Hi All,
I have started building a pulser to test how narrow a pulse can be generated. This will also serve as a test bed to see if the differential 12V output pulses can be boosted by means of an air-core autotransformer? Because I wanted to minimize capacity and wasn't interested in winning any beauty contests, I decided to wire everything in the air by mounting the IC with its back to a copper PCB groundplane. The ground, pin 7, is soldered directly to the groundplane while the Vdd, pin 14, is soldered to ground via a ceramic SMD capacitor. Therefore two corners are solidly anchored and there are another pin or two also going to ground to give mechanical stability. The picofarads start adding up with sockets and PCB dielectric material. This is only experimental prototype, so I am not going to worry how it looks, it will end up inside a shielded cylinder anyway. Didn't have enough time to test it, and am not sure if I have a 6V regulator, so will have to see what to use for power source. Sure is tough on my old eyes. Also will have to figure out a better way to hold things while soldering; at present I need at least three hands.
Regards, Earl
@Jason
here is the precision 90 degree phase shifting circuit. It is more accurate than using cross-coupled F/F's since every stage is synchronously clocked and also there are no propagation delays such as when using cross-coupled F/F's.
The circuit scanned, with 8 stages is for a 22.5 degree phase shifter. For 90 degrees less stages are necessary. Adding the OPTIONAL circuitry below the blue line gives 180 degree shift in relation to the upper SR.
EDIT1: Do you need only 0 and 90, or do you also need 180 and 270?
Regards, Earl
Jason,
here is how you use individual F/Fs instead of shift register.
Earl
Hi All,
here are the test results of my F/F pulser circuit.
Things turned out as I expected.
Next things to do:
try a coil of 3m wire connected to Q and Q/ outputs
and see what happens
hook-up to the 4426/4427 FET driver and see what happens
to the outputs.
Regards, Earl
Hi All,
Here is a scope shot showing both F/F outputs, Q and Q/, as well as the differential waveform. As you can see the differential ouput has twice the voltage swing of a single-ended output, which is to be expected.
The scope has 300 MHz analog bandwidth and the scope probes are 500 MHz rated.
I have been brain-storming a lot and came up with another alternate idea of feeding the excitation coils. Instead of my previous idea of using a parallel transmission line with a center-taped coil, my latest idea is to fed two coils with electrical and physical displacement of 180 degrees. The coils would be end-fed and the open end would be connected to nowhere (except maybe to "the blob" under influence of a bias field).
Making a center-taped coil exactly defined both electrically and physically is difficult, the end-fed coil is much easier to wind, and now that I can use differential outputs, I feel much better. I don't like nebulous grounds that are undefined - - using differential outputs with no relation to, nor necessity of, a ground pleases me.
Regards, Earl
Hi All,
here is the test circuit.
Regards, Earl
Hi All,
here is an animated GIF of how my thought experiment would have to function.
The GIF appears not to be animated when viewed in the forum, so it will have to
be downloaded and viewed locally.
Unfortunately, the FET driver ICs don't have 3-state output, so at the moment,
I don't see how to realize this idea in the form of a schematic.
For one coil pair, no problem, but to create a rotating field with FET driver ICs alone
in a differential manner now looks a bit difficult to realize.
Will have to think some more.
Regards, Earl
@Earl
Interesting concept. The quad coil concept is sound, as 4 phase can work as well as 3. I'm just not so sure about the capacity of flip-flops to handle driving the windings.
I know the point is as much potential switching with as little current as required, but you will find that there will still be some amount of current that flows, even in open ended windings. The goal is to fill and empty the potential as quickly as possible in the wndings. If the device you are using to do that limits inrush / outrush flow of potential by bottlenecking, it will greatly impact performance by slowing down the ramp-up and ramp-down of potential. So even if you are not planning on shoving loads of amps through that wire, you will still want the drive electronics to have the capacity to do the job fast in both directions. Both charging and discharging of the potential.
