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

Status
Not open for further replies.

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:
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.<HR></BLOCKQUOTE>

by generic 100uF 10V cap i suppose you mean an EL cap?...if it's running off of a power supply why is the cap necessary?...i would think a constant power source would make the input cap redundant...what's 595D, the package size?

so you have 3 seperate caps on the output?...all parallel?...what's the point of this?...do lower F caps mean less loss and somehow 3 low F caps are better than 1 big F/low R cap?...

full of questions...

it would be nice to see some eff. tests with actual batteries...and i'm curious as to any oddities your circuit has at low (<.9V) input voltages off of some dying batteries...

if 94% eff. is a reality, this circuit is awesome...
 

dat2zip

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

Yup, I knew there would be questions. OK, for the output, I tacked down capacitors as holding points like pogo pins to solder wires to. I originally had a 1uF ceramic and thought that would break away with the large wire to the load, so I added another 0.1uF 0805 soldered down next to it to add mechanical load. Later as I earlier explained I beefed it up to the 33uF. The 595D is an earlier LOW ESR that came out about 3-4 years ago. I'm not sure they make it anymore. The 150uF used on the Zetex probably has lower esr.

During the Ton period that charges the inductor, current is drawn from the input. With long wires the input looks very inductive and not a good solid DC power supply with LOW esr. The 150 makes it look that way. Hooking 2AA with short wires several inches long is very low impedance, so the capacitor is not needed.

I think I answered your questions. If not ask again. I will try to my best to answer your questions.
 

papasan

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

so when you took those readings you had two caps on the output?...a 1uF and a 33uF low ESR?...

some of the concepts i just haven't wrapped my head around yet...i'm not sure what you ment by adding mechanical load with the .1uF cap...

so to keep the switching noise down, therefore keeping efficiencey up, a fat low R low F ground is needed, is this correct?...

i suppose i'll request some samples of the 1675 ic and see how it compares with my readings from the 1674...perhaps the peak power of .5A is helping eff?...
 

dat2zip

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

Mechanical load. At the time I thought the 1uF 1206 standing on end with one side soldered to the copper clad board would not be able to solder the 22G wire and hang the Power resistor or the Luxeon on it without sheering the metal tab off the capacitor. Also, the 1uF capacitor I have only a few of so I didn't want to damage it so I added the 0805 0.1uF next to it to attach the load wire to.

Eventually I soldered down a 4 pin header and wired the output to it. So, the connector took away all the weight of the load.

The 0.5A limit. It turns out the one ZLT+ I was measuring had a peak current limit set to 0.4A. The triangle shape current waveform between the ZLT+ and the MAX1765 both had similar waveforms.

Under this condition neither IC is in regulation. Both are running wide open and the output current is now dependant on the input voltage. I was looking for the original Zetex thread but, it appears to have been deleted. I'm interested in seeing some of the data collected to see if it corrolates to what I'm measuring on the bench.

For the MAX1765 to be most efficient, it needs to perform one Ton cycle and one Toff cycle. More often than not, the MAX1765 generates many pulses during the TOFF. These additional short charge/discharge cycles are wasted energy cycles since there is lost energy turning on/off the internal switches. The MAX176X series do turn on/off much faster than the Zetex driving an external transistor so in theory the MAX176X should be more efficient. TonZetex-20nS, TonMAX176X-4nS.

The two major factors in comparing the MAX176X and the Zetex is that I think that the MAX176X with the internal syncronous PFET is more efficient than using an external schottky diode and that the MAX176X can switch the NFET and PFET on much faster.

The MAX176X with smaller leads, and higher initial cost is only worth considering if we can get better regulation or be able to do a design that does not require any adjustments. Ideally, a black box DC/DC converter to drive the Luxeon or LEDs in general would be a constant current source.

On my third layout I had ground exiting both sides of the ground pin. The inside path went up and to the right. Similar to the Eval board. It wrappped around the Output plane of copper. The other ground exited from the ground pin to the right and merged with the other ground to completely surround the output copper. This layout still glithes. The only one that doesn't glitch is the solid copper dead bug design.

Until the glitching can be stopped, I would not recommend making any prototypes unless you can look at the LX waveform to determine if yours is glitching or not. You will notice the efficiency drop if it is glitching. I suspect your 1st board is glitching.

