MC-E 2P2S to 2P autoswitcher, undervoltage cut off circuit feedback?

[uhg_]

Hello,

I'm planning to replace the stock LED in my Sigma PowerLED bike light with a Cree MC-E. I've come up with this tentative controller schematic based on a BuckPuck driver and the Picaxe microprocessor - which is the only such IC I can figure out how to program:

http://picasaweb.google.co.uk/ulrik...rsion?authkey=TnQPTJ5-nrI#5258585140467052290

Not being an electronics wizard I'd appreciate some input on whether this circuit is likely to work in real life - before I start inadvertedly toasting silicon. I have two questions in particular:

1. U1, Q1 and nearby components are supposed to latch the circuit in an off position once battery voltage drops below the cut-off for U1 (app. 4.3V; below this voltage the control signal from the Picaxe cannot fully dim the BuckPuck). Do you see any reason why this might not work?

2. Q2-Q4 are supposed to switch the MC-E LED configuration from 2 parallel chains of 2 dies in series (2S2P) to 2 dies in parallel (2P) when battery voltage drops below a point where 2P should theoretically (based on the MC-E data sheet) produce more light than 2S2P, or when battery voltage is simply too low to drive two LED dies in series. Again, do you see any reason why this might not work?

I realize that there could be many other approaches to this. My main aim in this project is to get more lumens while retaining or even increasing the light's versatility with respect to the kinds of batteries I've already got (6V and 12V NiHms), that are readily available for purchase here (AA alkalines) or that I'm planning to get (a 14.8V Li-ion). I've ordered MC-E specific lenses and am confident the cold climate where I live will be enough to keep heat management issues in check.

Once I get this working, I'll be happy to share the finished design, including the Picaxe software. For now, I just hope some of you wise people can help me evaluate the circuitry.


Thanks,

 

[Pig_Chaser]

My $0.02 (that's approx 1pence):

1) Assuming S1 is momentary, it should work just fine.

2) You may have a slight design flaw here. You are driving the Gate with the +LED output from the buck converter. When the FETs turn on, the +LED output of the Buck should fall from 6.6V to 3.3V, which in turn will lower the Vgs of the FETs turning them off again... this is fixed very simply by shorting the two LEDs connected the the -LED which should operate very near ground.

Oh and some might point out that shorting the LEDs will cause a momentary current spike through the remaining LEDs. The severity will depend on the Buckpuck's output characteristics (particularly capacitance).

For a non electronics "wizard" that sure is a cool looking circuit.

 

[uhg_]

Thanks,

I caved in and put in a DIP switch for reprogramming the Picaxe, so I can use its last remaining output pin to directly control the FETs. This helped me come up with a simpler circuit with fewer components:

http://picasaweb.google.com/ulrik.gade/DropBox?authkey=kK1hAMUiWAc#5259511013263099426
(Parts have been renumbered, sorry!)

Good points about the FETs. I've now moved them as you suggested and am going to have the Picaxe turn off the BuckPuck output temporarily during the switch.

One last question: The undervoltage cut-off sub-circuit relies on isolating battery minus from ground. I'd prefer placing the cut-off FET between the circuit and battery plus. I think this would require a P-channel FET but I can't figure out how to supply its gate with the required voltages to switch it on/off. Any pointers here would be appreciated.

Lastly, I did a studio of how I might completely rewire the LED dies from a 2S2P to a 4P configuration using a couple more FETs. As I've noted on the schematic (shown for 2 out of 4 dies; would be mirrored for the remaining dies), this has (at least) two issues that I don't know how to deal with. But a fun challenge nonetheless:

http://picasaweb.google.com/ulrik.gade/DropBox?authkey=kK1hAMUiWAc#5259513768763618098

I think I'm going to have a go at building this. Thanks for helping out

 

[Bandgap]

Some thoughts and questions.

Why not save a mosfet by permanently wiring the leds as two parallel pairs?
- then just use one mosfet to short out one pair, rather than two mosfets to short two individuals.

And could you save the switch by just connecting those points together during programming - just let the leds flash - although the gate capacitance might mess up the programming of course.
Or go to a pic-axe with more pins as this will only be slightly bigger than a 8pin version, and will save the space of the DIL switch or the extra components in the auto switch.
I spent a lot of time trying to 'stretch' an 8pin PIC bike light design. It was such a relief to move to 18pins

Is the power rail mosfet there because the "<500uA" of the BuckPuck is too much for your application?

I don't quite understand the U1 part and cannot make out the part number - does it switch off sharply and shut the circuit down?

I don't think even protected Liions are protected against over-discharge.

Your series/parallel switch will not work.
The parasitic reverse diode present in all power mosfets will be forward biased in your 'parallel' configuration - shorting out Vled+ to Vled-.
Also, the top pfet will always be pulled on in any mode by the forward voltage of the top led.
There is a way of doing series-parallel switching, but you have to be happy with losing a diode drop in the series configuration, or adding a few more components.

Steve

 

[Pig_Chaser]

...Why not save a mosfet by permanently wiring the leds as two parallel pairs?
- then just use one mosfet to short out one pair, rather than two mosfets to short two individuals.
...

