Gate Drive Transformer Design I am frequently asked about the design of the ferrite gate-drive transformers used in my Solid State Tesla Coil design. Particularly regarding the choice of core material, and the way in which it is wound. Unfortunately, the cores that I used for that project were ones that I had lying around in my junk box. Since ferrite cores never have any identification markings on them, I cannot say what the part number is, or even what the material grade is! A wide range of transformer cores, ferrite rings, beads and sleeves are currently available through the usual component suppliers. However, a large number of these are intended for high power applications or interference suppression purposes and are far from ideal for use in high speed gate drive transformers. The requirements of the gate drive transformer in this application are quite stringent. The transformer must pass pulse waveforms at a few hundred kilohetz with minimum distortion. An ideal gate-drive transformer would not adversely effect the rise and fall times, would introduce no overshoot or ringing after the rising and falling edges, would not limit the peak current delivered to the MOSFET gate, and would not cause any sag in the flat tops and bottoms of the waveforms. However, a real gate-drive transformer does all of these things! At least to some degree, but we can minimise these effects by careful design and construction. The purpose of this web page is to document some experiments that I have done to determine what cores, and what winding methods give the best performance gate-drive transformers. I wanted to be able to make a list of readily available cores that are suitable for gate-drive transformers in the 80kHz to 400kHz range. This is so that others could purchase a particular core type with some confidence that they will be able to make it into a useful transformer. I also wanted to investigate some of the clever winding schemes and "tricks of the trade" used to optimise the transformer performance once a suitable core has been selected. In short, this paper describes what cores and what winding techniques give best results. If you are only interested in seeing the results then click here to jump straight to the results and design tips. However, if you want to learn more about gate-drive transformers and how the testing was performed, then read on... Testing: I tested a total of ### different core types. Most of these were toroidal rings or sleves made of various material grades, but some E-I and planar cores were also tested. For the cores that performed well, I also tried different winding schemes to see what improvements, if any, they might contribute to the properties of the transformer. Each core was wound with one primary winding and two secondary windings. This is the most common case where one drive circuit is used to control two complemetary switching devices. Equal numbers of turns were used for the primary and secondary windings so that all transformers had ratios of 1:1:1 The actual number of turns varied depending on the core material and available winding space. I tried to make all of the transformers have a similar Magnetising Inductance, by employing more turns on those cores that had a lower Specific Inductance (Al) value. This is necessary to make a fair comparison between different core materials. We have two desirable properties for the pulse transformer. Firstly, we want high inductances for the primary and secondary coils, so that the magnetising inductance is high and the magnetising current drawn from the driver is fairly low. This causes the minimum of droop in the flat tops and bottoms of the pulse waveforms through the transformer. Secondly, we want to achieve the highest coupling coefficient between the primary and secondary windings that is possible. This minimises leakage inductance, and reduces voltage overshoot and ringing in the pulse waveforms at the secondary windings of the transformer. Therefore, I have defined a "Figure of Merit" for comparing different core materials and winding schemes. This Figure of Merit is defined as the magnetising inductance divided by the leakage inductance. Since we want a large magnetising inductance and a small leakage inductance, then this Figure of Merit should be as large as possible. I believe that this is a fairer way to compare cores and winding