Switching noise with LM2676-5.0!

[shadewind]

I just got my Rigol DS1052E so now I finally have an oscilloscope taking out some of the guesswork. I put it to test measuring the output of a little switching regulator I have on a breadboard, an LM2676-5.0. This is what it looks like:



The upper graph is the AC coupled output voltage and the lower is the unfiltered switching output. The current drawn is about 500 mA.

As can be seen from the pictures, this looks absolutely horrible with somewhere around 1V ripple voltage. The pulses seem to occur at the switching flanks. I've checked the data sheet for the LM2676 (http://www.national.com/ds/LM/LM2676.pdf) and on page 8, there's a graph of the typical output voltage which, of course, is miles better than what I have.

The output and input capacitors, from what I can gather, make a big difference in the performance of the circuit and this is the one I use for both the input and output:
http://se.farnell.com/jsp/search/productdetail.jsp?SKU=1219474

Looking at the data sheet, the ESR (I assume that is the impedance listed in the table) for the specified capacitor case (8 x 20 mm) is 0.030 ? which really isn't much higher than the 0.026 ? listed in the graph in the LM2676 datasheet mentioned above. From what I understand from the datasheet, the ripple voltage is the product between the ripple current and the ESR so a 0.004 ? difference shouldn't really make that much of a difference. Besides, using two caps in parallel (which reduces the ESR, right?) makes no difference either.

Could it simply be that I'm using a breadboard? Or is it a measuring error? Am I using the scope the wrong way? The weird thing is that the switcher supplies my little AVR circuit + LEDs and there's nothing wrong with anything... everything works even though the ripple voltage is (or at least seems to be) very large.

[Simon]

I think digital circuitry is less bothered about ripple unless of course it is so bad it causes confusion between lows and highs, computer SMPS are know to not be top quality but power very highspeed stuff, granted that there are ever more local regs on motherboards.

The bread board certainly won't help I guess, you have long connector leads and the part leads are also usually longer. Maybe try shortening everything ? of put it all on a veryboard but not the strip type (or cut the strips neatly). I've made 2 switchers on vero board with no issues, granted I have notr actually measured the output though

[hans]

"I put it to test measuring the output of a little switching regulator I have on a breadboard"

There is your problem. Breadboards are horrible for high frequency stuff. A LM2676 switches at 260kHz, which isn't one of the fastest, but you'll probably start to notice how bad breadboards can become.

For a good layout of a switching regulator, it's very important to make sure the path with fast alternating, high switched current are proper. Make sure you place the diode and the coil as close to the regulator as you can. Place capacitors as a second priority (well , except 100nF's ofcourse), and then start to lay around your feedback lines and in and output.

Make sure you don't make feedback lines cross straight through the high current zones. But because you're not designing a PCB but on a breadboard, that won't be that big of an issue I guess.