LTC3780 Driver

[Peufeu]

OK, so I want to put a light on my bike, it will have 3 White Seoul P4-U in the front and one Red Seoul P4-U in the rear. Same thing for the helmet. So, 6 LEDs total forward and 3 red ones in the back.

With that I should be able to see stuff, lol. Also I have noticed that my test-rig with only 1 Seoul P4 on the handlebars earns respect from drivers, so it will make for safer cycling.

So, battery voltage is more or less fixed between 10V and 15V because of the car horn (bough at scrap yard, much better than airzound, totally deafening, perfect).

Problem is, LED voltage is in the same range as battery voltage so I need a buck-boost regulator. Also, efficiency of drivers found on the web isn't that high, so out comes LTC3780.

The design is slightly complex because this circuit uses synchronous rectification and also the current regulation. So, I will use a ZXCT1051 current probe with a 0.1 Ohm current resistor. Hunting for the right MOSFETs was quite difficult, but DigiKey revealed the FDS8878 which looks good.
Due to the number of parts, this should be a bit expensive (especially the LTC3780) but the most expensive part is as always the PCB.

Features :
- Input voltage : 6V-20V
- Output voltage : 6V-20V
- Output voltage above, below or equal to input voltage
(actually the parts are 30V-rated)
- Output current : settable between 0 and 1A with potentiometer (actually, it would probably deliver up to 30W or something)
- Also configurable as a voltage regulator for powering the GPS etc
- Efficiency : 95% according to LTSpice
- Open circuit protection, short circuit protection, etc
- Hi/Lo beam : press a button and it goes to full power, release and it goes back to normal.
- Size : 40mm x 40mm

I am finishing the schematic and will post it shortly.

So, if anyone has tried this circuit before, do you have special advice ?
Thanks for any info

 

[snarfer]

If you haven't already, when you simulate in LTSpice enter the ESR and ESL for the capacitors you actually plan to use. You may be unpleasantly surprised at the results. Also be sure you are not exceeding the maximum junction temperature for the IC. I have spent quite a bit of time trying to get the LTC3783 to work in my application and have found that the Linear design notes are often lacking.

I am interested to see what you come up with, as I am in need of a driver that can handle some serious wattage. I looked into this IC before but concluded the parts count was just too high.

As far as FETs go you might look into Fairchild FDMS3672, or another from FDMS series. Mouser has them. There is a spice model which I have successfully gotten into LTSpice.

I get one-off double-sided PCBs with no through-hole plating, soldermask or silkscreen made for around 50 cents a square inch plus postage (no setup fee!), so if you need help with that PM me.

That said, I think for your app you could also get 95% or higher efficiency with a simpler and cheaper design employing a monolithic boost regulator in inverting buck-boost topology.

Good luck!

 

[snarfer]

-----------> whoops i hate when my back button does that!

 

[Peufeu]

Hm, I think I prefer the FDS8878, lower RdsON, lower price, lower capacitance, lower gate charge (and of course, lower voltage and current but it's OK in this application). The driver needs mosfet with low gate capacitance or else it'll overheat as you say.

Can you give examples of inverting bucl/boost ? I was thinking about that too, but didn't find any, perhaps I didn't look hard enough !

And since I'm playing with FPGA and other stuff, I'll have to order some boards anyway (probably from Malaysia or something) so this will just go with the rest. Thanks for the offer, though !

 

[snarfer]

I used the topology from the example circuit in LTSpice for LTC3783 LED driver then I substituted the LT1370 boost driver for the LTC3783 and added current sense amplifier LTC6102HV for input to FB pin. Depending on input voltage it came out to around 92% efficiency in my last simulations. I had gotten 95% efficiency but then I noticed that I was exceeding the breakdown voltage of the switch. For some reason the LTC1871 with external mosfet doesn't work nearly as well. I haven't figured it out yet.

Fairchild FDMS series seems to be updated version of FDS series. FDMS3672 is higher voltage. Might try FDMS8692. My cursory reading of datasheet indicates it switches faster with slightly lower gate charge than FDS8878. Of course I didn't check the price on FDS8878. The FDMS prices are something like 1.44 each.

 

[Peufeu]

Boards should arrive in a few days, we'll see if it works

 

[Peufeu]

There was a minor error in the Linear app note (you have to connect STDBY and RUN for the regulator to start). Apart from that, once soldered, it worked nicely. Dimming too.

Of course it's totally overkill, but I like it. Overkill is especially the 4 layer board, but I had to order some for a FPGA project, so it didn't add any extra cost.

 

[snarfer]

4 layer board, 4 mosfets, current sense amp, that is one expensive bike light!

Have you tried measuring the actual efficiency of the device? How does it perform?

Hey can we see some pictures?

 

[Peufeu]

I put the thing on my website :

http://audio.peufeu.com/node/66

I made a little efficiency measurement, it performs according to spec, between 90 and 95%. I couldn't find a 0.033 ohm resistor so I used a 0.1 ohm instead for the 3780's current sense control, so this limits maximum power and efficiency a bit. I'll just wire an extra 0.1 ohm resistor in parallel when I get one.

 

[Peufeu]

Photos of the helmet mounted light :

http://audio.peufeu.com/node/67



With 3 lights in series I will need to replace that 0.1 ohm current sense with a 0.03 ohm as it was originally intended. A bit above 6 watts, the chip's overcurrent protection trips because the resistor value is too large. (this is distinct from the current sense to set the current in the LEDs). Still, it provides a very intense light.