How to sense current in this circuit

[snarfer]

I'm working on a high voltage buck driver circuit for a long string of LEDs attached to a 400v PFC. Problem is that I need a fairly simple way of detecting the current through the LED string. Normally I would be able to just use a low side sense resistor, but according to my simulations that gives wildly inaccurate results, probably due to the capacitance in my circuit that is not referenced to ground.

Also at 400 volts it is not possible to use a conventional high side current sensing amplifier, as all the ones I've found are designed for voltages below 70 volts. Maybe there is one that goes to 100v, but that wouldn't help much anyway.

Right now the simplest alternative seems to be the IR2175 high voltage current sensing IC. But I wonder if anyone has a better idea.

LTspice schematic: http://www.bangkokgaffer.com/schematics/Driving_Circuit.JPG

Thanks for any input.

 

[SemiMan]

RC filter off your low side current sense should fix everything up.

Semiman

 

[snarfer]

Thanks I realized the problem with low side sense is just a matter of a delay while the capacitors charge. The RC circuit and a faster op amp cleans it up very well.


I do wonder though why I can't seem to find anyone supplying a driver like this commercially. I mean I can't be the only one trying to build a 400 watt LED light.

 

[SemiMan]

As a product I would imagine most people will shy away from a non-isolated solution. The second issue is making it tunable over a range of output voltages (not currents) makes for a range of trade-offs to go with it. A commercial circuit would have EMI filtering on the front end. Guaranteed you will violate conducted emissions.

There are lots of chips that will do this though. Any number of the single chip AC\DC circuits (ON, ST, Fairchild, Powertrends, IR, Infineon, etc) will do this and with less cost.

Semiman

 

[frenzee]

Please keep us updated on your progress snarfer. I'll be working on a vaguely similar driver (~40V).

 

[snarfer]

I'm not so sure that isolation is necessary. I haven't seen any LED drivers that are designed for that type of circuit. Even look at the datasheets for IR's IRS2540, or Supertex's HV9910B, which are both designed to run off mains voltages. Non-isolated buck drivers.

Maybe the reason is that generally LEDs aren't user serviceable items so there isn't the same shock hazard? I don't think that lighting ballasts generally have to be isolated anyway. Household dimmers aren't isolated. Some fluorescent ballasts aren't isolated.

The point regarding EMI is important though. Of course it's just an LTSpice simulation so I left out the PFC. It's easy enough to implement an entire PFC design from any of the big IC manufacturers. They post the whole thing on their sites, with gerber files for the PCB, bill of materials, etc... I just ordered a PFC evaluation module from TI actually. Supposedly needs no external cooling, 93% efficient, and so forth.

I spent a lot of time looking for an integrated solution for the buck controller and switches. Not as easy as it seems. Mostly the manufacturers make stuff for implementing flyback controllers under 150 watts. I want a design that I can integrate into the housing of the lamp, highly efficient and can interface to microcontroller for DMX control. I gave up on PWM dimming, just stick with analog for now. More efficient anyway. I ended up concluding that MCU controlled buck would work well because I could use a larger capacity FET driver than would be integrated in a controller, and because I could use a better FET as well. The integrated controllers are surprisingly expensive, so the costs would not be so different, especially given that I need to put in an MCU for the DMX. The dsPIC SMPS series can do all sorts of stuff, even frequency dithering for EMI reduction, etc...

That said, I have also been looking into a rather unique monolithic converter with zero voltage switching resonant converter topology. Problem is, they won't release SPICE model due to security restrictions, so there are some control details I'm not so sure about. Seems kind of fancy, but it's the lowest part count solution, if I can get it to work. Also, it seems like it would have far lower EMI as well.

If someone with experience getting things UL listed and so forth would be interested in collaborating with me on a potential commercial project I would be open to that so feel free to PM me and I can supply some more details.

 

[snarfer]

After some more experimentation with simulations I have to admit that SemiMan you are as usual entirely correct. I found some different ways to simulate isolated SMPS in the ltSpice user group on yahoo (a great resource!) and they are not only safer, but work better, with far fewer current spikes, and so forth.

Also, as you stated, it is entirely possible to buy ICs that will pretty much do it all, and cheaply too.

The only problem is of course that custom transformers are necessary, which is kind of a pain.

Evidently the only company that plans to come out with a high voltage driver like what I need is MagTech, and their product won't be ready until summer.

So I guess it's time to start learning about windings...