Gate drive transformer
From HvWiki
A gate drive transformer is used to drive the gate of a MOSFET transistor and has one or more of the following functions:
Contents |
Quick Tips
- Get a core of the correct type
- Calculate minimum number of turns required to avoid saturation
- Calculate the magnetising current and the gate current
- Select the wire diameter (based on the amount of current) and wire type
- Wind the transformer
- Test with a small series resistor, increasing or decreasing the value to control the ringing
- Repeat any of the above steps as necessary until it works correctly
Core
The best performance for GDTs can be achieved by using toroidal (doughnut) shaped cores. Toroids have the advantage of a continuous flux path, resulting in low flux leakage and therefore a lower leakage inductance. Toroids also are available in a wide range of sizes.
Other core shapes, like EE, EI, CI and pot cores, can all be used but with reduced performance due to increased leakage inductance. If these cores are all that is available, then they must be ungapped i.e. no space between the core halves. The cores should be fixed in place to ensure that they are held together firmly. Introducing an air gap dramatically reduces the AL, reducing the inductance and increasing the magnetising current.
Core material
Ideally, the material needs to be ferrite with a relatively high permeability / AL value, preferably higher than 4000 nH/N2, but the higher the better. Also, the material should be rated for operation into the MHz region.
Suitable manufacturers / materials include:
- Epcos - "broadband transformers, up to 3 MHz"
- Ferroxcube -
However, generic power grade ferrite can be made to work, albeit with lower performance.
Avoid powdered iron cores, from manufacturers such as Micrometals as these are designed for DC choke applications with a low AL and losses that increase significantly with frequency. These toroids are most often found in the outputs of PC power supplies where they are used for output filtering and coupling the outputs. Don't be tempted to use them!
Also, unless you can't find any other cores, avoid using EMI suppression toroids as the material is designed to have losses in the MHz region to reduce the amount high frequency noise.
Also avoid materials like Ferroxcube 3R1 which are designed to be used in magnetic amplifier (magamp) regulators.
Inductance
Inductance Formulae
Inductance of a certain number of turns on a ferrite core can be calculated by using the AL value found in the datasheet which is given in nano-Henries-per-turn-squared or nH/N2. The formula for calculating inductance is very simple:
- L = ALN2
Example
A core with 15 turns on it yields an inductance of 815 uH - what is the AL? Well, 815 uH is 815 000 nH so:
So for this same core, how many turns would be needed for an inductance of 4.2mH (4.2e+6 nH)?
![N = \sqrt{\frac{L}{A_L}} = \sqrt{\frac{4.2 \cdot 10^6}{3622}} = 34.052 \approx 34 [turns]](/images/math/8429e3ae211ac2e6095c643dbf8a70eb.png)
Design Tradeoffs
What does the inductance need to be in a GDT? Well, the tradeoff is between two choices
Minimise the the number of turns to reduce inductance and to minimise leakage inductance (making sure you don't saturate the cores) Increase the number of turns to reduce the magnetising current
Magnetising Current
Otherwise known as "magnetizing current" (for our viewers from the other side of the pond) this is the current required to maintain the magnetic field in the transformer. Taking the equation for voltage across an inductor...
...and rearranging it, we can see the peak magnetizing current is determined by
where Ipeak is the peak magnetising current, t is the period and D is the duty.
Because the current through an inductor appears as a triangle wave, the rms current Irms is a more meaningful result to us. For a triangle wave, this is given by
- Irms = 0.577Ipeak
where Irms is the rms magnetising current.
This current, added to the current required to charge and discharge the gate of the MOSFET, is the total current required for the gate drivers to supply.
Example Calculation
Assume we have a GDT and driver with the following characteristics: -12V to +12V, 100kHz drive signal with a 50% duty cycle being fed into a GDT with a 1mH winding inductance.
- = ( [ 12 - -12 ] * [ 1 / 100e+3 ] * 0.5 ) / 1e-3
- = ( 24 * 10e-6 * 0.5 ) / 1e-3
- = 0.12 A
- = 70 mA
If we double the winding inductance to 2mH, this will reduce the magnetising current by half to 35mA. Bearing in mind that the inductance is proportional to the number of turns squared, it is easy to add a few more turns to reduce the magnetising current.
