COP 17 Heater | Rosemary Ainslie

[poynt99]

Poynt - if you're still there - could you please post those details again - any that are appropriate. I hate following links because it kicks me out of this connection and then it's time consuming to cross reference. So - if it's not too much trouble.

How does this sound: Describe exactly what measurements you would like to see from the simulation, and I will reproduce them.

.99

[ashtweth]

We also have poynt's COP meter that he donated to every one and can add that to the testing of Rosie's if needed, i prefer REAL work and testing in the incubator circuit. Again my money is on Rosie guys.

Ash

[poynt99]

Ash,

It might be applicable to Aaron's and Rosemary's circuit, but a rectifier/filter circuit would need to be added. Their circuit presently has no DC output. I assume you are referring to the "Is COP>1?" diagram at the end of the "Cheap Oscillators" / MRA document?

.99

[poynt99]

In reference to my last post, here is a question to Aaron and Rosemary:

We agree that neither of you have any problems with the notion of using a flyback diode to help capture/recover/regenerate (insert your term here) some of the coil energy back to the source battery correct? Also you are ok with using the flyback output to charge a totally separate battery correct? And it is one of your claims that battery recharging does occur with or without the flyback diode present?

Will you possibly allow the following proposed method to determine COP>1:

Use a transformer and full wave bridge off the load coil to convert the AC voltage across it to a filtered DC voltage. We would use a transformer so that the DC output voltage may again be referenced to the source battery ground. Then this output will be applied to the source battery in the hopes that it will gain charge?

This of course means placing the transformer primary in parallel with the load coil.

The use of a transformer may make things a bit lossy (depending on your slant here), but if COP>17, this slight loss should be more than compensated for to keep the circuit alive and running indefinately.

Just a thought.

.99

[Harvey]

Hi everyone,

I forgot, the other thing I'm going to try is the rotation of the "Load" 10 ohm wire wound resistor ...... maybe the wire winding direction may have some gain/loss difference in the circuit operation ?

Glen

Wow...I don't know why I didn't think of that myself.

Rosemary, do you have a diagram that shows the physical orientation of the load resistor to the HEXFET and timer?

That load resistor will be producing a polar field straight through the tube with strong magnetic flux extending from the corners of the resistor where the last windings are on each end. It projects off the central axis somewhere between 20 and 45 degrees (I'd have to calculate the geometry). That field could be interacting with the gate-source inductance or possibly even creating a hall effect in the gate. Whatever is happening, it is of sufficient magnitude to charge the gate and turn on the HEXFET. And this is clearly occurring when the timer is effecting an output 'low' to the gate. This phenomenon seems to require a specific flow of current through the gate to exist. Whether this is related to the charge timing or to actual capacitive pass through current has not yet been ascertained.

But it would be interesting to discover if the magnetic field itself is doing the triggering. Hmmm, is the phasing right for that? If only there was a way to simulate an inductive feedback path and test the phasing...oh, well.

On the heating of the IRFPG50. The device can dissipate up to 190W if the junction temperature is kept at 25°C - in other words, with really good heat sinking it can function as a variable resistor. However, if the Tc raises to 100°C then the drain current is derated from 6.1A down to 3.9A and that's one way to end up with a piece of burnt toast like TK showed us. But if we are heating up the HEXFET, then we are wasting energy. The best thing to do with a HEXFET is turn it on, and off, very quickly and avoid any sloping or curved waveforms. Spikes are fine, sine waves are bad.

So, while Rosemary's circuit showed an extraordinary Coefficient of Performance, it still seems that undue losses were present in the IRFPG50 heating. This tells us that the self-oscillation was attempting to adopt a sinusoidal ring function during the on period of the HEXFET. Therefore we can see that the drive on the self oscillation is coming from the ring feedback. Why? Because if it were from the Timer, then the pulse would be sharper and result in less heating in the HEXFET. In reality, we all experience a fair mix of the two. Ideally, we want to hit the inductor hard and quick with a nice sharp rising pulse to full saturation and then just as quickly or better, turn off the HEXFET and allow all that stored energy to slosh back and forth in the inductor until it is all gone. Depending on how long that takes to do sets our off time. The on time is set by how long it takes to saturate the load - we don't need any more than that. The greater the initial spike after turn-off, the more energy we know we've stored in the field and the better the efficiency. Even at best operation, we are still going to get 6.86W of dissipation in the IRFPG50. 24V / (10 ohm load + 2 ohm Rds(on) ) = 1.85A and that times the 2 ohms is 3.7V across the IRFPG50 at full conduction and that is 3.70V * 1.85A = 6.86W (or you could just use 3.7V² / 2Ω) How could we improve that? We can't, not with this device. It is inherent to the drain source resistance (Rds(on)). Now, the interesting thing here is, that 1.85A is the amount of current that would be flowing if we measured it for 1 full second. That is the definition of an amp; the amount of charge that flows past a given point for 1 second with 1 volt applied across one ohm. If we only let it flow for 20µs, then it really isn't a whole amp. And if it really isn't a whole amp, then it really isn't a whole watt either. In order to be a whole watt, it has to be there for a whole second. Interesting, isn't it? That is the whole concept between PWM power circuits. The actual wattage is reduced to its fractional value thus allowing adjustable power conditions. So lets say our device is running at 3.7% duty cycle and 10µs on time. This means there will only be 3700 on times each second compared to the possible 100,000 on times available (100%) This also means that our 6.86W is really only an average of 0.25W for any given second and this also happens to be the energy in Joules dissipated in the HEXFET because that is the definition of a Joule, 1 Wattsecond.

