High current Step-down converter efficiency problem

Thread Starter

MrsssSu

Joined Sep 28, 2021
266
1638756610428.png
1638756660000.png
Hi People:), I am trying to build a synchronous step-down converter but my LT spice simulation shows poor efficiency. The picture above is at Here . I have attached my LT spice below for reference. May I know if this is the right way or anyone can correct my LT spice file and send it in the answers.

Thank you and Have a Nice Day :)
 

Attachments

Papabravo

Joined Feb 24, 2006
21,159
  1. What are your design requirements (goals)?
  2. How did you choose your inductor value?
  3. How did you choose the capacitor value?
  4. How did you choose the switching frequency?
  5. How did you select the MOSFET parameters?
  6. How do you account for switching losses?
  7. How much current in the MOSFETS while they are switching?
Look at the waveforms and tell me if you think the MOSFETS are switching from ON to OFF and vice versa in 1 nanosecond.
 
Last edited:

crutschow

Joined Mar 14, 2008
34,285
You are getting conduction through both MOSFETs being on during the switching interval.
Below is the circuit with non-overlap delays added to the two switching waveforms.
See how the M2 gate signal (yellow trace) doesn't go high until after the M1 gate signal (green trace) has gone low, and goes low before the M1 gate signal goes back high.
Note that the gate voltage needs to be 5V or 10V above the supply voltage, depending upon whether the MOSFET is a logic-level, or standard type, to fully turn on the MOSFET.
In a real circuit you could do that with a bootstrap MOSFET driver circuit.

With that, the efficiency is about 88% (load power divided by V3 power).

The output has a large overshoot due to the resonance caused by the output inductor and capacitor.
The is minimized is an actual circuit by using a slow start circuit.

(I didn't have your MOSFET model so I used one I had.)

1638764551378.png
 
Last edited:

MisterBill2

Joined Jan 23, 2018
18,176
Switching mode power supplies are NEVER that simple, and given the cost of components there must be a reason for the extra expense. And I am familiar enough with switcher design to know that every element of the circuit matters in the operation.
several websites of semiconductor manufacturers have excellent on-line design programs available, including T.I and S.T. , there are others as well. And I would ignore all of the cartoons on yootoob and bookface and those other entertainment venues.
 

Thread Starter

MrsssSu

Joined Sep 28, 2021
266
You are getting conduction through both MOSFETs being on during the switching interval.
Below is the circuit with non-overlap delays added to the two switching waveforms.
See how the M2 gate signal (yellow trace) doesn't go high until after the M1 gate signal (green trace) has gone low, and goes low before the M1 gate signal goes back high.
Note that the gate voltage needs to be 5V or 10V above the supply voltage, depending upon whether the MOSFET is a logic-level, or standard type, to fully turn on the MOSFET.
In a real circuit you could do that with a bootstrap MOSFET driver circuit.

With that, the efficiency is about 88% (load power divided by V3 power).

The output has a large overshoot due to the resonance caused by the output inductor and capacitor.
The is minimized is an actual circuit by using a slow start circuit.

(I didn't have your MOSFET model so I used one I had.)

View attachment 254296
Hi sir, I hope you remember me. Your answer to adding the dead time between complementary PWM signal using NAND schmitt trigger logic gate can be used to drive these mosfets with dead time right?
Furthermore, the 88% efficiency in this circuit is due to mosfet power loss which means we can achive more than 95% efficiency when we use mosfet that has extremely low Rds On (treat it as a copper wire only with no power loss) :)
 

LowQCab

Joined Nov 6, 2012
4,025
The biggest hurdle to jump-over here is that You are trying to make
the Circuit fit on the head of a Pin by running a super-high Switching-Frequency.
Every nuance of the Circuit starts to become ultra-critical, and unstable.
Just drop the Frequency down to ~20khz and you'll be much happier.
.
.
.
 

MisterBill2

Joined Jan 23, 2018
18,176
Two more considerations: First, switching mosfets require adequate gate drive to achieve that minimal on resistance, and so gate drive is important, and second, every bit of power delivered must pass through that series inductor, either just conducted, or else stored in the magnetic field in the core. So the demands on the inductor are considerable.. AND, as the frequency increases so do the losses.
Besides all of those challenges, the gate signals presently arrive "via magic." In the real world there will need to be a real circuit delivering those real gate drive signals. Also, in that same real world, at that selected high frequency, every PCB conductor will have inductance and resistance that matter, as well as magnetic coupling to nearby conductors, in addition there will be capacitive coupling where you really do not intend it. Other than those concerns, and heat transfer, the task is simple.
 

Juhahoo

Joined Jun 3, 2019
302
Do you use simulator to learn to use the simulator itself, or design something that will have some practical usage IRL?
In latter case buy some buck regulators that have a proper controller IC and start debugging it and make measurements. Simulators are...nah..just too far out from reality.
Analog offer great evaluation boards and proper explanations how to set up the component values. TI even have their own "switcher design" program which will output you a proper component values around their chips.
Then, if you design a PCB, you need to do it correctly. Switchers are challenging to do, and usually require many design iteration loops.
 

crutschow

Joined Mar 14, 2008
34,285
Simulators are...nah..just too far out from reality.
Beg to differ.
I designed and simulated a low-noise, buck regulator for a NASA satellite application.
The design would have been much more difficult if not for the simulations helping me tweak the design to get the desired performance.
And the results with the actual circuit were very close to the simulated results, including noise and efficiency.
If a simulator does not give accurate results it's often the result of poor component models and/or not properly including parasitics.
 

