Constant current boost mode circuit works with ground plane, but not without?

[SittingBear]

To give you insight this is a constant current boost mode driver for LEDs, original design taken from Texas Instruments here. Chip specifications can be found here.

The issue: The ground planed etched circuit works without error. It drives my test string of LEDs at 300mA (set by resistor) at around 36V from 24V input. The non-ground planed circuits, there's 4 of them total, all produce a quick and dim 'blinking' of the LEDs without ever keeping them on. I somehow doubt all of my chips could be bad. I have a feeling there's something I must have missed in soldering the circuit or misplacing something, or that it could be the removal of the ground plane.


Any help is greatly appreciated!

[nctnico]

Both circuits are terrible. The inductor and the diode must be as close as possible (and even closer) to the switcher chip.

[SittingBear]

Both circuits are terrible. The inductor and the diode must be as close as possible (and even closer) to the switcher chip.

Why would the diode and inductor need to be close to the chip? It's not using an internal switch to sink the inductor to ground - that'd be the big TO220 MOSFET you see doing that. I realize these circuits may look ugly, but it's homebrew prototyping and doesn't need to look pretty at all. Keep in mind these pictures are close-ups and may give a false sense of large distance from object to object.

[lapm]

Os someone already pointed: inductor and diode as close to chip as bossible.

Also im not suprised it docent work without ground plane. your ground creeps around like snake and that's usually bad idea when dealing with fast pulses like in switching power supply.

Also might want to wider a bit traces that carry hefty current.

Personally i would go with two sided pcb, dedicate on side for ground only.

[SeanB]

Think of each long trace as being an extra unwanted inductor, and then consider the long ground trace as being both an inductor and a resistor in series with the power line. With the high current pulses from the switch being applied to the ground pin of the switching IC no wonder it turns off, as every time it turns on the transistor you effectively disconnect the ground lead from the chip. Then every time you have a pulse you have to wait for the chip to do an internal UV lockout reset and then do a single pulse again which turns it off again.

Normally the chips blow up if used like that.

[Alex Eisenhut]

Think of each long trace as being an extra unwanted inductor, and then consider the long ground trace as being both an inductor and a resistor in series with the power line. With the high current pulses from the switch being applied to the ground pin of the switching IC no wonder it turns off, as every time it turns on the transistor you effectively disconnect the ground lead from the chip. Then every time you have a pulse you have to wait for the chip to do an internal UV lockout reset and then do a single pulse again which turns it off again.

Normally the chips blow up if used like that.

Like the kids say: "this". I more surprised he got even one board to work.

[Christopher]

If i remember I'll post a 12->80V boost PCB i designed a few months back with this IC. Large copper pours for the input, output and ground plane on the rear.

Think about your two current loops, when the switch is ON and when it is OFF. Think about your feedback back into the chip & keep all of this really tight.

[SittingBear]

Think of each long trace as being an extra unwanted inductor, and then consider the long ground trace as being both an inductor and a resistor in series with the power line. With the high current pulses from the switch being applied to the ground pin of the switching IC no wonder it turns off, as every time it turns on the transistor you effectively disconnect the ground lead from the chip. Then every time you have a pulse you have to wait for the chip to do an internal UV lockout reset and then do a single pulse again which turns it off again.

Normally the chips blow up if used like that.

Thank you so much! Definitely learned a big no-no from this mistake. I shall try it out with a ground plane and re-do the design a bit to fit vital components as close as possible to each other. My biggest limitation is that I can only do one-sided homebrew PCBs which greatly affects the layout of the circuit. I try to limit the jumpers which are necessary for some connections. I am also still learning the dos and don'ts of the PCB world.

I may or may not update the thread with the results of the newer circuit with the ground plane.

If i remember I'll post a 12->80V boost PCB i designed a few months back with this IC. Large copper pours for the input, output and ground plane on the rear.

Think about your two current loops, when the switch is ON and when it is OFF. Think about your feedback back into the chip & keep all of this really tight.

I keep the feedback, mainly the voltage reference comparator tied to the low-value current set resistor as close as humanly possible to the chip's pin with a thick trace to avoid any unwanted voltage drops. I can't do double-sided boards very well at all, though I guess if it is a big improvement I can try also creating a massive ground plane on the bottom copper side.

[T3sl4co1l]

The un-filled version might be acceptable with bottom side (contiguous) ground, but yeah.  The filled version is bad enough with no stitching whatsoever!

Tim

[nctnico]

I can't do double-sided boards very well at all, though I guess if it is a big improvement I can try also creating a massive ground plane on the bottom copper side.

You don't need thick traces for the feedback. The current through them is extremely small. Using double sided board is a good idea. When I make prototypes I mostly use double sided board where I keep one side as ground (using wires through a hole as vias). No hassle with getting the top and bottom side right but still a very solid ground plane.

[ludzinc]

<Best Hulk Voice>

Puny Ground!