LT1932 - Open Circuit Protection

[Xcandescent]

So I FINALLY got my prototype boards back for the LT1932-based GBA light circuit I'm working on, and amazingly enough they work! (Amazing, because that was my first time surface-mount soldering ANYTHING, and I was dumb enough to order 0402-sized components which I then had to solder in with two jumbo-sized Radio Shack soldering irons. DOH!)

I haven't gotten around to getting any real measurements yet (and I'm not sure I can actually get any accurate measurements for efficiency), but there's already a design issue which is causing me some concern. Because the LT1932 is a current-regulating part, any condition where an LED string burns out or goes open circuit could cause serious problems. The app notes for the LT1932 suggest dropping in a Zener diode to protect the LT1932 ... but also notes that the following will happen:

"If the device is turned on without the LEDs present, no current feedback signal is provided to the LED pin. The LT1932 will then switch at its maximum duty cycle, generating an output voltage 10 to 15 times greater than the input voltage. Without the zener, the SW pin could see more than 36V and exceed its maximum rating."

So while the LT1932 is just fine and dandy, you're sucking who knows how much power out of your source to generate an output voltage many times higher than you'd ordinarily see, which in turn is being consumed by nothing, because there's no load. This would also seem to be a Bad Thing.

My preference would be to auto-kill the supply voltage when the LEDs go open-circuit, but my meager comprehension of electronics is not sufficient to figure out how to do that. (Yet.) Any suggestions?

-XCN-

 

[remuen]

I already have built several LT1932 converters without using this Zener diode ( LED driver with the LT1932 from Linear Technology ) as it is not really necessary. It's used to protect the circuit and the LEDs.

OK, how does it work? Without connecting the load the LT1932 generates a higher output in order to reach the programmed current flowing through the (not existing) load as described in the data sheet. So the output voltage will possibly exceed the max value some other parts can bear (eg the capacitor) and they can get damaged. And also the output capacitor will be loaded up to its maximum. Connecting the load again this load in the capacitor could damage the LEDs - but I think this will not really happen with only a 1uF capacitor as suggested by LT.

Using this Zener diode the output voltage will not increase over the Zener voltage as then the Zener diode will act as a load.

You will find a detailed discussion on this topic under this link: ZLT+ additional construction tips and solutions to problems

Hope this helps you understand what will/could happen and why such a Zener diode should (

) be used for every current regulated step up converter.

 

[Evan]

I havn't had a problem without a zener, but I'm running off a single cell and 1.5x15=22.5; still under the 36 volt max.

You think soldering to a board is bad; my 1932 is wired point to point with fine wire.

I have been looking for a good circuit to run on one cell and have compared the LT1932, the 2 transistor "Brinkman" circuit, and a satcure. The 1932 has regulation going for it, and passes more power on a strong battery. It works on a broad range of inductors I've tried. But of the 3 circuits, the 1932 is the first to go out when the voltage drops, followed by the brinkman circuit. The satcure keeps going down to about .6 volt !

 

[remuen]

Originally posted by Evan:
I havn't had a problem without a zener, but I'm running off a single cell and 1.5x15=22.5; still under the 36 volt max.

You think soldering to a board is bad; my 1932 is wired point to point with fine wire.

I have been looking for a good circuit to run on one cell and have compared the LM1932, the 2 transistor "Brinkman" circuit, and a satcure. The 1932 has regulation going for it, and passes more power on a strong battery. It works on a broad range of inductors I've tried. But of the 3 circuits, the 1932 is the first to go out when the voltage drops, followed by the brinkman circuit. The satcure keeps going down to about .6 volt !

<font size="2" face="Verdana, Arial">Btw, I think you're talking about the LT1932 and not the LM1932 (LT is Linear Technology's designation, LM the one of National). Just mentioned this because we are already discussing about so many IC's which all have some numbers and a correct designation helps to talk about the same thing ...

The LT1932 is for sure not the right IC for only a single cell application as it operates only down to 1V. Some of the Maxim IC (eg MAX1674, MAX1675, MAx1797) will run down to about 0.7 volts input but are not current regulated as the LT1932 is. Also the Zetex ZXSC300 (the one used for the well known ZLT boost converter) is much better suited for this kind of application and it also is current regulated.

