counterfeit Xilinx ?

Surely you've sent one in to Xilinx for their appraisal. What did they say?

Bob

Reply to
BobW
Loading thread data ...

I got a batch of "Xilinx" Spartan XCS30 FPGAs from a Chinese seller, and am having problems with random failures at first power up. Sometimes it is a stuck I/O pin, sometimes a failure to configure. I first thought maybe we had an ESD problem, but I'm now thinking these may be counterfeit. They have white ink printed labels on the front, whereas other Xilinx chips have laser-etched labels. Also, these Spartan chips don't have the Spartan logo just below the Xilinx logo, like my other Xilinx chips. Anyone have any comments on this?

Jon

Reply to
Jon Elson

Unlikely to be "counterfeit" in the true sense, but I understand Asia has long had a problem with 'grey market' devices. Stories of truckloads on Intel Pentiums going into one alley in Hong Kong, and truckloads of higher spec'd devices coming out an adjacent alley abound :) With Silicon, there are also the post-package rejects to consider. What happens to them ? (XIlinx actually sell some as Easy Path- could these be Easy path devices, re-labeled ?)

-jg

Reply to
-jg

Photo.. ?

Reply to
sky465nm

OK, where should I put the photos?

Jon

Reply to
Jon Elson

I'm not touching that one...

Bob

>
Reply to
BobW

You can send it to me,and I will pass it on. snipped-for-privacy@xilinx.com

Reply to
Peter Alfke

Try this service:

formatting link
(Max 3MB/picture)

Reply to
sky465nm

OK, I will do that. I'll have to see how "macro" my camera will go. Do you have any comment on some date code 1999 Spartans that are printed with white ink (not laser marked) and have the Xilinx logo but no Spartan logo under it? These are XCS30-3TQ144C parts. I just got them late last year and am having what looks like a 50%+ failure rate. In the past, when I got an occasional ESD damage event, I usually observed the chip running EXTREMELY hot or totally shorting out the power supply. These chips all run cool, but many have what seems to be a stuck output pin or they don't configure at all, and the pins that should be doing certain things when the PROG/ line is cycled don't do anything. I've used several hundred of this particular Xilinx part in previous boards and never ran into any problems like this. That combined with the different markings make me think they could be counterfeit, maybe poor quality wafers that were supposed to be recycled, or something like that.

Jon

Reply to
Jon Elson

Jon, I really have no comments. I do not work in the Spartan group or even division, let alone building. I just promised to send your picture and comments to the people in the know. I hope the parts are fake, for I cannot imagine such a sudden surge in failure rate from good devices. Peter Alfke, who works in the Advanced Products (Virtex) division.

Reply to
Peter Alfke

Jon, I really have no comments. I do not work in the Spartan group or even division, let alone building. I just promised to send your picture and comments to the people in the know. I hope the parts are fake, for I cannot imagine such a sudden surge in failure rate from good devices. Peter Alfke, who works in the Advanced Products (Virtex) division.

Peter,

Can you tell us if you've ever seen any counterfeit Xilinx parts? If yes, were they authentic die that had been repackaged or relabled, or were the die themselves fakes?

Thanks, Bob

Reply to
BobW

I have no recollection of fake FPGAs.My educated guess is that the die inside were probably genuine, but either abused or not tested or test escapes.I have never heard that somebody had the money and the audacity to really design and make chips as complicated as ours. But I do not have any proof. And the outside world is getting smarter and has deeper pockets than in the past. Xilinx definitely would like to get their hands on those chips... Peter

Reply to
Peter Alfke

Think about it. No one is going to set up to create clone die. The costs would be massive, for tiny, short term returns.

Xilinx already have a path for defective (test failure/yield fallout) die to be sold, which is the easy-path. Those devices are tested/warranted for ONE or TWO bitstreams only.

Depending on their yields/sales mix, Xilinx may have a surplus, or shortfall of 'bad die' - (and as time progresses, you expect the yields to improve). So, it is possible EasyPath devices would work with many more test patterns.

The OP could try a series of pgm/verify on suspect devices, to check they can 'remember' a valid bitstream.

-jg

Reply to
Jim Granville

Oh, well, you can certainly have 6 that I have so far confirmed are bad in some way. If some Spartan expert at Xilinx can tell they are indeed counterfeit from the photo I sent, I guess I am going to have to scrap the lot of them, and so you can have units that have never left the tray. (I don't know if you'd have any interest in electrically experimenting with them.) The removed chips have been pretty heavily abused with our removal tool, the package was visibly flexing as I pried them from the desoldering tool. I've NEVER seen anything like that before either, but I haven't desoldered a TQ144 using this tool before, either. I usually just cut the leads off the body, but this time I wasn't sure if the problem was in the board or the FPGA until I had replaced the FPGA. The last one I did, I had to replace the FPGA twice to get a good one! I know that my procedures at my rework bench are pretty good, so I am really confident I didn't ESD-damage the first replacement.

