LED boost converter - final design

[Miyuki]

Hi I'm building small boost converter to power LED from 12V to 33-40V 2A

It basically works but at turn off time there is one short turn on and small oscillator cycle
What could cause it ? Any layout issue with induced currents to oscillator ?
Did you ever see it ?

Scope screens: yellow channel is Gate voltage, green channel is Output voltage / Oscillator voltage

I can clearly see that oscillator start new cycle but then end it soon and continue to discharge WHY ? 

[batteksystem]

you mean 33-40V?

[Miyuki]

you mean 33-40V?

yes, edited

[Siwastaja]

Switch mode supplies can be tricky, a PCB layout problem is a real possibility here. Can you post the layout image, or at least a high-res photo of the converter? A schematic would be nice to look at, too.

[bktemp]

Looks like subharmonic oscillations. When using >50% duty cycle this is a common problem. You need to add a slope compensation.

[Siwastaja]

Definitely use current-mode controller for a LED driver. Voltage feedback loop should be absolutely out of question. Current mode makes feedback easier and removes any need for any compensation and is the silver bullet for any control loop instabilities. Then you only need to think about layout and EMI related things like false triggering.

[Miyuki]

It have current feedback, it is current source to 2A 

But it not reach this current because of this switch glitch kill efficiency

[Miyuki]

Here is layout and schematics

Just oscillator capacitor have longer leads and are used two parallel because it was tricky to get desired frequency

PAD3 and PAD4 are boost coil

[Miyuki]

Solved by adding decoupling capacitors around UC3843 Ucc and Uref, there was spikes which disrupt oscillator

Now switching "nice and clean"

Green channel Input current 10A/div
Yellow channel Output current 4A/div

Just need bigger transistor I had bad calculated power dissipation

[T3sl4co1l]

Inductance is way too high.

What's the LM358 for? It looks like you're force-driving the internal error amp, but then doing exactly the same thing that the internal error amp does: negative feedback with a 2.5V offset!

Ed: well, it's probably not 2.5V, given that you didn't specify resistor values....... it's also a current feedback circuit, I guess.  In that case, an amp is still helpful, but it's probably better to simply amplify the current shunt signal, rather than force the internal amp.

The transistor has a considerable amount of stray inductance around it.  There should at least be an RCD clamp snubber across drain-source, for example like this:



(the 10n, 15 ohm, and BYV diode; best suited values will differ)

But it would be very much better to change the layout to fix the stray inductance.  Adding ground plane is a must.

Tim

[Siwastaja]

Yes, the layout is wrong on so many levels that you really should try to redraw it, thinking about current flow from the supply capacitors, through the parasitic inductances.

[Miyuki]

It was just a testing if I can drive this LED because it is powered from 200Hz PWM remote controlled source and layout was to be easy hand draw

I must use bigger transistor package and higher voltage ratting diode in final version
Boost inductor is calculated to 13uH

I agree layout is bad and wire wound shunt resistors don't help it

But biggest problem is to choose inductor core to not make that 200Hz noise

To use shunt amplifier and smaller value resistors is good idea

//edit:
Do you thing that Kool Mu / Sendust will solve audible noise ?

[T3sl4co1l]

200Hz???

Can you put component values in the schematic?

Tim

[Miyuki]

200Hz???

Can you put component values in the schematic?

Tim

This circuit works at 230kHz but is powered from remote controlled supply designed for common led strip 12V and have dimming capability > 200Hz PWM

[Miyuki]

Hi
I had some time to make final design of this boost converter

Replace transistor with TO220 package to keep temperature and other semiconductors change to surface mount, replace choke core from iron powder to sendust to make it silent and replace sense resistors with much smaller with amplifier
Thru hole capacitors and resistors I use because want to keep board single sided and I have them in stock

What do you think about this layout ?