The LT1932 does its best job with input voltages of 3 volts and higher. With less than 3 volts I would recommand it only for 3-4 LEDs as its efficiency drops to much with more than 4 LEDs and less than 3 volts input (see the results in my thread - I posted the link in my former post above).

Btw, I prefer soldering these DC-DC converters to a etched board except sometimes for prototyping. But especially the MAX chips are very sensitive to ground bounces and whatever else (the LT1932 isn't!!) and a well designed board helps to avoid this kind of problems - and you always can reproduce it.

 

[Xcandescent]

Originally posted by remuen:
I already have built several LT1932 converters without using this Zener diode ( LED driver with the LT1932 from Linear Technology ) as it is not really necessary. It's used to protect the circuit and the LEDs.

<font size="2" face="Verdana, Arial">Technically, as long as you've got a load, the Zener isn't necessary. It's when your load goes away (i.e. when an LED "burns out") that you've got problems. Since I'm powering this off of a Game Boy Advance Link Port, I would like a way to auto-kill the circuit in that circumstance to prevent possible damage to the GBA power supply. Remember that, even with the Zener, the part will be switching at its maximum duty cycle, and attempting to produce an output voltage many times higher than the input voltage -- which, in turn, taxes the GBA's voltage-regulated power supply, possibly causing Bad Things to happen.

It's assumed that the light and circuit are replaceable; the GBA less so (due to cost).

Granted, I've done plenty of stupid things while drawing power from the Link Port on the GBA, and the power supply seems to cut off properly in all cases. However, this is definitely not something I'd want to leave to chance. Again, my electronics knowledge is limited, so I'm not familiar with the types of parts or circuits that could be used to do the job.

-XCN-

 

[Evan]

Quite right, LT1932. I edited the post to change LM to LT.

The LT1932 is a warmup, I've got some Zetex 310 chips I plan to work with. But that's a Winter activity.

 

[Xcandescent]

remuen:

I completely forgot about that thread! Went back and read it ... those are some impressive numbers! (Am jealous of that board size too ... though I seriously, SERIOUSLY doubt I could hand solder a 10x10mm board!)

Regarding efficiency ... I had played around with Linear's SwitcherCAD 3 simulator, and plugged in some odd values to see what the effect might be. While I still don't have a decent RMS multimeter, the current readings I got with my crappy Radio Shack meters indicated really poor efficiency with just 2 LEDs (somewhere in the 65 - 70% range). In your opinion, what are the biggest factors affecting efficiency with this IC?

My current design (and don't laugh now ... I already feel really silly about this ...):

Cin 10uF
Cout 4.7uF
L 22uH (low DCR)
Rset 976 Ohms (= around 20 mA)
Schottky Diode IR MBRA140TR

I also threw in the "soft-start" circuit suggested in the app notes, with the following values:

Csoft .022uF
Rsoft 5.6kOhms

Also ... regarding overdriving the Nichias: how badly does that affect their life expectancy? I think I saw a thread about this earlier, implying that the effects were pretty significant, particularly without heat sinking.

-XCN-

 

[remuen]

Originally posted by Xcandescent:

Technically, as long as you've got a load, the Zener isn't necessary. It's when your load goes away (i.e. when an LED "burns out") that you've got problems. Since I'm powering this off of a Game Boy Advance Link Port, I would like a way to auto-kill the circuit in that circumstance to prevent possible damage to the GBA power supply.

<font size="2" face="Verdana, Arial">You're right, the Zener diode is only a safety precausion which you have to use if you want to stay on the safe side. Technically there is no other easy way to solve this problem than using such a Zener diode.

But if you're following this thread ZLT+ additional construction tips and solutions to problems you'll see that dat2zip had the great idea to connect the Zener diode not across the load but connecting it to the hot end of the Rsense. I've never tried it out but IMHO it also should work with the LT1932 by connecting the anode of the Zener diode to the hot end of the Rset (pin 4 of the LT1932). This would help you at least to use a much a smaller Zener diode.

As the effect of such an increasing output voltage happens only with current regulated boost converters you could use a voltage regulated converter instead, in your case eg. a MAX1722.

