Reducing SMPS noise

Exactly.

FBs do sometimes kill resonances in PCB traces and cables and such. They are cheap, so why not?

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

configuration, each step switched in or out by a SMT solid state relay. I t's essentially a decade box for a specific application. The relays are co ntrolled by a micro and all is powered by a boardmount enclosed CUI AC/DC s witching converter, supplying up to 400mA.

ee a lot of noise on the resistance output that can be from nothing other t han the SMPS. I'm assuming the power planes on the board are inducing all kinds of noise into the traces that connect the resistors and solid state r elays.

space required is a big problem. So I'm looking at filtering the output o f the SMPS. I'm a digital and DC analog guy, so this switching noise scare s me. Some research and an educated guess tells me to try an LC filter on the SMPS output. I'm looking for any first hand experience with this and d umbed-down advice that will hopefully get me a solution without days of hea daches.

ope confirm that. I've already tried putting several MLCC 1uF caps across the SMPS output but that didn't change anything.

a

zip

Huh OK. I should just download a spec sheet...but... Are the beads toroidally wound like the through hole ones.. and do some they have multiple turns?

That would be awesome, filter inductorwise.

George H.

e

Few bead spec sheets mention inductance, although you can usually estimate it from the Z/F curves. The LT Spice models include the baseline L.

Most are some sort of planar ferrite brick, surface mount. I don't know what's inside. They cost a few cents, so they must be bulk manufactured.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Huh? The Murata BLMxxPG series is rated for power applications. Some of the 0603s are rated for up to 3A.

e:

te:

der configuration, each step switched in or out by a SMT solid state relay. It's essentially a decade box for a specific application. The relays are controlled by a micro and all is powered by a boardmount enclosed CUI AC/D C switching converter, supplying up to 400mA.

I see a lot of noise on the resistance output that can be from nothing othe r than the SMPS. I'm assuming the power planes on the board are inducing a ll kinds of noise into the traces that connect the resistors and solid stat e relays.

of space required is a big problem. So I'm looking at filtering the outpu t of the SMPS. I'm a digital and DC analog guy, so this switching noise sc ares me. Some research and an educated guess tells me to try an LC filter on the SMPS output. I'm looking for any first hand experience with this an d dumbed-down advice that will hopefully get me a solution without days of headaches.

scope confirm that. I've already tried putting several MLCC 1uF caps acro ss the SMPS output but that didn't change anything.

or a

s 05.zip

z?

Have you (or anyone here) ground one down and looked at the cross-section?

You could lay down stripes of metal in the ferrite, you'd then need some edge metalization to connect 'em.

George H.

Larkin once again affirming his design process involves neither theoretical design, nor quantitative testing...

See Fig.8:

It's rare to actually find such a graph in a datasheet. Most of the time, they're hidden away in the manufacturer's database tool (i.e. TDK SEAT

2013).

When's the last time you had a product pass emissions tests, first time in the lab, zero mods?

When's the last time you had a product tested in a lab?

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

Show me where it says those impedance curves are valid at Idc = 3.0A.

You can't.

Some of Murata's parts have DC ratings (or they did, back when their data was accessible in an offline application), but the saturation is dismal.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

They certainly work.

They work fine as ferrites (frequency dependent resistors, really). I certainly don't use them as inductors.

Tim, I know next to nothing about the details of magnetic materials. (It might be a fun field to study.) But obviously a low Q means energy loss somewhere. I've got two knobs hysteresis and saturation to play with, so I don't see how it can be anything but a crappy inductor. But good for the task, like high k ceramic caps.

George h.

I did reduce the EMI suceptability of a temperature controller by about 30 dB by adding beads. The thermocouple front-end was rectifying RF. The sensitivity was in narrow RF bands where things resonated, and the beads killed the Q. The product from the competitor, the previous English vendor, could be shut down from across the room with a GR signal generator and a banana-lead antenna. Plus, they had lost the source code and couldn't make any firmware mods.

I use theory when it is available and necessary. We design and sell a lot of electronics. Instinct and a little experimenting or simulation are really efficient ways to get creative things done fast... if your instincts are good enough. Tennis players don't do a lot of calculus: they just hit the ball.

And we do test everything. Hard. First-article, then production test. We put a lot of effort into automated testing, tester boards and test sets and Python. It's a heap of work, but it pays off.

How do you test your designs?

We took all our Varian/Agilent OEM products through full CE/UL/etc NRTL certifications. We didn't have any EMI problems. We did have to modify some ground lugs and wire colors on the AC inlet side, and change some fuse types. We had to add one internal fuse once, for a small auxiliary power transformer.

The VME aerospace and semiconductor fab gear doesn't need certifications.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

It might, but when you add turns you increase the self capacitance which can act to short the inductor at high frequencies. Also, at parallel resonance it will become a very high impedance.