Conclusion
We can see that the magnetising current is relatively small but for GDT with less turns it can become significant. Therefore more turns for a higher inductance is preferred.
However, leakage inductance will increase with the number of turns and leakage inductance affects the performance more significantly than magnetising current.
Therefore, select the best winding technique to minimise leakage and use as few turns as you can without having a large magnetising current and without hitting saturation.
Saturation flux density
Saturation flux density is how much magnetic flux the magnetic core can handle before becoming saturated and not able to hold any more. This depends on several factors including ferrite type, temperature and electrical and magnetic conditions on the transformer.
When the ferrite saturates, the transformer no longer acts like an inductor with a linear increase in current over time. Rather, the magnetic field cannot increase further and current is limited by the source impedance of the power supply and the resistance of the transformer wire. This leads to very large currents and blown devices.
Ferrite typically saturates at a flux density (B) about 0.3 Tesla but this depends on temperature and the ferrite type. A typical design might have as a target a B = 0.25T at 125ºC which gives plenty of operating margin to the limits.
The formula for flux density is
where
- B is flux density in Tesla
- V is the applied Voltage to the winding in Volts
- t is time that Volts V is applied for in seconds
- Ae is the cross sectional area of the core in square metres - obtainable from the core datasheet
If this were an inductor carrying a DC component, the DC current would play a significant part in calculating the saturation.
The main thing to take from this equation is that for a given core and frequency there is a minimum number of turns that can be used without saturating the core. So, to avoid saturation, use more than the minimum number of turns.
Examples
We have a 50% duty cycle, 100kHz drive signal @ 12V peak. Core has a XSA of 50mm2 (= 50.10-6 m2) What is the minimum number of turns assuming a maximum B = 0.2T
A 100kHz drive signal gives a total period of
- t = 1 / f
- = 1 / 100e+3
- = 10 us
And at a 50% duty cycle, this gives an on time of 10 us / 2 = 5 us. Plugging the numbers in to the formula we get a minimum number of turns of:
- N = ( V x t ) / ( B x Ae )
- = ( 12 x 5e-6 ) / ( 0.2 x 50e-6 )
- = 60 / 10
- = 6 turns
The V x t product is referred to as Volt-seconds and is the voltage that can be applied for a certain amount of time. In the above case, the volt seconds were 60 V.us (not V/us, we are multiplying!). So how long can we apply 18V to this core, assuming the same amount of turns and same flux density?
- t = 60 V.us / 18V
- = 3.3 us
Wire and winding
Because one winding of the GDT is connected to the source of the high side MOSFET, it will assume the DC rail voltage when that device is turned on. Therefore, the insulation needs to be able to withstand 350V in a typical SSTC application.
Enameled copper wire (ECW) will give the highest performance but the thin insulation is easily damaged and it is difficult to get reliable results without flashover. The insulation varnish is specified in Grades, the number of which determine the number of coats of varnish that the wire has received. So Grade 2 wire has received two coats of insulation. Grade 1 insulation is good for several kV and is good enough for a GDT.
The advantage in using ECW, also called transformer wire or magnet wire, is that the insulation is physically thin. This enables high coupling between the windings, reducing the leakage inductance. You may be able to use some of the leftovers from winding your Tesla Coil secondary if it is thick enough.
Category 5 Ethernet cable has been used successfully by several people for their GDTs and is very reliable. It is best to remove the outer sheath, which will allow the wires to get closer to the core, reducing leakage.
Normal hookup wire isn't ideal, but can be used.
Wire Current
As for the current that the wire must withstand, there are two components to consider: Total wire current = magnetising current + gate charge/discharge current.
The gate current peaks on the switching transitions where the MOSFET gate capacitance has to be rapidly charged and discharged. To calculate the RMS current it is better to speak in terms of gate capacitance charge. We know that:
- Q = CV
and
- I = Q / t
therefore
- I = ( C * V ) / t
where
- I is current in Amperes
- C is MOSFET gate capacitance in Farads
- V is gate voltage change in Volts
- t is time that the voltage is applied for in seconds
So for a -12V to +12V, 100kHz system driving a 15nF gate, the RMS current would be
- I = ( C * V ) / t
- = ( 15e-9 * 24 ) / 5e-6
- = 72mA
Assuming a magnetising current of a similar magnitude, you could reasonably assume an RMS current through the wire of 150mA.
