Problem with voltage regulator LM338

[rumpfy]

With all semiconductors, it is important to read the data sheet.
It would appear that the device you are using is a LM338T.
Your application wants to reduce a 15 volt supply down to 6 Volt at a current of 5 amp.
The device dissipation in this case is 45 Watt.
For this device, the TI data sheet says:
"These specifications are for power dissipation of 50W for the TO-3 package, and 25 Watt for the TO-220 package."
The thermal resistance (j-c) is specified at 4 degree C per watt.(note NO decimal place). So I will take that to be 4.4 degree C per watt for the TO-220 package.
The maximum junction temperature for the LM 338 is only 125 degree C. (for the TO-3 package, the thermal resistance (J to case) is 1.4 degree C per watt.)
Under the operating conditions you want, the temperature rise of the device will be 180 degrees C. There is NO WAY that the TO220 device you have selected will come anywhere near doing what you want it do unless you cool the case to -55 degree C.
You have several options:
1. include a series resistor between the power supply and the LM338. My feeling would be to use say 1.2 ohm to an in/out differential of 4 volt. The dissipation in the first case would be 25 Watt and in the second case would be 20 Watt. but remember the data sheet values are given for an input/output differential of 5 volt.
2. Use an external pass transistor, although in the data sheet, TI use multiple LM338's for current greater than 5 Amp.
3. reduce the in/out differential to say 2.5 volt but I'm not sure how the LM338 will behave at this differential.
4. Give away the plan to use a DC regulator and use a switching regulator instead.

The thermal design for the LM338, shows that in your case, the power dissipation is 25 Watt. Since the maximum junction temperature is 125 Deg C, then the temperature rise above ambient (say 40 deg C) is (125-40) = 85 degree C. The required thermal resistance (junction to ambient) is 85/25 =3.4 degree C per watt. Since the device has a thermal resistance (junction to case) of 4.4 degree C per watt, then the device is unsuitable for the application unless the case temperature is held at 15 degree C. Using the series ballast resistor, the power dissipation is 20 Watt, and the required thermal resistance is 4.3 degree C per watt. Even with the 1.2 ohm ballast resistor, the device will need an infinite heatsink (zero Degree C per watt). This is not possible, so you are going to have to use more than a single LM338.
Using the TO-3 package, the thermal resistance (J to case) is 1.4 degree C per watt). The required thermal resistance of 4.3 degree C per watt allows the heat sink thermal resistance to be 4.3 - 1.4 = say 3 degree C per watt.
Looking at your photo of the heat sink, my guess is that it is probably about 8 to 10 degree C per watt. To get the thermal resistance down to 3 Deg C per watt needs a LOT of area AND black to boot.
Check what the suppliers RS or Farnell (Element 14) have to say. These suppliers have decent trustworthy data sheets that can be relied on for DESIGN.
Hope this saves a lot of heart ache.

 

[konradIC13]

Ok, i see it now, for now i dont need a perfect voltage regulator circuit, i just need something for now, i dont have nerves and time to make a switching dc-dc regulator and make PCB for it, making linear DC regulator is easiest option right now ( you know, i have the parts, chips while for switching one i would have to make an order for stuff because there electronic shop in my city is like "resistors and diodes only"). I might start designing a DC-DC switching converted after i finish this for future use, to have something more reliable.

So back to topic, how about this, i would use 2x LM338 in series, the first one will be set with R2 to drop voltage from 15 to 10.5 (that will give 22.5W with 5A (can drop from the 5A requirement) and second will drop voltage by another 4.5V (same power dissipation) resulting with 6V. Note, the current of 5A is absolute maximum but not constant, from my predictions the usual constant current might be in between 2-4A so the power dissipation will be a bit more lower.

Does resistors next to D2 and D4 should have higher power rating or they can be regular 1/4W ?

