Max 1674 Operating Temperature

I have a max1674 voltage regulating chip driving an LS/O. It's set up for variable voltage output. I'm measuring current across a 1.2 ohm resistor.

At 3.07 volts the LS is drawing about 250 mA and the max1674 chip is at about room temperature.

If I raise the current to about 300 mA, the chip starts to get hot to the touch.

This chip has a current limit of 1000 mA. What do you suppose is it's maximum safe operating temperature?

The data sheets talk about a maximum ambient temperature of 85 deg. C. (184 F)

Originally posted by jeff1500:
I have a max1674 voltage regulating chip driving an LS/O. It's set up for variable voltage output. I'm measuring current across a 1.2 ohm resistor.

At 3.07 volts the LS is drawing about 250 mA and the max1674 chip is at about room temperature.

If I raise the current to about 300 mA, the chip starts to get hot to the touch.

This chip has a current limit of 1000 mA. What do you suppose is it's maximum safe operating temperature?

The data sheets talk about a maximum ambient temperature of 85 deg. C. (184 F)

Click to expand...

<font size="2" face="Verdana, Arial">Well, as I'm sure you already know, the datasheet absolute maximum Tj is 150C. I would try to keep it under 125C for reliability reasons. The tough problem is that it is often hard to calculate what actual junction temps you are running.

jeff1500,

The Max1674 efficiency is very high. As such, if it is running correctly dissipates nearly no heat. The operative word is if. You can easily measure the input voltage and current and ouput voltage and current and determine if you are obtaining the data sheet efficiency. You should be able to get up to 88% efficiency at 300mA output current.

I have many cautions and notes regarding the Maxim167X series and how difficult it is to tame this beast.

You can look at my web pages to see the various revisions and some of my design notes to see if that will help you.

Maxim 1674 design notes.

If the ouput capacitor is not a ceramic, the output capacitor is not 0.05" next to the top of the IC, if you don't have a solid copper plane encompasssing the whole IC, you have not kept the switching current paths contained to the input loop and the output loop properly and the inductor you are using is not saturating. You will find that the IC doesn't get but mildly warm if done correctly.

I'll keep trying to tighten it up and try to measure the input current.

Is there a relationship between efficiency and capacitor size?

250 mA makes quite a bit of light, so I'm going to say, not too bad for my first surface mount project.

What would be a better chip? I've read all about the famous Zetex experiment. Is there another one that people like?

The LM2621?? if I remember correct is another good candidate for a boost converter. The LT1930 has been used successfully for 5mm (Nichia designs).

It really depends on your requirements. If size is not so important you can get many boost eval boards and they would work just fine.

Most of the Boost ICs out there are voltage regulators and you have to adjust the voltage and use a low value sense resistor to set the LED current. This is more than acceptable in a one time design. Making a whole handful of these you will want to figure out the best way to make a curent mode regulator and not a voltage mode regulator.

Linear technology has the free Switch Cad simulation software which is what I would start with and do some simulations first with different Linear parts.

Usually the Linear parts behave quite well and if you can find a good linear combination that works you are all set. The Switch Cad softwre is invaluable since you can try different inductors and capacitors of from real manufacturer part numbers. This will save you countless hours of frustration since no one person wants to buy dozens of different values and types from each of the component manufacturers.

Originally posted by jeff1500:

Is there a relationship between efficiency and capacitor size?

Click to expand...

<font size="2" face="Verdana, Arial">I tried them both a 22 uF and a 47 uF capacitor for Cout.

My meter isn't really all that accurate so I can't say much about efficiency differences.

However, at the lowest voltage where the LS just starts to light up, about 2.4 volts, I got a more steady light with the 22 uF capicitor. It blinked a little with the 47 uF capacitor until I got up to about 2.5 volts. Maybe the electicity goes into the capacitor instead of the LS.

Either that or it was a loose connection.

Originally posted by Doug S:
Well, as I'm sure you already know, the datasheet absolute maximum Tj is 150C. I would try to keep it under 125C for reliability reasons. The tough problem is that it is often hard to calculate what actual junction temps you are running.

Click to expand...

<font size="2" face="Verdana, Arial">Does this mean it'll run happily all day with a chip surface temperature of 100 degrees C?

Originally posted by jeff1500:
Does this mean it'll run happily all day with a chip surface temperature of 100 degrees C?

Click to expand...

<font size="2" face="Verdana, Arial">Hi Jeff

The temperature itself is no problem even though it reaches 100°C as these IC's can bear a temperature up to 150°C. The main problem is more that the IC gets so extremly hot with such a low current. Maybe there are some bouncing effects or whatever else. As Wayne Y. already pointed out this is almost for sure caused by your pc board layout as according my own experience all these Maxim chips ask for a very carefully layouted board. I myself usually need two or even three various layouts until I get the expected results. You should carefully read the design notes on the datasheet - and I would suggest you also should take a look at Waynes boards ....

Ah, just goop it with thermally conductive epoxy!

I've got a plan to tighten it up. We'll see if that increases the amount of current that'll pass through it.

Right now it looks kind of like a bunch of tangled up string. But it does work okay at 250 mA.

