Electronics > Projects, Designs, and Technical Stuff

(1/4) > >>

For the science fair this year, I wanted to do something politically correct because THOSE are the projects that always win. (Last year, the winning project was COLD FUSION from some dorks in California. (btw, it didn't work....)) So I tried to hook up a speaker with a full bridge rectifier to capture sound waves from busy intersections. Needless to say, it sucked at getting usable power and 99% of the power was RF anyway. (from the coils in the speaker). So then I started making these radio power modules.

Now that the recap is caught up with the present, I want to introduce everyone to the problem. I have measured the power a million different ways, and they all point to an exponential power curve. It's like taking a AA battery and putting it in series with another AA battery and getting 4 AA batteries worth of power. This is WRONG, but I can't prove WHY it's wrong. I simulated the circuit in LTSpice, and after 18 hours of simulating the high frequency circuit so I could have high resolution, I got a curve as seen in this screenie. (Also, the circuit is there at the bottom! :))

jpeg image hosting

So, to outline the other methods I measured it with: I charged a capacitor and used J=FV^2 to calculate the Joules produced, and divided that by the 5000s to get the watts. I got the same exponential curve with a correlation factor of .998. (sorry, quadratic... easily confused!) Also, I wired up an opamp with 1,000,000x gain and measured the voltage across a megohm resistor and simultaneusly measured the voltage. got the same curve. (R^2 = .97) I have not tried a scope yet though, because I live in a tiny town where nobody has one.

So I have only one theory: I think maybe GND is oscillating when I add them in series. I have all of the series circuits with common analog ground, but GND is not connected to the negative terminal of the modules. I think that the AC is passing straight through all of my caps and going out my rectifier's negative terminal and is oscillating. But, I don't know for certain. The simulation suggests it doesn't and all my measurements say it doesn't, but it is the one thing I can think of that it might be. I do not know a way to measure that without a scope though, and there is NOBODY here in Cruces with a scope.

Please please please help me! (I have tried using my computer as a scope through the audio ports, but it doesn't work because the frequency is too high and the voltage too low...)

Stephen Jenkins

I'm not totally sure what the question is...

But you are modelling a theoretically perfect voltage source (i.e. capable of producing infinite current) and driving a Cockroft-Walton multiplier with it. Yes, it would produce power proportional to the square of the voltage. Each one of those diodes will need 0.6V or so before it conducts. Possibly not what you are after when you are scavenging a few mV.

Alright Zad, I have some clarifications on the question. I want to KNOW what the power I am getting is. (It is NOT exponential because of the 2nd law of thermo.) The idea behind using the voltage multiplier is that the AC should pass through the capacitors without a wasteful diode voltage drop and should hopefully switch the diodes on without as much loss. (If you want, I can show with real world measurements why I'm convinced a x8 multiplier is best. That was about 20 revisions ago! :D) I am unaware of the voltage source supplying infinite current, so that may be a solution to my problem. After inserting a current limiting resistor, I should find out in about 18 hours! But I seriously doubt that is the problem because of all the real world measurements I've taken. But it still is to be seen what the result is. In the time since I posted about 4 hrs or so ago, I've placed the results of the simulated experiment into a spreadsheet. (which I most regretfully do not know how to upload and link to... :( If you want I'll email to anyone interested in seeing it!) On the simulation, notice that the voltage source is set to one volt. I couldn't find a nominal figure to work with when considering noise (which is what I'd like to harness. It's no fun just stealing from AM stations. I made sure my antenna length was in the IF, where there shouldn't be transmissions.) I just settled with 1V to see if I still get an exponential curve.

Here are the results of the real world measurements I took (forgive me, not enough time to use the table command, it's REALLY time consuming to use...):
# in Series         Max Voltage    Avg Current    Watts
1                          5*10^-3            1.1*10^-6         5.5*10^-9
2                          10.6*10^-3       .6*10^-6          1.2*10^-8
3                          104*10^-3        1.3*10^-6        1.4*10^-7
4                          200.1*10^-3     1.7*10^-6        3.4*10^-7

Excluding the anomolous 2nd current data point, the fit is perfectly quadratic. The regression yields:
4.97E-8x^2 - 1.36E-7X + 8.97E-8
with an R^2 of .999

This was the results from my opamp amplified current (because I can't measure uA with my meter. Dave's uA was the inspiration for my design, so I think it's pretty solid... if not, the systematic error should have canceled as I built confidence in my measurements. just something to think about...)

Then, this afternoon, I measured using dBm (thanks Dave for showing me that too!) and I got:

-44.7dBm
-44.4dBm
-39.7dBm
-36.2dBm

and converting that to milliwatts, you get an equation very similar to the last with an R^2 of .990. And to try and get rid of meter flucuations, I measured the noise. I had zero noise as I added more wire between the probes.

That's another good installment of information. I assure you that this has been MOST frustrating to me. I've spent weeks reading the Art of Electronics and watching Dave's videos, and other forms of reference trying to figure it out. I've asked numerous forums and engineers, so it's down to you guys! Thank you for responding quickly as well! I REALLY appreciate it because my deadline is March 29th.

edit:
I thought I should share my graphs, so you guys don't have to sit and look at numbers....

image hosting jpg
From that, I plotted the return of power versus cost and determined that x8 was the best because it cost the least per measured watt.

Because OpenOffice refuses to plot TINY numbers, I can't show the spreadsheet plots as more are added in series. But they all look very similar to the screenie of the last post except the amplitude...

Double EDIT:
I stopped the simulation in the middle of the run because the pattern was clear. (BTW, I am SO glad I know about RC time constants and didn't waste my time in the uS anymore!) So here it is:

image hosting png
The power was SIGNIFICANTLY reduced... but the numbers do not even begin to explain why my real measurements are what they are. I was measuring millivolts at microamps, not millivolts at picoamps. Clearly my current limiting resistor is off. Anyway though, it isn't working right still on account that my voltage is set way high on purpose, but I'm getting less out than I would expect based on real measurements. So if I DID do the simulation correctly, can anybody explain my REAL results because they don't add up and they consistently don't add up?

p.s. Sorry about the mega long post and the double edit. Plz plz plz reply! :D

Rerouter:
small point, most voltage doubler designs i see have the negative leg tied to then end of your C8, and the positive on the negative terminal of your voltage source,

would post a digram, but my simulator just crashed, so currently remaking it, also i was also noticing the slightly logramatic increase of wattate as the capacitor charged, but feel that might just relate back into time constants, as for where to measure the current, perhaps give your capacitve load your charging a 10u ohm resistor infront of one of its legs and measure on that, as that is a true reference of what current is entereing it,