Homemade Ferrite Core

[Desmond Hutchins]

Idecided to build a modern version of an old-school core memory, for fun and to learn how theywork. This article is about the initial testing I did with a single bit, and the measurements I tookof the response for a few different ferrite materials. As it turns out, you need a rather particulartype of ferrite in your core. Most ferrote torroids you can buy off-the-shelf today are not verywell suited to data storage.

When you pass a current through the core, it creates a circular magnetic field in the core. If thecurrent is high enough, the core will retain a permanent magnetization, in either the clockwise orcounter-clockwise polarity. So Voila! We can say that one direction represents a 0, the other a 1,and we have a single bit of storage.

So then how can we read back the bit stored in a core? The answer is that we can pass a second wirethrough the core: the SENSE wire. Then we can write a 0 to the core, and while we do it, we canobserve the SENSE wire to see what sort of voltage is generated on the SENSE wire1. Wheneverthere is a change in magnetic field, a voltage is induced in the SENSE wire. If we pick the rightferrite material, then when all of the little magnetic domains inside the ferrite switch direction-- i.e. when a current in the "0" direction is applied to a ferrite that is magnetized in the "1"direction -- there will be a large and sudden change in the magnetic field that we can see in thewire. If the ferrite was already magnetized in the "0" direction, then we expect to not see a largechange in magnetic field, and therefor we expect to not see a pulse on the SENSE wire

One of the most common ways to describe the behavior of different ferromagnetic materials is basedon the "hysteresis curve". Its also sometimes called the "B-H Curve". Instead of writing my owninferior description of this, I'm just going to point to a couple of articles that I think do a goodjob of describing what a B-H curve is:

For a good memory, we want a small current applied to a magnetized core to generate relativelylittle change in magnetic field, but when the opposing current exceeds a threshold, we want thefield to quickly flip to the opposite value. When looking at the B-H curve, this means we want it tobe relatively flat at the top and bottom, relatively wide (i.e. a high coercive force) with sharpcorners and a large slope when it transitions between +/- B values.

The datasheet for the two ferrites from digikey doesn't include a full B-H curve, but it does have atable that gives you some insights into its shape. For starters, you can see that the ratio ofsaturation to residual flux density is lower on the F25 than the F35, which suggests a flatter topof the curve. Secondly, the coercive force is much higher, so we can expect it to be wider. Justbased on that, I would expect the F25 to be a better option, and indeed experimentally I found it tobe significantly better.

The bulgarian ferrite is the smallest, and perhaps unsurprisingly, the best suited to core memoryuse as it was presumably design with that purpose in mind. I don't have any datasheet for it, butthey are available on ebay for quite cheap:

I used an AVR microcontroller to drive a small test-pattern that I used to collect all the datapresented here. The pattern is simply forward pulse, forward pulse, reverse pulse, reverse pulse,long delay, repeated indefinitely. Now I can trigger the scope on the first current edge, and seeexamples of flipping the ferrite magnetization and not flipping the magnetization next to eachother.

There were basically two types of current pulse I used: one with a slow edge, and one with a fastedge. The fast edge is more representative of how the core will be operated in the final memorydesign, but the slow edge is useful too because it allows a B-H curve to be plotted for the ferritesbased on the scope measurements. It's not perfect; I think a slower change in current would bepreferable for measuring the B-H curves. But for the moment anyway, this is what I have.

The main thing we want out of these sequences is to be able to clearly and easily distinguishbetween the response pulse when the ferrite is flipped vs when it is not flipped, and as you can seein the plots below, there is a pretty dramatic difference between the three.

This was the first ferrite I worked with, and I must admit that I was pretty excited to see any sortof different between flipping and not flipping. But, actually the difference is not as pronounced asone might like.

The purpose made bulgarian ferrite blows the others away though. There is almost no response at allto a current, until the magnetic polarity of the material is flipped, and then there is a strongpulse.

Since the SENSE voltage is proportional to the rate of change in the magnetic field, dB/dt, you canintegrate it to get the total magnetic field as a function of time, and then you can plot a B-Hcurve. These curve are a little odd in shape, due primarily to the fact that the current is changingtoo quickly with the current driver setup that I have. There is a time element to the response ofthe ferrite, because the internal magnetic domains can't all flip instantaneously, even if yourapplied current nearly does. None the less, the curves do still do a good job of showing how muchthe B field changes when you apply a second pulse in the same direction as it is already polarized.

