[PStechPaul]

I am working on an SCR gate drive board where I need a continuous DC current to two large SCRs connected anti-parallel for an AC switch on mains up to 480 or 560 VAC. That part of the design may not have much application for EVs, but it might be worth looking at the DC-DC converter I am designing to provide the gate drive. For SCRs of this size (up to 1000A 1600V), the gate may need 200-300 mA, and for my intended application of driving a highly inductive load (a power transformer with 10-20V secondary at up to 100,000 amps into a circuit breaker), we have found that it is best to keep gate current applied continuously.

Another important feature of the SCR trigger board is applying the gate drive at or close to the voltage peak, to reduce the DC offset and high surge current that will occur using a zero-voltage switch which is more common and ideal for resistive and incandescent loads.

I will show the overall schematic of this device, but mostly I will discuss the DC-DC converter that I made, which uses 12 VDC input and generates about 8-12 VDC output through a constant current circuit into the SCR gates. Two of these are needed, and the isolation must withstand continuous use on 480 VAC mains, which generally requires at least 4000 volts insulation test. Most DC-DC converters are not rated at this level, and the few that are, cost about $20 or more.

Page 1: Control section and drivers for DC-DC transformers:

Page 2: Transformers, gate power supplies, and drive circuitry:

 

[PStechPaul]

Here is a picture of the transformer bobbin with 30 turns of #24 AWG on primary and secondary. The coilformer is made for EI187 cores.

I had to open up the ID of the bobbin to fit the E20/10/6 cores I had. This one uses N87 material which has a higher AL and capable of 100 kHz and higher. The other one is identical except it has N27 material, recommended for 25-50 kHz.

Here is the output waveform with an open circuit:

With a bridge rectifier, 500 uF capacitor, and 47 ohm resistor:

I built the circuit on the PCB that I received Friday, and this is the open circuit waveform I got:

With a 47 ohm load:

I found that one of the SOIC-8 dual N-MOS FETs was actually an IRS2001, and both devices were drawing excess current and running hot. I replaced them with new FDS6930B devices, but it no longer had a waveform. I thought the IRS24530 was damaged, but I replaced it with no joy. Finally I discovered that I had an incorrect pinout with 1-2-3-4 as G-S-G-S, while it's actually S-G-S-G. Fortunately I was able to turn it upside down and it finally worked.

Pretty quick switching:

But when I assembled the board with Schottky rectifiers and capacitors with a 47 ohm resistor, it stopped oscillating. I thought perhaps the MOSFETs were fried because of the high current surges into the capacitors, so I added a 2 ohm resistor in series from the drive to the transformer primaries. It still had a problem, but I connected the SHUTDOWN input to GND and it worked OK.

Now I am thinking that I should add an inductor between the output of the bridge and the filter capacitors. Then maybe I can remove the 2 ohm resistor and not stress the MOSFETs and capacitors. Some additional design information:

http://www.smps.us/magnetics.html

http://ww1.microchip.com/downloads/en/AppNotes/01114A.pdf

 

[PStechPaul]

This DC-DC converter supplies about 9 VDC into a 47 ohm resistor, or about 200 mA and 1.8 watts. The IRS2453 is $1.48 in 10 piece quantities, and the two dual MOSFETs are $0.60 each for 10 pieces. The E20/10/6 cores are $0.38 each and the coilformer/bobbin is about $1.50. Other parts are about a dollar. So for about $5.94 you can put a high-isolation DC-DC converter on the PCB of a controller or charger. It's also repairable if something goes wrong. A rough measurement of efficiency showed about 97%. This DC-DC converter is unregulated, but should be OK for a gate drive. Changing the windings and adding a center tap could provide +/- 15 VDC for an IGBT drive.

I did simulations for the basic circuit with and without an inductor. With no inductor:

Adding a 100 uH inductor:

There are some serious current transients if I start the simulation from zero, but once it stabilized, There do not seem to be any extreme current spikes. Perhaps the leakage inductance of the transformer limits it somewhat. Indeed, with a coupling of 1.00, I see a damped current transient of 80 amps peak. Indeed, with 0.95 coupling, the initial input transient is just -12 amps. Apparently the source defaults to an ESR of 1 ohm. If I use 0.1 ohm there is an initial spike of 120 amps. It seems to dissipate in about 100 nSec. After the circuit stabilizes there seems to be very little current in the filter capacitor. The MOSFETs I'm using can handle peak current of 20 amps and the RdsOn is 38-50 mOhms. So for two in series and 0.2 ohms in each winding, the 12V supply would only be capable of 40 amps, and there may be some delay in providing the full 12V gate drive to the top devices, which may provide a soft start.

