only is the load inductive, but all motors develop a reverse EMF at the terminals while they are rotating. With that as a precaution you have to realize that the high side transistors aren't "fully on" until the gates are driven 10 to 15 volts ABOVE WHATEVER VOLTAGE THE SOURCE PIN OF THAT DEVICE HAPPENS TO BE AT. In addition of course there is a maximum PWM frequency that will make sense given the motor inductance and other considerations. All I'm suggesting is you get yourself a decent "diff probe" and take some waveforms from gate to source and see if you can still convince yourself that they're really being fully driven ON when you want them to be (and maybe even simulate the uC with a pulse generator if you have one), if you find you have a "clean bill of health" THEN I'd start worrying about coding issues etc. (it also doesn't make sense that you're reporting problems with "Q1 and Q2" because they should be an exact mirror of the other two devices if the circuit is built as drawn).
     
    Yin Min likes this.
  14. Yin Min

    Thread Starter New Member

    Jun 24, 2019
    7
    0
    Hi jeffl_2,

    Thank you for your suggestion. Now, after changing another new MOSFETs (same part no.) in place of Q1 and Q2 MOSFETs, it can be used with higher current load greater than 2.7A. Seems like it is because of the MOSFETs quality. But I am still questioning why this problem come only in this frequency with changing PWM. ( When duty cycle is 50% ON condition (Positive pulse only) , it can be used with higher current load ( >5A). But when the duty cycle is 20%ON and 30%OFF condition (Positive pulse only), it can be used with certain current load (means lower current) only.)


    Thanks,
    Yin Min
     
  15. jeffl_2

    New Member

    Sep 17, 2013
    4
    1
    These power transistors are "analog" devices until that aspect is "designed out of" the circuit, and they need to have their gates driven "to excess" before you can begin to observe the published specification for RDSon. I get the impression you're "assuming" if your drive signal goes into a drive chip designed for that function that the associated gate will be adequately driven but that's rarely the case. And if they aren't reliably driven fully into saturation then they dissipate lots of heat and if you haven't mounted them on large heatsinks their characteristics will degrade even further, sometimes it's temporary and reversible, often not. It doesn't take much of that 250 watts coupled through an inadequately driven transistor to completely "cook" the device especially if it's just in one of these surface-mount-only packages with a very high thermal resistance. I can only make recommendations but they are based on a lot of trial and error. I tend to think if you look at it you'll find the gate-to-source waveform will show that the drive margin is not adequate during some portion of the PWM cycle.
     
Loading...