another small breakthrough on our NERD technology.

[MileHigh]

PW:

I would think that only the Q2 DC bias current (likely around 200ma +/- 50ma) and less than 20% (possibly closer to or less than 10%) of the AC current would be unaccounted for during the osc phase if the FG common was connected to BAT- instead of the CSR.  The bulk of the AC current would bypass the FG via the Ciss reactance of all the MOSFET's which would be much less than Rgen at Fosc.

I think that you are missing a fundamental point.  I am looking at Q2-Q5 as a three-port black box without having to consider frequency and reactance.  I said, "We also know that the sum of the currents I_IN, I_OUT, and AC_OUT equal zero."

So if AC current is flowing out of the box from AC_OUT to the battery ground, then current has to be flowing into the box from either I_IN, I_OUT, or both.  Knowing the polarity of the battery power source, we assume that the current comes into the box from from I_IN.

Likewise, if AC current is flowing into the box from the battery ground to AC_OUT, then current has to be flowing out of the box from either I_IN, I_OUT, or both.  Knowing the polarity of the battery power source, we assume that the current flows out of the box via I_OUT.

Obviously the above two examples are simplified.

So either I_IN or I_OUT will account for all of the current associated with the gate capacitances of the MOSFETs and all of the AC current flowing through the CSR.  As a result, you simply don't have to consider the AC current flowing though the the CSR.  If the coupling on the AC_OUT port was DC and not AC, it would be a different story.

Ultimately the true net current flow is unidirectional.  Seeing a negative voltage across the CSR in negative oscillation mode is not an indicator that the battery is recharging.

I guess if somebody does work on the bench we might see this looked into.

MileHigh

[picowatt]

MH,

If we knew the drain current, we would know the total of AC and DC currents at that point.

At the source, the AC and DC paths split.  DC to BAT- via FG, AC to BAT- via FG and moreso via CSR.

We don't have Idrain for AC or DC, so we must rely on the source current paths for measurement.

I am not exactly sure what you are saying.  We can't ignore the AC paths as the AC is not symmetrical.

PW

[poynt99]

So.. what happens when NO FG is used at all, nothing hooked to the circuit at its points, and one simply "tickles" the Q1 gate/Q2 source  with a suitable resistor attached to batt positive? By tickle I mean no full contact, just a scratch and then removal.

Not much should happen TK. You will get a short burst (or multiple bursts due to contact bounce) of oscillation, but as soon as you take the resistor off battery--the output should go flatline.

The reason being; the MOSFET needs a slight DC ON-bias in order to oscillate.

You might want to ask yourself, "what would happen if I supplied ONLY a constant DC bias?". I know what will happen, and I assume you do as well.

.99

[TinselKoala]

Well, I haven't done it lately but I got the "no FG + tickle" idea from humbugger's work on that other forum a year ago. When I tried it the first time, it did what .99 said, and then when I really noisily and barely tickled, it went into an oscillation mode where there were oscillations across the whole trace which gained in amplitude until I grounded the tickled spot again-- like what MH described, I think. I think without a pull-down resistor or a zero or negative gate signal, a turned-on mosfet gate might just stay on until something drains the charge away.  But with floating gate a mosfet could do just about anything. (Except recharge a battery !) I'm going to wait until I have a few more spare PG50s before I try it again, just in case.

No comments about my most recent videos? I should have thought that the MOSDEF demo especially would have been interesting in light of the current (no pun intended) discussion. And don't forget about alt.snakeoil.... there are always some good video reports under that alias.

I've shot another video that I'm processing and uploading now, showing another interesting effect, that I predicted earlier but had not yet tested for.

[Rosemary Ainslie]

Picowatt

Regarding Q1 not turning on when the scope shots say it should be...

Not actually.  Late yesterday I was finally able to speak to someone about your questions related to the scope shots.  The immediate counter question was 'why is he applying DC?'  Here's the thing.  Apparently the coupling is set to DC -  throughout.  But the peak to peak voltages and 'bias' (not sure if that's the right term) of each channel is also shown.  Which means that 'correctly' you need to adjust to the values of the signal as AC - which is applied from the signal generator.  And that value is given.  You'll then see that the voltage is NOT the >10 volts but something considerably less.  Which also means that there is probably not enough voltage at the gate of Q1 to turn it on - as applied to that first test.

I have looked at as many scope captures as I could find of your testing and it appears that in all captures made after 2-22-11, Q1 is not functioning as it should be.  Scope captures made on or before that date indicate that Q1 is performing as one would expect from the data indicated by the gate drive and CSR traces.  It appears that on 2-22-11 there were some particularly "spikey" tests being performed as the duty cycle was being modified and I suspect that Q1 was being stressed at that time.  If you have additional captures to look at, you should be able to narrow down the date when something happened to the Q1 portion of the circuit.  I hope that helps.

Which also means that there is nothing unreliable about our data capture and there's no need to question the calibration of that machine.

Regarding the location of the connection of the function generator lead's ground clip, it does matter a bit.  When the generator output swings negative, Q2 is biased on.  With the generator set to its full negative position, if the schematic of the output stage .99 posted some time ago is correct, a maximum of -14.5V can be present at the FG output terminal when measured open circuit.  With the internal Rgen of 50 ohms, this would bias Q2 on at approximately 200ma.  This 200ma. flows from the battery through Rload, Q2 and the functon generator.

When the output from the generator 'swings to negative' as you put it - then it is applying a positive voltage at Q2.  How exactly do you propose that any negative current at all is then passed through RLoad?  Especially in view of the oscillations that then manifest which most assuredly are not unidirectional.

If the FG's ground clip is connected to the battery negative, this 200ma. of bias current will not be displayed on the CSR traces and hence not accounted for in the math calculations.  If the FG's ground clip were connected to the end of the CSR opposite the battery minus connection, this 200 ma. of current would be indicated at the CSR as a continuous +50 millivolts by the CSR scope trace for the duration of the FG's negative voltage output.  As the cycle mean indicates -28mv or there abouts, this +50 mv, if not in the CSR loop, represents a significant error.

Again. I appreciate your need to imply that there is an applied negative current flow from Q1 when the signal switches to negative.  Can you show us where this is evident on any of our waveforms?

I wish you could make more data available regarding the "flat battery" test you discussed at the beginning of this thread.  That test seems very interesting and using only one MOSFET in the Q2 location and three flat batteries makes things a bit simpler to replicate.

I'm reasonably certain that this could be tested - rather easily - by any one of you. And I'm also rather keen to see some data from TK's 'claimed' replication.  I'm not actually sure who here is working openly and who isn't.  There is absolutely NO information related to the NERD circuit that is being withheld - to the best of my knowledge.

Kindest regards
Rosemary