Bob
Hi All,
bit late with the update since had a lot of trouble.
Decided to try the pulse circuit with 74AC74 flip flops to see what happens. On a whim, I changed the circuit from last time just a little bit. Instead of clocking in a "0" on the D input and feeding back Q to Set/ this time I tried clocking in a "1" on the D input and feeding back Q/ to Reset/.
The result was oscillations on the F/F output, which I could not master.
I then went back to the original idea of clocking in a zero and feeding back Q to Set/. This worked. At around 5.8 V from a 3-cell lead acid gel battery, the pulse width was around 3.2 nsec with rise and fall times of around 1.5 nsec.
Very roughly about 3 times faster than the HC74. There is a slight delay offset of Q to Q/ , they are not really complementary, but we are talking here about very low nsec range. There are also more overshoot and ringing than the 74HC74.
I grabbed two pieces of plastic-insulated copper wire, each about a meter long and soldered one each to Q and Q/.
The result was different than I expected. The output swing did not decrease, but there was now a double pulse on the outputs. The second pulse actually has more amplitude than the first, actually swinging above the Vdd supply voltage.
The reason for this is most likely the fact that one output is fed to a Set input. If there is a reflection from the ends of the wires returning to the F/F, then this reflection could trigger the Set input. It will be interesting to see what happens with longer wires.
When bending the wires into a loop and shorting the ends together, the F/F outputs disappeared; too much load from too short a wire.
This double-pulse effect will probably disappear when using a buffer amplifier such as a FET driver.
So, I have lots of experiments to try. Will post some scope screen shots soon.
My suspicion is that I will end up using 74HC logic, together with a FET driver. At first I will try with just the FET driver, but no FET.
Regards, Earl
Hi,
Please see the attached PDF update showing test results of a 74AC74 pulse circuit.
Regards, Earl
Quote from: Earl on September 06, 2007, 06:46:51 PM
Hi,
Please see the attached PDF update showing test results of a 74AC74 pulse circuit.
Regards, Earl
Concerning your question: "The question is whether a voltage swing of 12V going to two, opposite coils is sufficient?"
According to Milan's setup at this page: http://www.keelynet.com/energy/milan.htm
He is using only 10V and via his open-ended Tesla/Avramenko hybrid setup, he is getting some impressive voltages built up...
Now in regard specifically to the TPU and it's wiring, I would still have to be asking the same question you did, maybe Bob knows ;)...
@Tao @All
I will also be trying the FET driver that is designed for quickly charging a capacitive load of 1 to 2 nF.
If there is a load of only very low pF the outputs should have fast rise and fall times.
If the FET driver can take 25V Vdd, then the output would be 50V p-p differential. If the shock to the environment (aether ??) increases as the square of the voltage, then 50V p-p would give about 16 times the bang of 12V p-p.
Just for fun, I will also be trying a multi-filar balanced to balanced HF step-up transformer to see if it is possible to step up the voltage without degrading the pulse waveform too much.
One thing I forget to include in the PDF was that the output waveforms hardly changed when I connected by entire body via a metallic object in my hand to the output. This means a lowly CMOS F/F has sufficient output drive to charge & discharge my body capacity with rise and fall times of ~ 2ns.
Therefore a 74HC / 74AC F/F has more than enough output drive capability to excite toroid coils with low ns impulses, at least open-ended coils.
Regards, Earl
Quote from: Earl on September 07, 2007, 04:50:21 AM
One thing I forget to include in the PDF was that the output waveforms hardly changed when I connected by entire body via a metallic object in my hand to the output. This means a lowly CMOS F/F has sufficient output drive to charge & discharge my body capacity with rise and fall times of ~ 2ns.
Therefore a 74HC / 74AC F/F has more than enough output drive capability to excite toroid coils with low ns impulses, at least open-ended coils.
Regards, Earl
Body capacity at low voltage is not very much of a load. Essentially unloaded operation will not be quite the same as operating in and perturbing the near field of an active HV bias potential field.
