SMPS Ringing Help

[mgbolts]

Hi There,

I have been building a bunch of different SMPS for a few months and are starting to get some OK results for my needs.  Still a long way to go, its certainly a whole bunch of electronics knowledge to learn for just one circuit!

My latest project is to build a ~0.75amp super small 5V reg.  I am pretty happy with the result but.....  There is minimal ripple a lot of ringing on the output.

I have tried various comp values will little effect.  Also tried more output capacitance, little improvement.  The biggest improvement came from changing the diode from a cheap Chinese ss34 to an ON-Semi number. 

Facts:

MP4459 Reg
Input Caps = 1000pF, 0.1uF and 10uF
Output Caps = 1000pF, 0.1uF and 20uF
MBR230LSFT1G 2A 30V diode
4.7uH 1.9A Inductor

The chip is running at ~1.8mhz with the ringing in the high 10mhz (+50mhz)

Whats annoying me is that the power appears clean but for the ringing....I was hoping to get it down to sub 50 peak to peak.  The power supply is 15v and output is 5V. I am feeding from a lipo to ensure clean supply.  The probe is using sprung wire for the ground ......  I am also running AC coupling and 20M BW limit.

Any ideas appreciated.

Thanks
mgb

[mgbolts]

Some quick labels on the layout....

[StillTrying]

"running AC coupling and 20M BW limit."

I'm surprised you're seeing 90 MHz ringing.

[mgbolts]

A good point, I will rerun and take a pic....

[mgbolts]

I had another look.  Definitely set to 20M BW limit.  I have the scope running at 10ns per division with 20M BW limitin the attached pic  The sinusoidal ring has a period just under 1 division which calcs to a frequency of ~80-100 mhz...

Also,  pic of the probe setup..

[exe]

I'd suggest start with the layout proposed in the datasheet. Basically, the idea is to minimize "loop area" where high current/voltage is switched. So, on your layout the diode should be much closer to the coil and ground polygon (see suggested layout in datasheet). There is plenty of documents in Google about this ("smps noise"). Perhaps, not a bad start would be this article: https://www.eetimes.com/document.asp?doc_id=1279232 . There is also "PCB LAYOUT GUIDE" in the datasheet.

BTW, I don't think compensation will help here, it does a different thing. I think there are only two ways to reduce the noise of a buck smps: reduce parasitics, and/or reduce switching speed. Hard switching edges will produce more noise.

Just my two cents, never built an SMPS myself.

[T3sl4co1l]

The layout is actually really good, assuming you don't have anything else tied into it (like an enable wire).  I would like the diode closer to the regulator, like wedged inbetween the reg and boot cap, but it's really not terrible where it is.  (It could actually affect compensation, in that the compensation network is laying over the regulator-diode switching loop, which makes it susceptible to its own noise.  Hopefully, it's made to deal with this, and will be okay.)

What's good about it?  You have the input and output at the same place, with caps stacked up around ground.  This shunts as much high frequency content as possible, to ground, at a single point, so that there is little difference in the RF voltages at the three pins.

There will still be some RF leftover, which I guess is what you're seeing.  Maybe it's quite a lot after all, since it's getting through the 20MHz BW limit.

The simplest solution is to follow up the input and output with an LC filter.  Not much will be needed, maybe 1uH and 10uF (ceramic)?

Tim

[exe]

BTW, what is the load? I'd suggest try with a purely resistive one for tests. With minimal wire length.

[mgbolts]

Thanks T3sl4co1l and exe.  I appreciate the input.  The reg will have a typical load of around 0.7amps at 5v although its designed to take up to 1.5amps with active cooling. 

My other switching reg designs have always put the reg to the left, inductor upper right and diode lower right..... not that different to the datasheet incidentally!  I did not have the same ringing issues with this approach. 

Given I got an improvement by 'upgrading' the diode, it does seem to be indicator of something going on with the diode. I will move it and report back in a couple of weeks.

BTW, part of the application is for RF TX/RX so its good hear your comments on RF characteristics of the caps.....

Cheers

[mgbolts]

Seems I misread your question.  It was tested under a load of 0.5amps with an electronic load, the ZKE Tech EBD-A20H. 

Have to say I love this thing, works a treat, great build quality compared to other cheap loads out there.  I am surprised that it has not been given more attention.  I think I paid around US$65 on taoboa

http://www.zketech.com/pd.jsp?id=18