Contact-Less LED lighting system for wet environments

[ocset]

Hello,
We have been tasked with lighting up  another huge system of  outdoor architectural water fountains with coloured LED lamps. 

In previous jobs, we find it’s a massive problem with moisture getting into all the connectors and contacts.     

Therefore, for this build, we are going to use our new contact-less LED lighting system. 
For this, there is a single power supply, which puts a high frequency (100khz) sinusoidal current into a 30 metre cable (twisted pair).  Its only the power supply we have to keep waterproof. All the downstream LED lamps have no metal contacts, and so they are waterproof !!!  Each  LED lamp  comprises  a ferrite coupler which “clips” over the 30 metre cable wires, and the secondary is embedded in this ferrite  as a PCB “printed” secondary. –No contacts! – each lamp can very simply be clipped or unclipped, from anywhere on the cable that we like. It’s a perfect system. Why is no one else doing this? 

The current in the 30 metre cable is a beautiful sinusoid due to the resonant topology chosen. The voltage waveform is also smooth. 

Can you see any problems with our new setup?  What about RF interference?    The 30 metre cable wire will be a twisted pair of multi-strand  conductors.

Schematic and LTspice simulation as attached.
 

[BrianHG]

 

[ocset]

Thanks, its a super efficient, soft switching resonant power supply aswell.
Where's the gotcha?
I dont think there is one.
The couplers are interesting. It has to be high permeability ferrite as the primary is just one turn.

[CopperCone]

can you get a contract for nuclear reactors to build in iran?

[ocset]

Thanks, well, this lighting would be good for in a nuclear reactor, because there's no sensitive LED driver  ICs in the bulb......just diodes and LEDs...so the lamps wouldnt be damaged by the neutrons in the reactor.  (the power supply would be in a lead enclosure for nuclear use) But can you agree with me that in the switching stage, the dead time setting is crucial..

[jbb]

Hi treez

This could be a fun project.  A bit different to the usual input + buck + LEDs.

I would like to draw your attention to the Inductive Power Transfer research of the University of Auckland, carried out by John Boys and Grant Covic (and a lot of masters / PhD students).  They have done some roadway lighting systems (very similar topology) using a 38.4kHz (I think) system.

Some safety comments:

• Protection for open and short circuits of the transmission loop is critical.
• I suggest you replace L13 with an isolating transformer (you want a little leakage inductance).  This way you can keep the 50 / 60 Hz mains out of the transmission loop.  Also some safety standards allow higher safety limits for higher frequencies.
• Remember to check what happens if someone leaves some steel object (sheet metal, spanner, pipe etc.) next to the transmission loop. This includes electrical reliability (i.e. wild shifts in impedance shouldn't make the primary MOSFETs explode) and thermal hazards; we had some surprises by accidentally induction heating steel desk legs, sheet metal and once a whiteboard 
• To protect against broken LED strings, you can add SCR clamps to the LED pods (i.e. broken wire -> over voltage -> short out) which vastly reduces the risk of fire.

Some more general comments:

• EMC compliance is possible.  I'm not sure if you'll be classed as an intentional or unintentional radiator.
• Adding an earthed center-tap to the transformer can help with EMC.
• Litz wire is your friend.
• The resonant capacitors are really important.  Select carefully and check for things like derating with temperature and applied frequency.
• Test gear like voltage and current probes tend to derate as frequency goes up.  Check their data sheets.
• Simple on/off control using M1 can be very effective.  However, it will impact the operation of the whole system.  This is because the power draw of D35-D44 is reflected back into the main loop via L6/L7 coupling as virtual resistance.  Changing the resistance of the loop will change the loop current unless you have closed loop feedback
• The suggested primary side control scheme (per University of Auckland research) is to run the primary FETs at fixed frequency and vary their on times (also known as having more and less dead time between M13, M14) to regulate the 'track' current.
• Superjunction MOSFETs (e.g. CoolMOS) devices look very tempting with their fast switching and low on resistance.  However, their body diodes have terrible reverse recovery properties.  CoolMOS half bridges can be good for soft switching applications (where reverse recovery does not occur) but have a tendency to blow up (sometimes literally) if you ever move into a hard switching region (as might happen when something goes open circuit).  If you value robustness (and time to market) over maximum efficiency, I would suggest you look at using ultrafast IGBTs (with co-packaged diode) instead.

