Maxim's 1674 IC as a LED driver...

Status
Not open for further replies.

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

Oops. papasan, Sorry I posted my first run with the MAX1676 which is the dual version of 1674,5,6 family. It can be strapped for either 500mA current limit or 1A. I chose this to allow testing both settings. In my configuration I have an tantalum on my input and output.

Later tonight I'll be hooking the Luxeon to it and I'll do some measurements with it.

As I posted in the other thread my next phase if this turns out to be efficient enough is to add a high side current sense in the feedback to the fb pin. This would give constant current mode which is what we want.
 

papasan

Enlightened
Joined
Mar 25, 2001
Messages
621
Location
Northern Virginia
Re: Maxim\'s 1674 IC as a LED driver...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:
As I posted in the other thread my next phase if this turns out to be efficient enough is to add a high side current sense in the feedback to the fb pin. This would give constant current mode which is what we want.<HR></BLOCKQUOTE>

how much room would this take up?...i've got a .5"x.5" board designed, but i'm waiting to get some resistors to lower the output voltage a bit and test everything before i etch it...it's got some room for a small tantalum input cap, but not a larger EL cap...i've found when using NiMHs an input cap is unneccassary...i haven't used alkalines enough yet to decide wether or not they're usefull enough to incorporate into the design...

all of this is just from 'playing around'...i have no electrical engineering training, just what i've picked up here and a little in books and on the net...
 

bikeNomad

Newly Enlightened
Joined
Sep 8, 2001
Messages
166
Location
Stanwood WA
Re: Maxim\'s 1674 IC as a LED driver...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:
As I posted in the other thread my next phase if this turns out to be efficient enough is to add a high side current sense in the feedback to the fb pin. This would give constant current mode which is what we want.<HR></BLOCKQUOTE>
How would you add the current sense without adding an op amp?
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

Lets see if I can get this image to post. Here's the schematic if you are interested. If you want the simulation file, email me and I'll send it to you. You will need the Linear SwitcherCadIII program. It can be downloaded from their web site.

lt1308.jpg

Dang it again. Hmmm, kinda difficult to figure out how to include a picture. Oh, well, the temporary web link is.

This doesn't work either. Hmmm, Some help here. I need someone for advise! AWKKK

The external 100K resistor sets the gain of high side current sense to 40. With the 0.085 ohm resistor scales the 350mA of current to 1.2 Volts to the fb pin of the LT1308.

Of course, the smaller the current sense resistor the less power or loss in the current sense resistor. There is a trade off with making the current sense too small and the offset and gain of the high side current sense amplifier. Also, increasing the gain of the current sense amplifer also lowers the frequency response of it. Combined with the load capacitor and output current the RC time constant on the Vc pin nees to be adjusted. I found this part particular finicky in changing output capacitors, load, and input voltage. You must try to keep the design to a limited range or it gets unstable real quick.
ADDED LATER: Using an external resistor incurs a 100 or 400ppm??? with temperature. See the data sheet for more info.

Edited later. Hey thanks Wayne. I remember now You can't use Geocities to reference to.
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

Here is some prelimary data from my MAX1676 Converter. I measured 92% earlier today on the bench with a LS simulator load. 4 1n4001 and 1 ohm resistor.

For the following data I'm using two Sanyo NiCAD batteries fully charged. Rated at 1.2A but I know these batteries are pretty crappy.

Anyway...

Code:
I double checked these using a power supply and got similar numbers. So, for now, I'm switching back to a power supply which means I can get more runs per component change. I should be able to test different inductor combinations and re-test in 1/2 hour or so and get a good idea of relative improvement or not.

UPDATE: Heres the schematic.

max1676.jpg


CSEL Moved to Vout to select 1A current limit: Cin and Cout different values than shown on schematic. L, tried different values. Run time table above was done with 3.3uH CoilCraft D03316.
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

I had serious ground problems. I finally got it stabilized and once stabilized I got consistent results. More like what the I'd expect from the data sheet.

Code:

There is no run time listed here since I posted in the last message I'm using a power supply. My measurements have show that the power supply and a set of NICADS gave similar results.

I think I can tweek the eff a bit more.

I'm using two 3.3 DT3316 with 0.04 ohms MAX DCR. My next step is try using just one.

