Building a battery powered coil

[euthon]

Hello everybody!

I'm trying to build a battery powered DRSSTC. The problem i'm trying to solve is common for all such devices, that is pumping enough power into the primary coil from a low voltage source (24 volt battery).

There are two obvious solutions:
1. Build a step up voltage converter to boost the bus voltage to some decent level.
2. Use transformer coupling between the output(s) of the bridge(s) and the primary coil.

However, both solutions are far from being perfect. A reasonably efficient step up converter that can can handle such power (~2kW) will be complicated, expensive, and possibly bulky. Transformer coupling is simple, but the cores and copper add bulk and weight (E55 cores are required, about 100 gram each core half).

My idea is splitting the primary coil in parts and connect each turn to its own bridge. Of course, the capacitor bank will have to be split accordingly as well. So, if there are eight turns in my primary, I can split it in 8 single turns primaries, connect each to its own invertor and get 8 times the power output.

What do you think? Will such approach work, or it will fail due to unequal resonant frequencies, couplings, etc?

[davekni]

Could you get a bit higher voltage?  48V batteries are reasonably common too.  With 48Vdc, I'd make a coil with single-turn copper sheet primary.  Run H-bridge directly from 48V.  At low voltage and high current, minimizing stray inductance will be important.  See this thread for ideas on low-inductance construction:
https://highvoltageforum.net/index.php?topic=1324.msg9795#msg9795

At 48V, FETs are likely better than IGBTs for the bridge, multiple parallel FETs per switch.

[euthon]

Hi, Dave!

Thank you for the answer.
I'd prefer to stay with 24 volts if possible. The main reason for that is the number of required cells. 24 volts is achievable with 10 LiFe cells, getting 48 volts obviously will require twice as much. Of course, 20 cells double the power output, but than there is no such thing as too much power, so one must stop somewhere. And than I need a battery charger to charge 20 cells...

Of course I'm going to use MOSFETs for the low voltage bridges. I've got few of these https://www.ti.com/lit/ds/symlink/csd18540q5b.pdf. In fact I have two bridges soldered and tested, but do not have yet suitable ferrite transformer cores to try running the inverters in parallel.

So, while the cores are on their way in the mail, I'm asking if someone has tried splitting the primary and what is the best way to do it.

[Uspring]

I believe Steve Ward used 4 separate primary windings in his Titan coil. I am not sure, my recollection of this has faded and I could not find anything documenting this in the net. AFAIK he used some sort of a current balancing network to achieve equal load on the bridges.
Currents differ in the primaries if the inductances aren't exactly equal and that becomes worse with increased coupling between them. Since the primaries will have differing resonance frequencies, you might also encur hard switching if you switch all primaries at the same time.

[euthon]

Hi, Uspring!

The only mention of the Steve Ward's Titan coil i found is this:

i've always used individual half-bridge sections that get tied in parallel with "long" output leads to act in a similar manner to "balancing chokes" found on variac stacks, for example. There's also options for splitting up the MMC so that each output can drive its own section of the total capacitance, enforcing current sharing.

It is not quite clear if he actually tried splitting up both the primary winding and the MMC.

I just tried winding two primaries on a 6 inch OD plastic pipe, one above the other. The inductances match quite well until the secondary coil comes into play. Due to different coupling coefficients, the resonant frequencies drift a bit (even when using the same MMC). That might not be a big problem by itself, but you are right, i'd loose soft switching. This is something i didn't think of. Thank you for pointing out.

So, I'll scrap the idea for now and will attempt paralleling bridges with transformers.

[Mads Barnkob]

TMaxElectronics started a similar project, just about a month ago: https://highvoltageforum.net/index.php?topic=1787.0

[euthon]

Hello, Mads!

TMaxElectronics started a similar project, just about a month ago: https://highvoltageforum.net/index.php?topic=1787.0

He is taking a more traditional approach - step up the voltage. Starting with much higher voltage (around 120V) also helps. As i mentioned, designing and building an efficient 2kW step up converter that would boost the voltage from 24 to 300-400V is quite a challenge (a two or four phase boost converter may be good enough for 120 -> 400V). I can easily build a brute force style push-pull converter with efficiency of around 85% and water cool it, but i strive for something more elegant. The only converter topology i can think of that may provide reasonable performance is the current-fed bridge, but i can't get it running in simulator yet, let alone built such a power supply. I may try an LLC in future, but they do not work particularly well for high step up ratios and are difficult to parallel.