Advanced operation: Selective Harmonic Elimination
If you run the primary switches near 50% duty cycle, you will get maximum power output.  But you will also get lots of 3rd harmonic content, which can be difficult to filter out.  3rd harmonic content may be subject to tighter emission limits and can make your life more difficult. 

It may complicate the soft switching considerations, but if you set the primary switch duty cycles to 33.3% (vs approx. 50% for maximum power), you have a conduction angle of 120 degrees (electrical degrees).  This reduces the amplitude of the 3rd harmonic to near zero (performing the Fourier Transform is left as an exercise for the reader) which can help with EMC.  This technique is called Selective Harmonic Elimination and can be very helpful.

You would adjust the duty cycle up and down a bit to regulate the primary loop current.

Advanced operation: Working with an Air Gap
I expect your current design concept requires quite high magnetic coupling factors.  This implies that you need closed ferrite cores (you mention a clip on ferrite part) with a very small air gap.  This is bad news for waterproofing and assembly.

Ideally, you want the only ferrite inside the LED pod.  This suggests a big air gap between the pickup and the loop, and therefore a low coupling factor.  Which will wreck your power transfer capability.

But all is not lost!  The approach developed by the University of Auckland involves adding a resonant capacitor (either series or parallel, there are properties and tradeoffs) to the output winding.

If you add capacitors to the pickup inductors you can form resonant circuits which vastly improve the available power.  This enables you to operate with very low magnetic coupling factors (e.g. k = 0.05 to 0.3) and could mean you don't need the clip-on ferrite structures.  Maybe you could just glue a ferrite plate to the top of the secondary PCB winding and then overmold the whole LED pod for waterproofing.  But it requires quite different circuit design principles.  For a start, series resonant and parallel resonant have very different behaviours.

If you want to go down this path, I strongly suggest you lobby your management team to get a little budget to buy some IEEE papers.  Don't laugh - there are some excellent ones which include a lot of the design information you will need.  I expect you could get a good set for < US$500 which is orders of magnitude less than the time taken to learn it the hard way.

[ocset]

I expect your current design concept requires quite high magnetic coupling factors.  This implies that you need closed ferrite cores (you mention a clip on ferrite part) with a very small air gap.  This is bad news for waterproofing and assembly. 
Ideally, you want the only ferrite inside the LED pod.  This suggests a big air gap between the pickup and the loop, and therefore a low coupling factor.  Which will wreck your power transfer capability.

Thanks. We were thinking of doing the ferrite couplers as in the attached, just using standard E cores. These will be embedded into the LED lamp….there will be a type of hinge mechanism so that the E cores can be separated so that the bus wire can be inserted  as shown. The Printed winding PCB comprising the secondary will be like in the diagram attached. (in fact it will be different as the secondary cannot be exposed to  atmosphere or water).

  The approach developed by the University of Auckland involves adding a resonant capacitor (either series or parallel, there are properties and tradeoffs) to the output winding.

Thanks, we thought about this. Certainly in the simulation  it behaves ok with a coupling factor of 0.99. We  realise  that wireless couplers need the resonant capacitor, and that our non wireless couplers can benefit from it.

[jbb]

I see what you're getting at there.  You should get quite high coupling factors, which should reduce the impact of the precise geometry.  With high coupling factors there's probably no need for capacitor compensation.


So I couldn't resist a brief digression into series capacitor compensation.  Maybe someone will find it useful.

Option 1 is a basic inductor pickup + load resistor (L2, R1), we see that the output power is crap.  This is due to the low coupling factor; as the load gets heavier, the reactance of the inductive pickup (Zl = j*\omeha*L) causes the output voltage (Vo1) drops away to nothing.  You can the voltage vs load in the top plot.

Maximum available power is approx. 300 mW.

Option 2 adds a series capacitor we can try to compensate for the pickup's reactance (Zc = 1 / (j*\omega*C).  Basically we choose C such that Zc = -Zl.  This boils down to choosing C such that LC resonates at the primary frequency.  Assuming your components are spot on, the output voltage (Vo2) is constant, irrespective of load. In theory, you can suck down as much current as you like: invite power is available!  However, there are component tolerances to consider.  Also the capacitor voltage stress goes up and up... (bottom plot)

Maximum power is infinity.

So, stepping back a bit, we see that Option 2 provides a constant voltage output irrespective of load.  We've turned an inductive pickup into a constant AC voltage source!  Be careful not to short circuit it - it will burn.

Also be aware of the peak capacitor voltages - they can get really high (see lower plot).

