ZVS driven LED array

[heatlocke]

As a long time lurker and newest member I would like to share my latest project. Most here are aware of the induction cooktop experiments of the past. Just about everyone has tried it. We wore it out ourselves 10-12 yrs ago and thought we laid it to rest. J Naudin  and others published work (GEGENE). Recently, another Tesla pancake coil arrangement came about and we found ourselves back at it again. With (4) 500 watts halogens just screaming, it appeared again like a big gain and it started all over again. Is it OU ? Can we loop it? Someone needs to put this to rest for good. ZVS drivers and batteries never seem to get us there. Cooktops are fussy and temperamental and in the end you are still with out much control over it.  I decided to build a circuit with total control over everything and thoroughly test it once and for all. Halogens (dinosaur bulbs) offer a pretty good platform for using Calorimetry. AI can identify possibly 20-25% increase in efficiency and maybe 5-6 ways its doing it. IMO, my own testing has disqualified most of it but Q effect, skin effect and disruptive discharge seem to be leading contenders for any gain.  Turns out, it a pretty dam good LED engine.  I can drive 80,000 lumens for 450 watts. No rectification,  (Anti-Parallel, 4+4 Branches ... 16 total) The circuit could easily handle double that.  I`m looking for comments, criticisms and where to go from here. I made a vid for some friends still working on the project. If you can stand me rambling on for  10 minutes please give me your opinion. I`ll share everything.

https://youtu.be/hYXf6Vs0xbk

[Chet K]

Heatlocke,

Welcome to O.U.R. Forum.

Tremendous gratitude that you share your Experiments!!
I remember plenty of interest in the induction/Hobs

Respectfully,
Chet

[Verpies]

The larger the area of the loops, which I have marked in red color, the more gate drive ringing occurs and the less stable and less energy efficient your circuit becomes.
Excessive gate ringing will also destroy your MOSFETs.



The scopeshot below depicts the voltage waveform of your primary winding.
Voltage measurements are usually inappropriate for inductors as they are current devices.  Next time please measure the current in the primary winding instead of the voltage across it.
Also, the region, which I have marked with the red ellipse, shows high frequency ringing (much higher than your 48kHz base frequency) voltage oscillation.  This is characteristic of high impedance ringing across an inductor.  This is generally an undesirable side effect which overvoltages the MOSFETs and wastes energy - unless you are aiming for it on purpose. If you are - what is its frequency ?



The triangular voltage output waveform of your secondary winding is devoid of this high frequency ringing.

Last, but not least, the output power measurement is trustworthy when using low-inductance incandescent light bulbs since a stable power-to-luminosity curve can be established for them, however not so much when using LEDs with their nonlinear characteristics and temperature coefficients.
...but most importantly, your input power measurement is untrustworthy.  The Kill-a-Watt plug-in power meter is unreliable at high frequencies and with high crest factor waveforms and multiplying the average current measured by your clamp-on ammeter by the line voltage is fundamentally flawed. 
average Amps *  average Volts = average Watts is valid only for DC.  This is because the product of averages is not equal to the average of products.


P.S.
Like an op-amp, I provide only negative feedback.
Positive feedback leads to instability and saturation.

[heatlocke]

"Like an op-amp, I provide only negative feedback."

I like all of it, thats why I`mm here.  Are you saying not to trust the kilowatt meter ? Its at the end of a long skinny extension cord and the entire circuit with everything plugs into it. The two amp meters are just to backup each other up because I don`t trust either. Because I`m at 34% duty cycle my bus only reaches 150vdc. Times 3 amps is this not 450 watts ?

[Chet K]

Heatlocke,

Sorry to but in!

Are you saying not to trust the kilowatt meter

I learned the hard way about down stream ( from kilowatt meter) accuracy when it was running "unusual" loads..( I was warned by Poynt99 at the time)
Ended up just doing simple water heating (caloric) tests against my Devices _claimed_ results.
Was a big enlightenment/disappointment

(In my case.
But definitely opened my eyes to how vulnerable meters are when deployed outside their design parameters.
Caloric is best whenever possible,and I have found it's almost always possible to do various quick forms of caloric measurements ( member ION had a "fixed loss to ambient " caloric test method he taught ...
Wildly accurate,and little more than a cardboard box,a thermometer and a timer!

( many cool versions)

Again
Sorry for interrupting.
Respectfully
Chet K

PS please remove this post if inappropriate

[Verpies]

Because I`m at 34% duty cycle my bus only reaches 150vdc. Times 3 amps is this not 450 watts ?

That's right, it doesn't have to be 450W.

Input power is notoriously hard to measure if it is not DC.  Even the caloric method fails for input power measurements. 
To do it correctly you have to multiply the instantaneous voltage and current at high frequency and then average the resulting products.  The Kill-a-Watt meter does not do that well.
Remember: The product of averages is not equal to the average of products (note the plural form of "products").  Really mull this over in your mind until you get it.  I can write it as a mathematical equation if you want - just say a word.

However, since your circuit does not really need AC, you can approach an accurate average input power measurement by rectifying the mains AC with a FWBR, feeding it into an RCRC ladder with huge final caps and measure the resulting DC amps and DC volts with regular meters AFTER the final caps - only then the product of the so measured current and voltage will approach true average input power without the need for high frequency i*V multiplication.

P.S.
Chet's resistive caloric method is accurate for measuring the average output power.
A PV cell in a dark box with an incandescent bulb, too.

