EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Chipguy on August 30, 2013, 08:44:06 pm
-
I was wondering why a 3.3V linear regulator in SOT-223 on one of my designs died the other day. So i started an investigation.
I got a TS1117 and tortured it until the thermal shutdown kicked in every couple of minutes.
It lasted 3 days than it was dead.
I did the same to an LT1117 from Linear Technology.
After 10 days of torturing i needed to PSU for something else and gave up.
In the x-ray images you can see the reason the one IC dies and the other does not:
The die in the LT is 4..5 times bigger and therefore does not need to take as much stress as the smaller one.
So the price difference of $0.27 to $2.00 indeed MAKES a difference.
Guess which of the 2 chips is more stressed dissipating the same amount of power.
(http://www.hollgi.de/FOTO/EEVBlog/LT-vs-TS-Die-small.jpg)
-
Still is 2$ really worth it? It would be interesting to see the xrays from some of the other more renowned brands ... the LT1117 is hideously expensive compared to just about any other XX1117.
-
Still is 2$ really worth it? It would be interesting to see the xrays from some of the other more renowned brands ... the LT1117 is hideously expensive compared to just about any other XX1117.
The question is not "is it worth it"? The question is how much is it going to cost when it fails?
-
Well, having the unit returned to your lab, doing a repair and shipping it back to the customer is definetly more than the additional cost for a LT1117.
I am going to investigate some other XX1117 as well:
NX1117
LM1117
TLV1117
Just need to get hold on the X-ray machine again ;)
-
Regular members of badcaps.net will know that Vizio TVs have problems with 1117-33 regulators failing on the main board. They must use the cheapest possible quality and perhaps the x-rays provided present a clue why?
-
How long does each one last when it's used within the data sheet spec? Are the specs for the two parts the same?
Do you actually know why the regulator you originally had failed in service? Or have you just observed that, when subjected to a torture test, a sample of one part from one manufacturer failed earlier than a sample of one part from another?
-
The brand name one also seems to have a bit more bonding wires, so they are probably helping spreading the heat out the legs.
-
If only the torture were more severe!
Both would fail immediately ;-)
-
First of all : Thanks for the many responses.
Ok some more additional information:
Bond wires : Yes, the LT has thicker bond wires than the TS. It is also one more than the TS.
The technical specs in the datasheets are the same.
There is just a maximum of torture you can give them when you stick to the specs.
I came to this issue because I designed something with it that failed after a few months of 24/7 use.
The regulator in that does not even get hot. It is around 40-50°C, roughly supplying 250-300 mA max.
The TS1117 just died from no obvious reason.
I made a circuit with a 4.7 Ohm load at the output pin drawing some 700 mA @ 3.3V
I had the PSU set to 9 Volt, so the regulator needed to take 2W, roughly double the amount a SOT-223 can take.
My torture was deliberately insufficient cooling i.e. no cooling at all to get the thermal shutdown kicking in.
Look at EEVBlog Episode #512. That's the most common thing to get wrong it seems.
Datasheet of TS1117: http://www.farnell.com/datasheets/50383.pdf (http://www.farnell.com/datasheets/50383.pdf)
The datasheet does not state any maximum power so I could have cranked it to the max.
However I used this datasheet http://www.farnell.com/datasheets/158283.pdf (http://www.farnell.com/datasheets/158283.pdf) from an NXP transistor to determine that a SOT can take around 1W. I doubled the power so the thermal shutdown would kick in every few minutes.
Giving hell to the thing with like 5 times what it can take would not be a realistic test.
-
There is no need to X-ray these. The datasheet tells you everithing you need to know. Thermal resistance of the TS is 110 K/W for the linear technology it is 15K/W. That means that the linear can dissipate around 7 times more power. At 2W it will be 245 degrees hot.
I hate these cheap Taiwanese semis. One time we hat to manually patch a production batch, because the quality TVS diode was replaced with one from taiwan, and it had 10 times the leakage current. There should be a capital punishment for using the same naming.
-
You can't compare "junction to solder joint" with "junction to ambient".
The LT datasheet does not state "junction to ambient" value while the TS does not state the "junction to solder joint" value.
If you look at the NXP datasheet tho it becomes clear that both parts should be roughly the same.
NXP states "junction to solder joint" with 20K/W while it states "junction to ambient" with 150K/W.
If you adjust that to the LT (15 K/W instead of 20) you come to 112.5 K/W for "junction to ambient".
Which is pretty much the same as for the TS.
I have already checked that out before I decided to x-ray them.
http://www.nxp.com/documents/data_sheet/NX1117C_NX1117CE_SER.pdf (http://www.nxp.com/documents/data_sheet/NX1117C_NX1117CE_SER.pdf)
I wanted to x-ray the NXP as well but Farnell delivered after 2 days this time (Murphy's law) which very rarely happens.
They usually deliever the next day.
Anyway that gives me a change to get even some more and have them x-ray'd as well.
It will take a few weeks until i get another chance though.
-
You can't compare "junction to solder joint" with "junction to ambient".
The LT datasheet does not state "junction to ambient" value while the TS does not state the "junction to solder joint" value.
Yes, you are right. Then they managed to give a totally useless information, since the did not define what they meant by ambient.