BTW: I just cranked it down. The LED stays on all the way down to 0.1 Vin. Holy smokes. Of course it is just a dim glow.
 

bikeNomad

Newly Enlightened
Joined
Sep 8, 2001
Messages
166
Location
Stanwood WA
Re: Maxim\'s 1674 IC as a LED driver...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:
Lets see if I can get this image to post. Here's the schematic if you are interested. If you want the simulation file, email me and I'll send it to you. You will need the Linear SwitcherCadIII program. It can be downloaded from their web site.<HR></BLOCKQUOTE>
I use SwCadIII. I wonder why you're using R2 (the 1 ohm resistor) if you have current feedback. R2 by itself would reduce the efficiency by about 2.5% or so.

Did you see my schematic where I got current regulation with the LT1308B without using the external current sense amp? I just used the 200mV LBI comparator, and ended up with what is basically a hysteretic converter.

See
my SwCADIII schematic here for details.
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

BikeNomad,

I saw your schematic. The diodes and 1 ohm resistor is the "LS Simulator" It's to model the LED. It's not part of the circuit. It's part of the load. Sorry if I didn't draw that very well. I was just doing some simulations and looking inductor waveforms. I was playing around with various inductor and capacitor combinations before firing up the MAX1676. It help clear up some of the possible combinations before hand.

If you compare my schematic against yours your 200mV current sense can be compared to 30mV for my current sense. The High side current gain reduces this by a factor of 40 and should thus improve efficiency a bit, but, if it works on this design then my plan is to add it to any Voltage boost converter and hopefully tame it into a boost converter (current mode).

I hope this clarifies the circuit.
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

papasan,

your board image isn't there. I'm curious since I see some problems with mine and have not been able to get the beast under control. I have stable waveforms but the IC is behaving incorrectly. I'd like to compare my layout with yours to see if I made any errors.

Here is my layout.
max1676_pcb.jpg
 

remuen

Enlightened
Joined
Nov 26, 2001
Messages
600
Location
a place surrounded by Europe
Re: Maxim\'s 1674 IC as a LED driver...

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

I have stable waveforms but the IC is behaving incorrectly.
<HR></BLOCKQUOTE>

These booster circuits behave sometimes very curious when the main components ic, inductor, in- and output cap (+ schottky diode if existing) are not really very very close together. It looks like there is still some place for improvements in your layout.
 

papasan

Enlightened
Joined
Mar 25, 2001
Messages
621
Location
Northern Virginia
Re: Maxim\'s 1674 IC as a LED driver...

i go through my server and clean it occasionally...sometimes i clean it too much =)...

1674%20board.gif


this layout is 1/2" by 1/2", nice and small...but looking over it it leaves no room for what i consider one of the nicest reasons for using this chip...using a pot for variable output...so it might have to grow a little...it also requires a home-made coil, as the coil pads are fairly small, and leaves little room for an input cap...size does have it's price...

what's your circuit doing thats odd?...first thing i would look at is the coil, that's probably your biggest single variable...you using an LS or a simulated LS?...also, i'm not very familiar with the 1676, so i can't really comment on it...i assume that R3 and whatever the through-hole part that's next to it are correct...
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

From the data that you've publisbed I see your efficiency is always above 80%. That's to be expected according to the data sheet of this part. I'll attach a scope waveform later when I get home. But, Looking at the LX pin I see a condition where during Toff the inductor runs out of current and it rings. This is the OK portion, The IC senses zero current and switches to the Ton state. Again OK so far. The Ton state is governed by two rules, one it reaches the 1AMP limit or two it times out at ~4uS. I see that 500uS later it reaches the 1Amp current limit trip. I'm using a 1uH so it should ramp up pretty fast. So far so good. Now here is where it gets strange, the IC switches to the Toff state and turns on the PFET to dump the Inductor current to the output. Again so far so good. But, 200nS later the IC appears to sense the the inductor current has reached zero and falses trips back to Ton and starts charging the inductor again, but, since not all the energy has been dumped, it does a short Ton and when it reaches 1Amp again, it reverts to Toff. This time the Toff goes all the way to zero. It kinda double pumps the Toff state.

