LED driver with the LT1932 from Linear Technology

[remuen]

Yesterday evening (CET) I've built my newest DC-DC converter with the LT1932 chip from Linear Technology with a hand drawn and etched pc board (this is my first pc board I've made since more than 20 years).

The LT1932 chip has the following main features:
<UL TYPE=SQUARE><LI>it is especially designed as LED driver which can drive up to 16 white Nichia's
<LI>it is in a ThinSOT package means very small
<LI>it needs only 5 external components
<LI>it is a current controller
<LI>it has a high output voltage of up to 34 volts to drive the LED's switched in series
<LI>it has an input voltage range of 1 to 10 volts
<LI>works even when the input voltage is higher than the output voltage
<LI>has an efficiency up to 80%
[/list]

I have choosen this chip mainly to drive a couple of white Nichias in a 2AA Minimag and also for some other mods with Nichias because of the wide variety of input sources.

I couldn't make a lot of tests up to now but here some first results:

The efficiency is much better than Linear Tech claimed on their data sheet and I would say it is excellent if you consider that it is a buck/boost converter.

I just started a run time test on two not fresh alcaline AA batteries. Here the first results:

With a clever layout the board could be smaller than 1 cm2 even one uses a 'huge' SMD inductor as I do.

 

[dat2zip]

As I keep looking I see new IC's coming out faster than I can hook one up to play with.

I received my samples of the LT1308 and LT1932 and I hope to make a similar board soon.

That's interesting you got such good efficiency numbers.

Maxim newest 21st issue of Maxim Power supply Analog Design guide that came with my Maxim samples shows the MAX1848 boost with > 85% efficient. One drawback is Vin 2.6 to 5.5Volts. I need to go back and compare this against the LT1932 specifications.

I think 2.6 Vinmin is too high for even a 2AA battery solution.

Also on the same page is a switched capacitor buck/Boost charge pump white LED driver. 1.6V to 5.5V.

 

[ElektroLumens]

Remuen,

Hey, that's great efficiency! Good job! Do you think it would drive a Luxeon? Well, the MAX757 circuit was only supposed to go up to 200mA, but I have seen it put out well over 500mA and even higher, by using different inductor/capacitor combinations. Have you done any experimenting in this manner?

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:

One drawback is Vin 2.6 to 5.5Volts. I need to go back and compare this against the LT1932 specifications.

I think 2.6 Vinmin is too high for even a 2AA battery solution.
<HR></BLOCKQUOTE>
There are some other interesting Maxim chips with the same drawback. I have contacted Maxim but it seems the don't have a good solution and that they don't intend to launch a chip similar to the LT1032 for the next future

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:

Also on the same page is a switched capacitor buck/Boost charge pump white LED driver. 1.6V to 5.5V.
<HR></BLOCKQUOTE>

As far as I've seen all the charge pumps do have a high min Vin

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by ElektroLumens:

Hey, that's great efficiency! Good job! Do you think it would drive a Luxeon? Well, the MAX757 circuit was only supposed to go up to 200mA, but I have seen it put out well over 500mA and even higher, by using different inductor/capacitor combinations. Have you done any experimenting in this manner?
<HR></BLOCKQUOTE>

It was never my intention to drive a LS with this chip. I can hardly imagine that one can do it except on a lower brightness (200 ... 250mA). But I think that's not exactly what you want.

IMO there a a lot of better chips to drive the LS eg. the ZLT (Zetex) or even the 'old' MAX757 or the LT1308B and many others more

But if you want to drive only a couple of Nichias I think the LT1932 is one of the best solutions - it is a 'general purpose' buck/booster for white LEDs - and it can be build extremly small.

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by ElektroLumens:

Hey, that's great efficiency! Good job!
<HR></BLOCKQUOTE>

Wayne, I have forgotten something:
To get such a good efficiency I have choosen an inductor with a very low DCR of only 0.074 ohms and low ESR ceramic caps. This low DCR inductor is quite huge for a SMD inductor so I've ordered some smaller ones with still a good DCR (0.13 or 0.2 ohms) but I am convinced the efficiency will drop with these ones.

The decision which inductor I will use depends only on the size the board has to have.

