choosing a power supply

If you mix two noisy signals whose spectra are bunched towards the high end, the heterodyne energy will be wideband, and the switcher LC filter will kill most of it.

I have six switchers near one another on a board, one Cuk and 5 of the little TI things, nothing synchronized, and everything is nice and quiet.

Given that waveform above, I was surprised how little noise I saw on the switcher outputs.

Even synchronized switchers can heterodyne. At zero frequency, which results in DC offsets.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

You synchronize switchers, and your stuff works. I don't, and my stuff works.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Now put an AM receiver next to it.

Sure, but you don't need the heterodyning to have that problem. It's there with just one.

Reply to
krw

But you don't know that the millionth one will still work, five years from now.

Reply to
krw

Does your synchronizer clock have an MTBF over 44 billion hours?

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

The Cuk is not spread-spectrum, so that will make a big spectral line. The little TI switchers will each make uncorrelated low-level wideband noise. Synchronized switchers would make one giant spectral line.

That board has two XOs, two FPGAs, two frequency synthesizers, one 600 MHz ceramic resonator oscillator, and five vicious-swing output driver stages. The switchers are minor EMI issues.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

They're not uncorrelated. They're not random, either.

Irrelevant. Everything else is a known. Synchronized power supplies are knowns, also. Unsynchronized and particularly SS power supplies are unknown, over lots and time.

Reply to
krw

A million of them do will have 44 billion hours.

Reply to
krw

The TI chips have some sort of spectrum spreaders inside. It seems to be shaped noise, not some simple triangle sweep or anything easy like that. I wonder how they do it. Pseudo-random shift registers or something?

How could the spread-spectrum modulation of different chips be correlated?

TI probably knows what they are doing when they design and fab chips. Not very "unknown." The TPS54302 is a great little chip. If you demand synchronization, of course you can't use it.

I've never seen a problem using unsynchronized switchers. Have you? What was the problem?

Where do you get your synchronization clock? Do all your switchers run at exactly the same frequency, or at ratios?

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Where does the clock come from?

What would happen if the sync clock failed?

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

It's usually just an FM signal. The modulation isn't random at all. We don't want to see that signal (re/demodulated by the power supply next to it, in a band we care about.

Yes, and they make sure they have a way of disabling SS because they know there is an important segment of their customers who don't want it and will reject a part if it can't be disabled. They actually do ask their customers what features are wanted before designing their parts. We have a TI power/analog guy assigned 100% to us. Collecting information is part of his job. He has the developers in for visits to hear first-hand what we want and don't want. Synchronization - yes. Spread-spectrum - no.

Yes. Birdies in receivers. We also modulate switchers so their harmonics aren't where the receiver is tuning.

Generally from microprocessor PWMs. All the same frequency, if possible. Harmonics, if not. Phase? Varies by design. Sometimes all the same phase, others spread it out. It's all done by UC timer/counters, so it's pretty easy to tune for minimum EMI (where it matters).

Reply to
krw

Microcontroller crystal =? Osc => TimerCounter => UC GPIO => PS.

Without the microcontroller, the widget doesn't work anyway.

Reply to
krw

It's not anything simple, or it would show up in the spectrum, or as ripple on the output. The spectrum is smooth and the output ripple is below my measurement floor; the switching frequency is the big ripple, and that's not very big. TI is surely capable of putting a serious spectrum-shaped random noise generator in a chip.

Sounds like some other TI people had different design goals.

Oh, you are doing RF. I work in time domain. Power supply ripple and switcher spikes translate into picosecond jitter in my world. Heterodyning would be a hazard to us if it caused low-frequency ripple on supplies. I've never had that problem.

I have had jitter from switching spikes crawling all over the place. I like the tiny TI part because it's very clean, low shoot-through spikes, and it's small enough that I can keep the critical current loops very local.

So until the uP starts up and makes the sync clock, the supplies run unsynchronized. That's OK, as long as they actually work that way.

Why not use spread-spectrum? It's there specifically to smear out the spectral energy to pass EMI tests.

Here's a home-brew SS switcher:

formatting link

formatting link

I added a triangle current into the comp pin of an ordinary switcher chip. The spectral peak dropped almost 20 dB.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Maybe you need better instrumentation. It's not below ours and it's certainly not below the floor of the radio sitting next to it on the same board. The fact that it isn't predictable is the other strike. It can't be tested.

Perhaps but on the parts they want to sell to us (a major customer), it can be disabled.

That you know of. We certainly have, so avoid it like the plague. Worse than having the problem is having it in the field and not knowing it. It can't be tested so it's a ticking bomb.

It does. They do. The same frequency resistor (usually) that you use to set the part is active when there is no clock. The external clock overrides the internal clock.

Because it causes problems. There are other ways to solve EMI problems. That's not usually a problem for us.

ICK! No thanks! There aren't that many birdies in an aviary.

Reply to
krw

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.