Current Controlled Boost Converter

[beenai2018]

I'm working on this switching Step Up converter for an RGB LED.

Input = 5V,2A+ (smartphone wall wart)
Output = 7-15V Variable,300mA Max
RGB LED is a 10 W Module and takes 300mA max. There are 3 LED in Series for each color.Each color has a different forward voltage to reach 300mA.
Brightness control is a1 KHZ PWM.
This is for US home/hobby use.

Circuit
100KHz OSC,Fairly simple Boost supply.Has bout 60 ma Current ripple that remains constant through the  PWM range.

The LM339 on Left is a 100Khz Oscillator. The LM339 on the right is set as a Current Limit Trip,which starts the process by going high. This allows oscillator pulses to drive the MOSFET gate which boosts Vsupply  and turn on the LED.
When LED Current reaches 300mA ,the voltage across 0.6 Ohm current sampling resistor goes beyond the 0.18V to switch comparator low ,which turns off the MOSFET . The PWM input (Brightness) is ANDed with Oscillator . So ,Brightness control is via POV and Color Accuracy is maintained as LED Current is 0 or (close to) 300 mA.


I need comments/review on this design  . It passes my sanity check and LTSpice Sniff Tests ,but I might be missing something(OC protection to be added,for ex.)







P.S :
>This project is my first rodeo with LTSpice. Usually I Just "buy and build/design on paper".
>I know there are IC's that do this better,faster ,and cheaper (Supertex HV99XX/Clones etc.).
>I was initially sampling MOSFET Current in the Fast  loop (for I Limit  Trip) , but correlating it with LED current was not linear .I switched to slow loop sampling in the  Led which worked better.


Update - Hysteritic Mode in Post 4

[T3sl4co1l]

Oh no, not even voltage mode, but fixed PWM, gated!  Yikes!

Current mode is the only way to go, because inductor current is the critical process variable and any other method leaves it as an uncontrolled variable.  Out of control means, it only takes one circumstance that causes inductor current to rise without limit, and the transistor is toast, gone and done.

No, it may not matter for a little LED driver and a fixed supply, but it's bad practice and that's why I will always call it out.

Correct, input and output current are not quite proportional.  So, big whoop, just control the output current, dual op-amps are free.   That's what I did with this LED driver circuit:
https://www.eevblog.com/forum/projects/building-a-simple-switching-circuit/msg1252706/#msg1252706

Tim

[Siwastaja]

A simple hysteretic current mode is not only better, it's considerably simpler with fewer parts. You basically won't need much else than a comparator and a gate driver. Possibly a current shunt amplifier to reduce losses while keeping accuracy.

Measure inductor current -> too high, turn the FET off. Current gets lower again -> comparator hysteresis turns the FET back on again. No overcurrent, no inductor saturation, no blown fets, 100% guaranteed over the complete intended and unintended operating range as long as you keep thermal issues and layout (voltage spikes) under control.

LED current is controlled by adjusting the comparator setpoint. For very low brightness, you need to PWM an additional enable signal.

So just keep the current limit in the "fast loop" and get rid of the oscillator part completely.

[beenai2018]

Tim,
Don't want to do Voltage mode with LED's.

The 5V input was an intentional choice ,as I could leverage  the safety factors(such as current limit) built in to them. . Having said that, the input caps may have enough juice to fry the mosfet.

Your  flash light circuit is very cool . Are the Copper plated enclosure ,just to throw together a box quickly or ,is it just a emi shield ?

Im trying to keep the circuit as simple as possible (Since I have to build x3 for  the RGB Led).

Siwastaja,
I'm refining the circuit to go to hysteritic mode. I will update with pics shortly.Im not doing the current shunt amp .Hoping my 324 clones don't have too much of an offset.

[beenai2018]

Hysteritic mode Circuit and Avg / RMS Current Graphs

(Graphs are Log in X axis ,linear in Y )

[T3sl4co1l]

Nice

FYI, ILimit1 needs ESR so the positive feedback can work.

Tim