I haven't done an exhaustive search, but I wasn't able to find any currently-on-the-market ferritecores that were more appropriate for a core memory than the F25. The bulgarian surplus ferrite issmaller, requires a lower drive current to flip, and has pretty much an ideal hysteresis curve forthe application. The current required is also an important consideration. It takes about 400mA totalcurrent to flip the BU ferrite, and about 1.5A to flip the F25 ferrite.

This Instructable is a work in progress for me as I will update and complete it with pictures "on the go", so every time I make a new batch or try a new recipe you will see the results here - so keep posted for updates.

* You can put it into any shape you need.

* It can be sanded or drilled when cured.

* It is a good shield against interference.

* Once mixed it ieasy to handle and form.

* You need no special equippment or expensive ingredients.

For high power applications, like an induction coil, you will need a thick layer of ferrite as otherwise the core will saturate or might heat up - it does not harm to use it too thick and you can always add more (around) if saturation is an issue, same for too much as you can sand t down or use a file.

Take a rough estimation of how much in final volume you need for your project and add another 20% be on the save side.

I go for the plaster version as this was my first way of doing it and because it the easiest.

Add the Iron Oxide in your mixing container (put gloves on now if you forgot about them ;) ) followed by about a third in volume of plaster.

Mix well while dry - for bigger batches using a jar with lid saves you a lot of black dust flying around!

Be aware that this mix dries a bit faster and can produce cracks in thicker layers, so working in small stages with new mixes to build up might be necessary (keep the dry mix and only use with water what you need).

To for moisture warm it up to about 30 celsius and place into a closed container that was in your freezer to cool down.

If not fully dry you will see a lost of mist condesing on the inside of your container.

It is very hard to give proper mixing ratios as the properties of the various resins differ too much but I work my way down from a 50/50 mix until I notice either mixing becomes a problem or the curing is not good enough.

I realised by the amounts of hits that I need to rush things a bit to get this instructable complete.

After experimenting with additives to reduce the risk of cracks forming during the drying process I decided to use this variation for a video on how to make the ferrite.

A bit of wall paper glue or if not available in your area use wood glue - this helps to keep the mix workable for longer and slows down the drying process also the finnished product does not crack as easy.

Add water (already prepared with glue) followed by your dry mix into a suitable mixing container and mix well until there are o lumps, keep the mix thick enough to work with but not so thin it runs off, unless you do a casting.

But do some small scale tests first to check if your mix produces cracks while drying.

Update 03/07/2015:

I experimented a bit more with a lot of different (possible) binders.

Nothing we would find at the local hardware store, the kitchen or local pharmacy (at reasonable prices) worked.

But then it hit me!

Grabbed my last few spoons of black oxide and mixed it with Sodium Silicate - Waterglass.

Of course, me being me, I did not take any pics or videos - shoot me...

Anyway I will try to explain:

Sodium Silicate is another "forgotten" chemical in terms of home use.

Some might still know it from the chemical experiment "Chemical Garden".

In the concentrated liquid form it is somewhere between full cream milk and warm honey in the consistency and glass clear.

Once dried it it goes rock hard - a feature used for repairs on wood, china ware and other things like heat resistant tiles.

If you know "Green Sand Casting" you are already familiar with just adding a tiny bit of water to that mix.

I did the same with the black oxide.

Started with a few tablespoons oxide and added the Sodium Silicate in tiny amounts.

Creates a lot of lumps and small balls, so doing this in a little ball mill might be a good idea (apart from the clean up bit).

Anyway, if you check the videos on grenn sand casting you will see the mix looks almost dry but keeps it shape when pressed - I tried the same but in the end just used a block form and small hammer to compact it.

(This reminds me to mention to get the ferrite mix out of the form after this step - I did not and it was impossible to remove the cured stuff from the form).

After this the testpiece went into the oven for about 90 minutes at full heat - this creates a nice and hard "ferrite".

To get is hard enough to be actually used it is placed into a kiln and is slowly heated to a glowing orange.

After that the cooling was done in the oven, preheated to full.

Oven was turned off once the piece was in and allowed to fully cool down over night.

The result was that

1. I was unable to get the cured ferrite out of the metal box I used.

2. It is so hard that I could not drill into it.

3. It does not break or crack.

I will try to find some more time and black oxide and make a short video of the process.

In the meantime, everyone still following can experiment as the only thing that matters is to just get the oxide moist with the Sodium Silicate so it binds together properly.

During the compacting a bit of excess might be pressed out indicating to use even less sodium silicate for the next mix.

The only downside is that you have to make a small furnace, metal melter or kiln so you can fully harden the mix, which is basically like a ceramic once cured.

3a8082e126