Hi-Q inductors

A question for the RF boys:

I've been making 50 MHz LC oscillators using some 150 nH 1008-sized surface-mount inductors. These are ceramic-core types, not ferrite.

They work fine, but Qs are low, in the 30 range. Coilcraft has "surface mount" parts with Q above 100, but they are either bare springs or springs glued into a plastic case...

formatting link

and

formatting link

(We have used this Coilcraft "maxi" inductor in the past, and found that the plastic cold-flows over time and slowly changes L. So we bake them for a day or so to relieve stresses. Coilcraft claims they don't cold flow!)

My concerns are...

For the highest Q parts, the bare springs, does L change as the FR4 board changes dims with temperature? I suspect Coilcraft doesn't test them soldered to a board... they apparently use some HP q-meter fixture.

What's the effect of a PCB ground plane on L and Q? Ditto.

Are these things significantly microphonic?

Any experience or suggestions for stable, repeatable TC [1], high Q inductors, preferable surface mount?

John

[1] so we can pad them with NTC caps and kill most of the temperature drift. Actually getting NTC caps is a whole nother story.
Reply to
John Larkin
Loading thread data ...

I thought that C0G caps were a good approximation to NTC?

I suspect that not many folks care much about TC, because they're assuming that you'll put this into a phase-locked oscillator. Which makes me wonder why you're not phase-locking your oscillator to a crystal...

Alas, though, I don't have much actual help for you.

--
Tim Wescott
Wescott Design Services
 Click to see the full signature
Reply to
Tim Wescott

NTC is Negative Temperature Coefficient. Inductors tend to have positive TCs, often in the +120 PPM range.

It's a triggered oscillator, starts instantly when it gets triggered. Phase locking is a nuisance under those circumstances.

John

Reply to
John Larkin

Man, I'm just asleep today.

Ah. No kidding. Using a DDS comes to mind, although the parts count and board space would mushroom.

--
Tim Wescott
Wescott Design Services
 Click to see the full signature
Reply to
Tim Wescott

For single-layer coils, the old formula from NBS Circular C74 (1924) is useful:

a**2 n**2 L(uH) = ---------- 9a + 10b

where n is the number of turns (counting the half-turn represented by the external circuit), a is the radius and b is the length, both in inches. It's good to a couple of percent over the usual range of coil dimensions.

Differentiating this formula will let you estimate the change in inductance with stretching and heating. (The little Coilcraft springs have wires whose radius is a significant fraction of the coil radius, so the formula will be less accurate for those, but it's close enough for an estimate.)

To get the stretching effect right, you have to take account of the reduction of diameter with stretching--the wire stays the same length, more or less.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

The ground immediately below the coil stabilizes the stray capacitances. When installing the coil directly on the uncoated FR4 material with a ground plane on the opposite side of the board, will form a stray capacitor with FR4 as the dielectric.

The dielectric losses should not be significant at such low frequencies, the changes in humidity may alter the FR4 characters and hence frequency.

Reply to
Paul Keinanen

1) I would guess that it is "obvious" that the L would change as the FR4 PCB's temperature changes - but not a whole heck of a lot. Why? The change of length is (guess without looking up CTE specs) maybe 0.1 mil for every 10C, and the length of those coils are 100X or more than that = = percentage change very small. 2) Ground plane would decrease effective L and Q; the Q would go down some due to PCB losses. 3) Microphonic? What is that? I am in Egypt near this big river..
Reply to
Robert Baer

I do not think the diameter changes as copper is too strong for that..i think that the coil tries to wind/unwind and in that process the coil instead of having a straight line center thru all loops, that the loop-center line would curve.

Reply to
Robert Baer

When you heat something up, all its dimensions increase by CTE*DeltaT.

If the board expands differently than the copper, one of them has to stretch. Do you think the poor little inductor stretches a big chunk of FR4 much? Because one has to stretch, or else the solder joint has to break.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

We're talking about 120 ppm/K here--very roughly twice the CTE of copper, assuming that the coil is longer than its radius.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

If the board shrinks more than the copper, the inductor may buckle a little, but it has to go fairly far to take up the length change--the height of the arch goes like sqrt(CTE*DeltaT). It's like the classical railway rail problem--if you take a 1-mile railroad track held very firmly at both ends, and then somebody comes along in the middle of the night and welds in a 1-foot extra length, causing the rails to bend upwards in a circular arc, how tall is the arch above the ground?

(Guess before looking below)

About 44.5 feet.

Cheers

Phil Hobbs

Gotta run to catch a plane--4AM Albuquerque time.

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

Reminds me of sagitta and my telescope mirror making days.

