LM2594 - what the hell is wrong with the device? Fake? Devil's custom pinout?

[Yansi]

Hello!
I have just finished soldering one of my relaxation projects, and proceeded to put some volts into it. But whatta?! Dead short in between GND and the circuits V+ supply line. After a small investigation composing of desoldering and resoldering various components, I have found the root cause: The devil's LM2594M-5.0.

The little bastard has dead shorted pin 5, 6, 7 and 8. After I cracked the chip open, the reason for the short is obvious. Pins 5 to 8 are connected to the substrate sheet metal.  At first, I thought well, a chinese piece of crap without a silicon (we already know the famous atmega328p slug with a lump of copper instead of a silicon chip) but it seems there really is some sort of chip.

After some more measurements, it turnes out that pin 3 against the substrate pins (5 to measures suprisingly close to 10k ohms. That got me suspicious again: What is the feedback resistor divider in the original part? The datasheet says 7.6k and 2.5k.  Thats a close shot!  Other pins against the substrate mease some tens of Mohm. There must be something inside. Maybe a LM2594 with custom pinout?

Have you ever seen something like this? What the hell is going on? I tried to look up some different vendor datasheets, but all of them share the same standard pinout. Not this devil's one.

Anyone interested would like to decap the bastard and put it into a microscope to identify it? I still have 9 pieces of this devil's chip. I may send you some for investigation. Unfortunately I have no such equipment nor the highly corrosive chemicals to decompose the package.

Here's the cracked puppy!

[xZero]

Have you got it from Aliexpress, Ebay or any other Chinese site?

[Yansi]

I don't know exactly where they are from, but it's quite possible they are direct sourced from china through ebay or so. They have been laying in my drawer quite some time. I have sourced many of these LM25xx regulators from china for hobby projects, never was a problem. Until now.
(Note: I have also spotted a HV version of LM2596 regulator, that is clearly not officially listed at the original manufacturers product pages from TI, NSC nor ONS)

I have also tried to search for any signs of a chinese "co-production" of these but no luck finding anything about them.

I know china produces huge amounts of these small switchers, for example the XLSEMI company is quite known... But their devices are clearly not intended to be a cheap fakes, they have their own series of quite respectable products...

[sleemanj]

Die and carrier rotated 180' in the package maybe (manufacturing or re-marking error)?

[wraper]

Just some other IC with marking replaced by counterfeiters. Image is not too clear but the top of the IC looks sanded. They either sand off the top or wash the old marking with solvents if possible.

[xZero]

I have recently got 10x LM338T for cheap as 4USD (eBay, 90k seller). However those seems to be either fake or damn low quality, since those suffers to output even low current as 1.5A... Two in parallel delivers 3A at max, while just one genuine TI LM338T delivers 5A (7A peak). Seems like LM317 labeled as LM338T.

Your case seems to be pure scam, I got at least barely  working parts. Parts you got are horrible. Too bad you cannot discover where you got them from... Pins attached to metal sheet? God, that's........ I have no words to describe..

[SharpEars]

If it's an IC or other active semiconductor and it's from China (or any other asian country), it's 100% guaranteed fake with zero exceptions. If it's from the US and is being sold on eBay, it's 95% guaranteed fake (if you know how to spot a fake and the seller takes pictures of the actual parts, you can filter for the 5% that are genuine based on the images - it is possible, but assumes you have access to 100% genuine parts and know exactly how all fonts, pins and even the color, alignment and positioning of the text and logo on the part should look).

If you want a high likelihood of genuine parts, buy them from a genuine distributor (at least 5-10% guaranteed fake). There is no sure thing other than sourcing from the manufacturer (0.1% guaranteed fake, due to them being manufactured in China) or no more than one hop from the manufacturer (i.e., direct distributor that sources from the manufacturer) (1-2% guaranteed fake).

You can fight these deceptive practices by buying asian parts on eBay and getting a 100% full refund after the parts arrive due to them being fake, but keeping the parts due to the draconian eBay buyer protection measures. In fact, I completely condone this practice to put the scammers out of business.

Counterfeit electronics should be made illegal!

[georges80]

...
If you want a high likelihood of genuine parts, buy them from a genuine distributor (at least 5-10% guaranteed fake). There is no sure thing other than sourcing from the manufacturer (0.1% guaranteed fake, due to them being manufactured in China) or no more than one hop from the manufacturer (i.e., direct distributor that sources from the manufacturer) (1-2% guaranteed fake).

5-10% fake from a distributor (e.g. Digikey/Future/Mouser/Allied/Avnet etc) - no idea where you are getting those statistics.