 

[rumpfy]

Konrad,
Can I strongly recommend that you get hold of the data sheet.
From my reading of the data sheet, you cannot dissipate 20 watt of power in this device unless the case temperature is held at 40 degree C maximum. Given 5 amps and a voltage differential of 5 volt there is 25 watt of power to dissipate. Your suggestion to use 2 LM's in series does nothing to solve your problem because each device will have a voltage differential of 5 volt. I dont know if you can operate the LM at a voltage differential of less than 5 volt. To get the device dissipation lower, you will have to either reduce the voltage differential OR reduce the current in the device. To get to 10 Watt dissipation, the differential voltage will have to be reduced to 2 volt and this is the minimum for most series pass regulators.
The maximum current that you can really pass in the LM338 is around 2.5 amp. The data sheet, page 17, shows the arrangement.
Reducing the current to 2.5 amp at 5 volt differential gives 12.5 watt. The temperature rise, (j to ambient at 40 deg C shows a thermal resistance of 6.8 degree C per watt. Given the LM338 has 4.4 deg c /watt allows the heatsink thermal resistance to be 2.4 degree C per watt. This is still a big heatsink but possible.
I checked the RS catalog. They have a stock nr. 507-5021 which has a 2.5 deg C /watt rating. It is described as "TO220 extruded 2.5 C/W". It has a flat section for bolting TO220 devices. It's size is 75 mm L x 70 mm W x 25 mm H. The local price is around 7 AUD.
You will still need the ballast resistor to reduce the input voltage to around 11 volt. It will dissipate 55 watt!!!.
Anything you can do to reduce the input power is going to help and you can see why switching regulators are so efficient. Incidentally, I built a switching reg using a coil based on the yoke ring from a 110 degree color TV. The switching frequency was around 10 kHz from memory. It was a 24 volt to 12 volt supply to run a fridge in a Toyota bus which had a 24 volt battery supply. Needs a lot of heavy wire!!.
Hope this helps.
ps forgot to say that there is a RS heatsink #263 - 251, which is similar to the one shown in your photo. This is 35 x 25 x 20 mm. The thermal resistance is 17.9 C/W

 

[ChrisP58]

Dropping that much voltage at that current will make a lot of heat. A switcher would be much better.

I would suggest ordering this demo board from TI. It isn't cheep, and you'll need to change a resistor to get the voltage you want, but should do what you need without making a huge amount of heat.

https://www.ti.com/tool/tps54540evm-558

 

[Mosaic]

This is the better part to use.
http://cds.linear.com/docs/en/datasheet/1074fds.pdf

Alternatively: you can roll your own switching reg with a TL494 driving any suitable driver bank (pref. NFETS) for much higher currents.

 

[konradIC13]

Ok, i get it now with the LM338, ill figure something out.
Switching regulators are nice, i always tried to design a schematic for them and make one but never finished it, usually got stuck at selecting parts. This chip LM1074 looks very nice and to my surprise i wont have problems buying it. The schematic looks simple too, the one "basic positive buck converter" but... i cant seem to find how you can set the output voltage._. the schematic is for 5V (while i need 6V), usually it was a resistor which value was function of other resistor value and some constant but i looked through and couldn't find it, this LM1074 is supposed to be regulated, isnt it?

From what i have seen in other switching regulatos i can suspect it may be R1=R2 *( ( Vout/Vfb)-1) and it seems right (datasheet Vfb is 2.21), am i right?

 

[Mosaic]

Have a look at EQ 23 on page 22.
https://www.thierry-lequeu.fr/data/AN44fa.pdf

 

[konradIC13]

I see it now, i like it. Im gonna buy that chip and make the circuit, and ill look through that design manual. This is a bit more complex that circuit on LM338 but its so much simplier than the DCDC converter i tried to make a while ago.
I got a hold of a few MAX15046 ICs and tried to make a DC-DC converter on them from (12V to 6V current from 3 to max10A) but i ended up on trying to select appropriate parts but somehow stopped there, this is what i made so far:

A schematic and i i looked through some parts and figured that:
Q1 - SiS478DN Rds(on) 30mΩm, Qgs 1.1nF, Vgs(th) 1.2v, Id 12A
Q2 - SiS436DN Rds(on) 13mΩ, Qgs 2.0nF, Vgs(th) 1.0v, Id 16A
L1 - DE1207 2.4uH, 10.3A, 10mΩ
I wasnt sure about diode, initially i picked ES2D-E3/52T 200V 2A diode but dont really know why, not sure what kind of diode it should be and what ratings.

