Use for the Tesla Switch

[nvisser]

My sentiments exactly.
The circuit Dave (DoubleD) posted does all that and is probably as simple as you could get it.
It took him and some other engineers about 3 years to get it where it is now.
Transformer drive circuits could maybe simplify it.
Like on some of the pdf,s that Seamonkey posted today

http://tonic-lab.com/pdf/reading_room/transformer%20isolated%20gate%20drive.pdf

[StevanC]

(snip)
Devising a driving method for Floating MosFets
isn't too difficult. But, it requires some understanding
of the principles.


Download This Document The discussion on Gate drive transformers
begins on page 30. The whole document is outstanding.

Study this web page. It has a TON of good info.

Downoad this one too. It is good technical information.

This one is good also.

That should pretty well cover it.

Then we can get down to putting the pieces
together...

Do note:
they consider 50Hz the slowest rate
a 2.0...4.0 mH xfrm is the one used in "so called" 300W PC power supplies
and that's the big one (3.3 mH)
looks like a bit of an overshoot? But is doable.
(still have to go trough the rest of the documents)

Best regards,
Stevan C.

[StevanC]

@SeaMonkey,
Are You mocking us?

I have trouble doing it in BJT right (kick) where i deal with a 4MHz BJT (=slow) and have slew rates of 100ns or ten times so...

And You throw at us TC (Tesla Coil) drivers with "roll their own" DIY high side MOSFET driver circuits with de-coupling transformers ("if...You will have to change the core" ) as we are Gandalf the white

Well, it can be done, but expect me to be "away" for a month in regards to this technique (=mastering the art) and such for each other sort of "magic"

Besides that, thanks for the good sources of MOSFET drivers (MAXIM and TI especially)


Now i go (hopefully) to troubleshoot the absence of the spike on my late PVAmp

Stevan C.

[Matthew Jones]

Vissie,
I would gladly provide a more obvious and practical circuit, but there are few basic considerations in regards to SPICE in general:
1. SPICE is practical to an extent, but one has to know to do without to be able to take advantage by doing with SPICE: there is really not much value in lanenal's design in terms of a blueprint - it is a model - an abstraction far from real world issues and drawbacks: like using a child's sketch to build a race bicycle.
2. lanenal's proposed "revelation worth" simple MOSFET switch has been known for long time and been a first consideration has it not been the problem of driving the supply of floating voltage to get the gates 10 (or preferably 18V) volts a top of the source.
3. Then he uses P-channel MOSFETS: less than optimal by twice the cost and half the performance. Not to mention driver issues possible
4. Then he uses transformer to drive a FET: shoot-trough, runaway, overshoot, failure conditions, a myriad of plaques and problems not even tried to solve, let alone design in the circuit: he could as well give six "voltage sources"; define them as pulses and "apply" them floating, each to a MOSFET, and call it "simplest". Yes model wise, but how does one implement that?
5. To drive a MOSFET's gate off a floating secondary one takes a high-side driver (such as IRS21851 already is, or LTC4440) and feed it's Vboost-Vs capacitor with a rectified (one diode sufficient) current. Therefrom You let the driver "clamp" the Vboost to 20V and let the driver take care of feeding the gate with "best diet" for it.
6. Once You do include those famous drivers to the circuit, You find the "neat simplicity" is all gone for now You see You ned not less than 30V (and more) of surplus voltage (from bottom "line") to feed the Vboost of the "upper" two MOSFETS or a AC generating circuit with two floating rectified 20V outputs.
7. To generate a AC one needs what's called oscillator - and this takes time to build, and power to run - so farewell "simple TS design" and the nightmare is all back again...

8. While we doing this all "so" to get 12V across a 10Ohm load (~1A), we need to feed the MOSFETS with 18V at least (10mOhm batteries considered) and that costs _power_, to get approx same results with BJT (namely MJL21194) we need just BD140 and optos (total six of each) - and just reuse the "Quer" of the PDF I posted.
And MJL will go up to 16A (dropping from 12 to about probably 8V across the switch)
that's at least 128W of usable power, at the cost of 64W of dissipation.
9. I did begin to consider a MOSFET Brandt-TS but I feel not ready to "ride the bull" yet...
I just don't know...
Stevan C.

