Planar inductors with cores on ICs

Folks,

In order to achieve much higher inductance than the usual few nanohenries on RF chips I am looking at MEMS structures, deposited core materials and such. It's been done at labs quite a while ago, for example:

formatting link

I am especially curious about the spiral inductor with core in figure 1. If you need very little in current handling and can exclude any DC current it should be possible to achieve a higher inductance in a given space than here. We'd have maybe 1/4th to 1/3rd of the real estate they had but tens of ohms in DC resistance would be ok. If needed one could also consider more metal layers. We'd probably have to run it at

13.56MHz. Also at a much lower frequency but there the inductance could collapse or be whatever it wants to be.

Did anybody ever see this in real life? Experiences? Foundry?

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg
Loading thread data ...

Intel is now building synchronous buck switchers on-chip, and they use some sort of magnetics. There are papers around on that. They use massive polyphase switching to keep the ripple voltage down.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

That is done since a while. For example, Enpirion specializes in converter chips with integrated magnetics:

formatting link

However, there are often inductors in the

Reply to
Joerg

Haven't viewed the document. BUT! Years ago, my partner and I approached Jim at Linear in order to encorporate MEMs inductors 'under' the SMPS chips, thereby all the customer need do is to add input cap and output cap and DONE! So just like a 3 terminal regulator the new family of chips would appear to the user as the same, but with SMPS chips [and built-in inductors] the user would have inherently higher efficiency.

My partner had worked out the nasty chemicals to do the high volume production of the inductor, I provided the electronic expertise and 'patterns'. From memory, we were up into the milliHenries within the footprint of an SMPS chip and standard package. And with high currents still above 100uH's

You have to realize the basic material had relative permeability about 1 million and practically zero coercivity, so still had high permeability at

200MHz, [again from memory, over 1000?] pretty incredible when you compare that performance to the fact that the fe molecule's moment of inertia ceases permeability around 1-2GHz.

Again, from memory, the permeability of his material at 10-20MHz was easily in the range of 10,000-40,000.

Now, [again from memory] ANY deposited material was PURE crap compared. Deposited material had terrible perm and even worse coercivity, which ate you alive per cycle. Historically, every deposited inductor structure I've seen was...well, a waste of materials.

He [my partner] had used deposited materials ONLY for filling in a gap, but NEVER as the main core material, and therein lay his success at the concept.

Interesting you mention how inductance can collapse at lower frequency. I once designed a structure that had almost NO inductance at DC, yet skyrocketed as the frequency went up. Now you're going to get me into discussing 'startup' problems.

The copper wires were 'deposited and in the original application we were running current densities through the 'wires' that were incredible. Comparable to 100A routinely going through a 36Awg wire, that's a decent comparison. Only when we went to densities on the order of 200A through

36Awg wire did we actually melt our wires. ...our wires were small, on the order of 25 microns square, but robust.

Also, from memory, I designed RFI/EMI filters for the RJ-45 telco connectors, built-in common mode chokes something like 80 mil square that had over 2mH inductance.

If you want, contact me offline, we can go into more detail. And I promise to quit trying to do this from memory and go get our actual numbers, models, etc.

Do you need the recipe? I lost contact over the ages here, he is/was older than me, but his son may know.

Reply to
RobertMacy

Uh,...magnetics are LOW impedance.

You want capacitance, now THAT's high impedance. :)

Reply to
RobertMacy

?

The inductors on power converter chips are of low inductance because they must carry large currents. We don't need to, I need as much inductance as we can reasonably get. So for us the design rules for the metal layers will probably be the limiting factors.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

We have no space under the chip.

What kind of miracle material was that?

Can't say that. I and another engineer in this company tried, by crushing a ferrite core, mixing it with epoxy, and then depositing it. Not ideal because we used whatever "runny" epoxy happended to be around but it worked. We can possibly get the real powder but MEMS fabrication is likely the better option here.

Ideally it should not collapse because it makes my electronics more complicated. But we could handle it.

I think we are talking about quite different things. My app is not power, it's signal processing, and a whole lot smaller. For example, the real estate I have to work with is realistically 0.007" wide and not a whole lot longer than 0.020", which has to also include the center leg for the core. The total height can also not be more than roughly 0.007".

We are not 100% sure that we need all this. But there is a realistic chance and if we do then I want to, as the scouts say, be prepared. At least have a concept along the lines of "Here is a possible solution and people have actually done it". That's why I was asking whetehr anyone has seen something similar to what's shown in figure 1 of the article.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

You should talk to Hasegawa, used to be at Metglas, bought by, bought by, bought by... He invented the manufacturing technique to make metglas, amorphic ferromagnetics. I've MEASURED relative permeability at above

1,000,000 on these materials. The stuff's coercivity is nil! Acts like a magnetic dead short.

