Conical inductors--still $10!

Am 14.07.20 um 15:01 schrieb Phil Hobbs:

The patent was one of Piconic's. IIRC they asked for $/? 30 > 10 years ago when I was with Infineon fiber optics. I then told the Mini Circuits rep that they might sell a bias tee for every ERA-1 if it was cheap enough and soon they had bias tees and/or 'flying' inductors. I don't claim it's been because of me. But the solution with cascaded Ls was still so much cheaper that we couldn't help but using them. It just takes some hours of experimenting with the VNA and you get it flat to 20 GHz, with an ordinary 100n AVX 0402 capacitor.

A friend of mine put some ferrite into a mortar, pulverized it, mixed it with epoxy and put it into a form made from clay? or whatever and a color pen. Then he wound the coil while it still was somewhat sticky. Ugly as hell, but it worked perfectly.

Cheers, Gerhard

With no added passives to avoid nasty resonances? As shown for example here:

It's definitely a curious thing. I've just used a pair of inductors to broaden a bias T, but those conical things would have been perfect. I really wanted 1MHz-6GHz but that's too painful to contemplate developing, and I don't have an extensive kit of the required inexpensive parts.

This looks reasonably competent and not stupidly expensive, but my Chinese assembler won't source them:

Please do share any combinations you know to work well. As you say, it's possible to do, just tedious.

I've been tempted to do similar. Also to series-connect a PCB spiral with a larger inductor, and try to tune that... despite the waste of PCB space involved.

Anyhow, my ears are open for hints.

Clifford Heath.

Nice. I'm not primarily an RF guy anymore, so I'm generally more concerned with pulse response and (especially) stability than flatness.

Cheers

Phil Hobbs

It was actually a Piconics patent.

The series connection, inductors and beads, can have resonances and ugly step response. Added parts can selectively kill Q, which is a mild nuisance to work out.

Coilcraft has a cool ferrite-filled wideband inductor

while only claimed good to 6 GHz, it's cheaper than conicals and physically tough and can handle a lot of current. One might hang a tiny smt inductor on the end of that and extend the bandwidth. I could TDR that combo when I get a slow day.

Most are #40 wire, maybe tricky to wind and fill, and I guess the market is small.

They are hard to use, and some don't wash well, so I'd avoid them whenever possible.

Here's an old PSPL bias tee:

MiniCircuits has cheap smt tees now.

This starts to remind me of the Barker & Williamson 'Christmas Treee' plate feed chokes half a century ago.

I don't think that I can find that in finite time. It also depends a lot on the layout.

Don't try a world record at the low end. Huge coils deliver nasty effects if you short the supply. Phemts do not like that.

I have found some pictures of a Piconics coil on a

light blue traces are the transmitting side of a 18 GHz differential TDR Agilent 54744A, yellow is transmission into a 54751A 2*20 GHz plug in.

It is probably possible to dope the Piconics by curling the hot lead in the air. I was not making a bias tee, I was just playing with the then freshly-mine TDR plug in.

Mini circuits: <

Piconics <

Inertia. Some high volume manufacturer may eventually find a product that n eeds them, and only then will it be worth the trouble and expense to tool u p to manufacture them cheaply in high volume. In the mean time, what market there is will pay the $10 unit price.

Conical antennas are known for being rather wide-band.

I suppose they are harder to make than we'd like, although I don't know why .

I have used Gowanda and Piconics. Yep--still pricy even though the patents are out.

My latest idea is to emulate a conical with a series of 2-4 "stepped sizes" of CCI ferrite core 0201,0402, 0603 inductors. I haven't had time to devel op a library of "favorite combinations." Some people are hesitant to use th ese coils above the first self resonance, but it is fine to do so.

hy.

s are out.

s" of CCI ferrite core 0201,0402, 0603 inductors. I haven't had time to dev elop a library of "favorite combinations." Some people are hesitant to use these coils above the first self resonance, but it is fine to do so.

btw, AVX also makes conicals now.

We've done series strings of inductors and beads, but they do have resonances that need additional damping parts. This is time domain, where we want very clean step responses. For RF, you wouldn't need to be so pickey.

ce

why.

ts are out.

es" of CCI ferrite core 0201,0402, 0603 inductors. I haven't had time to de velop a library of "favorite combinations." Some people are hesitant to use these coils above the first self resonance, but it is fine to do so.

Yeah, you're right. This is a nice paper giving the general idea to improve step response, although it is low frequency.

Yup, that's my trick too. Starting with the 0402 they're actually beads (Murata BLM15BA/BLM18BB). That helps control the effects of the pads in between the beads. I haven't spent enough time on it to optimize them, but they're pretty good medicine.

For stabilizing BFP640s, I like to use a single BLM15BB050SN1 in the base. Works like the bomb--I mostly use them for their studly beta, Early voltage, and low noise--the Infineon model has BF=450, VAF=1000, and RB=3 ohms, and the 1/f noise corner is pretty low considering it's a

Cheers

Phil Hobbs

The Murata datasheet is 277 pages long!

Surely a bead in the base kills Ft, which is why it's there. But does it also add HF noise? Seems like it would.

I guess a 5 ohm bead is no big deal.

Yup. But there are a lot, a lot of good uses for a BFS17 equivalent with a beta of 500, no Early effect to speak of, and 0.3 nV 1-Hz noise in the flatband. Especially at 17 cents in reels.

Not at 100 MHz. The peak goes much higher--the Murata BLM1xBB beads are pretty amazing actually.

Cheers

Phil Hobbs

Oh, and 0.08 pF C_CB.

Cheers

Phil Hobbs

I recently replaced a couple of BFT25s with BFS17s in a 120 MHz Colpitts oscillator. The '25s were just too hot. BFS17 is a nice stable part.

I might try your combo transistor and bead, just for fun.

We start the oscillator when we get a trigger, and it's the time base for making delays. But we phase-lock it to an OCXO within a few microseconds, so the long-term jitter goes away. What I want is picoseconds of jitter for those first few us. The passives in the circuit probably dominate jitter, I'm guessing.

I invented the phase-lock idea, and on a good day with a lot of coffee I can mostly understand how it works.

I could turn off the PLL and free-run it, and measure jitter vs time delay. OK, I'll do that.