This reminds me. A couple of days ago one of the other replicators called. Driving the primaries single ended with TC4420 drivers caused them to fail in the presence of the HV bias potential field. He said this did not occur with the HV bias potential turned off, unless loaded with 330K resistor.
Bob
Hi Bob,
What happens in a HV bias field is anybody's guess; that is why it is necessary to build, test, and measure.
There is a funny saying: "In theory, practice and theory are the same; in practice they are not".
And where we are going there is little practice or theory so most likely there will be a lot of learning and surprises.
Along with a trail of burnt semiconductors.
It would be nice to know if the FET driver failure is on the input or output part. Even on identical part numbers, manufacturers use different circuits and processes. And for non identical parts, the input and output circuitry is vastly different.
There is one universal circuit add-on that can protect inputs and outputs where the swing is between Vss and Vdd. I attach this simple and effective circuit.
Regards, Earl
Quote from: Bob Boyce on September 07, 2007, 07:20:38 PM
Quote from: Earl on September 07, 2007, 04:50:21 AM
One thing I forget to include in the PDF was that the output waveforms hardly changed when I connected by entire body via a metallic object in my hand to the output. This means a lowly CMOS F/F has sufficient output drive to charge & discharge my body capacity with rise and fall times of ~ 2ns.
Therefore a 74HC / 74AC F/F has more than enough output drive capability to excite toroid coils with low ns impulses, at least open-ended coils.
Regards, Earl
Body capacity at low voltage is not very much of a load. Essentially unloaded operation will not be quite the same as operating in and perturbing the near field of an active HV bias potential field.
This reminds me. A couple of days ago one of the other replicators called. Driving the primaries single ended with TC4420 drivers caused them to fail in the presence of the HV bias potential field. He said this did not occur with the HV bias potential turned off, unless loaded with 330K resistor.
Bob
Hi All,
here is the latest information, including a circuit schematic for a rotating-field pulser.
Feedback welcome.
Regards, Earl
Hi All,
it has occured to me that my analysis of the circuit was wrong. I did not see that even if the pulse generator was momentarily inactive, as soon as the inversion command is clocked through the delay flip-flop, the EX-ORs will flip with full voltage stroke. This can be considered a glitch and will mess things up.
So I am now back to the beginning where the task appears to be difficult. I know exactly what I want to do, but the schematic to achieve this is still out of my reach. Only the pulses may be inverted, but not any steady-state voltage levels. Not easy to see a solution at the moment. Back to brain storming.
Regards, Earl
Hi All,
I have been busy cleaning up the "big mess" in my laboratory and workshop, and this will continue to take some time away. Want to make another shelf and having some trouble how to fasten it to the wall. You all probably know that "big messes" have a tendency to avalanche unless they are cleaned up and organized from time to time.
As you remember, I had this idea to use both outputs of a flip-flop, each going to an opposing excitation coil. I found out that this meant reversing pulse polarity during rotation just like the commutator does on a DC motor/generator. The circuit I devised did invert pulse polarity, but also inverted the base signals of Vdd and Gnd. This glitch would always happen unless the pulse length would be increased until it is 50% square wave, at which point the glitch would happen at the correct time. However, I don't want to be fixed to a square wave where the pulse width varies with frequency.
I have been brain-storming about how to find a solution to this problem and although possibly with a tri-filar coil and switching I could do it, I have come across another idea where switching polarity is not necessary.
This idea is that the Q and Q/ outputs go to a balanced transmission line that is wound around 90 degrees of the core. Each coil is identically wound to the others. One thought is that the coil could be distributed around the full 90 degrees, ending just under the start of the next coil. Then the pulse width would be adjusted until the pulse ends just as the pulse beginning edge reaches the end of the transmission line. Now as the frequency input of the quadrature-generator shift-register changes the pulse repetition rate changes, but the pulse widths always stay the same. Each coil is excited in a rotary fashion at exactly 90 degrees.