techniques, than simply comparing leakage inductance. This is because some of the design requirements for minimum leakage inductance are in direct contradicition to some of the requirements for high magnetising inductance. Determining the Magnetising Inductance, Each transformer was first tested with both secondary windings open circuit, to determine the magnetising inductance. This is the inductance seen at the primary winding with no load connected to the secondary windings. The magnetising inductance was found by parallel resonating the primary winding with a 22nF low loss polyproylene capacitor. This resonant circuit was swept with an RF signal generator and the parallel resonant frequency was noted. The actual inductance value was calculated from the resonant frequency. Determining the Leakage Inductance, Two tests were performed in order to determine leakage inductance values for each transformer. Firstly, one secondary winding was short circuited as close as possible to the ferrite transformer. The other secondary winding was left open circuit. The resonant frequency of the primary was measured again, in order to determine the leakage inductance. In this case the inductance seen at the primary is much lower with one secondary short- circuited. It represents the inductance that is not common to both the primary and secondary winding and should be as low as possible. Next, the other secondary winding was also short circuited as close as possible to the transformer. The resonant frequency of the primary circuit was measured again, in order to determine the new leakage inductance. The inductance seen at the primary is lower again with both secondary windings shorted. It is now due only to the magnetic flux generated by the primary that does not couple with either of the two secondary windings. We also want this value to be as small as possible. Note 1: This parallel-resonant testing method is thought to be valid, since the 22nF capacitor swamps any self-capacitance of the winding being measured. The 22nF part was also selected to have very close tolerance to the intended value, and the same part was used throughout all of these tests. Note 2: Great attention was payed to minimising lead lengths around the transformer being tested, and those to the low inductance 22nF capacitor. This is important when you consider that an inch of wire posseses around 25nH of inductance. In the case of the better transformer designs, 25nH is a considerable fraction of the leakage inductance we are trying to measure! For this reason, the secondary windings were always shorted right at the transformer body when performing leakage inductance test. Note 3: I found it necessary to keep the amplitude of the RF signal source quite low in order to see the full Specific Inductance from some ferrite cores. This is due to non-linearity in the magnetising B-H curve for some of the materials being tested. This behaviour is perfectly normal for high permeability ferrite grades. It does not indicate core saturation, but mearly a gradual decrease in permeability as the magnetising flux density is increased. For this reason an amplitude of 1V pk-pk was used for all of the tests described here.) Orange ferrite toroid Ferroxcube 4330-030-34600 UK supplier: Farnell 305-6995 (£1.24) Dimensions: OD=25.8mm ID=14mm W=10.6mm Material: 3E25 Spec Ind: 5620 nH (Supplied by Alan Sharp) 9 + 9 + 9 turns of 0.5mm² multi-strand wires wound in seperate bundles spaced 120 degrees apart around the toroid. F0 = 46.45 kHz Lm=533.6 uH (Al = 6590 nH) F1 = 497.9 kHz Ll=4644 nH k=0.99564 (Lm ÷ 115) F2 = 558.7 kHz Ll=3689 nH k=0.99654 (Lm ÷ 145) 9 + 9 + 9 turns of 0.5mm² multi-strand wires interleaved around toroid. The combined interleaved winding was spaced evenly around the toroid. F0 = 46.91 kHz Lm=523.2 uH (Al = 6460 nH) F1 = 2.221 MHz Ll=233.4 nH k=0.99978 (Lm ÷ 2240) F2 = 2.401 MHz Ll=199.7 nH k=0.99981 (Lm ÷ 2620) 9 + 9 + 9 turns of 0.5mm² multi-strand wires tightly twisted together. The "twisted-threesome" was then wound evenly spaced around the toroid. F0 = 47.03 kHz Lm= 520.6 uH (Al = 6430 nH) F1 = 2.352 MHz Ll= 208.1 nH k=0.99980 (Lm ÷ 2500) F2 = 2.675 MHz Ll= 160.9 nH k=0.99985 (Lm ÷ 3240) 9 turns + 9 turns of 3mm lapped-screened wire interleaved around toroid. Screens paralleled and used as one primary winding, centre cores used as secondary windings. F0 = 46.68 kHz