[davekni]

Interesting project!  Sounds like a fun challenge.  Great to try unconventional solutions.  Will be extensive learning whether successful or not.

I'd prefer to stay with 24 volts if possible. The main reason for that is the number of required cells. 24 volts is achievable with 10 LiFe cells, getting 48 volts obviously will require twice as much. Of course, 20 cells double the power output, but than there is no such thing as too much power, so one must stop somewhere. And than I need a battery charger to charge 20 cells...

LiFePO4 cells are nominally 3.2V each, perhaps down to 2.8V under high load.  Are you planning to load them enough to drop to 2.4V/cell?  Also, high-discharge-rate LiPo cells will provide more energy density, at the expense of lower cycle life.  My neighbor uses two 6-cell LiPo packs in series, 44.4V nominal, for his larger helicopters.  Charges each half separately with a 6-cell charger.  Batteries similar to:
https://www.amazon.com/5200mAh-T-REX550-Airplane-Quadcopter-Helicopter/dp/B07JQ2Y78R
Some data from the manufacturer, for both LiPo and LiFePO4:
https://www.grepow.com/page/high-discharge-battery.html

The only converter topology i can think of that may provide reasonable performance is the current-fed bridge, but i can't get it running in simulator yet, let alone built such a power supply.

You could skip voltage regulation.  An H-bridge feeding a transformer followed by bridge rectifier is simple and likely sufficient.

My idea is splitting the primary coil in parts and connect each turn to its own bridge. Of course, the capacitor bank will have to be split accordingly as well. So, if there are eight turns in my primary, I can split it in 8 single turns primaries, connect each to its own invertor and get 8 times the power output.

What do you think? Will such approach work, or it will fail due to unequal resonant frequencies, couplings, etc?

My low-frequency (100kHz) QCW project uses four parallel primary windings, 7 turns each.  They are interleaved, wound as 4 parallel wires, each starting and ending at their own quadrant around the coil.  Each has its own MMC:
https://highvoltageforum.net/index.php?topic=1268.msg9305#msg9305
That could work with single-turn coils.  That's where my previous 48V suggestion comes from.  48V is likely sufficient for single-turn primary.  It is a shame to waste ferrite transformers for just a 2:1 voltage boost.  If you do go for ferrite transformers, I'd wire their secondaries in series, feeding one normal MMC and one normal primary coil.  With secondaries in series, each H-bridge is forced to have equal current.

Is "~2kW" your design average power?  Do you have a goal for duty cycle or peak power or ...?  Say 10% duty cycle of 20kW pulses?  Do you have sufficient bulk capacitance to spread out 20kW peaks into 2kW average battery drain?  I'm thinking ~0.1F to 0.2F capacitance to provide the 20kW bursts.  Battery performance will suffer if you expect it to provide 20kW bursts.

[euthon]

LiFePO4 cells are nominally 3.2V each, perhaps down to 2.8V under high load.

I got a couple of A123 26650 cells for testing, and they are around 2.7V under ~50A load when fresh and drop below 2.5V when half discharged. These are rated at 70 amps in official specs. I totally agree with TMaxElectronics (https://highvoltageforum.net/index.php?topic=1787.msg13577#msg13577) about the LiPo cells that are advertised as 50c current ratings. I'v got a number of them for copter fun, and very few batteries survived for a year. Even when not used and stored ~60% charged in a fridge.

You could skip voltage regulation.  An H-bridge feeding a transformer followed by bridge rectifier is simple and likely sufficient.

You are probably right, an "electronic transformer" will likely suffice, but when you consider that a short circuit protection is a nice feature to have, also it needs to charge bulk bus capacitors (which is kinda short circuit), and the full bridge requires a DC blocking capacitor anyway to avoid core saturation, you may as well design an LLC. As bad as the LLC is with its circulating current, it will be more efficient than a PWM bridge at capacitor charging. The current fed converter is inherently overcurrent-proof, does not suffer from the circulating current, needs fewer turns in the secondary, and does not require a DC blocking capacitor (needs a choke inductor instead), but has complicated control.

