FEMM - What's wrong with this picture? 2

[Nergus]

I'm open to being told what I've done wrong here!
Quite new to FEMM and self-taught - by an idiot.
I've modelled a static array of 13-off 3/8" NIB cube-magnets in N52, arranged alternate North and South at the top.
An analysis of flux density at a couple of mm above this array shows 12 'lobes' of flux (IBI) corresponding to the N-S joins between the magnets.
OK, so my problem is, when I scan the actual array, using a Redcliffe Magscan, the vector sum flux density shows 13 big lobes of flux corresponding to the centre of each magnet, with very low flux at the joins.
Why are these results so fundamentally different?

Thanks, Alan

 

[MJR2]

I suspect you are not actually measuring or reporting the vectors or vector sum quite the way you think you are. Similar to what would happen with a unidirectional probe that is not rotated to find a peak or perpendicular to the flux lines.

Mike

 

[ArtWagner]

Nergus:

I have to agree with MJR2. I might add that you need to pay attention to rms versus peak reported values; I have found them confusing in the past. Also consider that a physical probe averages the flux density over the probe's area whereas an analysis program can calculate a theoretical point value.

In general though, I think that you will be able to correlate the calculated and measured values.

Art

 

[Nergus]

Thanks Mike, I sort of see what you're getting at.
The IBI plot from FEMM is the instantaneous magnitude of the vector quantity with no directional information.

The vector sum output from the Magscan is.... I'm not entirely sure what? I can imagine the X-axis hall detector looks more like the FEMM plot as it straddles the N-S joins. The Z-axis detector registers zero above the join because the N and S fields are equal and opposite, hence cancel.

Trouble is, the Magscan is regarded as the 'absolute' hence FEMM is 'wrong' because the results don't correlate.
It's important to work out which result is most useful in terms of the application: attracting magnetic nanoparticles in solution above the array.

Any insights welcome! Alan.

 

[MJR2]

Do you want to move them or flocculate them, the nano particles.

If you look up the spec for the probe you hvae used and then make a running average of your FEA results you will find the answers closer together.

In general having learned how to use FEA to give me results I can measure I no longer can tell them apart. That is theoretical or real. However FEA and permanet magnets differ the most as compared to electromagnetics. We are talking about single digit percentages.

Mike

 

[IRstuff]

The highest flux density should be near the faces of the magnets, of which you have 13.

TTFN

FAQ731-376

 

[Nergus]

Thanks for the postings!
MJR2 - I'm trying to move the nanoparticles. The force on them is proportional to the absolute field strength and the magnetic gradient. FEMM says both quantities peak over the N-S join, hence 12 peaks for 13 magnets.
I will try to find out the physical dimensions of the actual hall elements and do a running average over the dimension... might need to ask how, later.

IRstuff - Magscan says that for a 9.5mm cube magnet, the peak is over the centre of the pole face, yet for a larger block magnet, there is clearly a dip in the middle (see attachment). So, more flux density at the side of the pole face, and clearly there will be lots of flux lines concentrated at a North right next to a South.

I've heard a lot of good things about FEMM and to be honest it would be far better if the FEMM result was (give or take) right. I still don't understand exactly why Magscan says 13 peaks and FEMM says 12.

Alan

 

[IRstuff]

Your data seem to be inconsistent; you claimed that the magnets were 3/8", which is 9.5 mm, and now you're saying they're not?


TTFN

FAQ731-376

 

[electricpete]

I think the FEMM is correct. The peak flux density occurs adjacent to the interface between magnets. 13 magnets, 12 interfaces, 12 peaks sounds right to me.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[Nergus]

Hmmm... I can see that in my last posting (gone midnight here) I lost track of the point I was trying to make, which was in essence agreeing with MJR2 (thanks).
If you have a relatively big hall probe passing over a small magnet, you get a single big rise in output. If you look at a big block magnet, where the probe is much smaller than the pole, you see peak flux at the edge, dropping down in the centre and rising again at the other edge.
I feel sure that a 'tiny' hall probe would see a dip at the centre of the small magnet. This would be like FEMM, analysing at a single point.

The difference is the hall probe integrating the field over its active dimension.

As the hall probe crosses the N-S boundary, part of it says 'big -ve output', part says 'big +ve output' and at one position you get full cancellation and no output.
FEMM is calculating the absolute magnitude of field at a single point, which is highest over the boundary.

OK, I'm convinced! Now I've just got to convince the people I work with to disbelieve the evidence of their own eyes (Magscan) and accept that I didn't waste 2 weeks optimising the characteristics of experimental magnet arrays, using FEMM.

Let's hope the nanoparticles back me up on this!

If anyone has an even more compelling rationale, or further insights, please post.

Thanks, Alan

 

[IRstuff]

But, YOU said that 9 mm magnets don't have that dip, and you're supposedly simulating and measuring 9 mm magnets.

If there was a dip in the middle, then you would see 11 large peaks and 2 smaller peaks, since the end magnets have no adjoinging magnets that would cause an increase in flux.

Moreover, your MAGSCAN is two dimensional, and would show same amount of walling on the two long sides, which is not evident. By your claim, the MAGSCAN's spatial resolution is high enough to see the dips, and so would also be good enough to see the peaks on the sides that have no adjoining magnets, but your scans don't show that.

TTFN

FAQ731-376

 

[Nergus]

Hi IRstuff,

Magnet size - 9mm is what's required, 3/8" is what I could buy 'off-the-shelf' to try on the Magscan. The intention was to back-up the simulation results... not what happened!

