Spikes on output of buck regulator

Try messing around with your ground or +V wiring. I would hazard a guess the pulse voltage developed across the output capacitor's leads is causing your error, and the same reason filtering doesn't appear to change anything.

Tim

-- Deep Fryer: a very philosophical monk. Website:

its very hard to measure this correctly...

for example, try connecting your scope probe and the ground clip both to the same ground point. you probably still see the spikes,,,,if so they are being injected common mode into your scope input..

there are special ways to connect the scope to the circuit to reduce this... instead of a scope probe, use a 50 Ohm coax cable, and make a small Rc filter.. and use a ferrite choke over the cable .... there are app notes about probably listed under Ripple measurment"

it also helps to use a magnetically shielded coil for the switcher......such as a pot core or a torroid,,,the open ended solinoid coils radiate magnetic field all over the place...

I cold type a book about this but I won't...

Mark

0.5v is a lot, although you need low ESR caps on the output to get rid switching ripple, maybe hard to get it low with perfboard, depends how youve wired it.

Colin =^.^=

I can imagine two sources of spikes at the output. One is caused by the parallel capacitance of the inductor, that couples the sharp voltage swing at the input side of the inductor to the output side. This effect is aggravated by the ESR and ESL of the output filter capacitor. I would try paralleling the 470 uF output capacitor with 1 to 10 uF ceramic surface mount (X5R or X7R).

The second cause in not having a single point that acts as star ground for the input capacitor, the output capacitor and the diode. Even a few millimeters extra distance between these 3 components will add inductive drop in proportion to the high di/dt through these three parts. The entire layout should be built around this single node.

Yes, the spike you measure depends on the bandwidth of the scope. With sufficient bandwidth you may be able to see spikes as narrow as 1.5ns corresponding to frequency components of above 300MHz. Just an LC won't be able to cope with them. The whole design actually has to take these frequency components into account.

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net

I've read some good things to try and I will try those. I've just thought of something else:

The regulator and diode are TO-220, and each have heatsinks. The tab on the regulator is connected to GND, and the tab on the diode is connected to the cathode (same as the regulator switch output). I haven't tried removing the diode heatsink yet, but could this be causing problems? Maybe I need to get a diode with the tab connected to the anode.....

You may also want to put an LC filter at the input. Consider this over simplified bucker:

L1 / L2 L3 ----)))))---+---/ O----+----))))----+----))))----- Vout ! ! ! --- --- --- C1 --- ^ --- C2 ! ! ! L4 ----------+------------+----))))--- GND

L2 is the working inductor.

L1 is the impedance between the power source and the switcher. It needs to be larger than the impedance of C1 at all frequencies so that the input ripple current flows through C1. The inductance of L1 is small, its SRF needs to be high. L1 should be lossy. Somwhere to the left of what I drew there may be another capacitor.

C1 needs to be a very low impedance at high frequencies so it usually is made from more than one capacitor. There should be a smallish ceramic in parallel here.

L3 is the output side impedance. It is more than the impedance of C2 etc. It may receive quite large short voltage spikes but other than that it is mostly dealing with DC.

L4 is the impedance between the ground of the switcher circuit and the ground of the rest of the system. It is not likely to be a real component. In theory, there is only DC in this leg because of L1 and L2. In truth, stray capacitances from the switcher parts to the system ground can cause currents here. If L4 is a real part, it must be very low valued and very lossy. You don't want the stray capacitance and L4 to resonate at all.

--
--
kensmith@rahul.net   forging knowledge

Perf board is not a suitable foundation for any switcher. Get something with a solid ground plane and make sure the caps are tied to that via the shortest possible path. The top lead should also be less than 1/10" with the oputput line entering from one side and leaving via the other. Keep all leads close to the ground plane. Provide a ceramic chip cap at the output.

--
Regards, Joerg

http://www.analogconsultants.com