Electronics > Projects, Designs, and Technical Stuff

Low quiescent current 3.7V to 170V boost converter

<< < (3/3)

SiliconWizard:

--- Quote from: RawCode on July 31, 2019, 10:28:38 am ---I used a BSP89(which is already inside LTSPICE), which is similar to that one you suggested, and i can get only 100V out of this circuit. Maybe 3.7V as Vin is too small :/
If i try to use a STN3N45K3 by ST and the situation is even worse.

--- End quote ---

(The STN3N45K3 has a much higher threshold so yes it won't work properly here @3.3V. All typical figures are given @VGS = 10V.)

Well, to begin with, the values of the feedback resistors you selected would not yield an output of 170V but ~249V. [ 1.24*(1+2e6/10e3) ]

I made the following changes to your schematic:
* More typical values for the compensation network,
* Appropriate values for the feedback divider (~168V),
* More realistic value for the series resistor of the input power supply, picked an inductor from a real reference, picked the output cap from a real reference too and changed to diode to a Schottky one (although this last point shouldn't make much of a difference here).

It does settle to ~168V in ~25ms.

I was unimpressed with the average current draw once settled though (~204mA), so an efficiency of ~25%. Tried to figure out where the excessive losses would come from. The suspect was the transistor with its relatively high RDSon. Adding the following trace helped showing that: (V(vd)-V(vs))*Id(Q1)
=> an average dissipation of ~404mW in the transistor (for a total power draw of ~673mW!)

You can then further try playing with the current limit (Rsense), try other transistors, consider generating an elevated voltage for the MOSFET driver...


--- Quote from: RawCode on July 31, 2019, 10:28:38 am ---Do you think that with a standard boost converter i can get a more efficient "conversion"?

--- End quote ---

Theoretically yes, as this simple charge pump topology is not very efficient, and you're cumulating (multiplying) the efficiencies of the two cascaded stages (boost converter + charge pump).
For significantly higher output currents, you would probably not even have a choice.

But in practice, at these high voltage, low output current levels and high conversion ratio, it's hard to find something appropriate that's reasonably efficient.
The typical compromise is between the switching losses and the gate charge losses. Higher gate voltage (RDSon will be lower) and faster rise and fall times (the transistor will spend less time in the linear region) will limit power dissipation (drain-source), but increase the losses to drive the gate... At this low output current, this compromise gets critical.

So the solution with a LT8364 + a charge pump looks like a simple an not too bad solution here.

Some people could suggest a flyback topology, but I think it will be pretty hard to do better than 60%.


--- Quote from: RawCode on July 31, 2019, 10:28:38 am ---How did you get the input current, though? I can only get a poorly readable graph for the current.

--- End quote ---

LTSpice can compute average and RMS values for any trace in the waveform viewer.
For average current draw, I take the average current from a current trace.

To do that, hold the CTRL key and left click on the trace's title (not the trace itself). A message box pops up wih the average and RMS values (or integral depending on the dimension/unit of the trace) of the portion of trace that is currently displayed (so zoom in/out to have it computed on the portion you want).


--- Quote from: RawCode on July 31, 2019, 10:28:38 am ---Anyway, I'm using a 380mAh lipo battery. I know, it's not very huge, but i plan to power on the Nixies only when needed.
How can i calculate the battery life out of this information?

--- End quote ---

Well, you simply have to compute the average current draw using a weighted sum based on the duty cycle(s) of your system and the average current draw of each scenario (display on, display off, MCU running or in sleep mode, etc)

Then divide the battery's capacity by the average current draw and there you get your approximate battery life.

RawCode:

--- Quote from: SiliconWizard on July 31, 2019, 03:06:24 pm ---Well, to begin with, the values of the feedback resistors you selected would not yield an output of 170V but ~249V. [ 1.24*(1+2e6/10e3) ]

--- End quote ---
Yes, i was trying to get an higher voltage without any results. I probably didn't save the file with the right feedback resistors :/


--- Quote from: SiliconWizard on July 31, 2019, 03:06:24 pm ---I made the following changes to your schematic:
* More typical values for the compensation network,
* Appropriate values for the feedback divider (~168V),
* More realistic value for the series resistor of the input power supply, picked an inductor from a real reference, picked the output cap from a real reference too and changed to diode to a Schottky one (although this last point shouldn't make much of a difference here).

It does settle to ~168V in ~25ms.

I was unimpressed with the average current draw once settled though (~204mA), so an efficiency of ~25%. Tried to figure out where the excessive losses would come from. The suspect was the transistor with its relatively high RDSon. Adding the following trace helped showing that: (V(vd)-V(vs))*Id(Q1)
=> an average dissipation of ~404mW in the transistor (for a total power draw of ~673mW!)

You can then further try playing with the current limit (Rsense), try other transistors, consider generating an elevated voltage for the MOSFET driver...


--- End quote ---

Thank you very much, i'll dig a bit more on this circuit.


--- Quote from: SiliconWizard on July 31, 2019, 03:06:24 pm ---
Theoretically yes, as this simple charge pump topology is not very efficient, and you're cumulating (multiplying) the efficiencies of the two cascaded stages (boost converter + charge pump).
For significantly higher output currents, you would probably not even have a choice.

But in practice, at these high voltage, low output current levels and high conversion ratio, it's hard to find something appropriate that's reasonably efficient.
The typical compromise is between the switching losses and the gate charge losses. Higher gate voltage (RDSon will be lower) and faster rise and fall times (the transistor will spend less time in the linear region) will limit power dissipation (drain-source), but increase the losses to drive the gate... At this low output current, this compromise gets critical.

So the solution with a LT8364 + a charge pump looks like a simple an not too bad solution here.

Some people could suggest a flyback topology, but I think it will be pretty hard to do better than 60%.

LTSpice can compute average and RMS values for any trace in the waveform viewer.
For average current draw, I take the average current from a current trace.

To do that, hold the CTRL key and left click on the trace's title (not the trace itself). A message box pops up wih the average and RMS values (or integral depending on the dimension/unit of the trace) of the portion of trace that is currently displayed (so zoom in/out to have it computed on the portion you want).

Well, you simply have to compute the average current draw using a weighted sum based on the duty cycle(s) of your system and the average current draw of each scenario (display on, display off, MCU running or in sleep mode, etc)

Then divide the battery's capacity by the average current draw and there you get your approximate battery life.

--- End quote ---

Thank you for all the effort and the time you spent for me! I really really appreciate it! :)

Navigation

[0] Message Index

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod