wideband inductors

Hi John. A set of series inductors ranging from small to large is the standard approach, as you have surmised. Along with a single series capacitor, this is what you get inside a wideband bias-T. For example the 12GHz 5575A, sold by Picosecond Pulse Labs at high prices (or grabbed on eBay at somewhat lower prices).

There may be additional components to make sure the series inductors don't resonate with each other, see the 2nd page of this app note:

BTW, you could purchase MiniCircuits bias-tee chips to get low-cost pre-made inductor strings, e.g. TCBT-2R5G to 2.5GHz at $8.95 qty 10, just ignore the capacitor,

Try the bias-T or bias-tee search phrase; let us know what you find.

Well, it *is* a Topic. Would you rather see a picture of my grandson?

Sure, but there may be an optimum strategy, given, I suppose, a fundamental relationship between L and distributed C for a given inductor technology.

The shunt capacitance is a killer. A drum-core 100 uH inductor we tried was 160 pF. A 22 uH surfmount 1812 was about 4. The best we've found so far is a 47 uH Vishay axial, whose SRF calculated out to 0.8 pF shunt capacitance. We'll have some tomorrow morning to verify this amazing number.

We considered a transistor cc source instead of the last big L, but were sorta surprised that no PNP exists with low enough Co and high enough power capability. Interestingly, the antique 2N3906 was about as good as anything else we could find, but still terrible compared to a real inductor.

Our latest simulations indicate that q-killing resistors (across the various Ls) are mandatory.

Our situation is applying power to the output pin of a distributed amplifier. We need extreme fidelity for shaped pulses up to 50 ns maybe, so 100 uH would be nice, and we need to keep the bandwidth in the 4-5 GHz range with no resonances or reflections or nasty things.

Our latest simulation looks promising with two 0402 ferrite beads + two axial inductors, various resistors, and some droop compensation elsewhere in the signal chain to make up for the finite L. The beads are interesting, circuit-equivalent-wise.

John

Thanks for the info. One problem with using a commercial tee is mounting it on the board, and the other is that most everybody brings in the DC through a feedthru cap, so if we add external L we'll get strange effects. Come to think of it, the Agilent 8133A pulse generator has huge SMA-box bias tees bolted down to boards all over the place, but they're huge boards.

It's looking like - as of 4 PM today - we can just fiddle with simulations and get an L-R string that will work. I was just wondering if anybody had done any serious thinking about this problem.

John

John,

Piconics has a conical inductor usable over a 1000:1 bandwidth (for example, 40MHz to 40 GHz). They claim no self resonance making them ideal for broadband applications such as yours. Here is some info:

Here's a better picture of the inductor:

Best Wishes,

Mike Monett

Oops! Here we go again!

Yes ;-)

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

I read in sci.electronics.design that John Larkin wrote (in ) about 'wideband inductors', on Tue, 14 Dec 2004:

ARRL or RSGB Handbook. This is the sort of thing that skilled amateurs develop.

Hi John,

This task is pretty common. The ARRL or RSGB handbooks are a great source as Senor Woodgate had suggested. The books are a 'must have' to me, at least one of them. Other than that it will be the data sheets. You could just pick beads, chokes, inductors and calculate out from where to where each goes for a certain desired Z. Capacitance determines the upper limit, inductance at 200mA the lower. Then place them in series but in a way that they can't 'see' each other.

A whole long stack of beads is another option. But 2k at a GHz or more greatly depends on shields, layout and how you place things. Ye olde EMI rules rule: Break the line of sight where you can, don't split grounds, keep things 'low'. A snippet of copper here and there can make more of a difference than a fancy extra inductor.

Regards, Joerg

Joerg,

can you please explain the "Break the line of sight where you can" rule?

Thanks

C = 1/(2 * PI * F * Xc) = 1/(2 * PI * 1GHz * 2K) = 0.08pF

This looks like trouble if you are doing a PCB but I suggest you look at Toko 33CS or CoilCraft 1606 inductors as the smallest real one. Depending on what the "couple of GHz" is, you may want to make the smallest inductor a PCB trace.

"John Larkin"

==========================================

I'm surprised it is necessary to ask such a question.

The obvious ideal, is a large number of series-connected coils progressively increasing in inductance.

But the limits must be pre-stated - the minimum impedance required and the frequency range to be covered.

The low frequency inductance is an elementary matter.

The self resonant frequency of a coil is not nasty. It may not be useful. But to damp it down with a shunt resistor degrades the minumum impedance performance of the whole string at all frequencies. Resistors are OUT.

In most practical wide-band cases only two, or at most three series-connected coils are necessary.