Originally posted by Xcandescent:

Am jealous of that board size too ... though I seriously, SERIOUSLY doubt I could hand solder a 10x10mm board!

<font size="2" face="Verdana, Arial">I've already designed a smaller board which is only 8x10mm but one could make it even smaller. If you can solder these parts on a bigger board you can solder them also on a smaller board. Just give it a try ....!

Originally posted by Xcandescent:

Regarding efficiency ... I had played around with Linear's SwitcherCAD 3 simulator, and plugged in some odd values to see what the effect might be. While I still don't have a decent RMS multimeter, the current readings I got with my crappy Radio Shack meters indicated really poor efficiency with just 2 LEDs (somewhere in the 65 - 70% range). In your opinion, what are the biggest factors affecting efficiency with this IC?

<font size="2" face="Verdana, Arial">To be honest: I've got these impressive efficiency numbers only with my very first LT1932 converter. The ones I've built since then have an efficiency which is about 5% less - but also these results are still very good and better than indicated by the data sheet. I really don't know why the first board is much better as for the second and third LT1932 Dc-DC converter I used the same board layout and the same parts????? I've measured this first version several times with always the same excellent results.

Btw, I wanted to add this information to my LT1932 thread but since it has been moved to the electronics forum it's blocked by an administrator or moderator so I can't update it!!!!!

What affects the efficiency most? I'm for sure not an expert for such circuits but as far as I have learned these are the main influences:

</font><ul type="square">[*]<font size="2" face="Verdana, Arial">Last but not least: The Switcher IC itself and the input voltage and the load. Some DC-DC converters have their best efficiency with the highest load whereas the LT1932 has quite a bit a different characteristc (see my results in the other thread). Using the LT1932 with only one Nichia LED it has a poor efficiency of only 74.5% at 3 volts input but with 2 LEDs already 93.7% at 3 volts input (with Rset = 560 ohms)
</font>[/list]<font size="2" face="Verdana, Arial">As said before, I'm really not an expert for DC-DC converters. Almost everything I've learned about these circuits in the last couple of months can be found in the original ZLT thread so I only can suggest you to study this extremly huge thing as I and some others did ......

------------
Added later:
------------

Originally posted by Xcandescent:

Also ... regarding overdriving the Nichias: how badly does that affect their life expectancy? I think I saw a thread about this earlier, implying that the effects were pretty significant, particularly without heat sinking.

<font size="2" face="Verdana, Arial">I only know that overdriving Nichia's does reduce the life expectancy. But who cares whether they live 20'000 or 50'000 or 100'000 hours

? It's in every case much much much longer than an incandescent lamp will work!

 

[Xcandescent]

Originally posted by remuen:
But if you're following this thread ZLT+ additional construction tips and solutions to problems you'll see that dat2zip had the great idea to connect the Zener diode not across the load but connecting it to the hot end of the Rsense. I've never tried it out but IMHO it also should work with the LT1932 by connecting the anode of the Zener diode to the hot end of the Rset (pin 4 of the LT1932). This would help you at least to use a much a smaller Zener diode.

As the effect of such an increasing output voltage happens only with current regulated boost converters you could use a voltage regulated converter instead, in your case eg. a MAX1722.

<font size="2" face="Verdana, Arial">While I could probably get away with using a voltage regulated part, I'd much prefer to keep using a current regulated part as this light is intended to be left on for extended periods of time, and will probably end up generating enough heat to increase the current draw (though I doubt it would come anywhere close to thermal runaway).

The suggestion for connecting the Zener diode to the Rset pin is brilliant! I did a quick sim connecting the anode of the Zener directly to the Rset pin (bypassing the Rset resistor), and the results look extremely promising -- current draw from the voltage source drops to less than 10 mA! I'll definitely have to try this in the next prototype! I suppose if I wanted to get fancy, I could toss an "open circuit warning" LED in there -- but since you're only gonna get an open-circuit if the light doesn't work, it would just be silly.