Option 3 looks at some component tolerances: the pickup inductor is 5% down and the capacitor is 10% down.  So we have less voltage from the inductor and our LC tuning is off.  We see that the output voltage is fairly constant above 10 Ohms load. Maximum power is approx. 1.8W

So, as a practical matter we got approx. 6x more output power by adding a capacitor. Not too shabby.  As a general principle, designs relying on 2x - 5x are fairly easy to manage, 5x - 7x are possible with care, and 7x + are only available with great care and some kind of adjustable tuning (oh yes it's possible).

General Comments
Adding a series compensation capacitor can turn an inductive output into a voltage source (within limits) and multiply the maximum power transfer.  This indicates that we can now accept lower coupling factors.  Which means the pickups can be further away from the primary loop and therefore leave some gap that can be filled with waterproofing material.

Please note that the output voltages of the series compensated pickups are only constant with respect to load current.  If the primary current changes, or the pickup coupling changes (i.e. distance to track changes) the output voltage will change too.

Secondary side regulation is probably required to deal with varying placement.

Also don't short one of these.


Parallel Compensation
Instead of putting the capacitor in series to form a voltage source, we can put the capacitor in parallel with the coil to form a current source.  Don't open circuit one of these; it will burn. And possibly shock you (voltage tends towards infinity).

[nctnico]

This won't work. First of all the loop area in the twisted pair is too small, the loads will interfere with eachother and third the capacitance of the cable will be a large load to the 100kHz signal. Think about impedance matching.

If it has to be wireless then you'd need to place wireless transmitters at the dry side of the fountain at pre-determined places and the lamps can be underwater.

[Marco]

Each  LED lamp  comprises  a ferrite coupler which “clips” over the 30 metre cable wires

What do you mean wireS? If it clips on both the feed and return wire there is no magnetic field.

Main problem AFAICS will be the tiny primary inductance, you might burn up the wire long before you get useful power to the secondary. You can make it resonant, but that doesn't solve the copper losses from all the current sloshing around.

[jbb]

This won't work. First of all the loop area in the twisted pair is too small, the loads will interfere with eachother and third the capacitance of the cable will be a large load to the 100kHz signal. Think about impedance matching.

If it has to be wireless then you'd need to place wireless transmitters at the dry side of the fountain at pre-determined places and the lamps can be underwater.

Loop area: Treez has added a bulk primary inductor. On the cable you need to spread the twisted pair apart at the sites you want an LED pod.
100kHz: quite possibly too high. 20 - 40 kHz may be more appropriate.
Cable loss: yes, high primary currents do lead to losses. Fortunately the losses are spread out. Use of Litz wire may be required to get AC resistance down.

[ocset]

What do you mean wireS? If it clips on both the feed and return wire there is no magnetic field.

Thanks, if you view the  inductive coupler diagram that I posted in post #6 you will see that  the fields from each wire in the coupler don’t cancel in the core.

This won't work. First of all the loop area in the twisted pair is too small

Main problem AFAICS will be the tiny primary inductance, you might burn up the wire long before you get useful power to the secondary. You can make it resonant, but that doesn't solve the copper losses from all the current sloshing around.

Thanks, ..Now I am going to be a sniffy little so-and-so and confess that I have already done one of these inductive couplers whilst in Germany.  At that time, I didn’t have anything to do with the power supply that generated the sinusoid (it was 50khz for their system). However, I did a printed PCB  secondary, and I had the setup shown in my post #6 above, and the coupling  was very good indeed. It did surprise me. I was pretty much getting the full Ip/Is = Ns/Np.
Incidentally, the inductive coupler that I did was for charging a battery.

and third the capacitance of the cable will be a large load to the 100kHz signal.

Thanks, the cable is loosely twisted. It has to be loose  because one has to pull the conductors apart so as to insert them into the inductive couplers as I show in post #6 above.
If you run the simulation in post #1, you can see that there is no place where you can add "stray" capactiance that spoils the performance.

You can make it resonant

Thanks, in wireless power transfer, it certainly  has to be made resonant..….though when you have tightly coupled inductive couplers like we have, there is no need for this resonant method I believe. In effect, all we have done is very simple indeed….we have just taken a series parallel resonant converter, and “ripped its guts open” and spread them out along 30 metres, and then simply connect secondaries along it. Its so incredibly simple that you wonder why its not a commonly spoken about technique, with off-the-shelf couplers available etc etc.