I've been using distributors for decades and have never received a part from them that didn't meet spec or work to spec. Occasionally a distributor may pull the incorrect part - which is generally dealt with expediently, but that is of course not counterfeit issue. A genuine distributor can track the part back to the manufacturer with lot codes, date codes etc. Parts are also stored/shipped appropriately (ESD/humidity etc etc).

It's the grey market that you are open to problems since many/most parts are bought as excess inventory or from more dubious sources. Buying components on ebay and various asian websites, well, I can understand folk trying to save a buck or don't have access to distributors locally, but in that case you need to assume the parts you are getting could be total garbage, 2nds etc. No point complaining or acting surprised. Don't want to order/wait weeks/build project and then find nothing works or works poorly then buy from a reputable distributor.

cheers,
george.

[amyk]

That looks like a standard SOIC-8 MOSFET. Are pins 1-3 also connected together, and pin 4 has (if the gate oxide is still intact) no continuity with the others?

[wraper]

If it's an IC or other active semiconductor and it's from China (or any other asian country), it's 100% guaranteed fake with zero exceptions. If it's from the US and is being sold on eBay, it's 95% guaranteed fake (if you know how to spot a fake and the seller takes pictures of the actual parts, you can filter for the 5% that are genuine based on the images - it is possible, but assumes you have access to 100% genuine parts and know exactly how all fonts, pins and even the color, alignment and positioning of the text and logo on the part should look).

If you want a high likelihood of genuine parts, buy them from a genuine distributor (at least 5-10% guaranteed fake). There is no sure thing other than sourcing from the manufacturer (0.1% guaranteed fake, due to them being manufactured in China) or no more than one hop from the manufacturer (i.e., direct distributor that sources from the manufacturer) (1-2% guaranteed fake).

You can fight these deceptive practices by buying asian parts on eBay and getting a 100% full refund after the parts arrive due to them being fake, but keeping the parts due to the draconian eBay buyer protection measures. In fact, I completely condone this practice to put the scammers out of business.

Counterfeit electronics should be made illegal!

What a BS

[Ian.M]

You can decap chips without strong acids or other highly toxic chemicals.
See: https://ms3c.wordpress.com/2010/07/15/boiling-chips-in-tree-sap/

[Yansi]

Okay I've finally tracked down where these are from: Aliexpress. The good thing is I already received a refund, due to extreme long delivery time. (Parts arrived after the refund). 

Also it seems it's not a mosfet, as there are no three shorted pins for the source and the pin 3 measures 9,98kohm  to ground (which is surprisingly close to what the voltage divider should measure in a genuine LM2594M-5.0 device). The rest three pins measures some megaohm-ish values.

Maybe I will play with them a little: Let's suppose they are a custom pinout of a genuine LM2594. So I'll try to find what the pinout is. Shouldn't be that hard, only 5 pins and not so much combinations.

I have also looked at these under my shitty microscope. The package seems not to be sanded down and the laserprint is pretty uniform over all of the devices.

The only problem pissing me of now is that I have to buy this convertor from some more reliable distributor. These old rusty switchers cost a fuckton of money. ($4+ each). That is not fucking acceptable,  never going to use these in any of my designs.

//EDIT: Look what we have on Aliexpress: They all look the same. http://www.aliexpress.com/wholesale?catId=0&initiative_id=SB_20151130045430&SearchText=lm2594m-5.0
With only one noticable difference: The pin 1 dent differs in size. Some products has small dent (like those I have), some has a big dent. All of them are the same marking, only different date/lot  code (four top characters).

[Fungus]

If it's an IC or other active semiconductor and it's from China (or any other asian country), it's 100% guaranteed fake with zero exceptions.

Complete rubbish. I've bought loads of chips/MOSFETs/etc. on eBay. So long as you avoid the people who sell big expensive parts for $1 each then you'll be OK.

I've only had a couple of bad lots out of hundreds of purchaes... and some lots were obviously second hand. They're the minority though, and they're half DigiKey price so I'm not complaining (you can get refunds anyway).

If you're manufacturing large batches of PCB then it's probably not a good thing to source parts on eBay. For a hobbyist it's not really a problem, and nowhere near 100%.

[Yansi]

So I went to a local electronics store (our local "Jaycar" equivalent  ) and bought two f...g pricey LM2594M-5.0. I can tell you, they look exactly the same as those from AliXpress. Stay tuned, I'll make some closeup photos to compare the prints and also will measure and test them.

