Output Transformer design
Or why is that nasty buzzing sound coming from the transformer? And, oh look at the pretty smoke.!
The output transformer needs thoughtful design and careful construction. I am not an electronics professional or an expert on transformer design - these are few things that I have learned from making transformers, reading and from others on the list.
Suitable Transformer Cores
These are both capable of 500 watts plus, available from Farnell, the ETD49 is also available from RS. What you are looking for if you are buying or salvaging transformers is: a core material which isn't too lossy at the working frequency and then get the biggest size available or affordable. Small transformer need more turns to avoid saturation in a small winding window. These transformers that I'm using are 70mm square, EC70 and 49mm square ETD49. You need to order 2 E cores, a bobbin and a pair of steel retaining clips.
Others have used toriod cores but these are more difficult to wind.
What we are trying to do is to match the impedance of the mosfet driver to the impedance of the coil for maximum power transfer. The impedance of the driver is simply V/I where V is the operating voltage and I is the maximum permitted current.
The impedance of the coil is harder to calculate ByLund uses the DC resistance of the coil as a first approximation. You can do cunning calculations but in the end it comes down to suck it and see. I buy extra bobbins so that I can try different ratio’s. Try something fairly conservative or build a multitap transformer – measure the current and try again. Counter to intuition producing corona and arcs actually increases the input impedance of the coil, so you will want a higher step up ratio than you calculate.
My first transformers were small and broke down internally as the voltage went up.
The output from the transformer may be 2 or 3KV and I observed corona on the wires leaving the transformer. Capacitance will induce a high voltage on the transformer core – which must be either grounded or insulated. (Don’t touch the core to see if it’s heating - ouch!).
Coat each layer with polyurethane varnish (not tinted or water based) or use a thin layer of clear silicone bathroom sealant, or epoxy resin (which conducts heat better).
Don’t wind the upper layers to the very edges of the bobbin, indeed start and finish each secondary layer a little further in. Don't wind from one end to the other and then wind back again - you get zero voltage difference between layers at one end and double at the other. Wind each layer separately take the connections out of the transformer and at the end externally connect the beginning of one coil to the start of the other. This also gives a multitapped secondary which is useful.
Put heat shrink insulation on every lead you take out of the core. Keep the connections as far apart as possible insulate well to prevent arcing. Insulation tape unfortunately is usually sold only in half inch width, which means you either over lap or have a gap in the middle. I am considering using several layers of parcel tape instead. You will need 2 or 3 layers at least.
Rule number 1: the bobbins are always too small for all the wire and the insulation that you want to put on.
If you use half the available bobbin space for the primary and half for the secondary (and the other half for the insulation – see rule 1.) then the power loss in primary and secondary will be about equal.
(You want proof: If the turns ratio is 1:n then in the primary the wire can be n times the area of the wire in the secondary, where the wire is n times as long. The secondary will thus have n squared times the resistance of primary. The current in the secondary is 1/n of the current in the primary. If the current through the primary is I and the resistance is r the power dissipation is I*I * r for the secondary we have (I/n * I/n) *r * n * n which is I * I * r. )
Of course in the real world the secondary windings are slightly longer than the primary windings, and there has to be more insulation between layers in the secondary.
Skin depth at various spot frequencies
The skin effect is what keeps us alive when we are accidentally zapped by a tesla discharge. High frequency current doesn’t penetrate.
The depth of penetration is given by: D = 66/SQRT(F)
D is the skin depth in mm, SQRT is square root of, F is frequency in Hz
This means that if say 2mm diameter wire is used for a primary at 200KHz then only about 20% of the wire area will conduct. Multi stranded cable is no better – it is a single wire with a rough surface to the signal. I therefore construct my primaries using a bunch of 0.4mm enamelled wires. At 100KHz the whole of each wire conducts, (well most of it - there is also the proximity effect) its still better though and also a bunch of 0.4mm wires is much easier to work with than 2mm gauge wire.
Work out the length of wire required (n*pi*dia or use a bit of string add 20cm for connections and 10cm for luck)
Work out the number of strands that can be fitted into the length (L) of the bobbin
Cable diameter DIA = L/n number of turns
Divide the cable area by the wire area to get the number of strands - subtract 25% to allow for packing limitations or make up small trail size pieces of cable and measure.
Now fix two posts the required length apart and wind wire from one post to the other n times. Release one post tension the wire and put in some twists – keep it even and don’t over do it.
Wrap your primary turns cut to length and the tin every wire. (I find it best to use a 150W soldering gun and to scrape of some enamel first (hold the wire between two pieces of sand paper and pull (if it breaks you held too tightly).
Making multistrand cables for secondaries where the length is greater and the width is small is less practical and I use the largest gauge of wire that will fit.
Ferrite (or any magnetic material) can only support so large a magnetic field before it is saturated. Typically this is 3200guass = 320mT.
N=V*10^8/(2*f*B*A) N is the number of primary turns, V is the working voltage on the primary, f is the frequency in Hz, B is the flux density in gauss, A is the cross sectional area of the centre pole in square centimetres.
I have tried to build transformers so that at most B will be less than 2000 gauss.
Saturation is more of a problem in push pull designs, the two sides are never completely equal so over a number of cycles the core becomes magnetised and saturated in one direction. It is therefore very important that some means of current limiting is employed. In half bridge and full bridge schemes a capacitor (e.g. 1uf 250v AC polypropylene) can be put in series with the transformer primary blocking any DC current.
Saturation can be avoided by having an air gap between the two core halves, naturally there is a down side to this and from what can gather ungapped transformers are the norm in push pull and bridge designs.
For maximum efficiency transformer manufacturers interweave primary and secondary layers. I haven't done this for ease of construction and insulation I wound the primary first and the secondary on top, taking the bottom layer to ground . This also reduces the interwinding capacitance which is another source of loss. But on doing a bit of reading the interweaving may be better.
Their is a limit to the amount of power that can be pushed through a single transformer - they get hot! One way round this is to use two transformers with either the secondaries or the primaries in series to ensure that the power is equally shared. I've also experimented with a using a double core - it works but I'm not sure if it any better (either from the point of view of efficiency or ease of construction) than using two transformers in series or parallel.
This is actually two transformer bobbins with the top protrusions removed and then glued together - a DIY large transformer which is an alternative to using 2 transformers in series for more power.
Be prepared to make several windings - till you get it right. Watch for arcing and heat - use an interrupter to reduce the power if it over heats - if it arcs its probably time to wind another one!