PopularFX
Home Help Search Login Register
Welcome,Guest. Please login or register.
2024-03-29, 08:47:07
News: A feature is available which provides a place all members can chat, either publicly or privately.
There is also a "Shout" feature on each page. Only available to members.

Pages: 1 ... 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 [20] 21 22 23
Author Topic: Graham Gunderson Energy conference High COP demonstration  (Read 211997 times)
Group: Elite
Hero Member
******

Posts: 3537
It's turtles all the way down
Dear Spokane1

Sorry to hear about all the problems on the home front. That you can get anything done under those conditions is amazing.

I agree partzman is a huge asset to this project and you both have my gratitude for your service and commitment to it.

Hopefully Graham G. will throw us some light on the subject.

BTW Poynt99 is quite good at the sims and got me started using LTSpice as well as some guidance, for that I am also grateful.

Regards, ION


---------------------------
"Secrecy, secret societies and secret groups have always been repugnant to a free and open society"......John F Kennedy
   
Group: Guest
But but.....

Using partzman's last .asc file and looking at a longer timespan at the beginning and the end of a powered run, we see some interesting things. First, during the power-on time the average input power is always greater than the average output power. Second, there is a big transient at the beginning of the run.

Does the stored energy released after power-off correspond to the energy stored in the system by this startup transient?

In the first screenshot below (showing the power-off) I've computed the output power by both V2/R and V*I for the 50R resistor. These expressions give exactly the same waveform graph, but the V*I formulation is correctly identified by the sim as "Watts" and so scales automatically with the same vertical scale as input power,  like it should. (The Green trace is exactly underneath the Magenta trace so you can't see it.)

The second screenshot shows the power-on transient.
« Last Edit: 2016-08-23, 00:23:36 by TinselKoala »
   
Group: Guest
So... if I've done this right.... in the Startup waveform set, the energy in the Input power waveform section shown in the plot above is about 4.89 mJ and in the Output power waveform over the same interval is about 3.43 mJ, representing a difference of about 1.46 mJ in favor of the Input power.
And in the End waveform set, the energy in the Output power waveform, from the time of cutting the input power to the final decay, is 468 microJoules, or about 1/3 the difference between Input and Output energy at the startup.
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
So... if I've done this right.... in the Startup waveform set, the energy in the Input power waveform section shown in the plot above is about 4.89 mJ and in the Output power waveform over the same interval is about 3.43 mJ, representing a difference of about 1.46 mJ in favor of the Input power.
And in the End waveform set, the energy in the Output power waveform, from the time of cutting the input power to the final decay, is 468 microJoules, or about 1/3 the difference between Input and Output energy at the startup.

TK,

Thanks for running these tests on the modified GG1 circuit and I concur with your outcome. Since I didn't state my overall objective with my post I'll quote one of my statements and explain.

"So, the effective starting and ending Lleak currents are 188ma and 163ma respectively which equate to starting and ending energies of 35.2uJ and 26.4uJ. This means only 8.8uJ of energy was consumed from Lleak to produce 172.8uJ in the load. Theoretically if this is correct, the energy lost in Lleak should be able to be replenished on a cycle by cycle basis for continuous OU."

What I was/am looking for is some clue for any anomalous energy in a detailed analysis on a cycle by cycle basis as is reflected in the above paragraph. I did not mention any general reduction in overall COP as that was not the goal and I was well aware the output dropped over time.

My approach is to understand Graham's circuit as fully as possible before building a bench version. There is something unique in his circuit operation IMO!

I'm not sure I understand your focus on the startup as nothing can be really be ascertained until periodic stability is reached which I know you are fully aware of.

pm
   
Group: Guest
Well, it seems pretty clear to me. During the steady-state running, the average input power is always greater than the average output power. This is easy to see simply by inspection of the average power traces. This difference in power represents losses in the system, probably mostly in the mosfets and the coil resistances.

Only after the power supply is turned off is there some "excess" energy delivered in the output, since the output power decays to zero over a few of milliseconds while the input power is zero. Integrating this power curve from the time of power-off to the end gives a value for this "excess" energy. The fact that the power decays to zero represents, again, losses that are not overcome.