Sadly, I just did a search for my work. Apparently it's on several of the

8 HD's that all crashed within a two year period. Yes, including the HDs that were the BACKUPs for the other HDs !! What's the chance of that? I have hardcopy somewhere, too, but...

The reason the inductor was placed UNDER the chip was that the inductor was 40-50 microns thick with bonding pads on top and being metal had excellent heat transfer characteristics. There is a lot of room in an IC package. I'm trying to remember the package that was used, but ??

Ok, brush away the cobwebs here: the 'inductor' pattern width was adjustable in the ranges of 5-10u inuslation, 15-25u copper, 5-10u insulation, PLUS MATERIAL, so repeat width on the order of 60-100 microns?, that gives what? 3-4 mils? So you can run two strips, down and back. total length two x 20 mils, or approx 1mm length. You have more height available than 50 microns so stack several. If you're doing a common mode choke the pattern becomes: Material, 5-10u, 25u, 5-10u, 25u,

5-10, Material. not very wide. [nothing sacred about the 25u width of copper, if you can stand the series resistance, make it 10u.] I also designed this thing to be able to use 'scrap' material from the vendor. This gave the vender a place to sell quality problems, and gave us a cheap source, almost raw foundry pricing at weight. For example, manufacturing would make several years supply for us in around 20 seconds. something like that. um, 600 fps about 4 inches wide and in 20 seconds you have a LOT of material. yea, a lot of material.

Depositing permalloy over the top of the 'open' structure was like adding the I core and GREATLY increases inductance per length, BUT no DC, please. Or, ONLY use as a common mode choke/transformer.

Tap into the expertise of the HD head designers. They make some cheap, small magnetics using deposited materials, too. They're might be a bit closed mouth about what they're capable of.

Reply to
RobertMacy

[...]

A million? Has that been published somewhere?

I always back up on DVDs and on flash drives. So far that has never failed me. Hard drives are a gamble, stuff can age in there.

There will not be any DC. There will be a low frequency on it but we can measure during the zero crossings.

I just wanted to test the waters for now. It'll still be a few months out. If we end up needing it, and there is a chance that we do, then I'll have to sit down with our semiconductor guy and possibly in a conference with some ferrite guys. The material isn't so critical but we need one that can either be "MEMS'ed" or sputtered or deposited in some other way.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I have some film made of the stuff, this version:

formatting link
...and it's marvelous. Works at LHe, too. By the way, my colleaques in our institution work on MEMS inductors, but I don't think they have deposited magnetic cores. And we're a research organization, so, even if the cores *were* deposited, I don't think we can be counted as a "commercial source".

Regards, Mikko

Reply to
reg

Thanks, Mikko. It looks like the properties begin to fall apart above a few hundred kHz though. We need to operate at 13.56MHz and that has purely regulatory reasons. We cannot prevent antenna effects which will cause the RF to leak out, and this is an internationally recognized ISM frequency. There is also 6.78MHz but that is not allowed in as many countries.

Ok, but isn't one of the jobs of your institute to bring this kind of technology to economic fruition, where it benefits either the people of the world or at least the country of Finland?

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Don't know but I ran across another guy on the net, irritatingly much more knowledgeable in modeling using both Jiles-Atherton AND ??, who confirmed the measurement/comment. He acted less surprised and more like, "Of course, didn't you know that?"

Be careful, I've heard that DVD's aluminum foil has problems over time too. It peels, or worse, corrodes. We have a 'strong' atpmosphere here, so aluminum 'rusts' and stainless steel appliances turn hazy brown, so I don't hold much hope for a DVD to survive either.

How about hardcopy? Is that still used? ;)

You may have to force the 'flip' and measure during the transition, but still possible.

Uh, just remembered. Any ambient magnetic fields to worry about? Fifty microTeslas may not sound like... And local AC mains or oldtime monitor fields can get into the milliTesla ranges.

Sputtered? Then watch out for the structure. You can't believe the idiotic implementations of some good materials I've seen. But for basics, and for easily obtainable materials/processes research the permalloy material, and see if it will work for you. The processes used in the MEMs approach are NOT compatible with semiconductors - IT WILL KILL THEM, the product is ok, it's the process that is potent. Has to be done in a completely different facility, albeit using semiconductor 'like' processes, still the chemicals devour the chips and gold and aluminum etc. is ok when finished though. Envision 'peanut allergy'

Reply to
RobertMacy

Have you gotten 'razor cut' on it yet? Careful to keep it sealed in bag with a dessicant, else it will RUST !