I'm not getting the same picture of a Monster device. I'll be a couple of more days before I order any parts but I don't envision the nightmare you do. I have already read on several drivers that can be used to simply. We don't need to do the traditional floating thing with transformers for gate power. They IXYS had several drivers that seem to be able to deliver anything you input into them, they just need a small signal to turn on, and they flip it to turn off, or supply that as well. All the while ramping voltage up and few other things.
But I am just talking right now. We'll figure it out. Then We'll set a few fires. The will get something that works.
I am still just worried primarily about how to protect while invoking the reaction I have seen in the past. Not as many docs on that.

Matt

[chainmailleman]

Where there is a will there will be found a way!

Do you have a rough diagram of how you are wanting
to apply power to your spot welder?

What sort of a battery bank do you have?

Rather than a commutator switch mechanism have
you considered the possibility of slow speed cam
operated really heavy duty copper contacts?
Only the contacts and the wiring would need to
be copper, the remainder could be wood, or
any other convenient material.

I have 8X 6v 230ah marine batteries.

I was thinking about a camshaft, since you seem to think similarly, it might not be a bad idea....great minds think alike after all.

Comparing the Brandt circuit, the only real difference would be the rectifier will be replaced with 2 turns of 000 gauge wire for the primary, and 20 turns of 6 gauge for the secondaries. Tesla's Patent on Electrical Transformers, figure 1. I already have this transformer made. It can handle some power, even though it's not a closed magnetic circuit. The core is a 3"x1" steel round.

I will have 1 relay disconnecting the transformer and the whole circuit so that way I can run the commutator without power applied if necessary.

My current progress is I have all my contacts made. I am constructing the cylinder for the contacts to mount on today, if I work fast enough I might even have the bearings put in.

If I can just run my computer using this, I'll be set. Lighting is childs play, the real goal is to be able to produce mains power from my shed...or garage. And then apply it.

The only thing I am really missing is a camera to show you guys. But I'll see what I can do about those 5 dollar disposables and put in on a CD.

p.s. I know an inverter does the same thing, but I'm that kind of guy that makes everything himself.

[StevanC]

I just dn't know,... one thing vories me:
I had 20% performance gain from 12V to 15V and still 10% from 15V to 18V on a 20V rated MOSFET in SPICE.
If I consider the IRF3205 (a decent model) (as I do), I just hope this won't be a lame chase?

It's a bit too acurate from my skill ATM...

I hold my fingers crossed for You all

Stevan C.
Maybe I should build a MJL21194 B-TS?

[ewizard]

I have 8X 6v 230ah marine batteries.

I was thinking about a camshaft, since you seem to think similarly, it might not be a bad idea....great minds think alike after all.

Comparing the Brandt circuit, the only real difference would be the rectifier will be replaced with 2 turns of 000 gauge wire for the primary, and 20 turns of 6 gauge for the secondaries. Tesla's Patent on Electrical Transformers, figure 1. I already have this transformer made. It can handle some power, even though it's not a closed magnetic circuit. The core is a 3"x1" steel round.

I will have 1 relay disconnecting the transformer and the whole circuit so that way I can run the commutator without power applied if necessary.

My current progress is I have all my contacts made. I am constructing the cylinder for the contacts to mount on today, if I work fast enough I might even have the bearings put in.

If I can just run my computer using this, I'll be set. Lighting is childs play, the real goal is to be able to produce mains power from my shed...or garage. And then apply it.

The only thing I am really missing is a camera to show you guys. But I'll see what I can do about those 5 dollar disposables and put in on a CD.

p.s. I know an inverter does the same thing, but I'm that kind of guy that makes everything himself.

I'm wondering if it might be an advantage to use a nylon plastic or PVC material as the cylindrical core for holding the copper contacts. I have this idea that electrostatic charges could come into play if brushes are dragging across the plastic in between copper contacts. Just a vague idea at this moment but wanted to mention it.