MisterBill2

Joined Jan 23, 2018
18,176
Beg to differ.
I designed and simulated a low-noise, buck regulator for a NASA satellite application.
The design would have been much more difficult if not for the simulations helping me tweak the design to get the desired performance.
And the results with the actual circuit were very close to the simulated results, including noise and efficiency.
If a simulator does not give accurate results it's often the result of poor component models and/or not properly including parasitics.
Certainly poor models and not adequately modeling ALL parasitic elements will lead to incorrect results. Bob Pease, far more experienced in circuit models than most folks, had a great deal to say about that.
Knowing every element that must be included in a model is not what I would anticipate from a beginner or even most casual users of simulators. The pain is most often found in the details.
 

LowQCab

Joined Nov 6, 2012
4,025
Make Your life easier .......
Use a "Push-Pull" Transformer Output.
Very low-Noise, Very stable, even if You don't know what you're doing.
.
.
.
LapTop Power Suppy FLAT .png
 

MisterBill2

Joined Jan 23, 2018
18,176
C
Make Your life easier .......
Use a "Push-Pull" Transformer Output.
Very low-Noise, Very stable, even if You don't know what you're doing.
.
.
.
View attachment 254766
Certainly true! A variable duty cycle inverter is far more forgiving than the typical switching mode supply. And the circuit shown can probably be very simply adjusted to serve as a typical 19 volt, 90 watt supply for a laptop computer for application in a vehicle. So thanks for publishing it. I am not sure just how clean the power for a typical laptop computer needs to be
 

crutschow

Joined Mar 14, 2008
34,285
Bob Pease, far more experienced in circuit models than most folks, had a great deal to say about that.
True.
He did not like or use simulators.
I think his may complaint was designers who used simulators and then built the final circuit from that without breadboarding it first, which certainly is a no-no.
Knowing every element that must be included in a model is not what I would anticipate from a beginner or even most casual users of simulators. The pain is most often found in the details.
Of course.
But that's not a good reason to say beginners shouldn't use them or condemn simulators in general.
I've had several times where a simulation of the circuit revealed the details of incorrect operation in an already built circuit, and that would have been very difficult to determine from the real circuit (partly because you can simultaneously probe all voltages and currents in a simulated circuit without disturbing its operation).

You don't like simulators, that's fine (how much have you used them?).
But I think you are doing a disservice to beginners when you say they shouldn't be used.
 

MisterBill2

Joined Jan 23, 2018
18,176
The big thing with simulators is that the users need to understand the limitations of what they are using. That seemed to be Bob's most common complaint about them. Incomplete models that left out important details and variables.A simulatr can be a handy tool for investigating and developing, BUT knowing what it is not doing is important. And blindly trusting the simulation is always a problem. And switching power supplies are seldom able to have all the variables needed to be considered.
 

ericgibbs

Joined Jan 29, 2010
18,766
hi,
Simulators are used successfully in many technical applications.
Civil and Military aircrew training, Scientific, Medical Space engineering,training and research. etc.

Most qualified personnel know the limitations of a simulation, but use a simulation to check for problems in a designed project.
Also, many parameters can be observed without resorting to a hand calculator.

To put down simulators is backward thinking, they have many advantages.

E
 

crutschow

Joined Mar 14, 2008
34,285
I'm curious as to how may of those making negative comments about simulators have had much actual experience with them.

I've used Spice simulators since they first became available and found them to be immensely useful in designing circuits.
I would never build an analog or digital circuit, even a simple one, without simulating it first.
It's located more errors in my designs, then I care to mention.

The idea that the only way to learn circuits is to physically build them is retrograde.
 

Papabravo

Joined Feb 24, 2006
21,159
I have been using simulation since SwitcherCAD (predecessor of LTspice). It was always an aid to investigation and education. It was NEVER a substitute for the breadboard or the bench. It probably reflects an inability to judge and use a tool on its merits, and that is incredibly shortsighted.
 

ronsimpson

Joined Oct 7, 2019
2,988
As crutschow found in post #3, the top MOSFET is not driven correctly. The Gate Drive should be applied from S to G. There was great heat loss in the top part.
1639255427400.png
If you layout a PCB, V2 should be connected to G-S of M2. Don't just connect to ground. At high frequency and high current ground does not equal ground.
 

crutschow

Joined Mar 14, 2008
34,285
It was NEVER a substitute for the breadboard or the bench.
Certainly.
A simulation is part of the design process, not to generate a final production design without breadboarding.
An exception was the low noise buck regulator I designed at work, which was so layout critical that my breadboard was the first PCB after the simulations and design.
I only had to make a few slight mods for the final PCB design.
 

Papabravo

Joined Feb 24, 2006
21,159
Certainly.
A simulation is part of the design process, not to generate a final production design without breadboarding.
An exception was the low noise buck regulator I designed at work, which was so layout critical that my breadboard was the first PCB after the simulations and design.
I only had to make a few slight mods for the final PCB design.
This brings up an interesting point. When I was a young engineer the cost of producing a PCB was quite large in relative terms (about 1/3rd of my yearly gross income) and required the dedicated labor of numerous professionals. This was ca. 1970. Now the cost of producing a PCB, is in relative terms, a fraction of the cost that it would have taken then, due to schematic capture, PCB layout software, and automated PCB fabrication. This means that today's breadboard could actually be the 1st PCB turn as @crutschow has indicated. In those bygone days, multiple PCB turns, despite the expense, was a common occurrence.
 
Top