You may also note that the material changes characteristics from 'top' to 'bottom' where top is the side that hits the nitrogen cooled roller so is a bit better.

For MEMS you cannot deposit this material, rather must deposit the 'wires' instead. Heat, flexing, anything 'ruins' the material. The deposited magnetic material has much lower perm and much higher coercivity, but does work in a pinch to fill in a gap.

Reply to
RobertMacy

Mikko just provided a link here in the thread. However, the relative permeability starts to fall apart rather quickly once you get above the audio range. A high value only at or near DC isn't very useful in most inductors.

You have to make a new copy once in a while. But it is less risky than a hard disk which can suddenly grind itself up or die due to external event such as very close lightning hits.

What surprised me was that my floppy disks from 25 years were still readable.

Very rarely. I try to minimize paper usage and also the clogging of shelf and cabinet space. Last weekend I cleaned up some more and, with a sigh, threw away the large binder with the manual for my ECA224 license. It's now so old that I'll never use it anymore anyhow. Of course, I separated out the sheets so they can be recycled cleanly. Can't yet bring myself to doing that with my old PSpice license because that stuff is in much nicer cloth-covered binders. But it'll have to be done. Some day.

If I have an error because of hysteresis or whatever that may be tolerable, as long as it is always the same.

Oh, there will be. The inductor should ideally be immune to that. If it picks up some of it I can ferret out my desired signla via a correlation method.

Well, we are doing something similar already but not at liberty to disclose. You can make things work but yes, have to make sure that processes aren't killing off each other's results.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

N O O O O ! ! ! !

First, you may NOT want to use this particular alloy since it saturates around 0.57, they have materials that go to 2T

Second, Those curves and those charts are based on 'old school' measurements and accepted techniques. They are observations taken using established techniques, using wires with the material in a 'bulk' form. Used that way, it's only good application is a high efficiency AC mains core. When you use it in MEMs, you get down into the molecular characteristics and believe me that stuff is still a dead magnetic short out passed 200-400MHz!!

For example, imagine a core that is 2 mils [50 microns] trying to run at

60 Hz, the conductivity is on the order of 1 MS/m so the skin depth is, look at that! around 2.6 mils, or 65 microns - a great use of the material. Now try to go higher in frequency and the skin depth eats you alive. The effective core area drops to nil and of course you have nothing there!!! See what I mean about old school thinking? Instead imagine the material used in the micron ranges. NOW the skin depth is better matched to your requirements.

We're stuck in the fifties here!

Reply to
RobertMacy

Yes it is, and we do it. I guess our "commercial outlet" on MEMS-related stuff is nowadays Murata

formatting link

now that they have acquired VTI Technologies. I just don't know the status of each R&D project in that direction, as it involves different guys with whom I merely share the coffee room (the only radius within which the information really spreads). Some MEMS-related R&D results end up to companies like Nokia, and those gadgets may never become publicly available as individual components.

I know there is work going on related to your needs, but I'd better just refer to a publicly available source

formatting link

Regards, Mikko

Reply to
reg

Luckily not, there was a warning on the package.

Oh, thanks for the alert. There was a warning on the package about that, too, but I may have taken it too lightly (it haven't rusted yet).

I can recognize the texture change across the strip, but I didn't realize its magnetic properties change, too. Thanks for the hint.

Regards, Mikko

Reply to
reg

Perhaps should have said, from TOP to BACKSIDE, the 50 micron distance.

from the shiny surface back to the duller surface.

Not, side to side, or along the strip.

You'll see the perm [huge] and coercivity [near zero] on the shiny side is incredible, however as you go 'down' into the material it deteriorates a bit, perm drops a little but the coercivity comes up. probably due to the 'slower' cooling and therefore slight crystallization occurring on the 'backside'

Reply to
RobertMacy

Oh, I see. My stuff is 16um thick, but I see what you mean. I have only used it as static shielding material in LHe, but was intending to make a tape-wound toroid out of it. Sounds that the stuff has so unusual characteristics that even making a toroid out of it would be sooo fifties...

So, if I just splat it over my circuit on PCB, I should land it shiny side down?

Regards, Mikko

Reply to
reg

The pole pieces of magnetic heads are made of molypermalloy (or at least they were circa 2005 when IBM sold its disc drive division to Hitachi). It has a bulk permeability of about 250,000 iirc, and the heads of course work out to very high frequency.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 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.