That could work with single-turn coils.

Well, as Uspring mentioned above, this is not going to work well due to potential loss of ZVS. If a bridge is switching below the resonance frequency, ZVS is lost and it gets pretty ugly due to reverse recovery of the intrinsic diodes. It is not so bad with IGBTs, since they have built in ultrafast diodes, but MOSFETs are very likely to fail quickly in this mode.

If you do go for ferrite transformers, I'd wire their secondaries in series, feeding one normal MMC and one normal primary coil.

Of course.

Is "~2kW" your design average power?

Yes. Mostly limited by the amount of cash i'm willing to spend on batteries. Commercially available battery packs (i.e. from cordless drills and the like) are grossly overpriced around here, mail order them from abroad seems like too much hassle (in particular due to Li-Ion shipping rules), so i'm only considering buying single cells locally.

Unfortunately, i have to leave for a business trip, and won't return to the project for the next couple of months.

[davekni]

and the full bridge requires a DC blocking capacitor anyway to avoid core saturation

For one-time hobby projects, accurate 50% duty cycle can be sufficient without a blocking cap.  Cap adds sometimes-annoying resonance.  With a bit of switching dead-time, DC core current alters rising vs falling edge delay to compensate for small duty-cycle errors.

Well, as Uspring mentioned above, this is not going to work well due to potential loss of ZVS. If a bridge is switching below the resonance frequency, ZVS is lost and it gets pretty ugly due to reverse recovery of the intrinsic diodes. It is not so bad with IGBTs, since they have built in ultrafast diodes, but MOSFETs are very likely to fail quickly in this mode.

Your transformer plans are likely better, but I still think 8 single-turn primaries could work.  The only imbalance is imperfections in construction.  With each primary being exactly 1 turn starting at a common height and ending at a common height and of the same diameter, spaced evenly in octants around the secondary, primary-to-secondary coupling factors are all identical.  H-bridges could be arranged in a radial array too, like boards in an old Cray supercomputer.  Agree that phase-lead is necessary for reasonable performance, though the FETs you picked out aren't that terrible for body diode speed and recovery charge.

Of course.

Sorry, I didn't read previous posts accurately, so mistakenly thought transformers were being paralleled.

Hope your trip goes well.

[euthon]

With each primary being exactly 1 turn starting at a common height and ending at a common height and of the same diameter, spaced evenly in octants around the secondary, primary-to-secondary coupling factors are all identical.

I was initially thinking of spreading multiple primaries vertically along the secondary. Like you'd normally wind the primary, than cut it in pieces and connect each piece to its own inverter. This should be equal to increasing the output voltage of the bridge. Your suggestion is similar to winding the primary with litz, and connecting each wire strand to a separate bridge. This is basically increasing the current in the primary. In theory, this should make no difference, since P = U*I anyway, but i'm not sure how the coupling will behave if you reduce the primary to a single turn. Seems like increasing the coupling is the trend lately.

[davekni]

Your suggestion is similar to winding the primary with litz, and connecting each wire strand to a separate bridge. This is basically increasing the current in the primary. In theory, this should make no difference, since P = U*I anyway, but i'm not sure how the coupling will behave if you reduce the primary to a single turn. Seems like increasing the coupling is the trend lately.

Yes, I think higher coupling is better, until coupling is high enough to cause racing sparks.  That's the only limit to high coupling that I've experienced.
My thought has one difference relative to litz wire.  I'm picturing 8 single-turn primary windings covering the same area as a conventional 8-turn primary.  If vertical (helical), each turn covers the full conventional-primary height (steep helix angle).  If planar, each turn covers the full ID to OD range of a conventional 8-turn primary (steep spiral angle, if that is a term).  And primaries are spaced at 45 degrees (octants) circumferentially.  I think that leaves the total primary-to-secondary coupling about the same as for a conventional primary.  It also minimizes primary-to-primary coupling relative to litz wire.  (Primary-primary coupling would still be fairly high, requiring bridges to have sufficient dead-time to cover any drive timing mismatches.)
Your ferrite transformer thought is likely easier.  Of course, transformer leak age inductance and wiring inductance adds to each other, 8x total given series connection, reducing effective coupling by adding non-coupled primary inductance.

Looking forward to seeing how this turns out eventually.