Now this doesn't ruin anything, and it still regulates the output quite well. But the switching takes time (~4nS) and each time it switches is lost energy. The more it switches the more losses associated with it.

So, when it behaves I get 82-90% efficiency, when it hiccups the efficiency drops to 75-78%. That's like a car engine misfiring. It's not good.

I've collected my schematic, layout and scope traces and have sent them to Maxim tech support for help. I'm still waiting for a response.

I also made my board spaced out somewhat so I can try different capacitors, inductors and such. This is my proto and if I can get it stable I'll make my 1st stab at a refined version that is small and compact. The ground is super important for this and most switchers. I've compared mine against the Eval board and found I made a few minor mistakes. 1st the ref bypass must go to "clean" ground along with the resistor divider network. I accomplished this by cutting my ground at the top and bottom Near C4 and C3 and then connected a small wire across the ground cut at C2. This mimicks more like the eval board. The cref and resistor divider are closer in ground of the output capacitor return.

The other thing to note, is that during Ton the IC ground pin pulls a lot of current. For the eval board the ground return goes by the output capacitor and then to the right where it picks up the input bypass capacitor.

The Toff period where the inductor is discharging to the ouput capacitor, the output capacitor return is along the same path along the top and to the right returning thru the input capacitor back to the inductor to close the current loop path.

Sorry if this is sorta too technical.

Look for the scope waveforms later when I get home.
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by papasan:
what's your circuit doing thats odd?...first thing i would look at is the coil, that's probably your biggest single variable...you using an LS or a simulated LS?...also, i'm not very familiar with the 1676, so i can't really comment on it...
[/QB]<HR></BLOCKQUOTE>


OK, here is the scope waveform of the LX pin and the output capacitor node.
lx2_annotated.jpg


The MAX1674/5/6 are all the same family of parts. The main difference is the 1674 has a fixed 0.5A switch current limit, the 1675 has 1A and the 1676 can be selected to be either 0.5A or 1A.

I chose the 1676 so I could build two models in one. A high efficient 1-N 5mm white LED driver (Nichias) or with 1A a DC/DC similar to the Zetex to drive the Luxeon.

In the waveform diagram, point A is where the inductor runs to zero in the Toff state and rings out. I removed the snubber resistor so this shows up. At point B, the IC switches to the Ton state turning on the internal NFET and you can see lower waveform rise since this is the I*R voltage across the NFET. As the current increase so does the voltage drop across the FET. It reaches the 1Amp in 500nS later in the center of the picture. It then switches to Toff, turning the internal NFET off and the PFET on. You can see the output (green) waveform rise as current is transfered to the output capacitor and load. At the end of B the IC for some reason thinks the inductor current is zero and switches back to Ton state. it quickly finds it hits the 1A limit since there is still inductor current and then at C performs a full Toff period.

I am using three different loads, A 15 ohm resistor, 4 1n4001 and a luxeon. I've been putting different loads and looking at the resulting efficiency. Currently there is the 15 ohm load on it. It seems to do this almost irregardless of load, input capacitor, and output capacitor.
 

papasan

Enlightened
Joined
Mar 25, 2001
Messages
621
Location
Northern Virginia
Re: Maxim\'s 1674 IC as a LED driver...

yep...over my head =)...

one thing, the 1675 is the .5A chip, the 1674 is rated at 1A output...

gotta get me a scope...what's your set-up?..a DMM with a computer hook-up?...gonna check out ebay right now =)...

MrAl, you out there?...in the zetex thread he was talking about odd waveforms with some inductors...part of the reason for the hand-wound ones (besides the efficiencey boost)...try a different coil, even if it's not rated the same, and see what happens...

i can't imagine why the internal logics of the maxim chip would turn on/off like that...

also, try removing the input cap all together, run a bare minimum set-up and see what happens...if you jump the output to FB with no resistors it's set to 3.3V out...i ran mine like this just to check it out...make sure you have a decent output cap tho to keep the ripple down...and cool your luxeon if you can...
 

ElektroLumens

Flashlight Enthusiast
Joined
Jan 5, 2001
Messages
1,565
Location
Cottage Grove, Oregon, USA
Re: Maxim\'s 1674 IC as a LED driver...