During the startup phase before steady-state is reached, the input power is quite a bit greater than the output power. Again, integrating both input and output power curves from the time of power-on to the time of steady-state gives a value for the respective energies, and shows that the input energy during this time is quite a bit greater than the output energy. This difference represents energy stored in the system, along with the inevitable losses.

The "excess" energy after turn off is only about 1/3 of the difference between input and output at the startup.

So it seems to me that the simplest explanation is that the system is storing some energy that was delivered in the startup transients, and returning some small part of this stored energy after power-off. During steady-state the input power is a little greater than the output power,  during startup some energy is stored, after shut-down some of the stored energy is released to the output. You get out what you put in, minus losses.

If one chooses to consider only the power in the primary tank as "input", that is, by measuring the voltage and current at the L1 coil terminals (by doing a differential voltage measurement Vecap2-Vecap to obtain the voltage across the coil for example) one sees that this power, in the hundreds of VA, is almost perfectly reactive and actually represents a small amount of stored energy sloshing back and forth between the coil and the capacitor, with losses being replaced by the power supply during steady-state operation, a tiny amount making it across to the secondary, finally to give it all up after power is turned off.
   
Group: Guest
Dear TK,

Nice summary.

Spokane1
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
Here are measurements from a bench test that is the result of my previous analyses of Graham's disclosure.

The schematic is generalized and is missing details.

pm

EDIT: Please disregard this post!
« Last Edit: 2016-08-25, 17:45:28 by partzman »
   
Group: Guest
Here are measurements from a bench test that is the result of my previous analyses of Graham's disclosure.

The schematic is generalized and is missing details.

pm

EDIT: Please disregard this post!

Dear partzman,

Tell us more about your bench test work if you can.

Spokane1
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4056

Funny thing (coincidence?) is that the control circuit used/designed by Spokane1/Tinselkoala (see light purple area in diagram 1 below) to drive the GG setup has the same
purpose as the control circuit designed by verpies (see light purple area in diagram 2 below) in the "sharing-ideas-on-how-to-make-a-more-efficent-motor-using-flyback-moderated"
thread on overunity.com here:  http://overunity.com/16167/sharing-ideas-on-how-to-make-a-more-efficent-motor-using-flyback-moderated/msg471611/#msg471611

Allthough it used different components (HEF4047BP), the outcome is the same, see screenshot 1 where i could copy the timing diagram originally presented in the beginning of this thread,
see upper part of diagram 3 below (i had to add a parallel 47nF capacitor to C7 to "delay" the pulse somewhat more).

 
I now used a 555 timer to supply the master pulse (A / yellow), but i could use a FG via an optocoupler to do the same.



Regards Itsu
   
Group: Guest
Dear Itsu,

I suppose that once we figure out the fundamental physics of this non-classical conversion process we will re-discover a lot of older approaches that came close or actually did function to some extent.

It is good that you noticed the timing similarities.

Just a note. In the Gunderson project there are two topologies under consideration. The simplified version is the one I'm exploring at the moment because of cost considerations (fewer component count). The more advanced approach is being explored by partzman via his simulations. The Graham notes you copied are much closer to partzman's circuit.

The paradox I have ran across is that the reverse engineering of Graham's logic board only yielded one signal for the H-Bridge and its complement. His notes describe two signals that are offset. I believe this would require another timer. So I don't know what is going on here. I believe that partzman uses a timing scheme that is different from what Graham's notes to Reyuki show.

Spokane1
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
Dear Itsu,

I suppose that once we figure out the fundamental physics of this non-classical conversion process we will re-discover a lot of older approaches that came close or actually did function to some extent.

It is good that you noticed the timing similarities.

Just a note. In the Gunderson project there are two topologies under consideration. The simplified version is the one I'm exploring at the moment because of cost considerations (fewer component count). The more advanced approach is being explored by partzman via his simulations. The Graham notes you copied are much closer to partzman's circuit.

The paradox I have ran across is that the reverse engineering of Graham's logic board only yielded one signal for the H-Bridge and its complement. His notes describe two signals that are offset. I believe this would require another timer. So I don't know what is going on here. I believe that partzman uses a timing scheme that is different from what Graham's notes to Reyuki show.

