Group project, to create a file of equations, rules of thumb and handy approximations

It's all relative Jim. I can design analog widgets too, though certainly not as well as yourself. I can likely run circles around you in a complex digital environment too. Digital isn't all about connecting two AND gates together. ..gotta make 'em work too. ;-)

Hey! Chimps are _NOT_ Monkeys! Chimps are Apes - I guess just Good Apes, because the Gorillas are the Great Apes. Humans are just the frail, hairless pink apes who don't even deserve to live in the same forest.

Cheers! Bobo

. We all must have helpful equations or rules of thumb that could be shared in this NG for the benefit

Ok, let's say you have a DC switching power supply with it's normal negative input impedance that requires an LC EMI input filter. Given: Ein = lowest input operational voltage, Pin = max input power, Fc= 2 pole filter frequency corner to meet your EMI needs. Then: C= Pin/(2*pi*Fc*Ein^2) and L= C*Ein^4/Pin^2 Now that's money, Harry

I have my own monkeyless forest, thank you.

I wish micro*hi* would get you (or somebody) to fix their SPI module. SPI has to be one of the simplest non-trivial digital circuits, one would suppose...

http://ww1.micro*hi*.com/downloads/en/DeviceDoc/80131d.pdf (page 2, item 2)

Now /SS (when enabled) should be level triggered, you'd think, and grounding it should work, but it seems it makes a difference whether you ground the CMOS input through a resistor (works) or not (doesn't work). But in the latter case if you drive the SCK so the protection diodes get a taste of current* instead of with a clean 5V square wave it sort-of works, even though SCK is on a different port from /SS. 8-( Perhaps some kind of dubious edge-detector with an internal floating node? Nice 'feature' anyhow.

• eg. by touching the input so it picks up some 60Hz hum

Best regards, Spehro Pefhany

I didn't have anything to do with the I/O.

But thanks for that pointer... I have a consumer product that I own that does funny lock-ups... I bet it has one of those processors in it.

I'm about to sue over it, since I have $thousands into this ;-)

...Jim Thompson

I don't get it. The LC will oscillate if it sees a net negative impedance at its resonant frequency Fr [1]. If the switcher loop does provide -Z at Fr, it oscillates, and whether it does depends on the loop dynamics, which don't appear in your equation.

John

The input -Zs of the switcher input is in parallel with the output +Zf of the filter. The sign of the Z (net Z) when connected goes with the lowest value so |Zf|

its right. but you still need to be wary of resonance in the filter itself, that can easily push |Zf| up over |-Zin|

Cheers Terry

Not really. Any LC goes to infinite Z at its parallel resonant frequency. So if the switcher Zin is negative at that frequency, the input filter oscillates. So the thing to do is make sure the LC resonant freq is above the switcher's -Z max frequency, which in turn depends on the switcher loop dynamics.

I think.

John

Now I get it! John L. and Terry Q. are correct. That is why you must De-Q your filters. The input filter we designed will be resonate at it's corner frequency causing it's output |+Z| to soar greater than the switcher input |-Z| and cause oscillation because the switcher's loop gain is >1 at Fc. So we must add a series Cs + Rs in parallel with C such that: Cs>3C and Rs =sqrt(L/C). Also note that the input leads my be a few ohms (50uH LISN) so that also must be included in your SPICE simulation of the complete circuit but the equations given are a good starting point. So Terry, have you ever designed a 1KW DC/DC converter with 28VDC input? Low line is 20VDC so -Rin

In article , John Larkin wrote: [...]

..or otherwise ensure that the input to the switcher has a positive resistance at the resonant frequency. You can do this by injecting some of the input voltage into the control loop. You use some regulation in the process.

yep. Its easy to plot the filter impedance in SPICE, and the smps input impedance can be calculated (and even simulated), but in practice using |-Zin_min| and ensuring the filter resonant peak is well below it is good enough.

yes.

you can even do decent models of inductor saturation, as well as including parasitic capacitance and skin effect in the spice models. Its not far from there to a first-pass at EMC by including a LISN model. Differential-mode EMI is actually pretty easy to model, its CM that is hard.

I designed the transformer for a 2.5kW, 16Vdc converter. Its the coolest thing I've ever done, literally. Its also tiny, and very cheap. The LVDC FETs took up twice as much surface area and including heatsink clips (but not the heatsink) were nearly as high as the transformer, which is

In practice as I gets very high, filter inductance tends to plummet as I^2R gets out of control, greatly limiting the number of turns practicable. Ripple current and lifetime requirements usually necessitate large amounts of capacitance, so the filters tend towards low-L, high C = low impedance. Reliably damping those can be tricky, especially over a wide temperature range, but it can be done.

High currents tends to lend itself towards higher-order filters, as minimising inductance automatically minimises I^2R losses.

Cheers Terry

I'm guessing that most switchers have a goodly distance between their switching frequency and the highest freq that has a negative input impedance, so it's usually possible to use a simple LC input filter whose resonant point is somewhere between.

And besides, it's usually spikes that need filtering, more than the fundamental.

Has anybody built a switcher with an input filter that did actually oscillate?

John

it varies widely with topology and control mode. CM control gives pure negative impedances up to at least 5% of Fsw, and depending on topology the input impedance can be pure negative resistance (if CM buck, CM buck-boost). These are the ones to watch.

Voltage-Mode control of some topologies gives totally different results. Dan Mitchell's Mathcad smps book has some pretty pictures for various control regimes.

Ive built one with an input filter that rang like a son-of-a-bitch, far worse than the damping factor I expected. But it didnt take long to find out what was going on.

I suppose I could modify one to oscillate.

Cheers Terry

Not even close. But then my forearms look like Popeye's.

Not quite.

For a simple RC low-pass filter:

-3dB frequency (Hz) = 0.16 / RC

Mark

And if it's critically damped 2nd order, or Bessel, bw * Tr = 0.35.

John

Four foot eighteen?