As for soldering on 10x10mm boards ... have you actually seen the soldering irons they sell at Radio Shack? =) I guess not if you live in Europe, but suffice it to say that it's a miracle I figured out how to solder 0402 resistors at all with them (basically, take soldering irons, have component sticking to one with solder, then fiddle around until properly stuck to board and hope it didn't get roasted. Oddly enough, none of the components did!) Admittedly, the slightly larger parts weren't that big of a deal at all; doing the IC was a no-brainer. My no-clean flux pen helped a LOT.

Now I just need to figure out a way to either spread the light from two Nichias onto a GBA screen without dimming the output, or just make a really tall light shield. I know people here do custom flashlights all the time, but somehow it seems the difficuly (and cost) of having to design and mold a plastic light shield prototype is considerably above that ... but I'm stupid enough to try anyway.

-XCN-

 

[remuen]

Originally posted by Xcandescent:

As for soldering on 10x10mm boards ... have you actually seen the soldering irons they sell at Radio Shack? =) I guess not if you live in Europe, but suffice it to say that it's a miracle I figured out how to solder 0402 resistors at all with them (basically, take soldering irons, have component sticking to one with solder, then fiddle around until properly stuck to board and hope it didn't get roasted. Oddly enough, none of the components did!) Admittedly, the slightly larger parts weren't that big of a deal at all; doing the IC was a no-brainer. My no-clean flux pen helped a LOT.

<font size="2" face="Verdana, Arial">I never tried to solder these 0402 parts - but IMHO it can't be much worse than a 10uMax IC

Btw, I use a more than twenty years old Weller soldering iron with the smallest possible tip. As you already wrote: Using flux is the secret!

Originally posted by Xcandescent:

Now I just need to figure out a way to either spread the light from two Nichias onto a GBA screen without dimming the output, or just make a really tall light shield. I know people here do custom flashlights all the time, but somehow it seems the difficuly (and cost) of having to design and mold a plastic light shield prototype is considerably above that ... but I'm stupid enough to try anyway.

<font size="2" face="Verdana, Arial">I only can say: Good luck!

 

[Xcandescent]

Quick edit: Suspected things were too peachy with the sim, so went back and double-checked the runs ... yeah, doesn't work so good when the LEDs ARE connected.

So yeah, back to the drawing board.

EDIT #2: Something was bothering me about the way I set up the circuit, so I double-checked AGAIN and realized that I had a trace in the schematic inadvertently connected to a point where it shouldn't have been. I guess my brain is just off today.

Going back to the suggestion of hooking the Zener diode from the output of the Schottky directly to the Rset lead (bypassing the Rset resistor), I went back and simmed it a few more times with and without the LEDs connected, and it does seem to work -- cutting current draw from the voltage source almost completely. The one issue that had bothered me about doing it that way was the voltage of the Rset pin with respect with ground -- but 100mV is low enough that there wouldn't be too much of an additional wait for the Zener diode to pass its reverse breakover point.

Sim graphs did show considerable power draw on startup when the LEDs are disconnected, so I took a cue from the app sheet and added a second transistor to the "soft start" circuit. The base of this second transistor is connected to the same point as the base of the transistor in the soft-start circuit, and the collector is also connected to the output of the Schottky -- but the emitter is connected to the output of the Zener diode, instead of the output of the LEDs.

Sims of that circuit show that the addition of a second transistor benefits both cases, spreading out the start-up time and reducing overall current draw on start-up considerably. Again, just a sim, but encouraging nevertheless.

Now I just need to magically learn 3D CAD/CAM design for that damned light shield. =)

-XCN-

 

[Xcandescent]

Hmm, didn't notice that the date/time of original postings don't change when edited. I'd assumed there would be a flag or something to note the update.

In any case, did a couple of run-time tests on the prototype I do have on hand yesterday ... a pair of NiMH's pooped out quickly (I suspect I've got a few bad ones), but a pair of Energizer alkalines powered a GBA with the light circuit powered from the Link Port and LEDs for around 6 hours. Current regulation held steady at around 20 mA the whole time. Definitely a good sign. =)

Need to replace a wire and redo some of my soldered joints today, then build and test a version using the reference specs in the LT1932 app notes, without the soft-start circuit. I suspect the initial power draw will overwhelm the GBA's switching power supply, but there's only one way to find out!

-XCN-