Eldarion and Bruce's build of Bob's Energy Converter

[MarkSnoswell]

I had a large bank of capacitors at the power supply, but I will try putting individual capacitors on the IGBTs to see if that helps.  I had assumed that I was getting some weird waveform mixing in the core and failed to look for more obvious problems.

not good enough. You have to have sets of capacitors 0.01u,0.1u, 4.7u  ... ideally low ESR and low inductances. The reason to use several sizes is that the caps impedance will drop untill they get to a certain frequency and then their impedance will rise! ... the lower value caps have their minimum ESR at higher frequency -- so by combining several different values you get the lowest ESR up to a good high frequency -- which you need to keep the pulse edges sharp. You need to repeat this supply bypassing for each driver and the output stages. The caps need to be right on (as close as practically possible) the leads to the devices. I tested -- a combo of 0.01, 0.1 and 4.7u SMT ceramics works best -- back it up with 4.7Uf tantalum... and then isolate any common lines between drivers with ferrite beads.

IXYS have some great application notes on their web site. Here is a link to the PDF for their low side gate driver evaluation board -- it's worth examining for general tips on how to drive mosfets fast http://ixysrf.com/pdf/switch_mode/appnotes/evic420.pdf   -- study that and take heed!

cheers

mark.

[MarkSnoswell]

Configuration:
NEV IL711 isolator.

IXDD414  driver. (22V)
2 IRF820 in parallel.  (450V)
Load = 5nH  30 ohms.  ( air cored primary made from Ni205 resistance wire).
The battery and driver circuit are floated down at - 450V.

22V for Driver from 2 x 11.1V lithium batteries.

Results:
Minimum reliable pulse input of 4ns. (+ 1ns rise and fall times)
Rise time........ 600ps  -- yes, no mistake, that's six hundred pico seconds = 0.6 ns
Pulse width..... 12ns
Fall time......... 1.5ns

Notes:
* No ringing, no bounce.
* Peak current in IRF820 = 7.2A each. 30% below rated max. Safely within SOA.
* IRF820 is readily available, cheap and very fast (faster then IRF840) with just 360pf gate capacitance.
* Highest operating frequency of output 41Mhz. - limited by drive pulse width.
* Highest operating frequency of ICDD414 driver is limited by power dissipation (833Mw in 8 pin DIP and 12.5W in TO220):
  ........in 8 pin DIP package: 700KHz.
  ........in TO220 package: ~ 10Mhz.
* 22V is slightly overdriving the 820 gates which have a rated maximum SG voltage of 20V.
* Il711 offers 2,500V isolation and level shifting = no problems driving the floating driver circuit at -450V -- still leaves 2,100V protection.



Driver Frequency limits:
The maximum operating frequencies can be increased by operating at a lower voltage of 11.1V = one lithium battery. The trade of is an increase of the rise time from 0.6ns - 1.7ns. The fall time is unaffected. At 11.1V driver supply the maximum operating frequencies can be doubled to:
8 pin DIP...  1.4  MHz
TO220.......  20   MHz



Load Power dissipation limits to frequency:

At 250KHz the load coil will dissipate 20W. The IRF840 output drivers are only dissipating 0.8% of this.

It is clear that the heat generated in the load will be the limiting factor to going to higher frequency. As it is not practical to generate shorter pulse widths the only way to reduce power dissipation is to reduce the output stage voltage or increase the load resistance.

Increasing the load resistance will slightly reduce the rise time but the fall time will degrade linearly with increasing resistance. If a high side load resistor is used to limit power then a mosfet switch could be placed in parallel with it and this would reduce fall times to the same duration as the rise times. Experimentation will be required to determine the most efficient and effective drive mechanism.



Resistance wire vs. high side load resistance.

There are two distinct types of load resistance:

a)    Resistance wire coil.

b)    Low resistance coil + high side load resistor

In (a) the pulse voltage will drop across the coil giving a distributed field in the device.

In (b) the entire coil will pulse up to the peak voltage.

Experimentation will have to be done to see which mode is beneficial/required for the device to function and produce excess power.





Spice analysis - software:
Analysis done with Linears free LTSpice/SwitcherCAD III software available from http://www.linear.com/designtools/software/



Analysis done from GND to +400V as the spice solver hates to have things reference below GND. IN the real world you just attach the single high side to GND reference and you have negative spikes... I have referred to rise and fall times in the negative going sense: rise = leading edge, fall=training edge.




Cheers

Mark.

PS.  Just had a bit of excitment as a big Koala walked around the house ... They are prety timid and you can get to within about 2 meters.

[eldarion]

Hi Mark,

Good work!  The only problem I see with that circuit is the 5nH primary coil--I have a hard time believing that my primary coils are anywhere near that low of an inductance.

When I was mentioning a capacitor for each IGBT, I was just writing sloppily.  I was thinking many bypass caps of similar values to the ones you mentioned with near-zero lead length directly across each IGBT and primary coil.

Also, my Earth ground is currently connected to the negative rail, so I should definitely connect it to the positive rail and try again.   Right now I am generating big, fat positive pulses, not narrow negative pulses...

Eldarion

[MarkSnoswell]

Ok -- it took a long time but I got it to work in the end.

Here is a circuit and results, for driving the 0.58mH primaries of the BB unit.

The red trace shows the voltage across the coil.
The green trace shows the current through the coil.

All voltages and currents are within the SOA of the parts. Overall power consumption is much lower than driving a lower inductance coil.

Leading edge transition is still 0.6ns. Falling edge transition is 20ns.  This is realy the best you can do with these standard parts and a 0.58mH coil.

I will explain the circuit more tomorrow -- exhausted now.

cheers

mark.

[Rich SAS]

Hey guys, Bob Boyce wanted me to pop in here and post a diagram of how my setup is configured, sec, bias cap blocking and load. Please excuse the crudeness of it. Did it freehand in ms paint, pain in the butt. You don't have to use both caps, just one is needed, that's just me.

Rich