I joking talking to a buddy of mine who is trying to help me I sent him an email saying:
"Now I know why Maxim calls this package uMAX, The u stands for u will never get this to work!".

grin.gif
grin.gif
grin.gif
grin.gif
 
D

**DONOTDELETE**

Guest
Re: Maxim\'s 1674 IC as a LED driver...

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by dat2zip:
I was looking for the original Zetex thread but, it appears to have been deleted. I'm interested in seeing some of the data collected to see if it corrolates to what I'm measuring on the bench.<HR></BLOCKQUOTE>

The thread hasn't had much activity lately, so now you have to dig for it.

It can be found at http://www.candlepowerforums.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=3&t=000983
 

dat2zip

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

OK, I got the IC under control today. My Rev3 is stable. There is a huge current spike from OUT to GND when the IC switches to the Toff state. My guess is that both the NFET and PFET are on for a brief period. I was seeing a 0.6V ground spike. Adding a 0.1uF on right side of OUT to GND did nothing. But moving it to the left side of OUT to GND above the IC got rid of it.

Tonight I'll post the fix. There is one critical layout constraint, that I think if you follow it will keep the IC happy.

Thanks Duggg for the link. I'll go back thru this and compare my readings with the thread.

A quick couple of measurements. This IC is wired with fb to output. Thus, if it should regulate to 3.3. Since my Luxeon is clamping it to 3.0-3.1 it is running full throttle and putting out maximum energy with peak inductor limit at 0.5A.

Code:

I've got my SMT 1uH inductor in there right now. I need to think about this some more about what is actually happening.

I want to experiment with many more options, see if the fb can set VOUT, and try different Output capacitor combinations. I've got a 22uF ceramic on the output and the P-P ripple is ~20mV or so. The LED current fluctuation is <10mA. SO, this combination provides more or less DC voltage and current with pretty small value components which is one of my goals.

Also, if this is stable and I add the high side current sense in the feedback, it should regulate current instead of voltage, but, we would need to use the MAX1674 since the 1675 cannot deliver enough to stay in regulation. Actually, the 1675 might be OK if you don't mind it running out of regulation early. With the current setup, it would maintian regulation of .35A at 3.0 Volts output from 2.8 - 2.75V Vin or so. Below that as Vin drop so would IOUT. At this point it would be just like the Zetex IC as far as I can tell.

Updated: Look at the MAX1676 Eval board and the component placement for C3. Look at the top side copper. If you follow this pattern for C3, you should be able to keep a clean layout.
 

dat2zip

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

OK, I made a web page with my observations and pictures, graphs and notes. I thought it would occupy too much to try to include all this on a discussion group.

The URL is:
MAXIM 167X design notes.
 

papasan

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

wow...alot of work...gonna take me awhile to digest it all...

so you hook your load up to the clean ground side?...and your power supply to the 'dirty' side?...

looks like you were trying to save some file space on a picture or two...if you convert the BMP files to GIF files they'll be alot smaller since they only contain 3 colors...

i'm a little confused (surprise!) on a couple things...looks like FB and REF aren't grounded at all?...and it looks like you have your coil going to out instead of in...

nine components...deffinatly less attractive than the five of the zetex...specially considering the cost and integral transistor and diode...
 

dat2zip

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

I think you are referring to the rev3 picture. The ground strip on the far left is connected via the trace running between R3. You probably didn't see this.

Since the high current switching spikes are contained locally to the capacitors, the load doesn't need to be connected to the dirty side, although it is recommended.

the load can be tied to any GND location, preferably in a way that doesn't draw any current thru the clean section.

The input power and ground should come directly to the input capacitor. Connecting the input ground to some other location will cause ground loops and that would not be good.

Thanks for the input on the pictures. For now it's there temporarily to help others. I'm not sure I'd leave it there. Do you wish me to make that permananent?

I think that a few additional % efficiency is worth the exercise. Whether or not the design gets used is rather up to the indivual.