 

[ronsimpson]

If you are headed to switching......
This being one of your first projects.....
Lets keep the parts count down.....

 

[Mosaic]

Where is the LM22678 sold in singles for $2.45?

 

[ronsimpson]

Where is the LM22678 sold in singles for $2.45?

No where. That the 2000 piece price. Sorry.

 

[rumpfy]

Hi konrad,
What Ron has put up looks OK. Check the data sheet for your self. Most, (all?) of these stepdown switching convertors are similar. There are 2 problems with ALL of them;
1. you need to get the inductor right; it must not saturate during the ON cycle. If the inductor saturates, the ON time is too long.
2. The diode MUST BE a Shottky diode. This type has a quite low forward voltage drop and this is important when you are looking for a 6 volt output.
Again, make sure your heatsink design is correct AND, dont try to run your design up to full power on a breadboard.
Many of the semi conductor makers have top shelf design briefs for switching regulators. They are well written. (Motorola AN719 is good).
The peak current in a switching reg is twice the DC current and the duty cycle adjusts to keep the input power in balance with the output power.
Good luck and keep posting

 

[konradIC13]

Unfortunatly i cant get my hands on the LM22678 chip ill have to make it on LM1074, most likely 7pin version (but not sure yet).

 

[picbits]

Do you mean the LT1074 ?

I've got a stack of those in stock - I keep meaning to have a play with a couple and build some PSUs as they look like really nice switching regulators.

 

[large_ghostman]

Do you mean the LT1074 ?

I've got a stack of those in stock - I keep meaning to have a play with a couple and build some PSUs as they look like really nice switching regulators.

do you still sell on ebay?? me and dad have had stuff from you before. sorry for the hijack

 

[konradIC13]

Yes, yes i meant LT1074, depending which version ill get (5 pin or 7 pin) ill try to make a schematic from the datasheet or make something on my own according to the design guide.

 

[rumpfy]

konrad,
Have a look at the On Semi application note AN920. This discusses this 34063 IC. It is a small 8 pin DIL package and is used extensively in small Dc-Dc convertors of the type that plug into car cigarette lighters for mobile phone charging and the like.
With an external transistor it will do what you want. Dont bog down in small details like which device should I use. Just get what you can get and design accordingly. Page 23 of the AN920 application note describes a dual output step down supply with a 5 Volt 5 Amp outlet and a 12 volt/300 mA outlet. Just dont worry about the 12 volt winding on the choke.
Scrounging is the way to go; "see what you can get hold of and decide the approach". Or, another way, "there is more ways of killing a cat than choking it"
Hope this helps.

 

[picbits]

do you still sell on ebay?? me and dad have had stuff from you before. sorry for the hijack

I don't tend to sell on eBay - I have an online shop (although it seriously needs an update). I've got millions of components at the moment but only a small fraction of them actually on the webshop lol.

 

[konradIC13]

Im still alive, just got my hands on 2 pcs of LT1074CT7 (7 pin version) and gonna make myself do the schematic, ill post result here soon.

 

[rumpfy]

OK you've got the device. Now get the application note 44.
I think you wanted to go from 16 volt down to 6, at 5 amp. I cant recommend strongly enough that you consider a tapped inductor. With a non tapped inductor, the peak switch current is twice the average output current. With a tapped inductor it is not like this, because you have transformer action in the inductor which reduces the peak switch current.
The LT 1074 is a fixed 100 khz switcher and the inductor design might be a bit tricky. If you buy one then OK, but if you want to design one then again, you have to find the core materials. I'm not sure where you are with this aspect. In some designs, a litz wire is recommended, but multi strands of a finer gauge will help minimise copper loss; 4 or 5 strands for example.
The device dissipation is probably around 6 to 8 watt at full load so a heat sink will be necessary. A soldered device can be tricky to solder on to the mounting base, and a bolt on is better for a home construction in my view. The problem with soldering to a PCB material is the possibility of the copper delaminating from the base material of the PCB. If this happens, the thermal resistance of the PCB is much increased.