The limit of performance at the higher frequencies is the distributed self-capacitance of a coil and its resonant frequency. The self capacitance of a solenoid-type coil and its resonant frequency are calculable from its dimenensions.

The overall response of a string of coils is just the sum of the impedance-responses of the individual coils. You can forget the mutual inductance and resonances between individual coils.

A single wideband solenoid coil is possible if the turns-spacing is tapered along its length.

In all cases length is a limiting factor. Especially if propagation effects are to be taken into account.

But no doubt there exists some academic person with the ability to produce a learned paper of doubtful practical use, even a book on the subject if a publisher can be persuaded to invest in it. Everybody has the right to make a living of some sort.

The foregoing follows from 3 minutes thought on the subject. I can always change my mind.

You are simulating ferrite? Doesn't that tend to vary a bit depending on the batch you get, especially over wide bandwidth?

I googled on "self resonant frequency solenoidal inductor" and got:

Had a similar situation 3 decdes ago trying to sum current into the output of a pulse generator. Only needed to be good to 500MHz or so. Used parallel resistors to kill the Q of the low frequency inductors. Never simulated it cause most of the issues in the last stage were parasitic. Never found any combination of real components that came close, so for the last stage ended up with a distributed network consisting of a single 10 Ohm wirewound resistor and flying lead to the output microstrip. That resistor had just the right combinatin or RLC to make it transparent at high frequencies. Second sourcing it was out ;-( mike

No, we're just using lumped models. The ferrite beads look like L || R (based on TDR measurements) and the axial inductors look like L || C, and we add a little node capacitance for PCB pads. This is an important problem, but it's still a small part of a bigger problem, so we're looking for a quick solution based on purchased parts. In other words, we can't be too concerned about things below the lumped-element model. Too bad, it would be an interesting problem if one had lots of leisure.

John

He looks a little like me. I'll give you some time to reconsider.

John

Hey, Win,

We just TDRd a couple of PSPL bias tees. Their thru performance is impressive - very clean path, apparently - and the side L seems to have virtualy no capacitance, but the Ls are pretty obviously there. One box has an effective impedance of 45 ohms, making the L leg appear to have about 500 ohms shunt resistance. The other one has a sort of impedance belly about 5 ns out. Given that we want less than 1% abberations over a 50 ns time span, they wouldn't be very good.

TDR pics to abse later. It's tempting to open one up.

I had the same experience with their expensive signal pickoffs. I was expecting superb performance, but they were only fair; I could do about as well with 0805s on a PC board.

John

All of my children look like their mother, except my youngest son, who is a facial blend, so not TOO ugly ;-)

My grandchildren are BEAUTIFUL!

...Jim Thompson

Just add to the L side on your own...

I imagine if you win one on eBay, you'll feel more comfortable opening it up. I've bought four on eBay so far, but they're used two each by a research group and myself, so no "spares" are available.

Hello Uwe,

It is an expression used in the EMC world and points to the containment of very high frequency noise, above several hundred MHz. For example, if you had a plastic enclosure or a Lexan window that just has to be there because you need a display of some sorts this is where RF leaks through. 'Line of sight' isn't meant literally, you just would try to put up (cheap) metal barriers that reflect RF back to the inside. Now if you did a thorough job here and then only took the metal skeleton, you wouldn't be able to 'see' any circuitry and you'd have a pretty good EMI performance. Just make sure there are no unwanted slot antennas. Those can be avoided by some overlap where there is enough capacitance between two metal parts. BTW, programs like SolidWorks are pretty good at demonstrating the level of metal shielding. Here you can take all plastics away and then rotate the remaining module around on the computer screen, before even building a single prototype.

In the OP's case of inductors there are two dominant mechanisms of coupling. Magnetic, which can often be mitigated by 90 degree rotation of cores, and radiation which can be dealt with via metal barriers that prevent inductors from 'seeing each other'. The capacitive path is usually weak unless you are dealing with a very small space such as in a cell phone. When you take apart the tuner of an older (1970's or so) TV you see that concept. They are housed in tin cans with several compartments. Except for very tiny holes to pass signals each stage cannot 'see' the others. Same in RF attenuators, cable TV amps and so on. It used to be that way in FM stereos as well but now they even skimp on that. The result is very poor performance the minute a strong signal shows up. Our 'new' stereo quits everytime a pilot announces entering the final approach :-(

'Seeing' is meant from an RF point of view. A fine metal mesh may let you see the inside but RF is blind to that if the squares are small enough. With plastics it is the other way around, a human eye can't see through it unless it is clear plastic but RF goes right through unless it is flame sprayed or otherwise metallized. Just think of it from a Radar operator's point of view. He or she can 'see' through dense fog, forests and other stuff.

Regards, Joerg