Maker Pro
Maker Pro

Isolated ground planes on a MCU controlled high voltage IGBT

stube40

Feb 9, 2010
85
Joined
Feb 9, 2010
Messages
85
I am putting together a PCB that has a CPU to control the on-off state of a high power IGBT. There are two power supplies associated with this:

1) A 24 DC lab PSU that is regulated on the PCB to both 5V (CPU) and 15V (TC427 MOSFET driver)

2) A meaty 150V / 40A DC PSU that flows through the IGBT and into a massive inductor

The trouble is, I'm nervous about connecting the grounds of both PSUs together due to my suspicion that in certain conditions in my application I'll get large negative currents on the ground plane and/or other nasty stuff that the CPU and related electronics will hate.

However, if I dont I have a concern regarding getting the correct 15V gate voltage to turn the IGBT on and also another circuit where I use a 2-resistor voltage divider to downscale the 150V to a meagre 5V so that it can be fed into one of the CPU's ADC inputs to measure the incoming voltage.

Can anyone suggest a way forward?
 

stube40

Feb 9, 2010
85
Joined
Feb 9, 2010
Messages
85
I had a thought - what if I used an opto-coupler?? The switching frequency is only around 10Hz, but reaction time has to be within 1ms - I think a good opto-coupler could achieve this.

If I used an opto-coupler, the only remaining problem is what to do with the ground for the resistor-divider voltage measurement feed that allows the CPU to measure the amplitude of the high-power voltage source. Maybe using the CPU's differential ADC inputs?
 

stube40

Feb 9, 2010
85
Joined
Feb 9, 2010
Messages
85
OK, I've been pointed to the existence of linear optocouplers to solve this problem! You learn something new everyday....

Having decided to go down the opto-isolator route, I'm struggling with the logistics of powering the IGBT-side of the opto-isolator.

To elaborate, there are actually 4x IGBTs and drivers in an H-bridge formation. They are routing the 150V / 40A power source through are large superconducting coil. We have found from previous experiments that all sorts of strange things happen when the coil is being charged and switched including zero voltage and negative voltages. Hence, this cannot be the source of the 15V for the IGBT side of the opto-coupler and the TC247 driver. Yet, to turn the IGBT on I need Vge to be 15V, but I'm worried about connecting the ground of the 150V PSU to the 15V PSU.

Does this make any sense at all, or should I upload a diagram?
 

stube40

Feb 9, 2010
85
Joined
Feb 9, 2010
Messages
85
OK, I've attached a diagram that outlines the schematic. I've replaced the TC427 and opto-coupler with a Vishay VO3120 isolated driver.

The 15V power supply can be created from the CPU side via a DC-DC converter with galvanic isolation. My main concern is getting the IGBT Gate-Emitter drive voltage to be negative rails of both the 150V PSU and 15V DC-DC are not directly connected together.
 

Attachments

  • H-bridge.pdf
    8.7 KB · Views: 1,324

55pilot

Feb 23, 2010
434
Joined
Feb 23, 2010
Messages
434
If you want to take your present approach, you need a galvanically isolated DC-DC converter for each of the four VO3120 module. You connect the negative output of the DC-DC converter to the Vee pin of the VO3120 and the collector of the IGBT.

You will basically have 4 completely isolated drivers attached to each of the 4 gates. You are looking at a lot of parts, but you will get the ultimate in robustness that will tolerate a lot of bad implementation, stray inductance, stray resistance and unwanted coupling in the high power path.

One improvement to this is to combine the power supply for the two lower IGBTs. But now you are obliged to pay a lot of attention to the layout/connection between the two collectors. Any stray inductance there can lead to the off IGBT briefly turning on, really ruining your day.

One way to generate 4 separate isolated voltages is to use something like a Versa-Pac transformer from COOPER-Bussmann. Use 2 windings for a center tapped driver coil and 4 windings for 4 output coils. You can drive the coils open loop, full wave rectify the output and use a linear regulator to get 15V. That will be much fewer chips and much lower cost than 4 isolated DC-DC converters.

On an unrelated note, analog opto-isolators have really loose accuracy specs. If you need to precisely switch at a specific current, you are better off doing the conversion at the IGBT side and sending back a digital feedback. If you need to measure the current using an ADC, you should really consider using a serial ADC, putting the ADC on the IGBT side and optoisolating the two or three digital lines that control the ADC.

Good luck.

---55p
 

stube40

Feb 9, 2010
85
Joined
Feb 9, 2010
Messages
85
If you want to take your present approach, you need a galvanically isolated DC-DC converter for each of the four VO3120 module. You connect the negative output of the DC-DC converter to the Vee pin of the VO3120 and the collector of the IGBT.

You will basically have 4 completely isolated drivers attached to each of the 4 gates. You are looking at a lot of parts, but you will get the ultimate in robustness that will tolerate a lot of bad implementation, stray inductance, stray resistance and unwanted coupling in the high power path.

One improvement to this is to combine the power supply for the two lower IGBTs. But now you are obliged to pay a lot of attention to the layout/connection between the two collectors. Any stray inductance there can lead to the off IGBT briefly turning on, really ruining your day.

One way to generate 4 separate isolated voltages is to use something like a Versa-Pac transformer from COOPER-Bussmann. Use 2 windings for a center tapped driver coil and 4 windings for 4 output coils. You can drive the coils open loop, full wave rectify the output and use a linear regulator to get 15V. That will be much fewer chips and much lower cost than 4 isolated DC-DC converters.

On an unrelated note, analog opto-isolators have really loose accuracy specs. If you need to precisely switch at a specific current, you are better off doing the conversion at the IGBT side and sending back a digital feedback. If you need to measure the current using an ADC, you should really consider using a serial ADC, putting the ADC on the IGBT side and optoisolating the two or three digital lines that control the ADC.

Good luck.

---55p

Many thanks 55p.

What do you think of this then? (half-bridge schematic jpg attached).

The question is, should I used a DC/DC that outputs -15V/+15V into Vee and Vcc of the VO3120, or should I go for a uni-polar 15V DC/DC that puts 0V into Vee and 15V into Vcc?
 

Attachments

  • igbt_-_1-2_h-bridge_1635.jpg
    igbt_-_1-2_h-bridge_1635.jpg
    22.8 KB · Views: 7,545

55pilot

Feb 23, 2010
434
Joined
Feb 23, 2010
Messages
434
The attached schematic seems right, however that does not mean there is nothing wrong with it. In other words, I can not guarantee that it will work. You also need to take proper precautions because you are dealing with VOLTAGES THAT CAN KILL YOU.

You need an isolated 15V DC-DC converter.

Why would you want to use a +/-15V DC-DC? That gives you 30V between the +15V and -15V outputs. You could use a +/-15V converters, but the -15V will go unused.

For the isolated DC-DC module, you need to pay attention to the following:

The DC-DC has to have enough isolation between the input and the output sides. If you do not have enough isolation, the input and output can short and you may find deadly voltages on the low voltage side.

Most isolated DC-DC converters have a minimum current requirement. You will need to make sure you meet that. If you are using a +/-15V output converter and leaving the -15V unused, you may also need to worry about maximum load current mismatch specifications.

Good luck.

---55p
 
Top