Any suggestions on taming the spikes from LM2596

[Rick Law]

I can use some advice...

I've been using an LM2956 buck converter for a while and got comfortable enough with it that I forgot an issue it has: Spikes when under heavier load.  Now I need to use it for a slightly heavier load (2-3A), I have to face the regular spikes (100KHz ish, +- 100mV to +- 150mV).

A low pass filter R+C should in theory take care of it.  But putting 2 to 3A pass a 10ohm will drop the volt to nothing.  Putting a large cap (1000uF) across it (V+ to ground) alone (without a resistor) did delay the occurrence of the spikes, but when I am drawing 2-3A at about 7.5V, it is basically useless.

Any suggestions on taming the spikes?

Thanks
Rick

[T3sl4co1l]

What kind of spikes?  Input or output side?

Switching noise is easily handled with CLC filters, use low-ESR parts (like big ceramics) if possible.

Tim

[mazurov]

Broadband spikes (on the order of 100MHz, generated during switching "instants", sitting on the top of each output ripple extreme, see Linear Tech. AN70  ) will couple through capacitance of parts, including interwinding capacitance of inductors. For small output currents ferrite beads work well; for larger currents you may want to thread output wire through something like clip-on ferrite cut from USB cable.

An output ripple which increases rapidly while the load is increases gradually may indicate inductor saturation. If inductor is not saturated at max.load but you still don't like the ripple - use physically larger inductor with larger inductance.

[German_EE]

There are many things that you can do to eliminate the voltage spikes but it may mean a board redesign:

1) If possible use a ground plane design around the regulator area, if you cannot do this then adopt a 'star ground' approach.

2) Leads should be as short as possible for the following components:

Input capacitor connected to pin 1
Ground on pin 3
On/Off connected to pin 5
Schottky diode across pin 2 and ground
Output capacitor after inductor

3) Use low ESR capacitors on the input and output.

4) Keep the pin 4 feedback signal away from the inductor magnetic field.

5) Make sure that you are using the correct value of inductor for the voltage and current demanded.

Some experimentation may be required.

[mzzj]

How is your layout look like and what other components you are using?

Or does this happen to be one of those ready made $2 ebay dc-dc converters?

[Rick Law]

Took me a few days to make the time to look into it in detail.

It is indeed a "cheap board", $10 cc/cv variety, $6-ish now, certainly not high quality stuff.  I am experimenting with digital control (which works well) to just play around - not anything serious to use but just to see what I think I know enough to make functional (is it so or not) and what I can learn from the exercise.

Finding the noise (which I knew but forgot), I connect up my other board (exact same board but unmodified) to get a base line to see if what I introduced made matters worst - yes slightly.  The unmodified one show that same noise at perhaps 10-20% lower noise peaks.  The trace below are made with an unmodified board.

The two attached photo are the noise (green trace on the posted pictures).  The first picture is the "fast" shot (scope too slow  to resolve the detail), the second photo is a zoom in by lowering the time/division.  The noise frequency looks like it is around 100-200KHz.

Looking at the noise closely,  I realize it may be the capacitor being drawn down, then for the rest of the switching cycle, it set at much lower level.  I added a 1000uF output cap.  It did a lot of good (say at 6V 100mA range), until the load increase to a point when the 1000uF got drawn down too.  By about 1A, the trace collapsed into a clipped triangle.  Just to validate, I upped the cap to a 4700uF, it delayed  the start of the collapse to a higher current.  Interestingly, when the power-draw is very high, the noise will go back down.  I suppose it is at a point where the charge is insignificant - which lead me to think perhaps I should lower the caps on the board.

I am not looking for a magic bullet.  I just like to hear some techniques so I can do more research into them, see how they work and see what I can learn.

[T3sl4co1l]

That looks perfectly reasonable for supply ripple from single stage filtering.  If you need a low noise supply, add some CLC filtering, maybe some common mode (ferrite beads or chokes) too.

Tim

[Rick Law]

That looks perfectly reasonable for supply ripple from single stage filtering.  If you need a low noise supply, add some CLC filtering, maybe some common mode (ferrite beads or chokes) too.

Tim

Thanks, Tim - now I have something to look up and read: CLC filtering next step, see if I can learn  something...

Rick