Hello papasan,

I have just designed a PCB board using the MAX1674. For the present, I'm going with a one inch disc, as that meets my immediate size needs. I have most of the components, but I want to use a potentiometer for R5. What resistor values are you using for R5 and R6? The specs indicate to use less than 260k for R6. I am looking at 200k for R6, and a 500k trimpot for R5?

There is a lot of room to downsize this circuit. Looks pretty simple, although the MAX1674 is really small!

Wayne

grin.gif
 

papasan

Enlightened
Joined
Mar 25, 2001
Messages
621
Location
Northern Virginia
Re: Maxim\'s 1674 IC as a LED driver...

R6 has to be <= 260K...then you figure R5 by taking what you use for R6 (i use 200K) and the desired output voltage...

R5 = R6 * ((Vout / 1.3) - 1)

or

261538 = 200000 ((3.0 / 1.3) - 1)

so if R6=200K and R5=261K then you should get 3.0V out...i haven't tried any of this out yet, the local radio shacks didn't have a very good selection of resistors unles i wanted to buy the various 100 pack thing...they're only like $.10 a piece from digikey, and i'm holding off until i get to $25 for teh free shipping, so...

as for a trim pot..increasing R5 makes the voltage go up...R5=230K is 2.8V, R5=261K is 3V...so i would probably get the R5 value for the max you want to send to the LS then subtract say 100K from that and get a 100K linear taper pot and put it in series with R5...so say you want 3.2V max, that's R5=292K, i would try an R value of 192K (which is 2.5~2.6V out) and put a 100K pot in front of it...
 

ElektroLumens

Flashlight Enthusiast
Joined
Jan 5, 2001
Messages
1,565
Location
Cottage Grove, Oregon, USA
Re: Maxim\'s 1674 IC as a LED driver...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by papasan:
R6 has to be <= 260K...then you figure R5 by taking what you use for R6 (i use 200K) and the desired output voltage...

R5 = R6 * ((Vout / 1.3) - 1)

or

261538 = 200000 ((3.0 / 1.3) - 1)

so if R6=200K and R5=261K then you should get 3.0V out...i haven't tried any of this out yet, the local radio shacks didn't have a very good selection of resistors unles i wanted to buy the various 100 pack thing...they're only like $.10 a piece from digikey, and i'm holding off until i get to $25 for teh free shipping, so...

as for a trim pot..increasing R5 makes the voltage go up...R5=230K is 2.8V, R5=261K is 3V...so i would probably get the R5 value for the max you want to send to the LS then subtract say 100K from that and get a 100K linear taper pot and put it in series with R5...so say you want 3.2V max, that's R5=292K, i would try an R value of 192K (which is 2.5~2.6V out) and put a 100K pot in front of it...
<HR></BLOCKQUOTE>


Hey papasan,

Thanks for the feedback (pun). After I made that post, I did the calculations,and pretty much came up with the same thing. My wife and oldest son are wizards at math, and they also came up with similar results. I just do not like buying parts, and then finding out I purchased the wrong stuff, then have to order again, and then wait again, etc.

If I get a chance, I'll post the PCB design I came up with. You can tell me what you think of it. It is one inch in diameter, as that is a good size for me. I designed the circuit in a way that a home made inductor could be used, and mounted on the opposite side of the rest of the circuitry. I am sure the entire cicuit could be shrunk quite a bit.

Before I attempt to solder this tiny IC, I need to get a different tip for my soldering iron, as the one I have is too big.

grin.gif
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

Ok, I haven't posted anything new for a week or so. But, I've been working real hard on this one. I've now built three different PCB layouts, and 4 boards. I've used oneMAX1676, and two MAX1675 ICs. I was actually able to move my MAX1676 from the 1st board to the second. I was so lucky too. As I was trying to apply lifting pressure with the tweezers and apply heat to the one side of pads it went twinggggggg. Soooo lucky it popped a few inches away. If it had jumped the work bench to the carpet that would have been the end of that.

From talking with insiders from another Manufacturer, ground pins can be very sensitive to spikes. Power IC's are mix of different logic. This small package and small leads are just begging for trouble. That and the almost impossible ground layout for a single sided board configuration.