 

[WolfTronix]

I would be cautious about the filter inductor...

Small PWM pulses will cause the current stored in that inductor (while charging the gate capacitance) to become a small boost converter, the resulting high voltage could damage the gate.

I typically use liner regulators after the bridge rectifiers, makes for a more stable (less sag) supply under pulsed loading...

Schematic:

I did not layout a board, just did point to point:

 

[PStechPaul]

I don't really understand how the inductor could act as a boost converter when the driving waveform is essentially two overlapping 12V square waves with only about one uSec dead time. The waveform is not really PWM, and I am not trying to control the output voltage. The output voltage from the bridge rectifier is the same as the supply voltage, or determined by the turns ratio of the transformer. There is really not much energy storage in the filter capacitor, and thus very little current except at initial turn-on. The inductor current will be just that which is supplied to the load.

If this were used as an IGBT gate supply the PWM would be done after the filter capacitor, and current would flow from the bulk storage of the output capacitor into the gate capacitance.

 

[WolfTronix]

I misunderstood how you where using the circuit.

 

[PStechPaul]

I have done some more work on these DC-DC converters. I had also posted the following on the "Another Homebrew AC Inverter" thread:

Here is an excerpt from my post on sci.electronics.design for a 12V-12V DC-DC converter that was able to produce 4 watts.
The parts are probably about $5.

I just received some EPCOS EE13/7/3 ferrite cores in type N87 and a matching bobbin for dual isolated windings. I was able to fit 12 turns
of #24AWG magnet wire for primary and secondary. I used some heat-shrink tape to hold the core halves together, and the open circuit
inductance when hot measured 156 uH, but then settled to 118 uH as expected at room temperature. Leakage inductance is 4.7 uH. Here it is in my simple test jig:

The output waveform with 13VDC input at about 30 mA:

The asymmetry is probably because of the ugly wiring. The IRS2453D has 1 uSec dead time, which you can see in the 90 kHz waveform.
I added a 15 ohm (actually 17 ohm) 2W resistor to see how much the output dropped under load, and I was surprised it didn't drop very much,
and looks like about 10.5 VRMS, or 618 mA and 6.5 watts. Input was 13 VDC at 590 mA, or 7.67 watts, and 85% efficiency.

Then I connected the output to a FWB of 1N5818 Schottkys, a 20 uF 25V CM capacitor, and a 33 ohm resistor. Here are the results at various input voltages:

13V 0.37A 4.81W 11.77V 0.357A 4.20W 87%
15V 0.43A 6.45W 13.56V 0.411A 5.57W 86%
17V 0.55A 9.35W 15.08V 0.457A 6.89W 74%
18V 0.70A 12.6W 15.90V 0.482A 7.66W 61%

I am pleasantly surprised at the performance of this little transformer. The EPCOS catalog rates this size core at 5W for N27 at 25 kHz, but 28W for N87 at 100 kHz.
I would have figured on 4x, but 5.6x is surprising. Probably because of greater surface area per watt for the smaller core.

The output voltage for a gate driver does not need to be tightly regulated, if the input is known and solid. It would not be difficult to wind the output for any
voltage you want, and perhaps even two windings for the +15 and -5 or whatever for IGBTs. You could use a 12V-20V and then a 15V zener on the output for the
positive voltage (which is more critical).

I also found that you can easily get 2 watts from a small common mode choke (less than $1 each). They have dual windings with inherently high isolation but also
high leakage inductance and poor regulation.

This is a TLF9UA202:

Waveform open circuit:

With 33 ohm DC load:

 

[Tony Bogs]

Thanks again, Paul!

I knew I had seen another IRF gate driver with a built-in oscillator. A short search on the Infineon website: AUIR2085S.
8 pin SOIC, automotive grade. HEV Auxiliary Converter is listed as typical application.
Frequency up to 500kHz, halfbridge with a capacitive divider (50%) as the other leg.
It also has the LC output filter in the "Typical Connection Diagram" on the first page of the datasheet:
http://www.infineon.com/dgdl/auir2085s.pdf?fileId=5546d462533600a4015355a82609133f

The 2085S circuit looks similar to a LLC converter. A chain of thought started with this topic and it finally dawned on me how I can simplify the LLC part of the SiC charger design.