To choke or not to choke?

is

It's not always "either L or RC" but often all three elements are combined to reach the desired end. Economics is a huge factor. Many of the modern L's are not recognizable as coils or toroids but are simply surface mount boxes or may be integral to or soldered under the connector housing.

And what used to be masses of discrete wires terminating in L/RC/LRC filters is now usually done through a few-wire serial bus that probably does have L's that you didn't recognize as such. e.g. USB jacks with ferrite beads and R's and C's on the PCB. Every USB chip vendor will show you sample schematics of LRC filtering.

And every high-speed serial bus used today has drivers where the transmitter does slew rate limiting to minimize and in some cases eliminate needed filtering. Actually this has been around since at least RS-232 and maybe before. Jim, want to talk about the MC1488? :-)

Tim.

red

oth

is

this

Surface-mount ferrite chips are nice. They are non-wound inductors, so the parallel capacitance is around 0.3pF. They are hardly ever listed by inductance - what you get is offered is an impedance at 100MHz, which often goes up to 1k - 1.6uH - and sometimes up to 2 or 3k. They are lossy at 100MHz, but work as perfectly respectable small inductors a lower frequencies.

-- Bill Sloman, Nijmegen

Chokes can be dangerous if the inductance is larger, because of inductive kickback. The uC is usually going to be ok because it has subtrate diodes but the stuff on the other side may not.

RC is just fine when the layout is clean. For super-sensitive apps (and that's hardly a uC) the 3-terminal EMI filters from companies such as Murata are nice because they efficiently suppress cell phone signal. But I have needed those only in analog settings.

The other designer knows SOME of the story but lacks some of the history. They simply allow plenty of current through. They were very popular for a while for power rails. Also they let fast signals through which 1kohm + 1nF will not.

Unfortunately they tend to resonate too. As the EMC standards increased the upper frequency you had to test to, and as mobile phones became more common, first at 800MHz then 1.6GHz then higher again, they became more of a problem than a help - they actually amplified noise into sensitive circuitry. So they fell out of favour.

The manufacturers caught up with the frequency requirements (1GHz) but now the standards bodies test to 2GHz or more... the problems are not apparent until you do a full EMC sweep over hundreds of MHz, so some people think they are great.

The key parameter when choosing whether to use them is the speed of the signal going through them. If you can use a 1k resistor and 1nF cap then you are filtering much more effectively than one of these composite filters can manage, until the RC begin resonating from their own self inductance at hundreds of MHz (so use small surface mount ones). Another reason not to use them is when you intend the product to still be manufactured in 5 years' time. If however you need to filter USB signals or get significant power through a line, then use a ferrite bead and a capacitor, or one of these filters instead of an RC. Personally I prefer a discrete L + discrete C instead of these filters as there are more options if you need a second source or a change to the characteristics, and because every time I've used a 3 terminal filter it has gone obsolete within a few years or resonated when the EMC regs were extended.