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

[MarkSnoswell]

Just a quick update on my efforts to get a very fast switch working... I now have clean rise and fall times of 11-13 ns. I am using TC4421/22 drivers and IRF 820A mosfet. I chose the 820 because of the low gate capactiance (only 360pf) and high voltage rating 600V. The only tradeoff is 8A maximum current pulse (2.5A average) .. but this is not a problem if we want to drive fast pulses into anything with even the slightest inductance.

Earl is right -- to get clean fast switch times (both rising and falling) took quite some effort... 1uF ceramic decoupling right on the power supply pins of the driver (SMT cap) with larger cap right on it's back, schotkey diode on the high side of the mosfet with reverse ultra fast recovery diode for handling the back emf. I have everything mounted on a small heat sink as one monothlithic block with absolute minimum lead lengths. Another tip is to run off battery to completly avoid any ringing or interaction back up into the power supply or through unintentional ground loops. Yet another tip -- you can have long runs of wire from your signal generator and to your devices and preserve the wave form if you use good quality twisted pair connections -- get some twisted pair network cable and canabalize the twisted pairs from that

The net result is an extreemly fast switch that runs cold at anything below 1Mhz. I have used the switch block up to 3Mhz and tested it to 11Mhz. To improve things further I am getting in some ultra fst pin drivers from Intersil (http://www.intersil.com/cda/deviceinfo/0,0,EL7158,0.html) these are rated at 40MHz continuous opperation and 8ns rise/fall times into nf's load -- so at least at the switch output I will be able to push transition times down to 4ns or so.

In tests driving primary coils on my toroidal ferrite core the fast switching makes a big difference! ... but her's one strange thing - now I can pull small sparks from both the core and other open coils (but only from one end of other open primaries)... but I cant see matching HV spikes on the scope!!!!   I will report when I know more about exactly what is happening -- perhaps I should just start testing with light bulb loads

cheers

mark.

[Bob Boyce]

Way to go Mark. Seeing sparks with nothing visible on the scope, that shows you are finding exactly what you are after.

You can ask Rich In FL about this, lighting 120 VAC 75 watt bulbs (even to the point of burnout - and the arcing beyond burnout) while nothing reads on a scope, voltmeter, ammeter, or RF wattmeter.

I plan to try mounting well bypassed drivers on tiny boards bolted right to the input terminals of some 1200V/400A bricks to see how well that works ;-)

@eldarion
What size LCD screen do you use for that FPGA board? I found some nice goodies here.

http://www.futurlec.com/LCDDisp.shtml

Bob

[eldarion]

First, good job Mark!  Are you driving those primaries open-ended or closed?  Also, are you using Bob's pulse sequence yet?

I have been quite busy populating the new driver board with parts.  Here is a pic so far, everything installed checks out except for the UCC27322s, as I have not yet fired them up.   Here is a pic: http://www.falconir.com/pics/DSC02301.JPG

I am currently using a standard VGA PC monitor to display all kinds of data.  I am not quite sure if I want to add a small character LCD on to the system yet, if anyone would like one in addition to the main VGA screen please speak up!  And also comment on what data you would want displayed on that smaller screen vs. the main one.  Also keep in mind that I will be allowing networked remote control of the complete system, so you would not have to have a monitor and keyboard hooked up all of the time.

Eldarion

[Bob Boyce]

I am currently using a standard VGA PC monitor to display all kinds of data.  I am not quite sure if I want to add a small character LCD on to the system yet, if anyone would like one in addition to the main VGA screen please speak up!  And also comment on what data you would want displayed on that smaller screen vs. the main one.  Also keep in mind that I will be allowing networked remote control of the complete system, so you would not have to have a monitor and keyboard hooked up all of the time.

Eldarion

If there was enough processing power to spare to drive a character display, it would be nice to have, at least for a benchtop testing system. I have looked over the information on www.digilentinc.com and it is unclear what the Spartan 3 units are supplied with, and what accessories would need to be ordered at the same time. That was my main reason for asking about LCD character displays, as that other site had a wide selection available.

Also, there are several models, at varying prices. It appears that the FPGA are directly surface soldered to the boards, which would mean that being able to upgrade to a higher gate count FPGA chip does not seem to be an easy to implement option. So, for a Spartan 3 based system, it looks like a choice needs to be made between the 200k - 400k - or 1000k gate Spartan 3 starter board,  the 500k gate Spartan 3E starter board, or the 1600k gate Spartan 3E-1500 development board. I guess I am just confused by the unfamiliarity with these products. What software is suggested for use in programming these? I would think that others here would like to know this as well.

Bob

[eldarion]

Hi Bob,

I am using the Spartan 3 -1000 version, which does not come with a character LCD display by default.  I will be happy to put one on the system, maybe display output power and frequencies or something?  I do have plenty of processing power left over.

There are several software packages available for programming the Spartan FPGAs.  The free one is the ISE Webpack, the more expensive option that I am using is called the Embedded Development Kit (EDK).  The EDK allows me to place a small 32-bit microprocessor into the FPGA fabric, whereas the Webpack does not.  (The Webpack will allow you to develop any logic you want on the Spartan 3/3E, but only has a small 16-bit "Picoblaze" processor available, and the Picoblaze is an absolute pain to work with).

Where I am going with all of this is to have a 4Kw+ rack-mount module that has some basic control functionality, but most of the configuration and set-up would take place with a keyboard and VGA monitor, then be saved to nonvoltatile memory.  At that point, once the system is installed, the network connection could be used to get power output, system status, etc.  I chose the standard network connection because it uses Manchester encoding, therefore small time shifts caused by the cores at full power should not cause a loss of communication with the controller.  In addition, I am thinking of having several of the modules converse over the network in order to balance loading, etc.  But that is far off in the future!

Hope this helps some,

Eldarion