[Matthew Jones]

I'm wondering if it might be an advantage to use a nylon plastic or PVC material as the cylindrical core for holding the copper contacts.

It will melt..

12 volt 5 amp at 800 rpms, melted piece of 2" schedule 80 that I used once. I have also used breadboard materials. It got hot an smeared the plastic onto the brush's I was using.

I had the best luck with a a piece of OAK. YouTube - TeslaSwitch_6_1.MPG

I could ad a mechanical load to the motor and generate extra power.

It ran great an still would if I put it back together. But I have long since scraped it for parts. It is a direct rotary copy of Ronald Brant setup

[StevanC]

going there (progress report)

Re:MOSFETs
(a side note)

A. to properly drive a MOSFET means getting it's gate above the source (Gate="base"; Source="emitter", Drain="Collector" on TO-220 packages) for a certain voltage: the closer the gate is to it's maximum rating the less heat the device will produce (almost) regardless of the LOAD it caries. Just keep it as high as practical.

B. The instant the gate is grounded the device ceases to conduct any current worth a mention - so no heat is generated.

C. The only heating that occurs is when both the voltage across the device and the amperage passing it are significant. This happens while we transit the gate-source voltage level.While we charge the gate, which can't happen instantly, we produce a "ramp" going from Source-level up to Source+18V (for example). All the while we are "not there yet" we have the device in state of crossing from insulation to conduction.

Transition:
While the device has from neglectible resistance and transitions to a significantresistance, we have both significant amperage passing trough and significant FVD ( Forward Voltage Drop ) across the device.
And that means [B][I][U]heat[/U][/I][/B].
There are practical ways of handling heat:

1. avoid heating

So, we either shorthen the time we generate heat (as short ramp up times as practical).
Or, we make the transitions so seldom we [U]recover[/U] from overheating through means of a massive heat-sink
Or, we do both (as we intend).

then there are problems feeding choosing devices:
It would be practical to choose a MOSFET with a P-channel for device above load configuration, as the N-channels MOSFETs take advantageously the FVD of the target LOAD when "below" (closer to "GND") to feed the gate with.
But the P-channel equivalents (and not all Nmos-es have one) are typically three times (x3) the Rds(on) of the Nmos (=heats mor while handles less)
Advantages can be found in sub 3A applications, maybe up to 10A top, but our intended 200A/36V (=55V rated mosfet driven allowing 20V+/- on gate) it's just about not quite there yet. AFAIK
Now, we could stick to Nmos-es only (N-channel MOSFETs) but then there is the problem where we need the switching to occur above a LOAD and all sorts of place in between.
This calls for "floating" gate drive.
And there we have two options:
1. boot-straping
2. transfromer-de-coupled floating feed

the boostrap has limitations: the Vboost has to be "floated down" occasionally to "fill the bucket" otherwise it won't work. And furthermore it has to be able to keep the Vboost above a certain threshold voltage to prevent under-voltage lockout that happens at 7...9V.

most low gate voltage MOSFETs touted to operate even from 4.5V already, in reality achieve actual full saturation only above 8V and still have performance increase when driving the gate above 12V.
This means there is no really low Rds(on) (=our goal) without a sufficiently high gate voltage regardless of the MOSFET in question

the transformers offer what look like most advantageous option:

Optimus-prime wins, decepticons are gone...
kidding,

The transformer offer:
a. freely floating voltage
b. feeding of arbitrary Pmos or Nmos with the exactly same WF (WaveForm)
c. provided adequate measures taken, clear oversight of the driving to power implementation

But there are traps:
I. Failure modes can include transients that lead all the powerstage to a early grave
II. it is possible to generate/intercept EMI right when we are "this close" for the sub 10ns transient really do propagate easily off wires... (=heavy grounding)
III. Good transformer are not easy to build (=skill), and then still require time and materials (=cache) to be built to a quality envisioned in the first place.