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by MrAl:

Looks like the 1932 chip wont drive an LS
because it only puts out 40ma tops.
To drive multiple led's, they are connected
in series, which increases the voltage, not
the current.
<HR></BLOCKQUOTE>

MrAl
The output current can be adjusted by choosing another value for Rset. And as the switch current limit is about 550mA at 25°C it should be possible to drive the LEDs with a much higher current. That's why I mentioned the 200 to 250mA to Wayne.

But as I already pointed out it is not a booster for the LS and it was never my intention to do it with this chip. So I don't want to test it with a LS.

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by MrAl:

Also, dont let the higher price of low ESR caps fool you; they should only be used when low ripple is needed. Keep in mind a low ESR cap draws more current then a non ESR cap when charging up, which in some apps can actually cause a drop in efficiency. The Nichias usually dont require ultra low ripple voltage operating 40ma or less. With a string of Nichias in series, the situation is even better. Two in series doubles the ability of the string to handle ripple voltage, three in series triples, etc.
With 10 in series, you can probably get away with two hub caps glued back to back :-)
<HR></BLOCKQUOTE>

I used only a 4.7uF input and a 1uF output cap (as recommanded from Linear) in this first prototype. This 1uF output cap should not cause a significant efficiency loss.

I used it because the inductor has only 6.8uH and one should be able to use this DC-DC converter not only with a couple of LEDs in series but also with only one LED without having to make any changes in the circuit design.

As soon as I have more time I'll try to optimize the circuit. I would like to eliminate the schottky diode because I think it causes the highest loss in this circuit but I'm not sure what will happen to the chip. This also means to remove the output cap.

 

[jeff1500]

Very good.

 

[jeff1500]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by MrAl:
Looks like the 1932 chip won't drive an LS
because it only puts out 40ma tops.<HR></BLOCKQUOTE>

How about the two transistor Brinkman circuit for an LS?

I've got it here in my notebook http://edusite10.tripod.com/led3/index.html

 

[remuen]

Here are the results of the runtime test with two already used alkaline AA batteries.

Here I stopped the test because the output current dropped below 1mA and can't be measured any longer accross my 0.1 ohm in series with the LEDs.

Even I stopped the test I let the booster run on the same batteries just to look how long the moonlight phase would last. Here some additional results:

At 6:00 AM Central Europe Time CET - almost 24 hours after starting the runtime test - this moonlight is still bright enough to read in book or find the key on the floor or to go to the toilet in the darkness. I can imagine that it will last another couple of hours. Let's see, I let it run ...

This test shows three things:
<UL TYPE=SQUARE><LI>A flashlight modified with a couple of Nichias and a booster like this has a completly different characteristic than a mod with a ZLT booster (Zetex) and a Luxeon Star. With the LS you have the brightness and with the Nichias you have the long runtime with a reasonable moonlight almost for the eternity ...
<LI>In this constellation (chain of 5 LEDs in series) it is not a single cell LED driver. I will soon make more tests with different numbers of LEDs.
<LI>The efficiency of the white Nichias driven with a low current is excellent
[/list]

 

[MrAl]

Hi there,

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by ElektroLumens:
Remuen,

Hey, that's great efficiency! Good job! Do you think it would drive a Luxeon? Well, the MAX757 circuit was only supposed to go up to 200mA, but I have seen it put out well over 500mA and even higher, by using different inductor/capacitor combinations. Have you done any experimenting in this manner?

<HR></BLOCKQUOTE>

Looks like the 1932 chip wont drive an LS
because it only puts out 40ma tops.
To drive multiple led's, they are connected
in series, which increases the voltage, not
the current.
Too bad there isnt a way to change the
output transistor; that's internal also.

Also, dont let the higher price of low
ESR caps fool you; they should only be
used when low ripple is needed. Keep
in mind a low ESR cap draws more current
then a non ESR cap when charging up, which
in some apps can actually cause a drop in
efficiency. The Nichias usually dont require
ultra low ripple voltage operating 40ma or less.
With a string of Nichias in series, the
situation is even better. Two in series
doubles the ability of the string to handle
ripple voltage, three in series triples, etc.

With 10 in series, you can probably get away
with two hub caps glued back to back :-)

Good luck with it,
Al

 

[jeff1500]

So, let's see, 5 leds for about 17 volts, or 17/5 = 3.4 volts per led. That stays about constant, then the current drops with the battery voltage as they wear down.

Looks like you've got a buildable design.

If you use nimh batteries it won't be as bright, or if you set it up for full brightness with nimh batteries it will overload with regular batteries.