By using a pulse width of 10ns in a rat race, the field would be rotating at about 1.5 billion RPM. Here is where my traditional brain half gets in trouble with the RE brain half. I can see where the 10ns pulse travels down the transmission line of about 3m, reflects from the open end and returns to the excitation source in the correct phase to get whacked again at just the right time. It seems most of the field would be between the two conductors of the transmission line. There would be relatively little current because the transmission line is voltage fed. My big question is whether all this happening in an open-ended transmission line accomplishes anything? Four of these excitation devices are arranged in physical and electrical phase quadrature. I can easily see that a single-wire arrangement with opposite coils being excited in a dipole fashion from Q and Q/ will cause an electrostatic field to become a rotating electrodynamic field when using phase quadrature. In effect this electrodynamic field is no different than an electric charge orbiting the nucleus and causing a strong magnetic field. But will this also happen when using 4 balanced transmission lines in phase quadrature ????
The general idea would be to use a 74HC164 shift register to do the quadrature phase shifting, which drives D-F/F pulse generators. Either these F/F outputs drive the transmission lines directly or the voltage level is boosted by a dual FET driver using both drivers for each segment. Another option would be to follow the FET drivers with FETS driving a tri-filar transformer. One winding for the FET and the other two to drive the transmission line. In this case, it would be better engineering to add a balun to interface the unbalanced FET switch to the balanced transmission line. This could boost the voltage shock up to 1 to 4 kV.
I am also thinking about using transistors or diodes in a pulse avalanche fashion to generate HV pulses with rise/fall times in low ns or even ps range, but there would be a lot of engineering to do since avalanche pulse generators are very phase unstable. The pulse repetion rate would probably top out at 10-100 k pulses/sec. I don't think I will put much time in this since FET transistors are so much easier and the engineering is routine. Depends on whether or not lower voltages work well with sharp rise/fall times at thousands, millions, or billions of RPM.
Regards, Earl
Hi All,
I have managed to invent a new way to mount shelves to vertical metal supports, so I finally finished installing the last shelf in the electronics laboratory. In next next days, will continue to clean-up and organize. Since I will be leaving for the FE congress in Switzerland this week, I won't be able to get that much done. I'll report if I see anything interesting at the congress. Erfinder, who is well-known and appreciated here at OU.com will be giving a talk. I am sure it will be interesting to meet him and chat in person.
I had another idea for symmetrically exciting a coil, which is attached, but my gut feeling is that my previous idea of an open-ended transmission line is better. I can't think of any other ways to excite a coil, both symmetrically and balanced.
I am totally convinced that very rapid rise and fall times are extremely important. Less important, but perhaps crucial is a small pulse width. I have various parts that will arrive in the near future, and I may just take a detour to investigate the behavior of diodes and transistors in avalanche breakdown mode. If it is not too much of a detour, I would be thrilled to be able to create pulses of a hundred or two hundred volts with rise times and pulse widths way below 1 ns, say 350 ps.
This implies that I can measure such pulses and my present Textronix scope probes, even if they are rated at 500 MHz do not please me. The only ground I have is a short ground lead. If I take the tip attachment off, there is no ground ring surrounding the probe tip, as I have had with other scope probes. I think I will fabricate myself some HF scope probes using SMD components, where I have the scope tip surrounded by a ground ring that also has a small probe. This will give me very short tip and ground leads. With only 2 resistors and one or two capacitors, I think I should be able to design and build my own high-quality 500 MHz 10:1 scope probe. It may not be a beauty, nor have exact DC calibration, but it should have good HF fidelity and shorter than short ground lead.
Earl
As you raise potential, you can narrow the pulse width and still get more response. So your reasoning is valid.