--

Rick C

What do you get when you saturate a crappy inductor?

A crappy piece of wire. :)

Not kidding -- Laird has an unusually helpful catalog:

e.g. p.18 for the 1206's.

Since it's a two-terminal device, you can use Kramers?Kronig to solve for R

• jX.

It would be handy if they had plotted log(|Z|), so that a 1:1 (inductive or capacitive) slope is apparent. But oh well.

In any case, they all start at low |Z| and go up along a nice curve, roughly exponentially (as one should expect for a ~pure inductance, on a semilog plot). Then the curve flattens out (as R dominates), and eventually peaks (pure R), and then it falls off again (capacitive).

For example, the LI1206H151R-10 is rated 0.15 ohm DCR and 0.8A. The 0 and

100mA curves are pretty close together (e.g., 100 ohms vs. 92 ohms at 40MHz), which is encouraging. But it's already -3dB at 200mA.

And at ratings, it's down to maybe 22 ohms (40MHz), equivalent to about four inches of wire.

Clearly, you aren't getting the impedance ("150 ohm @ 100MHz") you thought you were paying for, if you use this anywhere near rated current!

The dropoff is considerably worse than X7R caps, with respect to voltage. Though if you want an equally grotesque capacitor, Z5U is certainly very available...

And LI1206H151R-10 is one of the best here. Check out the nearby plots. MI1206L501R-10 has the peak frequency rising considerably with bias, and the bandwidth of that peak, falling equally fast. (That's not entirely fair, because the zero-bias FWHM is 170MHz wide, versus maybe 160MHz at rated current. The bandwidth is constant, so the Q is rising proportionally. So much for a lossy inductor!) Look at how sensitive it is: half its value, at just 10% of ratings!

Indeed, the saturation depends more on part size, than current rating.

If you need a lossy inductor, you're /far, far better off/ constructing one yourself. Not only are you guaranteed the inductance you need, under any condition, but you can design the resistance to be precisely what you need!

Related: how much flux does it take to saturate a ferrite bead? Simply: Phi = L * I_sat In turn, for a square pulse, say: Phi = V_pk * t_pw

So one of these 1206s that saturates at ~200mA, and has an inductance of maybe 0.52uH (take |Z| at a very low frequency, where R is small and X is dominant -- it looks like an inductor down here!), does so with about 0.1uWb of flux, which is equivalent to, say, a 20ns * 5V pulse.

The reason Joerg is fond of using FBs for gate resistors, from time to time, is this: _they're cheap saturable reactors._ The gate driver chip has completed its output swing, just as the FB saturates; upon saturation, a huge gulp of current flows into the gate, causing its voltage to rise quickly; then as the voltage overshoots (only slightly, because saturated inductance is small -- very little energy is stored), it pops out of saturation and dampens the ringing. Meanwhile, the lossiness is back, preventing the transistor from making naughty oscillations.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

Several of those parts hold up well at 1 amp, and even 3 amps.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Sometimes. Power supplies are super slow compared with signal nodes, so you don't have to worry too much about the additional capacitance and length of trace (i.e. antenna) due to building up a discrete RLC network. If you're living in that world, great, but many of us aren't. For example, the Murata 0603 beads are super nice for stabilizing microwave transistors in baseband circuits, where they can do some amazing things.

Beads are also useful insurance against board turns--you can replace an unneeded bead with a zero ohm jumper just by changing the BOM, and of course those are always useful for debugging since you can isolate circuit blocks by removing them. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Really? I'd love to see the VNA plots.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

Yeah, but you go from "power supply" to "small signal" in one paragraph. You know what I'm talking about, Phil, and I know you know better than to equate them.

The whole point, that I've apparently been a complete and utter failure at communicating, over the last four posts, is that power supplies are "large signal" and small signals aren't!

People can't even read, and this is a text-only forum... sigh.

Let's just put in extra tee and pi networks of DNPs and 0-ohms on every trace ever. You'll never have to rev the board again, hell, you can use the same board for a hundred other designs just with assembly variants. ;)

I try to avoid jumpers and terminators, if I can. It clutters the design, which causes real delays in layout, for customers who request 100% nodal coverage. Which is often. They have their place, but not every friggin' pin needs it. Use them prudently, not pervasively.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

Sure thing--I was agreeing with you about power supply filters, where you do have the luxury of slowness.

Sure. I usually put a space for one in the base of every transistor whose f_T > 2 GHz, and add places for extra filter components at the power entry, even though I may not need them.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

I just did a diff pair of BFQ18's driving a diff pair of BFQ149's driving, well, more stuff. I was shocked that everything just worked... no oscillations, no weirdness.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Interesting. Why cascade two pairs of 1-W-class transistors?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net