The best board I've built to date is my dead bug board with the IC pin 2.3 and pin 6 bent down and soldered directly to a solid copper board. (LBI, LBO, and GND) The remaining pins were bent straight and wires were soldered directly to the pins. SMT caps were soldered directly to the copper plane and the other ends were then wired to the appropriate pins or connections. Cout and Cin were soldered to right side of the IC and the REF bypass 0.1uF was soldered to the left side.

The total construction time was less than 10 minutes.

I've been fighting switching noise getting back in and causing the IC to switch states a lot more than it should. Although regulation is OK, efficiency is poor.

I've been trying to keep the components values small to make the smallest possible layout. But, I probably went to far and finally revisted the whole value scheme. With a 33uF LOW ESR capacitor on the output and using the ZLT+ ferrite bead core for the inductor I finally achieved one charge, one discharge state per cycle. I measured efficiencies

VIN Iin Vout Iout EFF
2.729 .503 3.187 .406 94%
2.43 .478 3.109 .348 93%
.922 .419 2.879 .113 84%

This is the 0.5Amp part and not the 1A part.

So, I'm not clear why it's driving the Luxeon so well. Probably because the efficiency curves are for 3.3 Volts output, with Vout @3.1 I can squeeze a little more juice out of it. I have the Iout equation graphed for Vout 3.1 and it shows it should maintain 3.1 and .35A in regulation above 2.4. Below that Iout starts going down with Vin. That's still not bad and might be a good attribute for a battery like NiCAD and NiMh since they don't change voltage that much. And when they do, they can't deliver as much as they get depleted.

Now if all the LEDs were 3.1 Volts I would be set.

This layout has fb tied to Vout for 3.3 Volt operation. Of course it never got to 3.3 and so it was putting out the maximum possible.

I'm working on my REV3 board. This one I hope will do the trick and keep the ground quiet.

I also tried my own hand wound toroid core. it to appears to work OK. The size of the toroid is still larger than the ferrite bead.

I still don't understand how the ferrite bead works so well. Maybe, I shouldn't think about it and just go with the flow.
 

papasan

Enlightened
Joined
Mar 25, 2001
Messages
621
Location
Northern Virginia
Re: Maxim\'s 1674 IC as a LED driver...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:
With a 33uF LOW ESR capacitor on the output and using the ZLT+ ferrite bead core for the inductor I finally achieved one charge, one discharge state per cycle. I measured efficiencies

VIN Iin Vout Iout EFF
2.729 .503 3.187 .406 94%
2.43 .478 3.109 .348 93%
.922 .419 2.879 .113 84%

This is the 0.5Amp part and not the 1A part.
<HR></BLOCKQUOTE>

nice efficiencey...can you give us more details on your component selection?...
 

dat2zip

Flashlight Enthusiast
Joined
Jan 5, 2002
Messages
3,420
Location
Bay Area
Re: Maxim\'s 1674 IC as a LED driver...

I have a generic 100uF 10V on the input since I'm running off a power supply the input capacitor is needed. A 0.1uF ceramic on the REF input. The output is a 33uF Sprague 33uF 595D type capacitor, 1uF ceramic and a 0.1uF ceramic on the output. The load was one of my 3 luxeon /O White. Source was a power supply. The inductor was the ZLT+ special.

My rev3 board with optimum ground and Cin, Cout butts soldered together on a common ground point near the GND pin of the IC and it still glitches. So, the solid copper dead bug is my best performing of all 4 layouts.

I'm getting close to a clean layout. But, I don't think I will get it perfect. The question will then be how much does it glitch or misfire and how low will the efficiency drop. When I see it ramble or misfire, the additional switching can be up to 3-4 addtional on/off cycles ~300nS in length and occupy 30-40% of the total cycle. So, a rough guess would be the additional switching losses will probably be around 10-15%. That is too much reduction. Especially for a more expensive solution to begin with. It would be ideal to get a clean layout and make the IC happy. Then we would get the max efficiency out of the part. It got promise, but, the higher speed switching, small package and singular gnd pin is going to take some more work. I'm already on rev3 layout. I'm thinking now double sided with total ground on the backside.

The only good news is my garage factory is setup and I am cranking out about one new board a day.
 
Status
Not open for further replies.
Top