LTC2977 Triming status when DC/DC load transient

Hi,

We use LTC2977 to trim DC/DC(LTM4650) with an external feedback network.

When the load transient, if LTC2977 exactly monitor the pulse (for example, 6mV diviation), if LTC2977 adjust its DAC to trim DC/DC ? Because the DC/DC will return back after some microsecond, but if LTC2977 trim the DC/DC during load transient, it will degrade the power supply performance.

Parents
  • 0
    •  Analog Employees 
    on Apr 24, 2019 9:23 PM over 1 year ago

    The LTC2977 servo loop acts to adjust Vout over relatively long periods, approx. 100ms.  There are two feedback loops in such a system.  The fast loop is the 4650 and associated top/bottom feedback resistors.  The slow loop is the 2977's ADC and DAC.  A DAC resistor adds/subtracts a small DC current into the feedback node of the switcher.  This acts to 'trim' the output. The servo loop updates at 100ms and is intentionally slow so as not to conflict with the fast loop.  For fast transients, the 2977 servo loop doesn't see pulses because the ADC takes 6ms to accurately measure Vout and moves the DAC one LSB, and is essentially 'blind' to small glitches. 

    To improve the load step performance of the 4650 (fast loop), you may consider adding components to the COMP pins and/or possibly consider layout changes if you already have hardware and are taking measurements.

    Mike

  • Hi Mike,

    So,

    1. What about LTC2977 fast servo mode? It will successively adjust the DAC one by one LSB until the target is reached, don’t need take ~100ms to monitor it again. So if fast servo is only applied upon channel startup, VOUT_COMMAND change and Margin command being issued? For other cases, 2977 only work at slow servo mode even though the “fast_servo_off” bit is cleared.
    2. Please see picture below, if ADC starts to monitor at point A (-4.8mV deviation), although ADC need take 6ms to accurately measure it, but it has measured one bit at least, so do we need to care this case?

     

    Best regards,

    Xuezhuan

Reply
  • Hi Mike,

    So,

    1. What about LTC2977 fast servo mode? It will successively adjust the DAC one by one LSB until the target is reached, don’t need take ~100ms to monitor it again. So if fast servo is only applied upon channel startup, VOUT_COMMAND change and Margin command being issued? For other cases, 2977 only work at slow servo mode even though the “fast_servo_off” bit is cleared.
    2. Please see picture below, if ADC starts to monitor at point A (-4.8mV deviation), although ADC need take 6ms to accurately measure it, but it has measured one bit at least, so do we need to care this case?

     

    Best regards,

    Xuezhuan

Children
  • +1
    •  Analog Employees 
    on Apr 27, 2019 12:58 AM over 1 year ago in reply to Xuezhuan

    Xuezhuan,

    1. Even in fast servo mode, the ADC is going to miss the 6us glitch due to filtering/averaging. When the output voltage is up and running (servo has already soft-connected) and a load step occurs, the output voltage will have a disturbance. The ADC samples this particular channel once every 100ms, and the user has no control over the time that the sampling occurs. It is asynchronous to the glitch. The ADC is taking a measurement that 'averages' the sensed voltage. 

    2. The purpose of the ADC is to make a very accurate voltage measurement. It is best to reduce noise artifacts and we suggest that you add a RC low pass filter in front of the Vsense pin. It is intentionally a slow servo loop, so it is more advantageous to remove high frequency components from the signal. Otherwise the DAC codes will be constantly bouncing back and forth between codes. The servo loop is intended to adjust Vout over relatively long periods of time.  It's the supervisor (comparator) that needs to react quickly (fault off), not the ADC.

    Mike