This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM3478 fails to switch with Vin < 7V

Other Parts Discussed in Thread: LM3478, LM3481

I am using the LM3478 in a conventional boost regulator with Vin nominally between 8V and 15VDC, Vout = 125V, 100mA load. F = 100KHz and nominal duty ratio is 33% but grows to 60% with Vin = 9V.  It is operated from 12V lead-acid battery.  It works to my satisfaction when input voltage is within specs (9V < Vin < 15V) but the circuit has the annoying habit of smoking front end components and switch transistor when battery runs low.  This should rarely happen but in long run is unavoidable.  If Vbatt runs down to about 7V then DR pin of LM3478 freezes HIGH holding the Xsistor ON and sucking the remaining battery charge down through a SB fuse, protection diode, inductor and switch.  This can amount to several amps for several seconds.  The fuse will eventually blow but not before being protected by the failing transistor, et al first.  Typical input current is 1.3ADC.

I understand that the chips changes bias mode at about 7.2V but it is supposed to keep operating down to 3.3V or thereabouts.  It doesn't.  On testing, the DR pin WILL go low at Vin=3.3V but not before it has drained all of the charge to be had from the input first.

Three specimens behave this way.

Why doesn't the chip keep operating below 7V?

Why does the Isen pin not shut the DR signal down at that point?

Could ton+toff be in danger of crashing 100% with low Vin?  Even if that did happen why would Isen feedback not kill DR?

Do I need to create low Vin detector and kill Vin at 7V approach?

Photo shows current through Q at vert = 2.5A/div and Voltage at drain of Q with Vin at vert = 50V/div, probably about 8V.

Any help would be appreciated.

Schematic follows:

  • Tim, I will move your question to the forum responsible for the part and notify the engineer.
    -Yang
  • Hi Tim,

    The device should take care of the low voltage operation although I think the duty cycle should be in the 90% range for the 8/9V operation mentioned. Can I ask you to look at WebBench on our website for component selection and best practice. I copy below the circuit I produced with your basic operating conditions.

    Regards,

  • Graham,
    Have made a few more measurements.
    "The device should take care of the low voltage operation" - yes that's the whole point to the question, but it doesn't. The chip stops switching when Vin falls to about 6.8V with the DR output freezing HIGH. This turns the FET on permanently and smokes the front-end components.
    The Vin pin has about 1V P-P of switching noise on it. When R2 is increased to 100R noise is reduced to about half but behaviour at 7V Vin is still the same.
    Once the unit stops switching with DR HIGH the FET current hangs at about 1.5A and is, of course dependant on Vin. When Vin is increased again until voltage on the 20mOhm resistor reaches 75 mV then DR is reset and switching resumes. On-time is about 55% when the chip stops switching.
    I don't need to know about conventional operational theory of a boost converter. What I need is to know about idiosyncrasies of the LM3478 specifically.
    Also datasheet (page 9) says that the unit continues in "hysteretic mode" once Vin falls below 7.2V. What is that? It is not defined or mentioned anywhere else in the document.
    About your circuit: 270uH for the energy-storage inductor is way too high for a switcher operating at 100 KHz. Mine is 6.8 uH and seems to work just fine for me.
    Anything you can tell me will be appreciated.
    Thanks
    Tim
  • Hi Tim,
    I have referred this problem on to design.
  • Hello

    I have encountered the same problem in my design. I designed a Sepic converter based on LM3478 with a dual 15uH inductor. Switching frequency is 0.8MHz, Vin range is 9 to 24V and Vout is 12V.

    The problem happens when I switch on my laboratory power supply (set at 12V) and power up the circuit shown in the picture above. Vin rises up to 2.5V and stays there because input current is 2A and the power supply will limit the voltage. DR pin stays at high level and therefore the mosfet is making a short circuit. That deadlock condition is kept for some time: the mosfet sinks a large DC current, the lab power supply turns the voltage low, LM3478 has too low Vin to work and keeps DR high. I have already burnt a prototype board due to the large current.

    DR voltage slowly goes down and after 2 seconds it becomes lower than the mosfet Vgs th, which turns off, then the input voltage can rise and LM3478 starts working normally. As I see it, the problem here is LM3478 drives a strong high level on pin 6 (DR) even though it is in shutdown mode due to low Vin. I tried connecting DR - gate trace to ground with a pull-down resistor but there was no change, either with 2k or 20k.

    Once the circuit has started to work normally it performes nicely, as expected.

    Thank you for your attention

  • Hello,
    One issue with the LM3478 is that is has 100% max duty cycle. This is OK for some applications but can be a problem in some cases. If the current limit cannot be reached, the system will keep the high side ON even if there is not enough voltage to drive the FET properly. In that case the system can hang up with the FET in active region and consuming a lot of power.
    If that is the issue, one option is to use logic level FETs that will stay in the ohmic region at lower input voltage. Another is to increase the current sense resistance to trigger the current limit earlier (if that is acceptable for the application). Finally another way is to switch to the LM3481 which is a very similar part but has 85% maximum duty cycle.
    Best Regards,Florent
  • Thank you for your reply. I understand it now. Maybe lm3478 is not ideal for sepic or boost configurations. However I have implemented an undervoltage lockout which solves the problem I had.