Maybe the optimum solution is a voltage regulated design that will provide a high constant output voltage so the leds can be in series with uniform current.

I wonder why the voltage regulators don't have higher output voltage and the chips like LT1932 don't regulate?

What's your next step? More research or are you ready to build a light?

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by Jeff1500:

If you use nimh batteries it won't be as bright, or if you set it up for full brightness with nimh batteries it will overload with regular batteries.

Maybe the optimum solution is a voltage regulated design that will provide a high constant output voltage so the leds can be in series with uniform current.

I wonder why the voltage regulators don't have higher output voltage and the chips like LT1932 don't regulate?
<HR></BLOCKQUOTE>

Voltage Ragulators are not designed to drive LEDs but logigal circuits which need a voltage of 3 to 5.5 volts.

I think it is a missunderstanding that the LT1932 does not regulate. Look at the output voltage and compare it with the input voltage. Something does obviously happen.

Ok, how does a current controlled booster work? I tries to increase the output voltage to that level where a certain current flows through the load according the formula I = U / R. In this case this current should be 38 mA as adjusted by a Rset. For this certain load of a string of 5 Nichias it must generate an output voltage of about 17.3 volts to reach this current. If you would take 5 other Nichias it had maybe to generate 17.8 volts or only 17.1 volts (depending on the tolerances of the Nichias) to drive the them with 38mA.

For all LEDs (Nichias and Luxeon Star) the current does say wether the LEDs are overdriven or not. So a current controlled booster will not supply a higher output current than adjusted and therefore not overdrive a LED even the input voltage varies. Exactly this is the biggest advantage of a current controlled booster! As example see my very first test measurment results in my first post. It has the same maximum output current from 3 up to 6 volts input.

Each chip has its limitation. With this booster, a string of 5 Nichias in series, an input voltage of 2.4 volts or less and a target output current of 38 mA we have obviously reached this limit because it can no longer generate the needed output voltage of 17.3 volts to drive the LEDs with 38mA. With 3 volts input the situation looks much better.

In this first test you have to consider that the Nichias are already slightly overdriven with 38 mA. If I would reduce this current to the recommanded 20mA the booster would have driven the Nichias for many hours with the full brightness. Look at the figures in the runtime test I have posted. I started the test at 6:30 AM and at 11:00 AM the Nichias are driven still with 20mA. You would have to add some hours more because in this time the Nichias sucked much more energy out of the batteries because of the higher current. And this happens with already used batteries with a start up voltage in this test of only 2.45 volts.

You have also to consider that the Nichias do not produce double the brightness with the double current. The increase in brightness is somewhere between 35 - 50% (I do not have exact figures for this but they can be found in this forum too) when increasing the LED current from 20 to 38mA. Using 8 Nichias driven with only 20mA would produce about the same brightness as 5 Nichias driven with 38mA but with a lower power consumption. With such a load of 8 Nichias driven with only 20mA the result with this LT1932 booster would be for sure much better.

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by Jeff1500:

What's your next step? More research or are you ready to build a light?
<HR></BLOCKQUOTE>

First of all: I did this first run time test as published not to show the best possible result but more to see where the limitations are. I will do some more tests to learn more about it and then poosibly try to improve it a bit. But I do have also some limitations - I am not able to do a similar good job as the Zetex guys did on their booster.

With this first prototype I could already make a fine mod. I have a Petzl Tikka headlight with 3 Nichias and 3 AAA batteries. I think using it in this headlight it would be a excellent solution as it would reach for sure more than 90% efficiency which is IMO an excellent result. Unfortunatly I didn't find out up to now how to remove the battery case without destroying the housing.

I could also finish the Minimag mod with eg. 7 Nichias driven with a current somewhere between 20 to 30 mA.

But I think first I'll do some more tests and as my time to work on this project is limited the mods have to wait.

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by jeff1500:

I wonder why the voltage regulators don't have higher output voltage and the chips like LT1932 don't regulate?
<HR></BLOCKQUOTE>

Jeff
Here some figures more just to give you an idea how the LT1932 works. I took 2 NiMH AA cells to make this test.

Using only 2 or 3 LEDs driven with 38mA gives a much a better result whereas the efficiency with only 1 LED drops - but this was to expext according to the datasheet. The 2 or 3 LEDs driven with 38mA are about the same as 4 to 6 LEDs driven with 20mA concerning the power consumption but 4 to 6 LEDs driven with 20mA will be brighter.