Spokane1

Spokane1,

Actually in my post #458 I show using two drive signals to the H bridge that are identical but with phase offset as Graham describes. That post also shows the timing relationship for the synchronized mosfets when a bucking condition exists in the secondary.

pm
   
Group: Guest
Spokane1,

Actually in my post #458 I show using two drive signals to the H bridge that are identical but with phase offset as Graham describes. That post also shows the timing relationship for the synchronized mosfets when a bucking condition exists in the secondary.

pm

Dear partzman,

So your excitation H-Bridge signals are the same as what Graham showed Reiyuki in the notebook? And it is your operation of the synchronous diode with bucking coils that has been slightly modified.

I need to get this straight so I can follow up on your approach later on.

Any ideas on how the logic chips need to be modified to generate that second timing pulse?

Spokane1
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
Dear partzman,

So your excitation H-Bridge signals are the same as what Graham showed Reiyuki in the notebook? And it is your operation of the synchronous diode with bucking coils that has been slightly modified.

I need to get this straight so I can follow up on your approach later on.

Yes I believe so.

Quote
Any ideas on how the logic chips need to be modified to generate that second timing pulse?

Spokane1

To generate the signals for the H bridge drivers using discrete logic, I would use two cross coupled D flip flops utilizing propagation delays and driven by a 555 timer. The drive for the synchronous mosfets would be derived from an appropriate edge of the previous D flip flops thru a variable OS. When you get to the point of trying the design, let me know and I'll draw a schematic.

pm
   
Group: Guest
When you get to the point of trying the design, let me know and I'll draw a schematic.

pm

Dear pm

I shall be happy to take you up on that offer in a couple of months or sooner.

Spokane1
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
It is slow here so I decided to post my progress in simulating a likeness of Graham's device. The attached sim is a model of a real transformer that is fabricated out of ferrite E cores in P material from Magnetics with a .010" gap in all legs. The ur is ~3000 on both E's.

The transformer pix shows that three windings are used but the two outside secondaries are left open in the sim so only the primary is driven. Each winding is 200 turns. I realize that G used U cores but I want to prove that I can accurately simulate a known core/coil arrangement before attempting simulation with additional unknowns.

The primary inductance of the actual transformer measured with a Genrad bridge is 5.1mH. The sim calculation from the plot is 5.17mH. In gyrator-capacitance modeling of inductors, the inductance may be calculated as L = N^2/Peff. Peff is the effective permeance of all core paths (in this case wrt to the middle leg of the core) and N is the number of turns. In this example, Peff = 131e-9 which calculates to 5.23mH. At this point Pleak is a best guess and will probably need adjustment when leakage inductance tests are done.

One may notice that I have used both B and H sources to model the gyrators. The B source is more useful as it's voltage can be determined by any function allowed in LtSpice where the H source is determined only by current. The non-linearity of the primary flux is determined by B1 and it's equation. Basically the voltage across B1 increases exponentially as the voltage drop across P2 increases. One may also notice the added bias to this calculation which I believe will be equivalent to PM bias on any portion of the core.

There is no core loss represented in this example at this point so there is no control over the hysteresis loop. This will be added later using a nonlinear resistance and then a B-H plot can be done and compared to the real transformer.

B sources are used to calculate the core flux and Pin.

pm 
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 4500


Buy me some coffee
It is slow here so I decided to post my progress in simulating a likeness of Graham's device. The attached sim is a model of a real transformer that is fabricated out of ferrite E cores in P material from Magnetics with a .010" gap in all legs. The ur is ~3000 on both E's.

The transformer pix shows that three windings are used but the two outside secondaries are left open in the sim so only the primary is driven. Each winding is 200 turns. I realize that G used U cores but I want to prove that I can accurately simulate a known core/coil arrangement before attempting simulation with additional unknowns.

The primary inductance of the actual transformer measured with a Genrad bridge is 5.1mH. The sim calculation from the plot is 5.17mH. In gyrator-capacitance modeling of inductors, the inductance may be calculated as L = N^2/Peff. Peff is the effective permeance of all core paths (in this case wrt to the middle leg of the core) and N is the number of turns. In this example, Peff = 131e-9 which calculates to 5.23mH. At this point Pleak is a best guess and will probably need adjustment when leakage inductance tests are done.