Not that I own a cadillac or Bently, but, you never know what someone would pay for the extra battery life. (not that it gains you much). Hmmmm, I hear the marketing telephone ringing...
shocked.gif
 

papasan

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

i see both halves of the ground ('dirty' on the right, 'clean' on the left) connected with a small trace that passes beneath R3...but the only way i see FB and REF getting to ground is through C5...a filtering cap?...but then why would a filtering cap be on the main out of the chip?...

ahh...is that a trace under the 'C' in C3?...

so any real tests with a luxeon and some NiMHs yet?...would be nice to see a run-time efficiencey chart like the ones here or in the zetex thread with your lay-out...
 

dat2zip

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

Run time. I suppose. I just measure efficiency. Yes, that is a trace to gnd under the silkscreen C3. Sorry about that.


This is phase 1 for me. I'm trying to add the high side current sense IC and make a solid current regulator output. I think the MAX1674 1A part will do the job nicely but with less efficiency.

Then I'm off to do a secret DC/DC converter. Can't disclose my 100% efficient gains energy back converter.
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, some real data from Sanyo NiCAD batteries rated 1.2A/HR. Not very good batteries, but it gives us an idea.
Code:

Looks like one of the batteries went reverse just below the 2V mark, The input voltage plummeted so fast it was hard to take measurements. Some of readings are bogus because of this.

The initial material list looks something like:
Code:
Total material cost not including PCB and labor would be $9.08. Using the M2308 Ferrite bead the cost would be $6.92.

The material list for the ZLT+ is
Code:
Total $6.81.

It's comparable, since the Zetex IC, transistor and schottky diode add up or about equal to the Maxim IC. Looking at volume pricing the MAX1675 is $1.81/1K and the Zetex parts add up to $1.72 or less since Digikey doesn't show 1K pricing.

This is just a comparison. Not biased pro or con one way or another.

EDITED LATER: Another battery run except this time with a fully charged set of Nexcell 1600mA/Hr batteries.

It ran almost full tilt for 2hours a tad more than the other set, but, it continued for another hour in diminished mode all the way down to 0.11V at which point I could no longer see the LED, but there was still output voltage of 2.2V with 0.11 Volts in.
 

dat2zip

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

I built Revision 4 now. I have one loaded with the 1674 and one with the 1675. I think both are OK. They do glitch quite a bit, but, not from the OUT to GND current spike. It appears to be input cap sensitive.

See my web page for details.

Tomorrow I'm going to flip C4 and see if that fixes the problem.
 

papasan

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

considering your observations that the grounds need to be isolated somewhat i redesigned my board...here's one i made up quick just to get some feedback on...

1674-5.GIF


the top left (FB) and top right (Vout) would be jumped by an outboard pot (SMD pots are hard to adjust when inside a case)...no place to solder the inductor's other lead onto the board, but i could expand it a little and add a floating pad...

i'm still not clear on why you use 2 output caps, dat2zip...the kemet cap has been shown to keep the ripple down to acceptable levels...are you doing this just because that is how the data sheets are or is there some other reason?...

so...any thoughts?...
 

dat2zip

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

I finally ran the numberes thru for the 1675 and the 1674. As I see from the data sheet and suspect I^2 R FET losses are worse at 1A the 1674 with the higher peak current is more lossy.

I measured a working 1674 and got very similar numbers to your 1st layout.

So, I think you can get away with your configuration with no input capacitor as with the ZLT+ project. But, if anyone is running input wires longer than a couple of inches they will need to put a 1uF ceramic or larger capacitor on the input to keep the IC happy.

Apparently your 1st board must be working fine! I got almost identical numbers to yours. My batteries are showing they aren't as good as they once were (NEXCell 1600mA).

putting the pot off board increases the fb node and risks coupling in noise and causing unstability.

If you build the board and experience similar efficiency numbers to your 1st proto, then, you are OK.

I think the ground spike from out to GND will be fairly large on your new layout, and your clean ground pickup is right in between the two path. for the MAX1674 adding the schottky diode helps reduce some of this spike current.

Look at my revision 5 on my web site. I detail some layout notes. See if they make any sense before making a board.

I just finished building the rev 5 and it looked very clean. The small 1608 inductor is sooo cute and small. The whole rev 5 is around the 14mm size of the ZLT+. Plus being adjustable I can turn the LED brightness up and down with a small small small tuning tool.

I found the 3mm SMT pots over at the surplus store.

The 1A part is much harder to tame, and I need to see if my rev5 with a 1674 in it is stable.