Bob
I have returned from the Swiss FE conference, which was very interesting even though I did not see any OU devices with certainty. One Belgian M.D. showed up with a new way of placing magnets and coils in an AC alternator and was claiming 4:1 OU, although his measurements were not scientific enough to give a conclusion. I liked his original way of thinking. He was a very funny presenter, quite a live wire and had everyone laughing a good part of the time. Mike Bradey from Prenderev canceled his Saturday evening talk with demo at the last minute. There were 50+ people who had paid him money and wanted to have a serious talk with him. My opinion now is that he is a bla-bla. I had an after hours discussion with Erfinder and others that lasted until 02 AM. In general, it was a very interesting conference, and I would not have wanted to miss it. Oh, there was a German researcher that built an uni-polar machine out of two alu wheel rims. He made the interesting discovery that he could mount the center and the rim on ball bearing and pass huge currents through the bearings with absolutely no negative effects. He was passing currents like up to 16kAmps through the bearings. And also very high RPMs. He said he had an anomaly whereby at high currents and high RPMs he could get in a fraction of a second a speed-up runaway. There was a very interesting talk by an older German fellow who worked on the VRIL flying saucer project and also with Hans Kolar. His brain and memory were still sharp as a needle. I am fluent in German so the conference was no problem for me.
So now I am back and have to finish cleaning up my electronics lab. Unfortunately, my notebook LCD decided to have problems, so I am limping along on a replacement that consumes many times more power. Since I am 100% solar powered, this limits my computer time, especially on cloudy days. I have ordered a new notebook to use while the defective one is repaired. As soon as my lab is operational again, I will continue my research with narrow, fast pulses. Will investigate transistors and diodes in their voltage breakdown region. If a transistor is just under its breakdown voltage and is pulsed on its base, it can avalanche on this sync. In order not to destroy the transistor, a choke or resistor is used to supply collector voltage to limit breakdown current. The collector current is provided by a small capacitor of roughly 2 pF and the emitter goes to ground through a 50 Ohm load. Will try both bipolar and FETs. I have some HV avalanche rated diodes so will also try these. I figure if the diode is just under breakdown limit and I hit it with a pulse that brings it into breakdown, then I can control the pulse timing. Another idea is to use the standard "short the Vsupply to ground through a choke coil" and feed the back EMF to the avalanche breakdown circuit. I.E. don't use a standard HV pulse directly, but pass it through a shaper circuit consisting of an avalanche device. Instead of limiting device dissipation by keeping the collector or drain capacitor low, this idea would use pulse width modulation to limit the joules dissipated in the shaper semiconductor. I believe this detour could pay big results if the circuit could provide HV pulses of 50 to 350 ps rise/fall times. Will be calculating the SMD R's and C's needed for a 10:1 probe and home-brewing my own 1 GHz scope probe, which will be needed for such fast pulses.
Earl
My progress has been severely delayed due to computer failure and many days of rain with no Sun (I am off-grid, solar energy only).
Hopefully in a week or two I will be cooking at full speed.
I have received a sortiment of SMD resistors, so am in good shape for a 10:1 HF home-brew probe.
Earl
My new notebook is up and running, but it will take another 2 weeks surely before it is running to my satisfaction. The seller didn't include the recovery disk so will see what happens when that arrives as the notebook was delivered with XP Home instead of XP Pro. I may decide to leave it the way it is. I first need to burn a HDD image to a DVD, but I'm not sure if it will fit on a DVD. Don't want to go gung-ho on reinstalling software until I've got an image backup. Can't even look at PDFs yet. And worse I can't generate PDFs.
I've received some more various avalanche-rated diodes and transistors so will have enough to blow up for my sub-nanosecond HV pulse experiments. I will need to build a transformer (DC/DC converter) to generate avalanche voltages and I had a bizarre idea. When I was welding the shelve supports for the laboratory, I grabbed one of my NEO magnets to hold pieces in place in order to spot weld on the ends, then remove the magnet before it got too hot, then continue. One time I forgot and noticed the (ex-)magnet sliding down the steel. So now I have a NEO unmagnetized core; it went way over its Curie temperature. I can't resist winding a transformer on this "just to see what happens". Will measure the AL of this and an identical magnet, wind transformers on both and will see. As far as I understand such a core contains mostly iron powder with small amounts of rare-earths pressed together under high pressure and maybe a very small amount of binding glue. If the powder is fine enough, it should have good HF & pulse characteristics.
