Coil and ground plane

Probably just better conductivity than the rough copper on the PCBs. If it were a perfect conductor, you'd get the frequency shift, but the Q would be almost unaffected.

Cheers

Phil Hobbs

Or maybe the 30mil was "Chinese copper" and didn't contain any metal other than a very thin layer of lead-based paint :-)

You put a shorting coil next to your inductor, and you diddle the series impedance of that shorting coil. When the shorting coil has either infinite or zero series resistance, the shorting coil accepts no energy from the coil and (in and of itself) creates no loss. When the shorting coil has a real resistance, the closer that resistance is to an "optimal match", the more it can dissipate energy from the tank circuit, and the lower the Q.

It's like a termination on a transmission line: Make a tank circuit that uses a 1/2-wavelength open transmission line as a resonator. Now short the end -- suddenly your resonant frequency drops by a factor of two, and you've got a 1/4-wavelength resonator. Now remove the short, and replace it with a resistor at the transmission line's characteristic impedance. Oops -- now you don't have a resonator at all.

Try a 30 mil silver sheet -- I bet your Q will be higher. (And, of course, you have some just lying around waiting for the test -- right?).

You may also want to measure resonant frequency at each one of those points -- I would expect it to be higher at each step in the chart. Old VHF amateur radio practice (and WW-II era 'real' VHF, AFAIK) was to use coils with brass shorting plugs, or copper disk "shorting turns" to tune them up in frequency.

Attached to plastic that turns to Roofies when it gets wet. ;)

Cheers

Phil Hobbs

axial leads on the inductor tells me you had the coil close to the ground plane. I wonder in which orientation you had the inductor aligned with the ground plane path?

Also, not sure if you had any formal coating on the copper but get the coil close enough to the plane and I do think you can start seeing a G line effect even if you have bare copper due to the glue on the other side. This just helps things along with thicker coppers and eddy fields. Maybe at 6khz G-Line effects are not noticeable but eddies are.

THere is also the magnetic coupling effects the ground plane can cause that works against you.

Have you tried extending the leads on the inductor to give it some space away from the traces? This was a common practice where you would leave longer leads on a axial inductor getting the reluctance of the field away from things it can couple to.

Just a thought

Jamie

I've already observed that for an unshielded radial leaded inductor like that one:

and could not come with any explication at all too.

Closed core (shielded) inductors obviously don't have that problem.

of

an

Are you 'dis'ing the board house tha made my pcb? The few times I measure the DC conductivity of a pcb trace is was 'pretty' much right on. (There is can be more than 1 oz of copper on the outher layers.)

I have a ~4mil piece of copper tape from 3M about the same as the 5mil pcb.

George H.

t -

of

es

be

han

Before I dug up the copper, I stuck a piece of ~30 mil aluminum panel next to it and got a similar answer... Q was a bit less, (19-20), delta F about the same.

George H.

of the

pcb

an

at

ce

Hmm OK...scratch, scratch.. maybe I can sleep on it.

No, but 30 mil aluminum had a bit lower Q... just as good?

I must admit I'm more confused by the low Q at some 'special' thickness.

Ahh, I did... see numbers in third column.... Resonant freqency kept rising with thickness, but the Q hit a minimum and then got bigger...?

George H.

=A0Old

ttdesign.com- Hide quoted text -

of

an

Yeah just sitting along the ground plane. The axis of the inductor parallel to the pcb.

No coatings. All the data posted was for the hunk of metal stuck right against the inductor body.

As you moved the conductor away the effect decreased.... I took some data as a function of distance,

It's perhaps exponential... but not enough range to really say anything.

Thanks Jamie. The problem is really solved. I just will leave off the ground plane.

But it seemed weird, so I posted it.

George H.

In order to get the copper to stick to the FR4, they put a 'tooth' into it, so the surface very rough. (Rip some foil from a piece of Cu clad if you don't believe me.) The conductivity of the surface layer is effectively reduced by a factor of 1/sqrt(1+(dy/dx)**2).

Cheers

Phil Hobbs

of

es

be

han

Yeah a pot core or toroid would be no problem. I'm glad at least one other person has seen this weird effect... I'd say something like surface scattering except that happens at a much smaller thickness

Why the large loss at less than the skin depth? There's gotta be some other scattering going on.

Color me confused yet again....

George H.

Because copper is very conductive.

You've built a transformer with poor coupling and a low resistance short on the secondary. The poor coupling looks like a large series (leakage) inductance, connecting the primary self-inductance to the secondary self-inductance.

Now, if the secondary were an open circuit, nothing would happen, because there's no path for current to flow through the secondary inductance. Q is high and frequency is normal.

If the secondary is a perfect short, all the voltage drops across the leakage inductance, putting it in parallel with the primary inductance. Q remains high, but inductance drops, so frequency rises.

If the secondary is the perfect value inbetween, so its resistance matches the leakage reactance, frequency rises a little (because the leakage inductance is not-quite-in-parallel, reducing total inductance) and Q falls (because the resistance is effectively transformed to the input side).

The resistance presented to the coil is roughly inverse with thickness, when thickness is less than the skin depth. Obviously, when thickness is zero, it has infinite resistance. As thickness increases, resistance decreases, and it just so happens, for the amount of leakage in your setup, the point of maximum power coupling occurs around this range.

Tim

Wierd.. I presume that case #1, "nothing" means DIRECT use of the inductor wires, and no EXTRA in the way of PC traces. The progression from 1/4 mil to 5 mil copper looks the opposite of what one might expect on a pure conductor basis (lower thickness means higher resistivity means greater losses resulting in lower Q). So..maybe what is happening is that the RF wave travels more above the conductor surface for thinner conductors (PCB material being a capacitive "relector" perhaps)?

I would guess that if you tried 60 mil thick trace then 120 mil thick, that the Q would approach that of the inductor alone (all else being equal).

of

an

Yeah nothing was the coil in free sapce without a piece of copper next to it.

I think the 'two Tims' of the SED nailed it.

But thanks,

George H.

on

OK, I'm not quite getting that second line... But I can make the coupling weaker in the Spice model that Tim Wescot posted for me.

Q is

=A0Q

ls

This looks like the critical idea... but I'm having a hard time 'seeing' the leakage reactance. But will this way of looking at it work....? So the copper sheet is a one turn secondary... with a small series resistance. The primary has hundreds of turns so this small resistance can look a lot bigger to the primary. If the 'transformed' resistance gets as big as the charecteristic impedance (omega L) then the Q would be one.

Thanks Tim and Tim, I think I've got it!

George H.

hen

ro,

es,

of

an

Second that. I usually simulate PCB copper, and plated copper less, as

e of the

e pcb

be

than

ing

that

rt

nd

lace

cottdesign.com-Hide quoted text -

You have got to use femm 4.2 and do an analysis of a coil adjacent to copper.

During development of our NDE Eddy Current Instrumentation, we were confronted with eddy current losses as a function of conductor thickness as looking like a very low frequency 'spatial' wave. In other words, losses as a function of thickness of conductor were NOT a smooth function rather looked like a standing wave in the material. As the material went from zero thickness to infinite, the losses 'wiggled' in value and looked like a 'resonance' with the structure. You could envision it as a 'reflection' from the truncation of the material. We actually used that characteristic to measure the material thickness at any distance far more accurately than you would using normal techniques.

nce of

ing

comes

to be

r than

s a

ker

s.

e Q

How thick is this surface layer?