OU/COP>1 switched cap PS cct like Tesla's 'charge siphoning'

[poynt99]

Hi Sandy.

hi Poynt

yes, removing R1 does increase max stored voltage achieved - did a trial run, couple months back i think, to see if worth following up & decided it will be

think efficiency was high but dissipation was focussed in DC res of L and therefore inaccessible, so just parked that for future test & continued with full RLC branches

It makes sense that there will be some losses in the inductor itself, although removing R1 will shift the total circuit impedance largely back to a reactive one so most of the energy should be transferred into C2 from C1. Hopefully most (if any) excess energy tapped in the process will end up in C2 as well.

In my sim, C1,C2 settle at a voltage of 5V with R1 present. Without R1, they settle at about 5.4V. In both cases, the presence of the inductor is bringing the transfer efficiency above 50%, because without it, the final C1,C2 voltage would be 4V.

original tests (using my handdrawn schematic you just re-posted) were showing efficiencies round 125% - BUT - this was on the assumption that the external energy converted in R whilst charging was equivalent to the final stored energy in the output cap, as claimed by textbook treatment

my subsequent tests showed this relationship was not constant - though for my particular tests it didn't reduce the overall efficiencies below 100%

latest switching control uses PC parallel port o/p thro' SFH618 opto-isolators, re-shaping test-side signals thro' CMOS Schmitt devices, driving straight into MOSFET Gates

opto i/p LEDs powered by PC port drive; all isolated test circuitry powered by C1, as you noted from earlier tests

most recent change to test circuit is replacement of P-chan MOSFETS (caution: partnos. from memory!) FDN304P replaced with IRF5305(?) because circuit chewed up the P-chan every month or so & wiped out my stock (i'd salvaged a few from scrap I/O boards)

all the best
s.

Thanks for the info and the circuit. I'll maybe draw something up showing the whole or most of the complete circuit.

Regards,
.99

[AbbaRue]

One point that causes confusion when dealing with capacitor charge and discharge times.
A capacitor charges to two thirds the input voltage in the first time constant.
Then it charges to two thirds the remaining voltage in the second time constant.
eg. If a capacitor takes 5 seconds to charge to 99.6 volts from a 100 volt source
It will charge to 66.6 volts in the 1st second
Then to 88.8 Volts in the 2nd second.
Then to 96.2 Volts in the 3rd second. (This is considered fully charge for most circuit applications.)

And when you discharge that same capacitor it follows the same scheme. 
It will discharge two thirds it's stored voltage within the first time constant.
So a capacitor charged to 100 volts will discharge to 33.3 volts within the 1st time constant.
Then to 11.1 Volts in the 2nd time constant.
Then to 3.7 Volts in the 3rd time constant.

[nul-points]

thanks for the TC info, Abba - useful for measuring the cap values in these circuits - the switching operation between caps in these circuits is mostly taking place at a very small fraction of the circuit TC


hi Poynt

In my sim, C1,C2 settle at a voltage of 5V with R1 present. Without R1, they settle at about 5.4V. In both cases, the presence of the inductor is bringing the transfer efficiency above 50%, because without it, the final C1,C2 voltage would be 4V.

i found i had a copy of the reduced Rload test run i mentioned - not exactly fitting your suggested variation but possibly enough of a pointer to see if it's going in the right direction - i can try a closer test when i get access to my setup again

input power was being provided by my original C1 (0.31F) so the test run covered approx 430  charge/discharge cycles of C2 (196uF) to discharge C1 from 8V to 7V

i need to confirm this following aspect when i can retest, but i believe the trace shows the voltages on Rload (1ohm) and C2 at the point of mid-energy input (approx 7.53V on C1)

the max stored voltage achieved on C2 is 5.33V

the results show slightly higher Rload energy dissipated in charging C2 than in discharging it - DC res of L is 0.5ohm so the total dissipated energy is at least 50% higher than that recorded on Rload

hope this is useful

all the best
s.

[poynt99]

Perform the testing you would like Sandy.

I recommend however simplifying as much as possible, which is why I suggested ( http://www.overunity.com/index.php?topic=4419.msg146734#msg146734 ) the simple C1 to C2 discharge/charge test through L1, and sticking with your 190u and 196u capacitors. The circuit is operated until VC1=VC2, and the voltage noted.

I don't think it can get much easier or simpler. If you get higher than about 5.6V then chances are you got it

.99

[nul-points]

Perform the testing you would like Sandy.

I recommend however simplifying as much as possible, which is why I suggested ( http://www.overunity.com/index.php?topic=4419.msg146734#msg146734 ) the simple C1 to C2 discharge/charge test through L1, and sticking with your 190u and 196u capacitors. The circuit is operated until VC1=VC2, and the voltage noted.

I don't think it can get much easier or simpler. If you get higher than about 5.6V then chances are you got it

hi Poynt

yes, i was just mentioning some values from that previous test since i'm not able to try your suggested test for a few days 'til i get back at my setup and my previous test was close to your suggestion of removing the Rload

i thought this test might give you some idea if it's in the right ballpark - it appears that the average input energy was approx 5.4mJ per charge/discharge pulse train and the associated total output energy converted was around 6mJ on the 1ohm Rload (with around 1.6mJ extra charging energy lost on DC res of L)

i'll let you know as soon as i've been able to run your suggested test

all the best
s.