Pulse (Re)Magnetizer for neodymium magnets
Neodimium magnets (NdFeB) come in a huge variety of shapes and sizes — blocks, cylinders, buttons with mounting holes, to name a few. Yet one often overlooked thing that may also be customized is orientation of magnetization. Yes, you can have the same cylinder magnetized axially or in transverse direction, or even at some angle. For example:
Ok, you say, but magnetization is done in the factory and we just order the one we need, and if it happens to be wrong for our next project, then we can’t do much about it, right? Or can we? And how they are doing it in the factory, by the way? And wait, here is my bucket of HDD magnets with their weird magnetization pattern, may we repurpose them for something?
Those were the thoughts that led me to starting this project. In theory changing magnetization direction in material is simple — you just expose it to an external magnetic field that goes in the needed direction and is strong enough to overcome the initial magnetization of the material. What makes it tricky in practice is that this “strong enough” for NdFeB is somewhere around 2T (Teslas that is). A few practical ways exist for creation of fields of that intensity and the simplest one is to build an electromagnet (coil) and pass a high enough current through it. Usually electromagnets draw not-that-high currents due to the iron core that does most of the magnetical heavy-lifting. But the iron core would struggle to get fields above 1T because of saturation and there are other problems with it too. So we are left with an ‘air-core’ electromagnet where the field is only produced by the current in the coil. How much current is needed depends on the coil’s geometry. For magnetization of HDD magnets we need a coil of ~50mm diameter and about the same in length. There is some math that goes into optimizing coil and system parameters, but we’ll leave it for another post. For now what is important is that our target is to drive ~10kA (kilo-Amperes) of current through 20 turns. And if you’ve just thought that its a lot of current, that’s because it really is (your mains outlet may deliver merely 25A on a good day). The only way for our coil not to turn in a globe of molten copper at this current is to keep the timing short, and that’s where the wording ‘pulse magnetizer’ originates from. Limiting the pulse time to about 1/1000 s will keep the coil cool enough.
If only the amount of current was the goal, then a pulse welding apparatus would do the trick. Yet, it is not only that. The thing is that you have to create the magnetic field in (and around) the coil, the field has energy density. It means that you have to pump energy into the coil to create this field. For our case the amount of energy is about 1000J (Joules). This means that you need to push that 10kA during 0.001s under voltage of 100V. Taking in account losses and some nasty coefficients that we leaved out, it makes in practice something more like 500V which corresponds to peak power of 500V x 10000A = 5MW. Not great, not terrible — that’s power of a ‘village-class’ power station.
At this point you might be inclined to think that with those specs the thing is just not going to happen. But quite on contrary, with proper choice of components the required parameters are quite achievable. First let’s get a storage for that energy. An electrolytic capacitor of 470uFx400V can store 38J. Taking 9 of them makes it 340J.
It was possible to connect them to a single pair of bus conductors, but ensuring even current distribution between individual caps on the bus is a bit complicated. The easier solution is to take equal-length conductors to each cap and try to impose at least some symmetry into their layout. Now let’s prepare 4 of these blocks to get 1400J of storage capacity.
And start assembling them with nuts and rods:
The second critical component is one that discharges those capacitors into the coil in a 5MW pulse. And for that we take a beeeefy SCR (thyristor) that is rated for 250A of continuous current and ~6kA in pulse (but it has some room above it):
Connect those with bus bars (a thick copper foil) and add “flyback diode” in parallel to the coil.
Two parallel blocks of capacitors are here connected in series to achieve a total voltage rating of 800V. Higher voltages (with appropriate coil matching) allow for somewhat less current to be used for the same final result, which makes SCR’s job easier.
Now we need to provide charging for the caps and firing control for the SCR. I had several slightly different implementations of those circuits along the development, but the general idea is more-or-less like this:
Caps are charged with a voltage multiplier from mains (!!VERY UNSAFE!!) and SCR is fired by a push-button from a Li cell. In operation one must keep in mind that during the charging (SW1 closed) all the components, including the coil, are connected to the mains.
The voltage multiplier consists of diodes D1-D4 and high voltage film capacitors C1,C2. A 60W lamp LA1 is connected in series with the charging circuit to limit charging current. TVS diodes D7, D8 are used as a crude means of limiting the charging voltage of each half-tank C3/C4. The voltage between tank halves should be distributed equally and to keep an eye on that a balance meter is provisioned. The flyback (freewheeling) diode D10 prevents recharging of tank caps to reverse polarity on the second quarter of the oscillation cycle. Firing button SW2 provides a short pulse through cap C5 (few uF) to the gate of SCR.
Now let’s have an overview of the resulting device (one of the intermediate implementations):
A piece of HDD magnet has been demagnetized by heating it on flame over the Curie temperature. After that this non-anymore-magnetic piece has been placed inside the coil:
Now charging:
And finally Fireeee!
Yes, it did magnetize the NdFeB! I was not convinced though that I did reach the full magnetization on these parameters and tried firing it at full 800V. This has definitely pushed the magnet to full saturation (as well as sent few objects left nearby into rapid movement), but it has also destructed the coil:
At this moment I have realised that I need a new, sturdier coil design and this served as a start of the project, described in my previous post.
A curious reader might also wonder, why would I bother remagnetizing HDD magnets in the first place. And believe me, there were reasons, for which just stay tuned for the next posts! To name one.
