Post Go back to editing

LTC1778 maximum inductor value

What is the largest practical inductor value (10uH, 100uH) to be used with the LTC1778, specifically when used with light loads (10mA to 250mA)? I understand that the LTC1778 will drop back into discontinuous mode with light loads but has anyone studied what the threshold of inductor size is where the LTC1778 can no longer regulate or operate within an SOA, given that a larger inductor value will likely yield higher self-induced transient voltages, which may exceed the absolute maximum ratings of the LTC1778 (36V). Is there any issue with using a snubber network from SW to GND to dampen such transients, should they occur, when using an inductor greater than 22uH?

Parents
  • Hello,

    before going to your question I would encourage you to choose a more modern part. LTC1778 was introduced before 2005. Its direct successor, the LTC3878 was introduced in 2009. In the meantime there was quite some progress and growth in our Buck controller section, which lead in general to more bang for your buck, depending on your requirements.

    And if your asking for 10-250mA of output current you can select from many monolithic buck regulators with no need for FET selection or snubber or complicated layouts.

    And normally customer would like to optimize the design by using smaller inductors as far as possible to reduce foot print and save boardspace.

    To check the impact of your selected inductor I recommend using our LTSpice tool, which will show what happens with inductor ripple for example or any other interesting signals as well.

    kind regards

    Markus

  • Markus,

    Thank you for your comments.

    Yes, I am aware of the LTC3878 but this inquiry was regarding an in-production design, which was initially created in 2003. There has already been some consideration to update the design with the LTC3878 at some point. We are currently using the LTC1778-1. 

    The reason for the question is that we have a requirement to use a different 'load' on the current design, which reduces the demands on the LTC1778 from 10W of output to perhaps 3-4W. The LTC1778 will likely operate in discontinuous mode throughout this operating range. I've seen the pulse repetition rate drop down to 12.5kHz, depending on the inductor and the load. 

    I would assume that at some point, the LTC1778 cannot switch over a short enough period (<300nS) or run the repetition rate out far enough to maintain regulation. To address the issue, we've used chokes from 33uH to 100uH (even though the math suggests much large values). Since the design is powered by 24VDC, the ringing from the larger inductors values creates a peak voltage exceeding the 36 Vmax of the SW input, likely resulting in damage to the LTC1778 or the MOSFET G-S region.

    We've considered using a RC or RCD snubber across D1 (the Schottky diode), but the discharge current into the low-side MOSFET seems to interferes with the Isense mechanisms in the LTC1778, resulting in other issues, such as the over-current latch tripping. We normally defeat the overcurrent latch (in the base design) by pulling up SS with a 10K to 100K resistor. However, since the LTC1778 regulates in current mode, I would assume that any false (non-inductor) current in the low-side MOSFET would cause indeterminate behavior.

    Thus, it appears that the inductor must be sized so that the turn count is small enough that the self-induced currents do not result in the SW input exceeding 36V. So there appears to be a practical limit to the inductor size (or at least the turn count), when operating with a 24VDC rail.

    We started out using 68uH and 100uH inductors in the new circuit WITHOUT including an RC/RCD snubber but the results were mixed results. Some boards will work, some work for a while, even in the field for a year and some never work. We have fielded product since 2004 with 10uH and 22uH chokes driving a 10W load without issue. 

    So again, I think there is a practical limit in this case.

    The load profile is not constant which makes the design a further challenge, which is why I chose the LTC1778. We actually vary the output from the LTC1778 by pulling the FB pin low, over a linear range under microprocessor control, plus the load current can vary over a wide range. Historically, we have required the LTC1778 to start with a 1V output driving less than a 300mA load, then drive the FB pin to provide a 12V or 15V output, with up to a 1.5A load. So the numbers are generally worked for 80% power, which is where must of our customers operate for the bulk of the time. I understand there are impacts to efficiency operating in this manner, but again the LTC1778 can drop back to discontinuous mode and it has worked fine, with loads under 1W, even though the efficiency may suffer. 

    And yes, at these new lower power levels (10mA to 150mA), a linear-regulator would make more sense but we were trying to avoid redesigning the product for this one application and 30 units.

    We may use a 10uH or 22uH choke (a we know that has worked) and as you mentioned, just accept the ripple.

    Thank you.

    Lauren

  • Hello Lauren,

    huh, quite a long and elaborate answer. Thanks for clarification, although this weird circumstances/requirements will not help really for an answer. In my opinion/experience the maximum 'possible' inductor value depends on the overall design. Increasing L beyond typical values reduce the current slope and makes it harder for the part to detect a clean limit. In the end the part falls back to a pure voltage feedback loop as the current loop will not work anymore. And you have experienced the impact of larger L to the SW voltage as well. So this is an 'general' statement in terms of quality, but no quantities given as these will vary heavily on the rest of the circuit.

    For your DAC control on the FB pin: Please have a look on our LTC7106, with its current DAC output easier and more stable/accurate than your voltage DAC approach.

    www.analog.com/.../ltc7106.html

    And I didn't mean a LDO as replacement, but moreover a part like our LT8606 with two FETs inside and a maximum Iout of 350mA. With 42V Vin,max more headroom than any LTC1778 circuit. Easy to design.

    kind regards

    Markus

  • Hello Marcus,

    Thank you for the additional comments.

    To clarify my point about the use of the FB control pin to adjust the output voltage --

    I actually employ an optical isolator driven by a 16-bit PWM signal, which is integrated by a RC filter and that resulting voltage (about 1V to 3.5) is buffered by a LM358 which drives the base of a PN2222 (with a series resistor). The collector of the transistor pulls down the FB line, proportional to the PWM signal -

      

    A 470 ohm resistor is in the emitter lead to limit the full power range.

    It's not entirely linear, especially at the bottom, but the MCU corrects for that, by measuring the output voltage at the 'load' and correcting the PWM drive to meet the required target. About the first 4000 counts have little effect until the transistor wakes up. The remaining 60000 counts result in a pretty linear response.

    Also, this circuit/system was designed back in the 2002-2003 timeframe. The product is still shipping today but an update is planned. Your comments and recommendations will be taken into consideration.

    Lauren

Reply
  • Hello Marcus,

    Thank you for the additional comments.

    To clarify my point about the use of the FB control pin to adjust the output voltage --

    I actually employ an optical isolator driven by a 16-bit PWM signal, which is integrated by a RC filter and that resulting voltage (about 1V to 3.5) is buffered by a LM358 which drives the base of a PN2222 (with a series resistor). The collector of the transistor pulls down the FB line, proportional to the PWM signal -

      

    A 470 ohm resistor is in the emitter lead to limit the full power range.

    It's not entirely linear, especially at the bottom, but the MCU corrects for that, by measuring the output voltage at the 'load' and correcting the PWM drive to meet the required target. About the first 4000 counts have little effect until the transistor wakes up. The remaining 60000 counts result in a pretty linear response.

    Also, this circuit/system was designed back in the 2002-2003 timeframe. The product is still shipping today but an update is planned. Your comments and recommendations will be taken into consideration.

    Lauren

Children