The point though is to go from 2s/2p to 2p... you're proposing going from 2s/2p to 2s. However, that won't lead to a lower Vf and the buckpuck will try to maintain it's current through a single series string so there is no power savings. In fact higher current through the single string will raise the Vf increasing the power and possibly damage the LEDs.

...Your series/parallel switch will not work.
The parasitic reverse diode present in all power mosfets will be forward biased in your 'parallel' configuration - shorting out Vled+ to Vled-.
Also, the top pfet will always be pulled on in any mode by the forward voltage of the top led...

I noticed that too, the p-jfets depicted won't work at all. However, i looked up the part number and it's calling for an n-channel mosfet.

 

[uhg]

Many good points, thanks! About the permanent wiring, I'll definitely try that. Picaxe is a PIC range preloaded for educational purposes with an interpreter that allows it to be programmed in BASIC rather than assembly language - which cuts me into the PIC game. More info here: http://www.rev-ed.co.uk/picaxe/ I'll give upgrading to a larger PIC another thought. Been trying to use stuff I've got lying around. But as you point out this may not be a smart place to economize. The voltage supervisor is an MC33164-5. Its output follows input down to app. 4.3V where it drops sharply to ground, thus permanently shutting off the circuit via the FET. And it's tolerant of in excess of the 16.8V peak from a 14.8V Li-ion. Not sure I understand re. the Buckpuck. My rationale for the hardware shutoff is that below 4.3V the Picaxe can no longer produce a strong enough signal to force the Buckpuck to fully dim the LEDs, which it would have to do to protect the battery or the LEDs (the software tries to protect against both battery over-discharge and LED overheating). I'd love to do without it, and to shut off the positive rather than the negative power rail. But can't figure out how. I'm planning to buy a Li-ion bottle battery from Batteryspace.com. They're claimed to have various protections, including over-discharge. Just out of curiosity I'd be interested in a pointer to a better series/parallel switching circuit. The attraction would be the efficiency gained by spreading the load over all four, rather than only two, LED dies. Going down that road would probably amount to gross over-engineering but, hey, I'm not doing this for a living. I can be wasteful if it's more fun! Thanks again, uhg

 

[Bandgap]

The point though is to go from 2s/2p to 2p... you're proposing going from 2s/2p to 2s.

Oh no I'm not.

Would it clear anything up if I said:

Wire the central points of the two series strings together on uhg_'s diagram, then delete one of the mosfets?


Steve

 

[Pig_Chaser]

Oh no I'm not.

Would it clear anything up if I said:

Wire the central points of the two series strings together on uhg_'s diagram, then delete one of the mosfets?


Steve

lol, i was just logging on to recant... i just figgured out what you were saying. And that would've been more clear but i did jump to a conclusion without thinking it all the way through.

 

[Pig_Chaser]

Here's an idea as to how you might go from 2s/2p to 4p using a 2p2t relay. It kind've borrows from Steve's suggestion to connect the central points.

 

[uhg_]

That's exactly what I'm trying to do, except with a FET to reduce power consumption and b/c I'm concerned a relay won't withstand the vibration on rugged singletrail.

Judging from other posts here, the MC-E dies seem quite well matched: might there not otherwise be a potential issues re. different forward voltages in this design?

PS: Sorry about the lack of line-breaks in my previous post. Don't know what I did wrong.

 

[Bandgap]

Cant do diagrams so:

For series parallel switching, this is the idea that I came up with for my bike lights:

Assuming you have a two leds wired in parallel at the top, and the same at the bottom.

- connect a diode between the top pair and the bottom pair, pointing towards the bottom pair.

Have a look at Pig Chacer's diagram - a diode permanently betwen the pairs instead of the relay contacts

Wire a PNP or pfet from the bottom of the diode to the + rail

Wire an NPN or nfet from the top of the diode to the - rail.

Short the gates/bases together with an NPN to switch the thing to parallel

Don't underestimate the high efficiency available from Zetex NPNs and PNPs (base current only needs to be collector current/200 - and they will happily work form the 3.6V you have available across one LED) - you will, of course, need some base resistors if using bipolars.

If you want to draw it for checking - feel free.

Good luck

Steve

 

[uhg_]

Like so:

http://picasaweb.google.com/ulrik.gade/DropBox?authkey=kK1hAMUiWAc#5260408275424350210

Would prefer FET over diode to reduce power loss (heat) but this solves the problem with the reverse diode on the FET in between the two clusters. Nice, hadn't thought of that.

Thanks!

 

[Bandgap]

Like so:
http://picasaweb.google.com/ulrik.gade/DropBox?authkey=kK1hAMUiWAc#5260408275424350210
Would prefer FET over diode to reduce power loss (heat) but this solves the problem with the reverse diode on the FET in between the two clusters. Nice, hadn't thought of that.
Thanks!

That's it. I used something similar.

You have to be a bit clever to get a mosfet to short out that diode without creating a leakage path or short somewhere.
You have to use one upside down (mosfets are actually bi-directional, but have a diode across the channel in one direction) and get the gate connected in the right direction
How about accepting the 0.4V loss in 7.6V to begin with.

Steve

 

[uhg_]

You're right, better to go with the small drop than further complicate the design. Thanks again