That's what's happening here, regards, Earl.
I had a lot of trouble trying to measure AL of my 10mm NEO cores. Didn't have the proper size plastic to turn on the lathe, but finally got a coil form for the 10mm dia NEOs. With 10 turns the air core had a L= 2.0 uH. Didn't seem to make much difference whether NEO core was magnetized or not: L dropped !!!! to about 1.48 uH. A steel tube raised the L slightly, but a 8 or 10mm threaded steel rod dropped it a bit. A solid ALU 10mm rod dropped the inductance the most, down to 1.2 uH. Used an L meter where a uprocessor most likely uses frequency of osc to measure either L or C. At turn on it calibrates itself.
So I am very confused, nothing makes much sense. I can try winding a lot more turns and see what happens - or take a CMOS gate and build a quick and dirty Collpits oscillator. I know that a brass core can decrease inductance, and therefore maybe ALU also, but anything ferrous whether NEO, steel rod, or threaded stock should raise the inductance. This did not happen.
I found a ferrite bar and have some steel or iron fence wire, so will continue testing tomorrow. I am very puzzled.
Earl
Hi All,
Have had my lab time reduced due to bad weather and zero Sun for some days. The Sun is now back and tomorrow will put another 4 solar panels up to boost my energy resources.
Used some more turns on the coil and found out that a NEO slug does not have extraordinary AL value compared to METGLASS or ferrite antenna rod. It looks like I will be using a ferrite pot core to make the HV pulse transformer. Would like to have the minimum turns as possible; e.g. just one single turn for a 12V primary winding. Therefore a secondary with only 10 turns gives 120V out and 20 turns gives 240V out. Will be starting off with the primary being driven by just a CMOS IC with AC coupling capacitors with a stroke of 12V. If necessary, will migrate to a MOS driver, and if this is still too weak, then driver and power FET. I don't need nor want much power since the secondary will only be zapping a diode or transistor with a HV pulse to cause the semiconductor to go into avalanche. My idea is that by keeping pulse width down I can keep the number of transformer turns down while keeping current very low and core flux level also minimal.
To keep fall time from increasing too much, I suspect that an output resistor of 50 Ohms will have to be used. I may try somewhat higher like 200 Ohms since this would give 4 times the output voltage.
Should have time to start soldering and testing this week. I would like to have a HV sub-nanosec pulse generator as soon as possible. Will see how difficult it is to cook this soup.
Earl
I can certainly relate to the no sun. I recently upgraded to a new OutBack GVFX 3648 inverter/charger and have only had 1 partial day of sun since. Best day since then was only 6 A/H into my battery bank from all day. Typically had ony 3 or 4 A/H per day since installing it. Figures huh? I have a set of 5 panels, 125W 12/24V nominal ea. Still in storage, for lack of a decent mount. During summer I can lay them out, but the rest of the year the shading is too high.
If current requirement is real low, can't you use some really fine wire in your ferrite pot core? This should allow you to keep drive requirements low.
Bob
Hi Bob, All,
I'm surprised about your Florida comments, I thought FL was sunny almost all the time.
I am going to separate computer, lab, and frig and put each on their own string. I do have an emergency 1kW gasoline generator with load-controlled RPM, but I don't like to use it unless absolutely necessary. My hot water is solar heated. In Winter a bit more difficult, but I really try to be 100% solar. It teaches one to conserve energy. I won't change even if FE becomes available; don't like to waste. My fridge has no side door, it opens on the top. I have plans in my head for water-cooled fridge and freezer using Peltier elements and 8 cm thick insulation on all sides, but didn't have enough time this year to build them.