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>
Originally posted by DSpeck in another thread:

remuen, your results with the 1932 look excellent. This is what I was planning when I began this thread. I put the wrong chip number in the thread title, but don't know how to change it...
Do you have enough components to do a test run for me? I would really like to to know how the circuit runs with 1, and then 2 branches of 8 series LEDs, in parallel with each other. I am planning to run my flashlight with 3 AAA Alkaline batteries, in series for 4.5v as Vin. If you could try this in your test, that would be great.

I'd do it myself, but I haven't got the components, nor a simulator. If you haven't got the parts (I realize 16 LEDs is quite a few!), don't worry about it.

BTW, which circuit configuration and component values are you using? You don't mention either in the other thread.

Doug Speck.
<HR></BLOCKQUOTE>

Sorry to all of you interested in this chip but the last two weeks I had absolutly no time to work on this project.

DSpeck, for my ver1 board I am using the following components:


Cin 4.7uF low ESR ceramic cap
Cout 1.0uF low ESR ceramic cap
L 6.8uH low DCR inductor (0.074 ohms)
Rset 560 ohms (= 40mA output current)
Diode Zetex ZXCS1000


In the meantime I made a ver2 board with a 10uH inductor. But there was not the slightest improvement in the results but an efficiency loss of 1 - 3% compared to the ver1 booster. Maybe this is caused by the higher DCR (0.093 ohms) of this inductor.

Here are some more results with my ver1 board which will give you an idea how the LT1932 works in your configuration.

Rset is 560 ohms means the output current is set to 38mA. For the test I used a regulated power supply. I don't know wether 3 AAA batteries in series will be able to deliver the necessary input current.Maybe you would need a bigger input cap because the measured input current is only a average current means the current peaks are about 2 times higher. But in every case an input voltage of 4.5 volts is a good choice for this booster.

For me these results are bit dissapointing even they show excellent efficiency values. In my certain case this is not exactly what I have excepted because I wanted to use the LT1932 booster for a 2 AA Minimag mod. But it will be a excellent choice for my Petzl Tikka mod (3 AAA batteries and 3 white Nichias).


The more LEDs I'm using the higher the input voltage must be to get good results. In the low input voltage range the output current drops very fast. With 5 LEDs in series the output current already begins to drop between 3.0 and 2.4 volts input. So an input voltage of 4.5 volts (three 1.5 volts cells) or higher is in every case a good choise.

If you drive your LEDs on 4.5 volts with only 20mA instead 40mA the result will look much better.

In this case you could compare the above published results with your configuration with two chains of LEDs in parallel. Eg 5 LEDs in series driven with 40mA will be about the same load as 10 LEDs in 2 chains of 5 LEDs in parallel (not exactly but will give you a good idea how it would be).

Btw, LEDs driven with only 20mA will have a better efficiency than driven with 40mA: 2 LEDs driven with 20mA are brighter than only 1 LED driven with 40mA.

DSpeck, I hope these informations will help you to come to a decision.

 

[DSpeck]

That's exactly the info I was looking for, remuen.

It looks to me like 2 parallel chains of 7 LEDs, each driven at 20 ma, would be a great way to go, and a damned bright flashlight, too.

Maybe it would be better to have both chains on all the time, and just vary the current output for a dimming function. If we tied the dimming control input to the Vin rail, and had a voltage divider (as on pg 11 of the data sheet) with either 2 discrete, switchable values, or a minimum setting with a potentiometer between Vin and the Rset pin, we could have a nice variable brightness flashlight. Is there a reason why we couldn't tie directly to the Vin for this function?

Also, would using 4 cells be a noticeably better choice? I'd expect higher efficiencies yet, and the form factor of 4 AAAs side-by-side (like the PT Attitude) would still allow for a fairly compact flashlight. Remuen, could you check and see? I hope you don't mind, but it should also give others some more useful data as well. Thanks!

DSpeck

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by DSpeck:

Maybe it would be better to have both chains on all the time, and just vary the current output for a dimming function. If we tied the dimming control input to the Vin rail, and had a voltage divider (as on pg 11 of the data sheet) with either 2 discrete, switchable values, or a minimum setting with a potentiometer between Vin and the Rset pin, we could have a nice variable brightness flashlight. Is there a reason why we couldn't tie directly to the Vin for this function?
<HR></BLOCKQUOTE>

Up to now I didn't look into the dimming function because I don't need it for my applications. It is an interesting feauture. If you download the additional design notes from the Linear website you will find some more information about this topic.