One may notice that I have used both B and H sources to model the gyrators. The B source is more useful as it's voltage can be determined by any function allowed in LtSpice where the H source is determined only by current. The non-linearity of the primary flux is determined by B1 and it's equation. Basically the voltage across B1 increases exponentially as the voltage drop across P2 increases. One may also notice the added bias to this calculation which I believe will be equivalent to PM bias on any portion of the core.

There is no core loss represented in this example at this point so there is no control over the hysteresis loop. This will be added later using a nonlinear resistance and then a B-H plot can be done and compared to the real transformer.

B sources are used to calculate the core flux and Pin.

pm

PM

How closely do you think this is starting to resemble EMJs partnered output coil setup?,as i am seeing some resemblance here.


Brad


---------------------------
Never let your schooling get in the way of your education.
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
PM

How closely do you think this is starting to resemble EMJs partnered output coil setup?,as i am seeing some resemblance here.


Brad

Brad,

I'm not sure. I have to admit that I'm not that familiar with EMJs setup so will have to take a closer look.

pm
   
Group: Elite
Hero Member
******

Posts: 3537
It's turtles all the way down
PM

How closely do you think this is starting to resemble EMJs partnered output coil setup?,as i am seeing some resemblance here.


Brad

pm said:

Quote
I'm not sure. I have to admit that I'm not that familiar with EMJs setup so will have to take a closer look.

The problem is that there is no setup or circuitry. Just a non inductive or bucking coil arrangement with no attached circuitry. This is being re-branded as "partnered coils". Lately any device utilizing two coupled inductors is being claimed to have the "magic" partnered coil approach. You would have to read some of the 484 pages at OU.com to see where this is going.

Brad

See my post #464

In all the pages 484 of the partnered coil threads here and on OU.com, and all the cut and paste, there is not one circuit offered by EMJ that actually produces the goods. In view of this I find it a bit false when he claims this old non-inductive winding technique to be his invention by renaming it "partnered".

EMJ, Chris I know you will probably read this, so prove me wrong by posting one circuit using "partnered" coils that can actually produce a COP>1 that you have tested with actual measurements and with good instrumentation.

To the claimant goes the burden of proof.

We can agree to disagree, and still be friends and gentlemen.

Regards
ION


---------------------------
"Secrecy, secret societies and secret groups have always been repugnant to a free and open society"......John F Kennedy
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
Being the impatient one that I am, I jumped ahead of what I had intended and did a model of a 1/4" square U core to somewhat represent Graham's device but certainly not even close to being identical. The results are interesting as can be seen in the attached sim below.

There are some best guesses in the model that include the leakage inductances, the 5 ohm core loss resistance, the core bias and the nonlinear exponent used in B1. There is a small gap of ~.005" and the core ur~2500 and identical for both halves. The load is 20 ohm and the secondaries are bucking during the synchronous mosfet "off" phase due to M5 and M6 alternately turning off every other cycle.

Pin is measured with B3 and is the differential of Ps-Pf times the primary current i(V6) and shows 2.36v on the plot but is really 2.36w. This represents the total reactive power across the entire cycle period.

The output voltage across the 20 ohm load R2 is -9.824v avg which equates to 4.83w. This is a real/reactive power COP = 2.05.

The input DC current i(L1) is seen to be 324.4ma resulting in an input DC power of 12.97w. The reactive peak power is ~ 230w.

The model needs adjustment and tuning in many areas but overall there is promise in Graham's design. What I envision is a unit that operates at a digitally synthesized frequency of 50-400hz thus allowing the negative energy portions of the reactive input to the primary to be returned to the DC supply thus creating a true OU device.

pm
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
Dear partzman,

Tell us more about your bench test work if you can.

Spokane1

Spokane1,

I am sorry that I originally missed this post and just now discovered it so my apology. :-[  I am willing to share my bench work but at the moment my only documentation contains measurement errors so I will repeat the tests and post when complete.