So if the HEXFET is really off, and all the BEMF is contained in the load resistor for dissipation, then the transistor (including the body diode) should experience zero current and zero watts during the off period. So mathematically we can deduce that 24V at 1.86A is 44.64W 3.7% of the time which results in 1.65W average circuit power through the HEXFET during 'On Time'. If 0.25W of that is in the HEXFET, then that leaves 1.4W to be dissipated in the load resistor. Now in this example we have about 3.7KHz. What happens if we lower the frequency down to 2.4Khz? Well, if we keep the same duty cycle, our energy will be identical. Why? Because we have fewer pulses but they are longer and the off-time is longer too. So we find that if the duty cycle is preserved the power ratio is frequency independent. And this is from 1Hz to 1THz and beyond, it doesn't matter as long as the on-time is 3.7% the total cycle. So how can we possibly get 17W of dissipation in the load resistor?

One way is to increase the current through the HEXFET temporarily. This device can handle pulses up to 24A as long as we keep it cool. But that would mean that the resistor must drop in resistance while not dropping in inductance. Is this plausible? Or even possible? We know that the inductor will have an impedance based on its internal resistance and inductive reactance that is frequency dependent. But the impedence will never drop below the resistance. So the 10 Ohms stays unless there is some magneto resistive action. So if the 10 ohms stays, there would need to be some miracle to increase the voltage to about 597V across the circuit. Clearly this isn't happening.

Another way is to change the duty cycle. And this is exactly what happens when the device enters its aperiodic operation. The on-time remains the same (and you will recall that if this were 100% we would have 44W dissipated in the load), but the off-time shrinks considerably. I would have to go back and check to see what Aaron and TK had registering for duty cycle during aperiodic operation, but in my case it can reach 50%. Well, 50% would be 22W, all we really need is 39% and we're there. But what about the 'shunt'? It measured 1.13W, properly averaged data dump from the scope across it for a long run. How can this be explained? It has to do with two simultaneous currents flowing from the battery at the same time in opposite directions. In this case we need to change our reference to the drain-resistor junction. After the HEXFET is turned off and the magnetic field collapses, this point crashes into a hard negative voltage more than 40V below B(+) and 28V below ground. So in this case, the B(-) looks positive relative to the drain. During this period, current flows through the resistor from the B(+) terminal and begins recharging the inductor, but at the exact same time, at least 28V worth of current is flowing from B(-) post, through the 'shunt' resistor and body diode to nullify the negative spike. That 28V BEMF not only does not register the positive inflow from the B(+) but it also works to negate mathematically, the previous forward current that has already flowed. In other words, there is stuff happening when the IRFPG50 is off, and part of that 'stuff' is messing with the power readings. What to do...???

I have suggested that when Aaron has time, we get 3 sensing resistors in place for each leg of interest. B(+), IRFPG50 Source, Timer circuit ground. Then we can run some aperiodic tests with all three being scoped and see how the current is flowing. The only problem with this test, is that his rig is not producing the 17W we are looking for...yet.

It would seem, that to reproduce the effects Rosemary experienced, we will need a lot of ringing with a strong negative component. My rig barely offers one cycle excursion below zero. I find this perplexing, because it necessitates a very vertical falling gate signal and a fair amount of inductance to resistance ratio to encourage a longer ring time with multiple passes below zero. Both of which are not present in my system. Aaron did have a 1 ohm resistor but this divides the voltage between the HEXFET and the resistor in a way that produces excess dissipation in the HEXFET. But it may be the price we have to pay to get enough energy stored in the field to do what were discussing here.

I think I'll stop rambling for a while.

[ashtweth]

Yup here is poynts one
http://www.panaceauniversity.org/Mag...0Amplifier.pdf

Let me know how you need it set up poynt, great another diagram you need to draw . Our one will utilize Rosie's circuit with solar power

[witsend]

Poynt

Also you are ok with using the flyback output to charge a totally separate battery correct? And it is one of your claims that battery recharging does occur with or without the flyback diode present?
Yes

Use a transformer and full wave bridge off the load coil to convert the AC voltage across it to a filtered DC voltage.
Not sure what a 'filtered' DC means. Presumably just that the AC changed to DC? I know from personal experience on AC tests that there's a marginal loss through those diodes on the bridge rectifier. Doesn't matter much on high voltages but not sure of the effect on lower values.