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by DSpeck:

Also, would using 4 cells be a noticeably better choice? I'd expect higher efficiencies yet, and the form factor of 4 AAAs side-by-side (like the PT Attitude) would still allow for a fairly compact flashlight. Remuen, could you check and see? I hope you don't mind, but it should also give others some more useful data as well.
<HR></BLOCKQUOTE>
I've already made similar measurements with 6 volts input. I will post these results later on. As far as I remember there is about the same efficiency as with 4.8 volts. But 6 volts will give you a longer runtime on a higher output current so it will be the better choice.

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR> Originally posted by DSpeck:


Also, would using 4 cells be a noticeably better choice? I'd expect higher efficiencies yet, and the form factor of 4 AAAs side-by-side (like the PT Attitude) would still allow for a fairly compact flashlight. Remuen, could you check and see? I hope you don't mind, but it should also give others some more useful data as well.
<HR></BLOCKQUOTE>

DSpeck, as promised here the additional results. I'm going to post the whole row again so it's easier to compare.

 

[DSpeck]

Those efficiencies look great, especially in the middle range. I still think 2 rows of 7 LEDs in parallel look about the best, although if I dropped down to rows of 6, I'd get better regulation, right down to 0.6V/cell, which is excellent battery usage.

On another note, if I just have one setting for the current, it would simplify the light. I'm not sure of the value in dimming, although it would extend the battery life. With this circuit, the runtime should be pretty long anyway.

DSpeck

 

[remuen]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by DSpeck:

Those efficiencies look great, especially in the middle range. I still think 2 rows of 7 LEDs in parallel look about the best, although if I dropped down to rows of 6, I'd get better regulation, right down to 0.6V/cell, which is excellent battery usage.
<HR></BLOCKQUOTE>

The efficiency is really impressing and much better than I have expected. I'm quite sure it is the result of the very low DCR inductor I am using. I will later on make a comparison with another inductor to see what happens with the efficiency.

IMO with 5 or 6 LEDs per row driven with 20mA each would be a better choice than 7 LEDs per row because of the better efficiency.

You can run the batteries down to about 1.8volts (0.45volts/cell if you use 4 cells) with still a reasonable brightness. But then the output current will drop very fast and the light will stay for an 'eternity' in the moonlight mode (see the runtime test on page 1 of this thread).

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by DSpeck:

On another note, if I just have one setting for the current, it would simplify the light.
<HR></BLOCKQUOTE>

That's exactly why I don't consider the dimming mode for me. It would need an additional switch or a pot or whatever else which makes a flashlight mod much more complicated.

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by DSpeck:

I'm not sure of the value in dimming, although it would extend the battery life. With this circuit, the runtime should be pretty long anyway.
<HR></BLOCKQUOTE>

You will find all information concerning the dimming control on the pages 7 and 8 of the datasheet. I think the easiest way to dim the LEDs (eg. down to 10mA) in a flashlight is using a DC voltage on the Rset pin. The formula to calculate the necessary Radj with your values of Rset an Iout is shown on page 8 of the datasheet. With it it should be no problem to calculate this resistor.

Concerning the runtime: I don't know the runtime with 4 AAA batteries. I made only one with 2 AA alkaline batteries and one with 2 AA NiMH cells as posted on page 1 of this thread. 2 fresh AA alkalines have about 5600mAh, 4 AAA alkalines about 4400mAh so you can make a rough estimation of the possible runtime based on my results. But there is still one question: What is the max current the 4 AAA alkalines can deliver? For the moment I have only AAA NiMH's at home and no battery case for the AAA's (sorry, but I don't want to solder the NiMH's together for making a test). If you have some AAA you could measure the voltage accross a low resistor of eg. 0.1 ohm (don't make a short cut with your DMM) and calculate the current.

 

[DSpeck]

I haven't got any very low value resistors, so this test will have to wait. With the 4 batteries in series, the light should still last several hours at good to excellent brightness.

Now to order some parts and do something with them. Who is a good online distributor to buy all the parts at once from?

DSpeck