Since NEO powder makes such super magnets, I thought an unmagnetized NEO slug must by definition be very ferromagnetic. I was surprised to find its lower AL value. Just for kicks I will also measure a ferrite ring magnet to see how it compares.
I am going to grab a pot core, which I should have in my junk box, and start winding. Will start engineering from only gut feeling and change course depending on measurement results. Most avalanche pulse generators use 2 to 4 pF (or a short coax stub) on the collector, charged through an R or L. I am going to use no C, but instead the energy from a magnetic field. It is my own idea; don't yet know how things will turn out. Will write a detailed report on it, whether it works great or not.
Earl
Hi Earl,
Just looking in. I have some circuits and simulations to add soon also. Looks like fastest transition times are achieved with IXDD414 driver and IRF820 for output. I just have to do some final experimental tests to confirm the latest designs and see if I can stop blowing driver chips.
You should also check out the IL710 isolator chip from www.nve.com -- very fast (fastest availble) and does level shifting as well -- great for driving floating supplies.
cheers
mark.
@Mark, @All,
My present avenue is to "jump around" the usage of FET drivers and FETs.
In other words, go from a coil and switching element over to an avalanche breakdown. Switching is at best in the lower nanoseconds or tens of nanoseconds. By using avalanche breakdown, there is a possibility to drop to say 50 to 350 picoseconds and this in a real-world practical result. IF and this is a big IF, I could generate sub-nanosecond pulses with amplitudes of say 100 to 400 Volts and IF these have phase stability without lots of jitter, then this would be a quantum step forward. When a semiconductor is near breakdown, the electric field gradient is so high, that when an electron smashes into a neighboring atom, it knocks several electrons loose, which then proceed at high velocities to smash into other atoms. It becomes a classical chain reaction and the semiconductor becomes a short circuit in a matter of picoseconds. This is what I desire to master and harness in order to shock the environment. I have decided it is worthwhile to pursue this route and will follow it until I decide that the results are not beneficial enough for the time invested - or until I have my beautiful pulses. Since I want to admire a FE generator before the end of this year, I am going to force this with all my possibilities. The desire is there, don't know whether a rendezvous with destiny is on the agenda though.
Earl
Earl @all
I think I am getting a handle on what thee minimum pulse time and rise times may be. I have done some simulations of the coils as transmission lines -- much more accurate and revealing than lumped analysis. In summary the results are:
Pulse width and amplitude: The primary outcome of varying pulse width and amplitude is to control the amount of power delivered into inductive load (straight wire or coil). The pulse width is varied such that the back EMF rises to the maximum permissable by the driving device. The longer the pulse width the more energy and the higher the BEMF. You dont want the shortest pulse width possible -- that yields to small a BEMF. You want just the right length pulse width to achieve maximum dv/dt and peak BEMF.
Pulse rise/fall time: The primary target here is fast enough such that the rise/fall time exceeds the capacity of the driven device to respond. When driving transmission lines there are maximum slew rates that are easily determined. A I show below for the particular device I am testing that looks to be 6ns transition time as an absolute minimum required... and faster wont help at all.
Practical considerations:
1. When driving very fast pulses into the sorts of loads we are considering (air cored and iorn powder core coils) the peak currents required to generate BEMF's of 500V or so are tiny - under 1 amp. This permits use of lower current mosfets which switch much faster then high current devices.
After a lot of looking around I chose the IRF820A -- this has simillar ratings to the IRF840 many people are using. The difference is that it's peak current capability is 8A but it is *very* fast and has only 360pf gate capacitance.
2. Drive the mosfet gate as hard as you can. The IRF 820A can take 30V gate drive. Even with only 360pf gate capacitance the limit to switching speed is peak current from the driver. I am using IXDD414 which can work up to 40V and deliver peak current of 15A -- I only need to drive at 22V to reach peak current limits. It is the peak current that is limiting speed -- you need very good bypassing on the driver supply. Faster switching speeds could be obtained with high gate drive voltage but would require going up to IXDD430 to deliver more gate current -- and analysis shows that we are probably far exceeding the shortest switching time required.