Lm=528.4 uH (Al = 6520 nH) F1 = 3.038 MHz Ll=124.8 nH k=0.99988 (Lm ÷ 4230) F2 = 4.128 MHz Ll=67.6 nH k=0.99994 (Lm ÷ 7820) ________________________________________________________________________ Large TVI suppresion ring. UK supplier: Maplin ##### Material: #### Dimensions: Spec Ind: #### nH 26 + 26 + 26 turns of 0.5mm² multi-strand wires tightly twisted together. The combined interleaved winding was spaced evenly around the toroid. F0 = 67.83 kHz Lm=250.2 uH (Al = 370 nH) F1 = 1.396 MHz Ll=590.8 nH k=0.99882 (Lm ÷ 425) F2 = 1.634 MHz Ll=431.2 nH k=0.99914 (Lm ÷ 580) ________________________________________________________________________ Grey ferrite toroid NMG Neosid 28-795C36S UK supplier: RS 232-9561 (£0.99) Dimensions: OD=22.1mm ID=13.7mm W=12.7mm Material: F9C Spec Ind: 6110 nH 9 turns + 9 turns of 0.5mm² multi-stranded wires tightly twisted together. The combined interleaved winding was spaced evenly around the toroid. F0 = 54.50 kHz Lm=387.6 uH (Al = 4790 nH) F1 = 2.042 Mhz Ll=276.1 nH k=0.99964 (Lm ÷ 1400) F2 = 2.374 Mhz Ll=204.3 nH k=0.99974 (Lm ÷ 1900) 8 turns + 8 turns of 3mm lapped-screened wire interleaved around toroid. Screens paralleled and used as one primary winding, centre cores used as secondary windings. F0 = 58.06 kHz Lm=341.6 uH (Al = 5340 nH) F1 = 3.309 Mhz Ll=105.2 nH k=0.99985 (Lm ÷ 3250) F2 = 4.319 Mhz Ll=61.7 nH k=0.99991 (Lm ÷ 5540) ________________________________________________________________________ Grey ferrite toroid NMG Neosid 28-780C36 UK supplier: RS 232-9808 (£0.88) Dimensions: OD=25mm ID=15mm W=10mm Material: F9C Spec Ind: 5100 nH F0 = 61.07 kHz F1 = 2.089 Mhz F2 = 2.385 Mhz ________________________________________________________________________ Small blue ferrite toroid NMG Neosid 28-794C36S UK supplier: RS 232-9555 (£0.39) Dimensions: OD=13.9mm ID=7.5mm W=7mm Material: F9C Spec Ind: 4250 nH 9 + 9 + 9 turns of 0.5mm² multi-strand wires tightly twisted together. The combined twisted threesome was spaced evenly around the toroid. F0 = 57.44 kHz Lm=349.0 uH (Al = 4310 nH) F1 = 2.376 Mhz Ll=204.0 nH k=0.99971 (Lm ÷ 1710) F2 = 2.729 Mhz Ll=154.6 nH k=0.99978 (Lm ÷ 2257) ________________________________________________________________________ Large blue ferrite toroid ####-#### UK supplier: ####-#### Dimensions: OD=##.#mm ID=##.#mm W=##.#mm Material: #### Spec Ind: #### nH 9 + 9 + 9 turns of 0.5mm² multi-strand wires tightly twisted together. The combined twisted threesome was spaced evenly around the toroid. F0 = 88.11 kHz Lm=148.3 uH (Al = 1830 nH) F1 = 2.118 Mhz Ll=256.7 nH k=0.99913 (Lm ÷ 580) F2 = 2.435 Mhz Ll=194.2 nH k=0.99935 (Lm ÷ 765) ________________________________________________________________________ E-E core set, Ferroxcube ETD29/16/10 UK supplier: #### Dimensions: #### Material: #### Spec Ind: 2100 nH 15 + 15 + 15 turns of 0.5mm² multi-strand wires wound in three concentric layers around the centre limb. A standard ETD29 bobbin was used to facilitate easy winding. F0 = 48.28 kHz Lm=494.0 uH (Al = 2200 nH) F1 = 864.3 kHz Ll=1541 nH k=0.99688 (Lm ÷ 320) * F2 = 1.349 MHz Ll=632.7 nH k=0.99936 (Lm ÷ 780) ** * Outermost secondary winding shorted ** Both secondary windings shorted Inner secondary "magnetically shields" outermost secondary winding ________________________________________________________________________ Winding the primary and secondary coil on different parts of the core gives relatively poor coupling and results in high leakage inductance. This causes slow rising and falling edges accompanied by overshoot and ringing which is difficult to damp when used as a gate-drive transformer. Therefore windings should never be separated like this unless it is absolutely essential. For example, to minimise inter-winding capacitances or withstand a large voltage between windings. Interleaving the primary and secondary windings gives a big improvement in coupling, and dramatically reduces the leakage inductance. In this test, the leakage inductance was reduced by a factor of twenty, just by rewinding the transformer with interleaved windings. This translates to much faster rise and fall times, and far less series resistance required to damp overshoots and ringing. It is essential that primary and secondary windings be at least interleaved to obtain good performance in gate-drive applications. However, we may be able to make further improvements. Twisting the