Hello there Jeff,

Your original question was very interesting,
in that you are basically trying to correlate
the case temperature to the junction temperature.
If it works out that the junction temperature
is less then 150 deg C, you are ok, or else
your design is not ok.
Now normally, when a manufacturer specifies
a max junction temperature, it has practical
significance in that you can calculate it
from the case temperature via the thermal
resistance of the package.
Since i cant find any mention of the thermal
resistance of the package on the Maxim site,
i cant calculate the junction temperature
from the case temperature. Since i cant
calculate the junction temperature, i cant
verify the design no matter what current you
are operating at.

Anyone else have any luck finding this info?

Good luck with your LED circuits,
Al

I think I've got it going. It's only slightly warm at 300 mA.

1. The max1674 is mounted on a SurfBoard adapter and plugged into a breadboard.

2. The other components are regular through hole items and are also plugged into the breadboard.

3. An LS/O is mounted on a 1/8" thick aluminum disk that's soldered into a 1-1/2" chrome/brass sink drain tube. A short piece of the brass tube is used as a retainer ring since solder doesn't stick to aluminum.

4. I got a 1.0 ohm resistor and measured it at something like 0.98 ohms. I'm measuring a steady .305 volts across the resistor.

5. I'm adjusting voltage with two 50k trim pots and a 10k potentiometer connected to the feed back pin on the chip. The 10k pot dims it from full to zero.

6. According to Brock's LED web page, an Arc LS provides 300 mA to the LS, so that was my target current level.

7. I just had it running at nominally 300 mA for about 20 minutes and the chip is only warm to the touch. I don't know what temperature, but it's in the luke warm dishwater range.

8. The sink drain tube is only slightly above room temperature. There's a heat sink pad under the LS and I think I've got reasonable thermal contact between the disk and tube.

9. I tried putting a 1.0 ohm resistor in series with the input line to measure current but the resistor started to get hot and I couldn't get reliable meaurements. It's a 1.0 watt resistor. Output was about 3.2 volts x .300 amps = .96 watts, so ya, I'm going to be more than that for the input.

Looks like something good is happening.

jeff1500,

That sounds much better. What ever you did seems to have made the circuit much better.

I'm not sure what inductor you are using but if the inductor saturates then the IC can get real hot real fast. The inductor tends to get hot too, so, if the inductor and IC are geting hot together might be an indication that the inductor is going into saturation.

What inductor are you using? It should have a safe operating current greater than 1A. It needs to be a good ferrite inductor. I've used the ZLT inductor with 3 or 4 turns with good success.

If you are going to sweep the input to see how it will regulate you will need to remove all the 1 ohm resistors as they will artficially raise the point where the IC can maintain regulation.

With the 1 ohm resistor in series with either the output or the input will force the output to deliver more power Vout = VLED + Vsense_resistor.

This would mean that if the the low Vin regulation was 1.8V without the 1 ohm resistor, you might find it comes out of regulation at 2V and not 1.8V.

Having the resistor on the input would just lower the voltage as seen by the IC. Let's say again that it drops out of regulation at 1.8V, with the 1 ohm resistor on the input there would be a 1V drop. The IC would drop out of regulation at 1.8 + 1V or 2.8V battery voltage.

Also, you should be using good size wire for the battery connection and the output to the Luxeon connections. Small fractions of a volt drop in a battery operated device have a heavy penalty cost on efficiency and battery longevity.

To do a fairly good accurate efficiency calculation you should try using two 0.1 ohm resistors. One for the input, and one for the output.

I have SMT version of the 0.1 1% accurate SMT resistors. I've been selling them 2/$1.00. That's a real steal since they go for over $2.00 each at Digikey.

WayneY
Wayne

http://edusite10.tripod.com/led3/smount/parts1.htm
The parts I'm using came from this list. The inductor is 39 uH, 2.4 amps. It stays cool.

I think I had two problems: 1) My meter wouldn't measure the 1.0 ohm resistor accurately on the ohm scale. I used another meter for better readings.

2) The breadboard connections are loose. I can make the current jump all over the place by wiggling things.

My most optimistic theory about the chip getting hot is that loose connections when near the 350 mA level made the output voltage jump and sent the current off the deep end.

I made a tight circuit board layout drawing so I think my next step is to solder it all together.

I put a 1 ohm resistor on the output and measured the voltage at the led. Then I removed the resistor and reset the voltage to the value that gave me 300 mA through the led.

I've got ten 1 ohm reistors, so I guess I can solder them all in parallel and make a 0.1 ohm set.

Sounds good. Everything you just said makes good sense.

You are absolutely correct in removing the 1 ohm resistor and reseting the voltage to get the correct current.

The 1674 has a poor internal voltage reference which changes with input voltage. So, you should expect to see some variation in output voltage with change in input voltage. The IC is regulating correctly, it's just referencing a changing reference voltage.

This is one of the most unfortunate things about most voltage regulators that are targeted for powering PDAs and digital logic. The devices they power don't care if the 5V varies +/- 0.25V over the battery life. The LED on the other hand varies it's current dramatically with a small change in voltage.

Remember to try to keep the Luxeon heat sinked well. Letting the LED heat up will change the amount of current it draws. If I recall correctly the VF has a temperature coeeficient of -2mV/C.

It's best to make your current measurement after the LED has thermally stabilized.

Wayne