Lm317 vs LM338

[queennikki1972]

So I played around with the LM338 5 amp variable IC in the T0-220 package and I literally wasted 3 of them and still cant get them to work the same as a Lm317.. It would immediatelty short them out. According to Google they have the same pinout so I dont understand why..

Both were being powered by the same setup.. 40VA transformer - Bridge Rectifer - Smoothing Capacitors - Resistor and Variable Resistor (Pot) and one works and the other doesnt.
 
Am I missing something here... The circuit below is the basic circuit used for learning.

[queennikki1972]

But that would short it out like that?

[Bud]

Try reducing the output capacitor to the value recommended in the datasheet. Also the datasheet recommends using a prote tive diode across the 317.

[Ian.M]

But that would short it out like that?

Nope.
However overvoltage at the input could be an instant failure.   

Try reducing the output capacitor to the value recommended in the datasheet. Also the datasheet recommends using a prote tive diode across the 317.

A large output capacitor shouldn't be a problem.  The regulator will simply current limit till its charged.  An anti-parallel diode to protect the pass transistor is only required if the input can ever be shorted to (or rapidly dragged down to) ground with a large output capacitor.   Its internal protection is good for a reverse surge of up to 25A.   There is no way short of rectifier failure for the voltage of a conventional secondary => bridge rectifier => reservoir capacitor unregulated supply, not powering other large loads, to collapse rapidly enough to damage the LM338 unless you physically short the input.

One possibility is that it may be acting as a RF power oscillator due to poor layout.   That can cause excessive dissipation and cause the internal thermal limiting to misbehave.  Post a photo of your actual circuit!

Also, where did you get them from?  If it wasn't a major distributor you could have fakes. 

[queennikki1972]

I tend to order bulk from china so its a possibility of fakes. I dont always get bad stuff from china. Ill try some US ones and i will consider all of the advice. The layout is a mess do to soldering and desoldering in frustration but was basically like the drawing attached. As I said I hooked up a LM317 and it worked just fine so I doubt the layout was the issue.

[floobydust]

You absolutely have to have the two protection diodes, and the LM338 minimum load is 3.5-5mA 3.5mA typical 10mA max. so make sure you have that.
In another thread your power transformer was 24VAC 40VA, which under light load can go over 35VDC, especially if it's an import. It could go too close to the IC's ~40V limit. Otherwise, I would guess you have fakes that can't take high voltage input or high current.
edit: corrected LM138/338 mixup

[Zero999]

You absolutely have to have the two protection diodes, and the LM338 minimum load is 3.5-5mA so make sure you have that.
In another thread your power transformer was 24VAC 40VA, which under light load can go over 35VDC, especially if it's an import. It could go too close to the IC's ~40V limit. Otherwise, I would guess you have fakes that can't take high voltage input or high current.

Yes and if the output capacitor is oversized, there's more chance it will overvoltage the LM338, since the rectifier capacitor will charge up much more quickly, than the output capacitor.

I think the more likely cause of failure is the LM338 being fake.

[magic]

Test them in some simpler/safer circuit.

Connect ADJ to ground through 100Ω.
Connect OUT to ground through 100Ω.
Connect IN to 5V through 100Ω.

ADJ should be at 0V, OUT should be at 1.25V.

[Audioguru again]

You absolutely have to have the two protection diodes, and the LM338 minimum load is 3.5-5mA so make sure you have that.

Nope. The minimum load for the more expensive LM138 is 5mA when the input voltage is fairly high. For the LM338 it is 10mA. Therefore the value of the output to adj resistor is too high. Use 120 ohms and change the pot to match.
But it affect only the output voltage rising with no load and will not cause the IC to short.

Why do people buy cheap Chinese fakes anyway?? 

[queennikki1972]

Why do people buy cheap Chinese fakes anyway??

Because not everyone can afford $0.25 resistors.

[Mr. Scram]

Nope. The minimum load for the more expensive LM138 is 5mA when the input voltage is fairly high. For the LM338 it is 10mA. Therefore the value of the output to adj resistor is too high. Use 120 ohms and change the pot to match.
But it affect only the output voltage rising with no load and will not cause the IC to short.

Why do people buy cheap Chinese fakes anyway??

Probably for the same reason we don’t all have a lab filled with R&S and Keithley gear.

[exe]

Because not everyone can afford $0.25 resistors.

A fake component can cause a lot of damage, consume a lot of time to troubleshoot, and cannot guarantee performance. For these reasons I suggest not to buy ICs from unauthorized suppliers.

I do buy things from China when I can verify performance, or I'm willing to take the risk, or just for fun, or I believe the part is hard to fake, or when a part is not available elsewhere.

[floobydust]

I used to use Linear Tech version LT138A/338A which was in a beefy package TO-3 or TO-3P. That thing was tough and had higher SOA output current.  But too bad it's obsolete now.

[Audioguru again]

A real LM338 will shut down safely if it is overloaded and gets too hot. A fake one might simply melt into a short circuit.
EDIT: An LM338 at Digikey costs less than a cheap McDonalds hamburger.

[Mr. Scram]

A real LM338 will shut down safely if it is overloaded and gets too hot. A fake one might simply melt into a short circuit.
EDIT: An LM338 at Digikey costs less than a cheap McDonalds hamburger.

If you include shipping the difference can be 1000%

[David Hess]

I have never had any problem with genuine LM350s or LM338s.  They have always operated as higher current and power LM317s like they should.

[mikerj]

A real LM338 will shut down safely if it is overloaded and gets too hot. A fake one might simply melt into a short circuit.
EDIT: An LM338 at Digikey costs less than a cheap McDonalds hamburger.