There will be some current due to the interwinding capacitance, but we shall ignore this for our purposes.
Wire Diameter & Current Density
The formulae for current density is
- CD = I / X
where
- CD is current density in A/mm2
- I is current in Amps
- X is cross sectional area of the wire in mm2
A typically quoted maximum current density for wires is 5A/mm2. So, for a current of 150mA, the minimum wire area would be:
- X = I / CD
- = 0.15 / 5
- = 0.03 mm2
Given that:
- X = π * r2
the diameter is given by:
- = 0.195 mm
where d is the diameter of wire in mm
So thin wire should be acceptable for GDT applications give the currents involved.
Leakage inductance
Leakage inductance is a parasitic component of transformer design caused by poor coupling between windings. This leads to small amouns of magnetic flux "leaking" out and not being transferred to other windings. It appears electrically as a separate inductor in series with the windings and causes problems.
Primarily, in SSTC applications, we are concerned with the amount of ringing caused by the leakage inductance resonating with the gate input capacitance of the switching device. With ringing, there are a few main concerns:
The ringing overshoot on waveform edges can be enough to breakdown the gate insulation of MOSFET devices. Most MOSFETs are specified to +/- 15V of gate drive voltage (relative to the source) with some having +/- 20V or higher. Back-to-back 15V zener diodes are seen in many designs to act as protection in the case of excessive ringing to prevent the gate being damaged. With good GDT design, these should not be necessary.
The STE70NM60s MOSFETs have a rating of +/- 30V with internal back-to-back zener diodes to clamp the voltage. This seems to be a common feature in ST power MOSFETs and one that is appreciated.
Excessive leakage inductance, combined with insufficient damping resistances can distort the waveform beyond all recognition and in this case (opposite) the ringing is sufficient to change the operating state of the MOSFET between on, linear and off modes.
The high frequency ringing could couple into other circuits and affect performance, cause false triggering, etc.
Some of the ringing could be reflected back to the GDT primary driver, causing damage to the device. This needs to be avoided as the intended input signal is totally overridden by the ringing, making the waveform useless, as shown in the waveform on the right. In fact, red waveform, if appearing on a MOSFET gate, wold turn the device on and off multiple times during one switching period.
Leakage inductance also causes a time delay in the signal which can be seen in the above waveform as the phase difference between the input waveform (green) and the output waveform (red). This is because the voltage across an inductor cannot change instantaneously - the larger the leakage inductance, the larger the time delay. This is examined below in relation to the possibility of it causing shoot through conditions.
Leakage Inductance Calculation
To reduce leakage inductance, we must increase the coupling between the windings. It's that simple.
Well OK, it's not that simple. We'll start by defining the formula for interwinding coupling as
where
- k is coupling coefficient (no units)
- Lleak is leakage inductance (Henries)
- Lmag is magnetising inductance (Henries)
Magnetising inductance remains constant for x number of turns, so to decrease the leakage inductance we need to increase the coupling between windings.
To get maximum coupling, the windings would need to occupy the same physical space as each other for a coupling of 1 (unity), which is clearly not possible. The next best thing is to get them as close as possible. Various suggestions have been made as to the best way of doing this, but there didn't seem to be any data to back it all up.
Trace Inductance
Trace inductance is the parasitic inductance of the traces connecting the driver to the GDT and the GDT to the MOSFET gate-source terminals. It has a similar effect to leakage inductance in that is causes ringing and time delays. This can be minimised by using short, wide tracks as connections between driver, GDT and MOSFET and also by keeping the current loop area small.
Ribbon cable, with alternate + and - wires also has quite a low inductance, as does tightly twisted wires. Double sided PCB, with the gate trace on the top layer and the return on the bottom will also have a low parasitic inductance. Keep those loops small!
Driver
It should be noted that the GDT merely couples the drive power to the gate of the transistor. As a result, whatever devices used to drive the GDT primary need to be able to supply the high peak currents required to rapidly charge and discharge the transistor gate.
MOSFET gate drivers are ideal for this task, such as the following devices:
- Texas Instruments UCC37322 non-inverting and UCC37321 inverting drivers are the most popular choices, with a 9A peak drive current capability.
- For the more difficult, higher capacitance IGBT gates, Ixys make a selection of high current gate drivers, including the popular IXDD414
- Maxim also make a selection of gate driver ICs