(/end side note)

I was in my "lab" last night,
1. I made the 07301 Quer and got it on the bigger "daughter board"
2. I made the IRS21851 (high side high voltage high speed single gate driver with undervoltage lockout (@9V) and overvoltage 20V clamp) daughter boards. They get plugged to the motherboard
3. I made the rough part of the supply (LM7812 feeding a LM7805 that feeds the 07301), so I have 5V and 12V and the full supply voltage (on PVs it goes as high as 24V)
and run few basic tests:

a. The 21851 won't Que (on output) until a real MOSFET is connected: I thought it was broken now i know it was not )

b. It is easy to test the outputs of a 5V providing "Quer" with a blue 3mm LED: no need for a resistor and modes current trough

c. The Vboost (still my intention to test) is a bit low when fed from 12V, maybe I will feed it directly off the raw input (trough the 4007 diodes)

d. When run from a Walwart "12V" (=16V off duty) teh instant the MOSFET (IRF3205) lights the bulb, the PSU sags, and as soon as the Vboost goes below 9V (10uF20V tanatlum cap)

more info to come later...

Stevan C.

[StevanC]

current progress (preparations continued)

on the pictures:

ATmega48
portB
2x IRS21851 (backed with 2x 10uF tantal (below the boards - not visible)
on the small heatsink: "back to back" LM7812 and LM7805 (no thermal grease)
on the black big heatsink: IRF3205 (one not connected (yet))

The two black diodes: UF4007 (substitutes 1N4007)

questions

Stevan C.

[wwdotme]

Hey chainmailleman
are going to use the 4 switch or 6 switch set up?

And Matt did you try the 6 switch or just used your 4 switch setup?

Guess I'm going to build the six and try both.

thanks guys

[ewizard]

It will melt..

12 volt 5 amp at 800 rpms, melted piece of 2" schedule 80 that I used once. I have also used breadboard materials. It got hot an smeared the plastic onto the brush's I was using.

I had the best luck with a a piece of OAK. YouTube - TeslaSwitch_6_1.MPG

I could ad a mechanical load to the motor and generate extra power.

It ran great an still would if I put it back together. But I have long since scraped it for parts. It is a direct rotary copy of Ronald Brant setup

Thanks for saving me from a mistaken idea. I've got oak aplenty Nice video you did but surprised you scrapped it. Did it not make any excess energy?

[Matthew Jones]

......And Matt did you try the 6 switch or just used your 4 switch setup?.......

The easiest is the 4 switch Patrick Kelly version. The OLD version with diodes going the correct direction. If you don't want to use diodes there are options but its a big commutator. I have drawing somewhere of how it would lay out. Let me know if you wanna go all out.

Thanks for saving me from a mistaken idea. I've got oak aplenty Nice video you did but surprised you scrapped it. Did it not make any excess energy?

I did a run on it that lasted almost 2 months. I was not sure of the total battery capacity though. I had it drive a small motor that put out 12 volt 1 amp and ran light bulb.
I still have everything I just took the diodes and motor out.
I have several better versions, and I am not nostalgic. It was the first real good one I built.

Matt

[SeaMonkey]

The easiest is the 4 switch Patrick Kelly version. The OLD version with diodes going the correct direction. If you don't want to use diodes there are options but its a big commutator. I have drawing somewhere of how it would lay out. Let me know if you wanna go all out.



I did a run on it that lasted almost 2 months. I was not sure of the total battery capacity though. I had it drive a small motor that put out 12 volt 1 amp and ran light bulb.
I still have everything I just took the diodes and motor out.
I have several better versions, and I am not nostalgic. It was the first real good one I built.

Matt

Has anyone involved with this project put together
a verbal description or explanation of what is
desired in the way of performance objectives or
operational parameters for the final goal?

Apart from being as 'simple' as possible and
as efficient as possible?

Something along the lines of an RFP?

[Matthew Jones]

Has anyone involved with this project put together
a verbal description or explanation of what is
desired in the way of performance objectives or
operational parameters for the final goal?
Apart from being as 'simple' as possible and
as efficient as possible?
Something along the lines of an RFP?

It would not make difference. Everyone is going to try there own thing as it stands, especially the first time. After that the goal is to extend the life of the battery while doing work. In some cases maybe even recharge the battery.
Its not that huge of list.
Easy the way to go no matter what your build , your first time.

Matt