One situation I see is that when an inductive coupler is suddenly clipped on to a live bus, then the bus suddenly “sees” the  inductance of the  inductive coupler, and this causes an overvoltage across the  primary. There is not much that can be done about this because the bus isn’t accessible to put capacitors/snubbers across it etc.
LTspice sim of  this overvoltage as attached.

[Marco]

AFAICS with say 10 uH primary inductance you'll be looking at a matched load in the 100 mOhm range (referred to primary). Say you want 10 Watt, that's 10 Amperes RMS you got churning around in that cable.

Okay, you won't burn it up but a reasonable cable will get warm to the touch.

[ocset]

Thanks, as you know, these guys are alreay doing it
http://www.heicolighting.com/

The simulation shows how it works, i re-attach it here, we must be wary of the simulations, but in this case, i believe the simulation has called it right. There isnt such big current in the bus.

Essentially the system i show here, as you know, is absolutely nothing knew...its simply a modified Seris Parallel resonant converter with its 'guts dragged out' so that the multiple ferrite couplers can be clipped on. It really is no more than that. -Just a standard topology.

[Marco]

They seem to use a closed core though (they are threaded on). A toroid can have much higher effective permeability than an E-core for the same size, there's always a bit of an airgap, which makes life easier.

[SiliconWizard]

This is a nice approach.

You should take cable capacitance and linear resistance into account.
For long cables they will be significant and will reduce efficiency.

As for EMC, you'd really have to have your design tested in a lab in nominal conditions (which could require a big chamber to accomodate for the cable length), and it would depend on the local regulations somewhat. That said, if you get a clean sine wave @100 kHz, radiated emissions shouldn't be a big problem. Near-field emissions may be though.

But there is another thing to consider: once you're going to load the network, the resonant frequency may shift and it could get much less efficient then.
Have you measured overall efficiency in different loading conditions?

Also, in your simulation, you're assuming a coupling coefficient of 0.99, but wouldn't the actual transformers be considered as air-core transformers? (I have no exact idea of your setup around the cable.)
If so, 0.99 seems very optimistic.

[Delta]

If you are using this in the UK, it will have to be able to cope with mains voltages of 292v RMS for periods of several minutes.

[Marco]

I was wrong, for a 10 uH primary the matched load is 6.3 Ohms (referred to primary, 2*pi*10e5*10e-5)). That's reasonably high, so cable losses shouldn't be too bad., assuming the load is matched. The matching is rather important, for a given cable current the power transfer will quickly diminish as the load impedance changes. Load matching minimizes copper losses.

[ocset]

Also, in your simulation, you're assuming a coupling coefficient of 0.99, but wouldn't the actual transformers be considered as air-core transformers?

Thanks, The couplers are as in post #6 above. They are a lot more tightly coupled than air core transformers.

As i mentioned in post #11 above, i have already  worked on a system like this and got   Is = (NP/NS) * IP
-That was without putting in any  impedance matching networks.

In wireless coupling, i would say as you know too, that matching networks are essential.
Though we are not doing wireless coupling here.

The matching is rather important, for a given cable current the power transfer will quickly diminish as the load impedance changes. Load matching minimizes copper losses.

In the attached LTspice  simulation, there is no matching networks.
I dont see any excessive current in the cable.
Do you declare that the simulation is inaccurate?, or that there are parasitics or something that we have not accounted for?

[Marco]

You don't need a matching network, you just need to be in the ballpark of the right impedance ... which you already happen to be. I wonder to what extent the rather blocky voltage waveforms across the primaries matter for emissions.

If you use the 14 AWG wire Heico uses you'd be burning 5-10W in the wire.

[ANTALIFE]

Have you though about using Qi wireless power transfer? There are quite a few IC's to choose from to make the design stage easy, most of them can do fancy things like foreign object detection (WPC 1.2 & 1.1) and have relatively high efficiencies for a wireless system (~75%)

Also here is a good review paper that looks at a bunch of different wireless charging techniques:
https://arxiv.org/pdf/1509.00940.pdf

[Marco]

That won't let you attach armatures at random points on the wire, which is the entire attraction of this scheme.

It is quite an attractive technique and I'm surprised how even with a relatively low cost E-core you can still pull decent power from a wire, wouldn't have expected it.

[ANTALIFE]

That won't let you attach armatures at random points on the wire, which is the entire attraction of this scheme.

It is quite an attractive technique and I'm surprised how even with a relatively low cost E-core you can still pull decent power from a wire, wouldn't have expected it.

Ah that's true