As far as two ceramic output caps. I put two footprints since some DC/DC converters need a large output capacitor to control ripple and a small one to suppress the transient spikes. So, I kept both which allows me to experiment with different values and see the tradeoff of ripple and transient noise suppression. (As it turns out I ended up with only one on rev5)

With larger 1A current the output ripple is much larger on the 1674 than it is on the 1675. I'm not sure if you are shooting for overdrive but, I'm starting to like the 1675 more and more.

By design or by luck it tends to run full tilt for the Luxeon and cannot overdrive an LS and is thus self limiting.

Plus, by not switching peak current of 1A, smaller inductors can be used making a smaller design and more efficient as well.

I've got a post going trying to gather LED voltage drop at .35A to see what voltage distribution the LS actually have. Seems they are around 2.9 - 3.2 Volts which is better for the 1675 since it cannot deliver as much current as output voltage goes up from 3.0 to 3.5V. As output voltage goes up MAX Iout goes down and vice versa.

The other real nice thing about the 167X series is they regulate voltage. Initially I was going to convert the Voltage mode to current mode, but, one major problem with current mode regulators is open load condition. The output voltage rises to infiinty or to some high value and typically destroys the output capacitor or blows the LS if you hook it up with the output storing such a high voltage charge.

No load on the 167X regulate to Vout. So, if you had it dialed to 3.2 Volts, a no load condition would be 3.2 Volts. Not 20 or 30 Volts.

There are those that want MAX brightness, and the other extreme of people want max battey life. For MAX battery life, the 1675 is better than the 1674. For overdriving one LS the 1674 is way better.

I think the rev5 can serve both purposes depending on which IC you load and the associated inductor. (1A peak VS .5A peak).
 

dat2zip

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

On my web page I'm starting a desciption of the details of the Zetex and MAXIM 167X series. It's based on some formulas and measurements I've taken on different units.

It's mostly for my reference and I'm sure some of the CPF members might get a headache reading this. Others will find it useful.

I've computed the MAX Iout equation for the Zetex IC and will be updating the web site with more details and booring graphs.
 

dat2zip

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

Well my direct CDRW that I transport back to and from work decided to puke on me today. I've lost all my work. Calculations, formulas, graphs, notes.
mad.gif



The good news I've done a MAX1676 with all the features. It can do the 1A, 0.5 mode, be voltage regulated or run wide open and it has emargency mode that can be enabled that switches in below 2V (resistor settable). The emergency mode is like the ARC moon mode and I don't want to infringe on ARC termalogy so I'll call it emergency mode. I'm currently making 4 of these eval boards with a customer list of 3 + 1 for testing by me.

What details I have are:
DC to DC comparison Zetex, MAXIM 1675, MAXIM 1674


I've got a total of 7 samples so, if anyone is interested I'll have 3 more with jumpers, switches or something to configure the bugger. It's not intended to be used on a flashlight mod since I have my rev 5 that is much smaller and I plan on making more of them. This 1676 is an eval mode since you can test different configurations to optimize your flashlight mod, duration, brightness, emergency mode etc. There are just too many variables to try to make individual ones. I don't have a price yet, but am thinking of $25 for the eval board.

I'm also thinking of sending three inductors with it. The ferrite bead, 10uH, and a 22uH inductor. The 10uH will handle the 1A peak switch current and is the actual part I would be using on the 1674 board. The 22uH can handle the 0.5A and would be the actual part used on the 1675. The ferrite bead is an option but increases the board height.

I've got some data and graphs of full throttle and regulated mode on the side by side comparason on my web site if you need to figure out which one will last xxx long for your particular application.

I enjoy making the boards, testing them so I'm offering the service to CPF members that don't have access to solder the uMAX tiny footprint and components.

My current list is Elektrolumens, Jollyroger, and Lambda express interest.

If CPF members want me to make the 1676 a regular item my arm is sticking out to be twisted... I just need to know if there is enough interest so that I can buy from Maxim at the 25 piece qty to get the lower price break. Plus the jumpers or switches need to be added to the physical layout.

Also, the 1676 I expanded the size since I wasn't concerned about making it small, so the board is .8" in diameter and several inches square after adding the headers or switches.


My earliest ship dates would be mid to late March for those not on the first 4 list. Maybe sooner if I get my Coilcraft order in.

I would prefer you request via email to
[email protected]
 
Status
Not open for further replies.
Top