Yan

[georges80]

Looks means nothing... I had a run of 250 boards assembled in China before I requested all parts to be purchased in the US from Digikey/Mouser/etc and shipped to China. That run of boards, all did not function. There was an NS switcher IC on it that wasn't working unless I floated a couple of NC pins - never had to do that on previous use of that chip/design.

I live near NS (or where NS was before it was absorbed by TI). I contacted their failure analysis lab and sent it some bad parts and some 'authentic' parts. They x-rayed the parts and confirmed that the bad parts did NOT contain actual NS dies, even though the external markings all looked authentic. The assy house confirmed the bad parts had been sourced on the grey market in China...

The assy house reworked all the boards and used authentic parts, all boards then worked correctly.

So, no idea what was in the fake parts, it was something 'close' to what was meant to be in there - but not the real thing.

Live and learn. The counterfeit semiconductor market is real.

cheers,
george.

[amyk]

NC really means "do NOT connect", you could've just as well received the message "these are real parts, we changed the silicon slightly but you're using them wrong"

[Monkeh]

A lot of datasheets specify NC as no connect or ground.

[bktemp]

A lot of datasheets specify NC as no connect or ground.

I have seen datasheets specifying NC as do not connect or as not connected or even as connect to GND. It depends on the manufacturer and the specific part.
Many power ICs use the GND pins as heat sinks, therefore suggesting to connect them to the GND plane. Precise analogue ICs use NC pins for post package trimming (many voltage reference ICs do this), those pins must not be connected. Even small leakage currents into those pins make the output voltage drift out of specs. I have also seen datasheets recommending NC pins for signal routing to simplify the single layer layout.
There are often improved versions of ICs with additional features connected to the NC pins from different manufacturers. They often are compatible as long as the NC pin is floating.
If the datasheet does not tell you exactly what is connected to NC pins, do not connect them.

[Fungus]

There was an NS switcher IC on it that wasn't working unless I floated a couple of NC pins - never had to do that on previous use of that chip/design.

NC means "No Connect" - they're supposed to float. Connecting them to GND is wrong (and so is tying them to Vcc!)

[Yansi]

Based on what?

Look in the LM2594 datasheet. See the pins 1 to 3?  They are all NC.  They are suggested to be soldered to board. Guess where they are connected on the reference PCB design: Ground! 

Y.

[Fungus]

Based on what?

Look in the LM2594 datasheet. See the pins 1 to 3?  They are all NC.  They are suggested to be soldered to board. Guess where they are connected on the reference PCB design: Ground! 

Which manufacturer? The reference circuits of TI and ONSemi don't have them connected to anything.

http://www.ti.com/lit/ds/symlink/lm2594.pdf

http://www.onsemi.com/pub_link/Collateral/LM2594-D.PDF

TI says "No internal connection, but should be soldered to pc board for best heat transfer" but not GND.

The people who made your "fake" part obviously didn't see your "reference PCB design". They only read the datasheets.

[Yansi]

Page 30, look, carefuly. 

The people who made your "fake" part obviously didn't see your "reference PCB design". They only read the datasheets.

I don't get the context of that... There's nothing wrong with grounding or whatevering pins 1-3.  Genuine part will work. The chinese crap has shorted pins 5 to 8 and seems to use pins 1 to 4 instead. Jeeeez, I already explained.

[georges80]

There was an NS switcher IC on it that wasn't working unless I floated a couple of NC pins - never had to do that on previous use of that chip/design.

NC means "No Connect" - they're supposed to float. Connecting them to GND is wrong (and so is tying them to Vcc!)

If you don't know, don't write... Mine was an example of a NS part, the LM2674/75 to be specific.

The pins are marked NC which means 'not connected', and NS TELLS YOU TO GROUND THEM FOR BETTER THERMAL PERFORMANCE.

DNC means 'do not connect'.

cheers,
george.

[bktemp]

The pins are marked NC which means 'not connected', and NS TELLS YOU TO GROUND THEM FOR BETTER THERMAL PERFORMANCE.

DNC means 'do not connect'.

In this case yes. But NC can also mean do not connect. The datasheet for the classic REF02 voltage reference says:

NC = NO CONNECT. DO NOT CONNECT ANYTHING ON THESE PINS. SOME OF THEM ARE RESERVED FOR FACTORY TESTING PURPOSES.

If the datasheet explicitly says NC pins are not connected then you can connect them to GND or use them for signal routing, but without any further information NC pins should always be treated as do not connect.

[georges80]

Yes, though I presume that most folk can read the datasheet to determine what to do if anything with NC/DNC pins, I can at least manage that minor skill

cheers,
george.