We would use a transformer so that the DC output voltage may again be referenced to the source battery ground. Then this output will be applied to the source battery in the hopes that it will gain charge?
Not sure why you need a transformer if the inductor itself generates the required.

This of course means placing the transformer primary in parallel with the load coil.
Ok. We've tested this variation but the current flow was quite extreme.

The use of a transformer may make things a bit lossy (depending on your slant here), but if COP>17, this slight loss should be more than compensated for to keep the circuit alive and running indefinately.

Poynt - are you proposing this circuit to be simulated or actually built? Quite frankly I'm having difficulty understanding the rectifier arrangement and where you'll return the energy - another battery? - the source? Whatever. Can you give us a schematic? And why do I get the impression you've done this? Indeed I'd be interested in the results. I'm a sucker for punishment.

But I would very much appreciate the data from Aaron's last test and from the published experiment posted here for easy reference. Please oblige.

[witsend]

Yup here is poynts one
http://www.panaceauniversity.org/Mag...0Amplifier.pdf

Let me know how you need it set up poynt, great another diagram you need to draw . Our one will utilize Rosie's circuit with solar power

Ash - you've got to be one of the kindest people in the world. I just hope your optimism here is rewarded. If you ever want an example of aperiodic oscillation you should check my own optimism gauge. Definitely in resonance.

[witsend]

Harvey - the load resistor was always placed within 6 inches of the MOSFET and more or less parallel.

I get it that you're trying to determine what actual energy is lost to the battery during the 'off' period? Presumably this question can be resolved by checking current flow through the MOSFET. In one of Luc's test he put an LED - or it might have been a torch light - at the source to determine this. But he worked at lower frequencies. In any event - I think it's worth doing this test with the shunts at the three legs.

Just so comforted to see you're still grappling. I had an idea you'd signed off.

[witsend]

simulations

How does this sound: Describe exactly what measurements you would like to see from the simulation, and I will reproduce them.

.99

Sorry Poynt - I missed this. Many thanks. Ok. Can you firstly do a repeat of the published circuit. Presumably it may require those amendments to the switch? Or do you even need a switch on a sim? Then at 3% ON - waveforms over the shunt, wavforms over the load resistor, waveforms over the battery. And some quick reference to their average values and the wattage levels of energy dissipated/delivered.

Then the same on Aarons new circuit.

Many thanks

[witsend]

Hi witsend,
speaking from different Resistors, what did the scopeshots look like what you got by your Tests,
they been simialer to the one we have here?
Thanks.

Joit - they were all over the place with a frequency that we could only gauge on an average. Actual frequency ranged between about 10Khz higher and lower. The spike values also varied but never exceded the voltage of the spike generated at the switch. In effect it looked like bad copies of the switched cycle. At the actual switch one could see a periodic waveform. Then these 'ghost' waveforms between each switched cycle. Hope that helps.

As mentioned repeatedly - I am just so sorry that I never photographed this.

[Joit]

Hi witsend,
Thanks and yes it does, especially this 'Ghost' waveforms sounds interesting.
But my Point is, they been at the Quantum article similar to this one we got?
Because every Load gives a specific Waveform, even more at pulsed Circuits.


@Harvey it sounds good at all, i guess, i go look for some loose Wire to play around with it.
FEMM4.2 is a Magnetic/Eleectrostatic Sim, when you are interested,
but i doubt, it can simulate this inductive Influence you did mention at the Gate from a Transistor.

[witsend]

Hi witsend,
Thanks and yes it does, especially this 'Ghost' waveforms sounds interesting. But my Point is, they been at the Quantum article similar to this one we got?
Because every Load gives a specific Waveform, even more at pulsed Circuits.

Not sure which waveforms you're referencing. If it's Aaron's 'sine wave' number - definitely not. If its the waveform on your video - yes, close. If it's Aaron's earlier waveforms they're similar.

I like the fact you're winding your own resistor Joit. You must try and use thick wire - I always found it helped. I don't think it needs a core. I'd be most interested to see this.

[Groundloop]

Done soldering

Hi Group,

Finally, my RA circuit is soldered and ready to be tested.



The heat sink to the right is my 0,25 Ohm 1% accuracy shunt resistor.
(This resistor is not wire wound.)

At the back you can see my 80 Watt 10 Ohm wire wound load resistor.

BTW, GOTOLUC has approx. 19 of these PCBs to give away for free.
Just PM GOTOLUC is you want a PCB. (The PCBs are without parts.)

Regards,
Groundloop.

[witsend]

Groundloop - it looks really, really pretty. Be most interested in seeing your numbers when you start testing.

And thanks again for going to the trouble. it's much appreciated.