If these units are NOT counterfeit, then they must have been salvaged in some manner that leaves them incredibly "new" looking, but stressed in some way. Absolutely no sign they've been "fiddled with", though.

Jon

Reply to
Jon Elson

I only have a couple bitstreams for this particular Spartan size. Several of the chips would not go through the basic steps of the load sequence, such as clearing the addr register of the serial PROM. The others looked like they were mostly working, but had stuck output pins or failed logic internally driving those pins. I really didn't want to go into a great detail diagnosing the problem once I was able to prove it was a bunch of bad chips.

Some time ago I got a tape of Analog Devices op-amps that about

60% of them went up in smoke when the board was turned on! On closer inspection, the date codes were all over the place on the same tape, at least an indication that the chips were salvaged off some production run or something.

Curiously, these troublesome Spartans have a number after the date code that varies. I wouldn't expect a real counterfeiter to bother with that. All the chips would have the same logo, part number, date code, etc. on them. I'm still stumped. I'm not doing anything different here, and I've only had one or two bad chips in years of making these. I can EASILY attribute that to random ESD problems. (I got 12 V into one when a wire touched, and a couple units were taken out by lightning, too. I expect that.)

Jon

Reply to
Jon Elson

I have XCS10XL in TQFP100 from around 1999/2000 and they also have printed labels. They were obtained from the official German distri at that time (Metronik/Unique). So I guess that white ink labels are no sign of unoffical chips...

--
         Georg Acher, acher@in.tum.de
         http://www.lrr.in.tum.de/~acher
 Click to see the full signature
Reply to
Georg Acher

Hi Jon,

I am an independent distributor of electronic components. Most of the product I supply comes from fortune 500 OEM's , but occasionally we do have to go to the broker market. Only in very extreme cases would we buy product from China. I would guess that about 95% of the product that is coming out of China brokers/traders is counterfeit. Who is the company you bought them from? Are they in Shenzhen? It is a very bad sign if they did come from that area.

We have a very strong quality department and know most of the tricks these companies are using. I would be more than happy to assist in verifying if you have authentic product. From what I have read, I doubt it, but to be fair I typically have a negative attitude towards product coming from China. Please give me a call and I will be more than happy to give you my insight and maybe be able to help you avoid having future problems with buying from China.

I mentioned that I supply product from fortune 500 OEM's. I would like you to know that normally the product is in my stock. I am able to do this as we hold inventory for these OEM's on a VMI or SMI agreements that we have with the OEM's. They allow me to borrow some of their material to fill critical needs of other customers. Just about all my inventory is purchased direct from the factory or franchise distribution.

Jon E. Hansen Strategic Sales Pyramid Technologies Inc. snipped-for-privacy@pyramidemail.com

formatting link
(949)864-7745 Direct (888)288-8913 Ext. 205 (949)636-1416 Cell (949)864-1869 Fax CERTIFIED ISO 9001:2000

Reply to
jon

Yup, I have some test fixtures here with Spartans in them from VERY close to the same date code as the lot in question. They also have a white printed label. The most obvious difference is the one in the fixture has Spartan(tm) in big letters right under the Xilinx(tm), while the questioned one is missing that Spartan line. Every onter Spartan chip I can dig up has the Spartan marking on it, too, in whatever marking technology was used. Another difference is the questionable one has a "weaker" label printing, under a microscope it looks "speckled" while the good chip looks to have very dense ink. The speckles look a lot like looking at laser printer toner under a microscope. Of course, I'm just jumping to lots of conclusions here, with no real basis to stand on.

When I pulled the removed chips out of the desoldering tool, they bent noticeably. I thought these package materials were supposed to be thermosetting, and shouldn't get soft when heated to desoldering temperatures. Maybe my desoldering tool was getting a lot hotter than I thought, but I just ran it up until I could lift the chips.

Jon

Reply to
Jon Elson

But, I have no other examples of Spartan chips which DON'T have the Spartan(tm) marking right below the Xilinx(tm) logo. That was one of the things that made me curious, although I don't know why a counterfeiter would miss something that obvious.

Jon

Reply to
Jon Elson

While I have never had to use them for counterfeit detection, I have use Process Sciences for process control, inspection and rework for several years now. I have always been very happy with their work. They have some information about their counterfeit detection services here:

formatting link

One of the trade journals I get has been running a series of articles about counterfeit parts. I think it was either EE Times, or EDN. Some of the things that they mentioned include remarking to a better speed grade, a lower power version, or a larger memory size.

Regards,

John McCaskill

formatting link

Reply to
John McCaskill

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.