I was initially using E cores but will re-run the tests with fabricated U cores instead.

pm 
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
Spokane1,

Here are some bench results of a 1/4" square "E" fabricated core in Magnetics 'P' material with a ur (initial) of ~2500 and Bsat~.45T.

The core center leg is gapped at .010" and the primary is located on this leg. The secondaries are on the outside legs with no core gaps. The PMs are .25"dia x .12"L neos from Bunting in N30 material. The primary is wound in 5x34 litz and the secondaries in 12x34 litz.

The PMs are positioned such that the N poles face the center leg forcing the primary winding's mmf to alternately aid and buck the PM flux in the upper legs of the primary core half.

The input signal seen on CH1(yel) is from the output of an EV 7300A power amp driven by an Agilent 33120A in burst mode. There is no resonance capacitor in the circuit. The output load resistor RL is four paralleled 100 ohm 1% non-inductive Caddock power film resistors. The current sense resistor Rs is a 1 ohm 1% non-inductive Caddock power film resistor.

The schematic shows the circuit and scope connections.

Scope pix PM3_A shows the circuit operation when the phase of the input signal to the primary creates an mmf in the primary wrt to the PMs that creates a higher than normal core permeability during the time the input signal is in the clamp phase. Alternately, the core perm is lower than normal during the sine portion of the input drive waveform.

The Math(red) channel indicates the input power of CH1 x CH2 with offset correction applied to CH2. The load is connected directly to the secondaries with no mosfets or capacitor as can be seen. The output/input COP is 1.09 as is calculated from these measurements. One should not be fooled by the fact that only one cycle is measured as the scope has taken many averaged waveform samples that allows only one displayed cycle to be measured quite accurately.

Scope pix PM3_B shows the same circuit with the connections reversed at the primary with the resulting primary MMFs operating exactly the opposite as compared to PM3_A in respect to the PM core flux. Note the COP is now ~.70.

I plan to generate signals that will allow the secondaries to switch alternately thus creating bucking mmfs in the secondaries to compare the performance with the AC driven output.

pm   

 
   
Group: Elite
Hero Member
******

Posts: 3537
It's turtles all the way down
Good work again pm, I'm all in favor of using tools such as the amplifier and signal generator in the research phase.

I'm sure your amp has a low enough output impedance to effect the drive and clamp phase.

Regards
ION


---------------------------
"Secrecy, secret societies and secret groups have always been repugnant to a free and open society"......John F Kennedy
   

Group: Professor
Hero Member
*****

Posts: 3354
The load is connected directly to the secondaries with no mosfets or capacitor as can be seen.
What?!  No back-to-back MOSFETs in the secondary circuit?
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
Good work again pm, I'm all in favor of using tools such as the amplifier and signal generator in the research phase.

I'm sure your amp has a low enough output impedance to effect the drive and clamp phase.

Regards
ION

Thanks ION.  Yes I agree on using generators and amps when possible. I am always amazed what one can do with the newer arbitrary waveform generators!

The EV amp has a DF >75 at 20kHz and probably rolls off at higher frequencies. It replicates the burst sine very accurately although the power level is very low.  I would like to try the bridge mode to double the drive signal to the PM3 as the gain increases up to the level the single channel can provide so it is not operating at the most efficient point on the B/H curve yet. However, this requires differential measurement on the current sense resistor which I don't prefer. I have found that Hall based current probes (at least mine) are not accurate around the neos.

pm
   
Group: Experimentalist
Hero Member
*****

Posts: 1665
What?!  No back-to-back MOSFETs in the secondary circuit?

Yes I know. :o

Seriously, I expect the efficiency to go down with the addition of the synchronous mosfets but in all fairness to Graham, I'm not sure how he has connected or timed his secondary output. I'm not even replicating his topology but something does seem to be going on here.

What I see when closely looking at the circuit in operation, the input current waveform is displaced downward with the proper phasing as compared to no pms or reversed phasing. The result of this is slightly lower input reactive power resulting in apparent OU. Could be worth pursuing.

pm
   
Pages: 1 ... 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 [20] 21 22 23
« previous next »


 

Home Help Search Login Register
Theme © PopularFX | Based on PFX Ideas! | Scripts from iScript4u 2024-03-29, 08:47:07