The original 1 mile track represents the chord of a (large) circle (you did say a circular arc) after the 1 foot of extra track is added. This involves a sin(x)/x term that is supposed to equal 5280/5281. I get _about_ 0.0337 radians for x (iterated a couple of times.) That leaves the sagitta as (5280/2)*(1-cos .0337)/sin .0337.

I get 44.488', which as you say amounts to 44.5'.

Yes, I solved before looking.

Jon

Reply to
Jon Kirwan

Have you ever used silver plated glass inductors?

Reply to
Michael A. Terrell

To avoid creepage or flow you'd need ceramic inductor "cores".

Not too much IME.

It lowers Q, depends on proximity. One has to be smart about plane voids under high-Q inductors.

Oh yeah :-(

Talk to an engineer at Johanson. That's one of the last RF component strongholds still standing. They have ceramic core inductors but AFAIR their Q is high only in the gigeehoitzes. Don't know where your circuitry runs.

Can't you use a coax resonator?

NTC caps? In the year 2010? I wish your purchasing folks good luck, and provide them with a complimentary package of Maalox ;-)

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

P.S.: Pulse makes ceramic core ones as well:

formatting link

For best Q you need to use larger sizes, 1206 yields a better Q than

1008 for example. The ones from Pulse ride high so more clearance from the board.

Also Tamura but forget their web site, it didn't find their own stuff and luckily Digikey had the datasheet (but no stock):

formatting link

Maybe someone offers a lift kit for inductors, with chrome-plated Bilstein shocks and all that :-)

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

We use them in some other situations, especially when we can sic a phase lock loop on them very soon, like 800 ns, after oscillator startup. Their TCs are very stable, +100 or so, unaffected by pcb tce, but the Qs are low, in the 40 range. I really want a Q over 100, preferably 150, and that seems to come from big lazy coils of copper wire.

I want to use a triggered oscillator to make time delays without the phase-lock overhead. Such an oscillator piles up jitter fast, and the lower the Q, the less jitter.

Is it the copper proximity that lowers Q? I can cut clearances under the inductor, within reason.

Experimenting could get tedious here.

50 MHz is my target here.

I've done that too. The problems are

  1. Expensive

  1. Low impedance, 10 ohms range, so the "initial conditions" current for a triggered oscillator is high.

  2. Their low limit is around 500 MHz, so the kickoff driver has to be very fast and we'd need an ECL divider to get it into our FPGA. More $ and mA.

In some places, they're great. High Q, very low TC.

We do stock some surface-mount 18 pF N750s, but they are made to order so you have to buy lots of them. Big reels, > 10 weeks.

John

Reply to
John Larkin

The Pulse parts look nice, especially the standoff, but Q is still

40-50 at 50 MHz. Looks like I need a big ugly coil of copper.

I wonder if putting a few surface-mount inductors in series would help any. Probably not enough to make it worth the nuisance.

John

Reply to
John Larkin

Probably meant higher Q :-)

But it also makes for a sluggish start.

The lion's share is probably the eddy currents. I rarely advocate compromising solid ground/power planes but this is one of the situations where you have to. Strategically applied slits might be sufficient if you can't whack a hole in it for some reason. The ideal scenario would be a real hole, plain air. Or whatever city folk call air :-)

Not really. All you need is some copper clad, a Dremel, a solder iron, an impedance analyzer (or VNA or spectrum analyzer with track gen). Make a resonant circuit and see how high you can get the Q.

Do you need the inductance as low as 150nH?

Yep, AFAIR around $5 unless you need large qties. They are made to order, not really a la carte.

You can get lower in frequency but things will start to require real estate.

You could probably make some profits on the side by offering them in small qties :-)

The 150nH from Pulse should get you in the vicinity of 80 for Q. Sometimes a 330nH is better at 50MHz.

You could talk to companies like this, see whether they know a source or can get you higher Q inductors. I have never dealt with them but could be worth a call:

formatting link

Or contact coil winders directly:

formatting link

Winding coils and making sure they remain stable is an art unto itself. There are tricks, like pre-stretching and pre-twisting the wire.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Probably.

Nope, I can start a high-Q LC oscillator instantly.

That does sound tedious. But I might have to do it. I was hoping for free consulting here.

I could also build an oscillator and measure jitter, or look at it on a spectrum analyzer. May as well lay out a board with several test circuits.

That gives Xl around 50 ohms, which works pretty well for a triggered oscillator. Lower L probably gives higher Q, so 100 nH might be OK too.

And post-baking. We bake some of the Coilcraft parts, 150F for 24 hours, to relieve stress and stop cold flow of the plastic form.

John

Reply to
John Larkin

No, but if they were fused quartz, they could be pretty stable--roughly

0.2 ppm/K from dimensional change.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

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.