[heatlocke]

OK, so I agree with comments about the scope shot, circuit and the calometric method I used. I`d like to push back on the 450 watts issue though. At 4:40 in my vid I show my line filter. I made it from (4) T106-52 cores and (2) 1 uf  X1/Y2 caps. This concerned me, so I put my scope at the same outlet as the circuit, less the kilometer and find no remnant of my waveform at the line. I`m satisfied the filter is working. I think the 450 watts is pretty close to accurate. All the other known numbers match up. I think the absence of heat anywhere speaks for something. The 5 sec. start up delay and the 1.25 min. shut down delay show the circuit consumes 5 watts with no bus power. No DC charges for rectification either. What am I missing? 

[Verpies]

I show my line filter. I made it from (4) T106-52 cores and (2) 1 uf  X1/Y2 caps.

That is just a high frequency EMI filter.

...find no remnant of my waveform at the line.

It doesn't matter what voltage you see on only one channel of the scope.  Voltage is not the average input power nor input energy.

Show us the traces from two synchronized channels of your scope - one showing the input voltage and the second one showing the input current - measured at one point of the circuit and simultaneously and with the channels zeroed-out perfectly. Provide the calibrated vertical V/div factors too and endeavor for the two waveforms to be maximized on the screen vertically ...and horizontally containing 1 < cycles < 2.  Draw a crude diagram showing the probe positions.  Do not blow up your ground leads nor scope.

A cycle is as long as it takes for both waveforms to completely repeat.

[heatlocke]

" .... the second one showing the input current - measured at one point of the circuit and simultaneously and with the channels zeroed-out perfectly."  ...... OK, Just to be clear so you know who you are dealing with  .... I haven`t a clue how to do this and thats why I`m here. Really, I`m a rookie on a scope. I have an old analog scope with its own personal iso trans and never without it. I treasure my old scope and had a bad experience with 2 probes years ago and have never tried since. What you seen was my full, complete knowledge base. I have several 600v probes but only one P4100 HV probe. Could I do this with one probe at separate times? Will my old scope measure amps ? Really cause I don`t know. Not playing dumb here but for real.

[Verpies]

The most convenient and safest way to measure instantaneous current is with a clamp-on current probe.  Itsu has many videos with it, so I hereby ask him to post a link for you.

If you don't have a clamp-on current probe for your scope then all is not lost.  You can use a non-inductive series current sensing resistor (CSR) to substitute for the clamp-on current probe. 
There are two ways to connect it: 1) before the voltage measuring point and 2) after the voltage measuring point.  Only one of them does not include the CSR loss in the measurement. 

I posted a schematic for doing exactly this on this forum somewhere and Itsu made many videos about measuring voltage and current simultaneously with the CSR.  If he does not come through, find my schematic for simultaneously measuring i & V with the CSR here.

[heatlocke]

I just so happen to have a 15mOhm current sense resistor (15FR015E) on the board already inline at LOW source and Gnd. and not using it. Can this work for anything ?

[Itsu]

Well, i have several video's about using a current probe or using a csr (as backup), but it would take me some time to find a nice example.

But as i understand that there is no possibility to use a current probe here, i won't dig into it.

Also when using a csr on the AC (like live grid?) it would need some extra care and equipment like isolation transformer and a 2 channel scope with math function to make accurate input measurements.
Best thing is IMO, as verpies already suggested, to make some DC first and measure voltage and current there.

The 15 mOhm csr mentioned would be usable, but i do not see it on the shown PCB and as there is no schematic it is hard to say if it can be used.

Itsu

[heatlocke]

OK, Its right on top of the board and easy to reach. Its direct in line with LOW fet source and the Neg pin on the bridge

[Itsu]

Ok, that might work, but we need the complete schematic to be sure how your half bridge is build up.
You mentioned a LTspice schematic in the video of an earlier design, so do you have one of this present one?

But even if that csr is OK where it is, you still need a second probe to simultaneous measure the input voltage to be able to calculate the input power.

Also, what is the load, i see some pancake coils and later on some heavy duty? leds.

Itsu

[heatlocke]

This is the last one I was keeping up with and I sure hope its correct. The correct fet part # is IMW120R030M1HXKSA1. Some TVS diodes are in there but missing on the pic. Better FWBR also. Thanks

[Itsu]

Looking at your Half Bridge setup, i do not think the csr is in the correct position for this schematic.
IMO it will only measure the current going through the lower MOSFET, thus at 50% duty cycle will show about half of the real current.

Anyone correct me when i am wrong?

Itsu

[heatlocke]

OK, this is my attempt to follow up on Verpies post #7 suggestion "Show us the traces from two synchronized channels of your scope". Channel 1 probe  (X 1) is direct connect to a little torrid with 100 turns of AWG 26 magnet wire and a 22 ohm burden resistor and a single wire thru it on one wire leading to the primary and at  .2 volt setting ..... Channel 2 is  as before with P4100 probe at primary connections. Voltage setting at .5      Did I get these positions right ?



 
https://www.youtube.com/watch?v=y2bT5mv8X34

[Verpies]

CTs don't output voltage proportional to instantaneous current. Depending on the ballast resistor (which better be non-inductive), they output the time derivative of current and have serious phase and frequency response deficiencies (not only near DC). This depends on the CT's core and inter-turn capacitances of its windings.

For this reason I have not recommended using a CT for wideband phase-coherent current measurements.  CSR is so much more accurate...