The simulations show that driving an IRF20 hard results in switching times 4x faster than IRF840 in our applications.
The simulations show that transition times are just below 1ns at the coil... however, we are limited to the rise time of the real world transmission line we are driving -- in the circut I am about to test the BEMF rises to 500V happens in about 25ns in the simulations -- a slew rate of 20V/ns !!
In the full simulation of the transmission line there is a full 6ns from the time the IRF840 switches of to the time the BEMF starts to rise. This sets a conservative maximum rise time of 6ns -- something that is not hard to achieve at all.
In the representative circuit I am about the test (lab experiment) the simulation shows that with a 44V high side supply (I am using 22V for the gate drive) a 600ns pulse into my 69uH coil will result in a BEMF peak of 450V 55ns after the IRF840 is turned off.
Although every circuit is different the guidelines are the same:
1. Drive a low gate capacitance mofset hard.
2. Start with a shor pulse width (50ns) and keep increacing it untill you reach a safe maximum BEMF.
This simple aproach maximises dv/dt and peak voltages while minimizing power input at the same time.
I'll post the sims and experimental results soon.
cheers
mark.
Mark @All
My present line of thinking is not to go with BEMF to generate the environment shocking potential. Instead, I will use avalanching in semiconductors to produce the pulse directly by discharge of a HV-charged capacitor (or magnetically-stored energy). Instead of using a coax stub as a capacitor, as some people use, I intend to use microwave-rated hi-Q ceramic chip capacitors. For an idea of a home-brew avalanche pulse generator see:
http://www.holmea.demon.co.uk/Avalanche/Avalanche.htm
With an avalanche pulse generator, you are continually and repetitively "blowing up" the semiconductor, but it survives because the avalanche energy level is being controlled.
My first experiments will use a pulse-by-pulse HV generation with no rectification. I will control the avalanche energy by means of pulse width. Voltage will be set by means of turn ratio. Thinking of using 2 independent secondaries to avalanche 2 semiconductors at the same time: 1 for one side of the parallel transmission line, the other connected with reverse polarity to the other side of the parallel transmission line.
Attached is a very preliminary idea.
Earl
@All,
here are some more ideas concerning avalanche HV pulse generators.
The thought about tying the base resistor to cold end of emitter resistor does not please me 100%. It might be better to tie it to the emitter.
In one idea, I have moved the output resistor to the collector side in an attempt to achieve bi-directional output around zero volts.
For the idea with the step-up transformer that is connected to the parallel transmission line, I have a special idea for the construction of a one-turn primary with low loss. Even though the pulse may be extremely short, its amplitude could be, say, 1 to 10 amperes. This means the primary must be low-loss. Will draw this up tomorrow morning since it is already 1AM here.
Don't know how important it is, but some of the semiconductors that I will be testing are avalanche rated.
Earl
Hi Earl,
I love this avelance idea -- I had no idea you could abuse semiconductors like ths and get such amazing short pulses! -- excellent.
mark.
@Mark
If you liked the avalanche idea, your ears will wiggle with this one.
Especially for you.
Transformers will alway degrade transition times, so best to get
HV just with resistors. This is harder to do than to say.
I have more tricks up my sleeve.
Earl
@All
thinking out loud with another circuit
Earl
@All
it's getting late, so here is the last thought of the day.
Earl
Hi Earl,
I am going to try the Avalanche pulse generators. Quite apart from the fast pulses I like the efficency of these -- nearly all the energy goes into the pulse.
One power loss that people overlok is the Mosfet gate drive. At high frequencies this can be the dominant power drain in an efficent output switching configuration. A triggered Avalanche pulse generator uses almost no power for the trigger.
Inspired by the avalanche pulse generators I am also going to try the same RC supply scheme on my regular Mosfet output stage. Doing this I can drive with square wave and control pulse width and energy with the pulse capacitor (pulse forming network - or coax).
cheers
mark.