What I'm saying is, if you Magscan close to a big block magnet (100mm, like the previous attachment) with a pole-face at the top, or do a FEMM simulation, the results are very similar. Big field density at the edges, dip in the middle. I don't see why that characteristic should be any different if the block is only 9mm across.

Attached is a FEMM analysis above one (N) and then two (N-S) 9mm magnets.

The Magscan probe is too big compared to 3/8" to resolve the dip in a single 3/8" magnet and it shows a dip rather than a lift above the join line since equal and opposite North and South contributions are integrated across its detection area. This would be true of the Z-axis and Y-axis detectors. I'm not exactly sure how it calculates the 'vector sum' output, but whatever contribution comes out of the X-axis detector appears to be less than previously came out of the Z-axis, hence a net dip instead of a peak at the N-S join.

I think I'm clear enough in my own mind that FEMM is better at spotting where all the flux lines are concentrated than the Magscan in this situation.

The acid test is the effect on the nanoparticles over the centre of the blocks or over the joins, a test we will do in a few weeks....

Thanks for making me think!

 

[IRstuff]

I think you are reading into the data what you want to see and are ignoring the facts.

If your hypothesis were correct, you'd see the same thing across the narrow dimension, i.e., a trough shape or shoulders that would straddle the gaps in your MAGSCAN results. But you don't, so the postulate is incorrect.

Moreover, your second scan output contradicts your hypothesis as it shows that the other two lobes are same polarity. However, since your outer two magnets are of the same field orientation, their boundaries with their adjoining magnets are opposite in polarity.

TTFN

FAQ731-376

 

[Nergus]

Don't worry about it, I'm going to see Prof D next week, he knows all about this stuff!

Alan

 

[ArtWagner]

I have a question. What is the dimension of the active area of your probe tip? What orientation are you holding the probe when you are taking the data?

I looked at your plots of the single magnet, 9 mm I believe, and then a pair of magnets adjoining one another. The FEMM plots looked reasonable to me.

It's worthwhile to remember that B is a vector quantity and has values in three dimensions (although only 2 appear here) and that you are plotting the magnitude of this vector.

Art Wagner

 

[Nergus]

Hi Art,
The Magscan is a flat bed with a half-metre square active scan area. The 3-axis probe is on stepper-motor driven slides for the X and Y axis movement - which is a raster scan under software control. The Z-axis 'control' is manual probe height adjustment via a micrometer. The most useful output is a .csv file with raster matrix locations for the probes X, Y and Z-axis components. In addition there is a 'vector sum' matrix, presumably each value is the root of the sum of squares of the three orthogonal axis components (TBC).

I have a few questions for the technical expert from Redciffe, when he returns from holiday next week, probe dimensions being one.
Clearly the probes for the 3 axial components can't occupy the same physical space. I'm guessing there are 3 probes lined-up in the detector head, which has a longer dimension along the X-axis. The 3 outputs would be sampled and stored as the probe scans the X-axis, deskewed for position and popped into the appropriate matrix location.

Meanwhile: I've done a Magscan of a single 3/8" x 3/8" x 7/8" magnet and two of these magnets arranged the same as the FEMM simulation. I'll post the results as soon as I get chance to process them.

Alan

 

[Nergus]

OK, here are the Magscan outputs for the 3/8" square x 7/8" long N38 magnet with the North face up. Next are 2 magnets side by side, N then S up...

What was the question again?

Oh yes! Are we all agreed why it looks nothing like the FEMM simulations attached last time?

Alan

 

[IRstuff]

I don't know if we're agreed on why.

Your two tests confirm that your vector sum peaks should exactly correspond to the number of magnets present. Your first test shows a single peak, and your second test shows two peaks, corresponding to the number of magnets in the test.

The y-axis would then be aligned with the long axis of your assembly, and shows the two expected direction changes that would occur in the middle of the magnet faces.

So, your original MAGSCAN results are consistent with your current MAGSCAN results.


TTFN

FAQ731-376

 

[Nergus]

Hi IRstuff,
Unfortunately, Magscan consistent with Magscan has never been a problem. I really do get the N magnets = N peaks thing, it has a kind of 'get what you pay for' simplicity and natural justice.

It's the FEMM (N-1) peaks that demands the burden of proof. Yet instinctively, FEMM is right! You have a N right next to a S, of course there will be a bunching of flux lines, shortest / easiest / path of least resistance - it's the way the universe works... which equals a peak in flux density over the join-line.

Tell you what, I'll try the ultimate arbiter - a piece of paper and some iron filings! Can't do it until next week though.

Alan

 

[bz2005]

Hi Nergus,

I agree with your hypothesis that the MagScan measurement is the net sum of the flux through the Hall element area. Therefore when the probe scans across the boundary of two alternating magnets, it will miss the real peak and gives you two "peaks". Did you notice your MagScan B sum peak value is sutstantially lower than your model predicts? That's a hint it cancelled out a great amount of the z-compnent, and only gives out the net outcome.

I've repeated your model in Lorentz-EM. My result agrees well with your FEMM result; see attached picture. There is only one peak of the B sum across the boundary, not two.

I used Hall probes from Lakeshore. The active area of the Hall element could range from 0.5mm diameter to 1.0mm. Also note the element are often a certain distance away from the probe tip, which also explains in part why the MagScan number is much lower than what the model predicts.

As to the nanoparticles - they should move toward the magnet boundaries, not the centers of the magnets. The boundaries are where the product of field strength and field gradient is the greatest.

Bo