If you include shipping the difference can be 1000%

Who orders a single cheap component?  You order a bunch of stuff you need to make the most of the postage.

[dcbrown73]

There seems to be a disconnect here between people who do EE for a living and those who only do it as a hobby.   Your normal hobbyist doesn't use 100 LM338 transistors in any type of reasonable time frame nor can all of them afford to buy all possible components they may need in their hobby in bulk.  (not to mention finding a place a to store a huge inventory of components in any type of organized manor in a corner of your bedroom or whatever.

I try to order in lots of 10, but my problem is that today's learning experiment has me ordering 5 different components and as I learn something new.  It leads me to another learning experiment that has me ordering three more components again two days later.  (a major reason I buy resistor / transistor / cap / etc bulk kits, but alas those are normally Chinese kits)  All the sudden I find my self needing to order 3-4 times in two weeks as my learning moves forward.   At $9-$12 per shipment.   All the sudden I'm getting close to $50 is shipping alone in two weeks.

It would be nice if I knew everything I would need for a month or two in advance.  I could save a lot of money over the course of a year for maybe a new Keysight scope!

[schmitt trigger]

I am also a little dumbfounded by these failures. I built many power supplies utilizing the LM338 without problems.
The only points you should be well aware of are:
-Ensure the minimum 10 mA load current, as Audioguru explained. But that in itself won't cause a failure.
-Use the antiparallel diode, as recommended in the datasheet, to prevent the output capacitor from discharging into the regulator.
-Keep the ground (common) wiring short, use thick wires. A star-grounding system also helps.

Of course, if the devices are fakes, this will not prevent the damage.

[exe]

All the sudden I find my self needing to order 3-4 times in two weeks as my learning moves forward.   At $9-$12 per shipment.   All the sudden I'm getting close to $50 is shipping alone in two weeks.

It would be nice if I knew everything I would need for a month or two in advance.  I could save a lot of money over the course of a year for maybe a new Keysight scope!

That's why I tend to buy a lot of stuff that I don't use in the end to save on shipping and waiting.

I used to be upset about it, now I just call it "price for education". Unless you are really short of money, I don't see $50 as a big expense. I myself can save much more just by stop buying junk food, alcohol, etc.

Anyway, I'm yet to finish a project without a need to order missing parts last minute. I just accept that I'll have to do an additional order or two. My annual budget for this hobby is $XXXX (amount removed not to shock people) (it's ok to spend less, it's the amount I allow myself without hesitation). It's big money, and people tell I'm crazy, but I think it's fine to pay so much for own education. Who said it should be for free? Last year I didn't spend much because most of the time I already have parts and equipment I need. I spend my money, but when I was a child my parents just gave pocket money so I could buy spare parts from the flea market because they thought that would facilitate development of my tech skills that would make me more competitive on the job market.

So, I suggest optimize for speed of learning and development, and not put unnecessary budget constraints that limits you. Save on beer, but not on parts.

[Mr. Scram]

That's why I tend to buy a lot of stuff that I don't use in the end to save on shipping and waiting.

I used to be upset about it, now I just call it "price for education". Unless you are really short of money, I don't see $50 as a big expense. I myself can save much more just by stop buying junk food, alcohol, etc.

Anyway, I'm yet to finish a project without a need to order missing parts last minute. I just accept that I'll have to do an additional order or two. My annual budget for this hobby is $2000 (it's ok to spend less, it's the amount I allow myself without hesitation). It's big money, and people tell I'm crazy, but I think it's fine to pay so much for own education. Who said it should be for free? Last year I didn't spend much because most of the time I already have parts and equipment I need. I spend my money, but when I was a child my parents just gave pocket money so I could buy spare parts from the flea market because they thought that would facilitate development of my tech skills that would make me more competitive on the job market.

So, I suggest optimize for speed of learning and development, and not put unnecessary budget constraints that limits you. Save on beer, but not on parts.

I don’t think “just spend $2000 a year” is useful advice.

[floobydust]

IF a LM317/LM338/LM350 fails short, it burns up the potentiometer.
So replacing the IC may not seem to work, the output stays at top voltage. You have to check the potentiometer with an ohmmeter that it can still dial down to zero.

A very long time ago, my lovely 10-turn pot up in smoke. So I added a tiny fuse in series with it.

[exe]

I don’t think “just spend $2000 a year” is useful advice.

That is not what I advised. It's not about spending a specific amount. It's about not going too cheap when it doesn't make sense. I'll remove the amount as other people can also trigger on it.

[Mr. Scram]

That is not what I advised. It's not about spending a specific amount. It's about not going too cheap when it doesn't make sense. I'll remove the amount as other people can also trigger on it.

It's not about the amount or "triggering". It's about advising people to spend what could easily be an order of magnitude more on parts where the cheap alternative is not only functional but for a lot of people even the only real option. Your junk food and beer budget may well exceed the disposable income of others. Fakes are an issue but not hugely so. I'm not saying you'll get original parts but most are functionally equivalent enough as not to make a difference for learning purposes.

[exe]

I'm not saying you'll get original parts but most are functionally equivalent enough as not to make a difference for learning purposes.

As I said before, I believe this is a wrong approach doing more harm than good. I advise buy new and original parts, esp. when it comes to jelly-bean components. Why gambling and guessing why it doesn't work? How do you know if it's "functionally equivalent" and not going to break any moment in the future? If one wants to save on delivery, buy, e.g., from lcsc.com.

You can't have original parts, fast delivery and cheap price.