primary and secondary windings together would seem to improve coupling. And indeed it does, but the improvement is not as dramatic as our first change. In this test, twisting the primary and secondary windings together and then rewinding the transformer only gave a 15% reduction in leakage inductance. Whilst not as dramatic an improvement as I had hoped for, it is still an improvement. Perhaps the improvement is somewhat offset by the fact that twisting the wires into a spiral bundle moves some portions of the wire length further away from the ferrite core. It should also be noted that it is easier to wind a twisted bundle through the ferrite core than winding several individual wires by hand, especially if the transformer has numerous windings. It is also possible to fit in more turns, or use thicker wire, when the wires are twisted into a bundle. Three wires twisted together take up less of the inner circumference of the toroid, than three wires laying side by side. Using twin-screened cable appears to be the ultimate in achieving tight coupling. The two screens are paralleled and the two inner cores are used as the secondary windings. In this case the primary windings totally enclose the secondary windings due to the construction of the screened cable. This ensures very good coupling and gives a worthwhile decrease in leakage inductance. In this test, the leakage inductance was approximately halved by rewinding the transformer with twin lapped-screened cable instead of twisted wires. There was also a dramatic reduction in the leakage inductance with both secondaries shorted as opposed to shorting only one. This is because of the very tight coupling between each primary and its associated secondary winding. This method of construction gives the fastest rise and fall times possible, and requires the minimum of series resistance to damp any overshoot and ringing. Therefore it is highly recommended. The only possible problems I see with this method, are that inter-winding capacitance is quite high due to the very close proximity of the primary and secondary wires. This is an unavoidable consequence of striving for the minimum leakage inductance. However, it is thought that even a few hundred picofarads of stray capacitance in the gate- drive transformer would not be significant in this application when compared to the many nanofarads that large die MOSFETs exhibit at their gates. It is also worth remembering that small diameter screened cable is often targeted at small signal audio and RF applications, and sometimes does not have particularly thick insulation between its centre core and screen. It is important that suitably rated screened cable is used here to prevent breakdown of the insulation between the centre cores and the outside screen. Such a breakdown would invariably damage the control circuitry! Finally, I had a quick look on the internet to see what figures are typical for "off-the-shelf" pulse transformers available from places like RS and Farnell here in the UK. I was shocked to see leakage inductance figures in the micro-henries and tens of micro- henries for a range of small ferrite pulse transformers stocked by RS. The primary inductance figures are also much higher than required for our application, so these are clearly intended for operation at much lower frequencies. Those offered by other suppliers are equal in their unsuitibility for use as gate-drive transformers in SSTC inverters. It is unlikely that anyone will find a gate-drive transformer available off-the-shelf with sufficiently low leakage inductance for driving MOSFETs at several hundreds of kilohertz. There is a reason for this. Most commercial power electronics designs either use direct drive to the MOSFETs, or use lower frequencies in very high power applications. In the few instances where isolated gate-drive signals are required in the hundreds of kilohertz range, the designers simply choose a core and design the transformer to suit there requirements. With this information, some relatively simple test equipment, and a little time, you can easily design and build gate-drive transformers that are perfect for driving MOSFETs in SSTC applications. Such home-built pulse transformers will easily out perform off-the-shelf devices by a long way. You will also learn a bit about the design of transformers in the process. This work is unfinished.