ATINY87 failures in production

Go To Last Post
14 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi,

We have a series of 5mm LED display boards that use the ATTINY87.  This series of boards has been in production for about 1.5 years.  In the past 3 days, we have had a roughly 25% failure rate of what appear to be nonfunctional/corrupt MCU’s.  The series uses the internal 8MHz clock.  To keep things simple, the focus will be on 2 failures I found:

Here is a photo of one of the failures:  (the 6 pin ISP header is on the right, C11 is a bypass cap)

a failed MCU

 

 

We have 5 stations in production where the MCU’s are programmed.  At one of these stations, I looked at 6 boards with Atmel studio.  (these boards had not been programmed before)  The programmer is a MK-II programmer modified to supply 5V to the MCU.  On two of the boards, I could not read the device ID.  After some probing with the oscilloscope, I could see the clock and data going to the MCU from the MK-II, but no data was coming back to the MK-II.  Here are some things I have tried:

     -The 5V to the MCU seems stable.  I added a 0.47uF cap across the 5V lines, no change.

     -In case the fuse settings had been changed to an external clock, I added a 1MHz, 5V  square wave signal to pin 14, no change.

     -A debug wire  connection (using the AVR dragon) was tried.  No connection could be made.

     -The reset pin of the MCU seems to be stable at 5V.

    -I asked the tech to replace the MCUs and once that was done, the boards could be programmed and were testing normal.

 

Some of the other failures from production are rather interesting.  (Production uses a batch file on the network to program hex file and fuses in one step)  Several boards had the fuses set to an external clock.  Once I set the clock back to the internal RC oscillator, the flash data had the repeating pattern of “FFFF0000”.  Others had a hex file or parts of a hex file I could not identify.  (all of the hex files we have are on the network, so they are easily searchable)

The only thing I can think of is bad parts from manufacturer or that we some got counterfeit parts.  (The parts and reel do not appear to counterfeit) 

This is an image of the reel sticker:

attiny87 reel sticker

 

Right now, I’m out of ideas on the possible cause of failures.  Any thoughts would be helpful.

Matt

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

matt6ft9 wrote:
-The reset pin of the MCU seems to be stable at 5V.

/RESET should drop when you do e.g. Read Signature.  If it doesn't...

 

matt6ft9 wrote:
Several boards had the fuses set to an external clock.

Interesting how you determined that, if you are having trouble contacting the chip.

 

What ISP bit rate are you using?  Have you tried slowing it down?  Perhaps a batch is borderline different and needs to be slow?

 

What is the datecode of your chips?

 

[AFAIK you will need to record/photograph all the markings on your chips for Atmel support to verify being genuine]

 

 

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Theusch,

-Looking at the reset during device read was something I had not tried. The reset line does go low a couple of times during the attempted read.

-Anytime we have a bricked tiny or mega MCU, it has been a rule of thumb to try an external clock.

-Several different ISP bit rates have been tried, all the way down to 2.152KHz rate.

-The markings on the chips are:

   TINY87-SU

   1402C                    PH

   3X2510-1

 

This is before the Microchip buy out.  (the reel in production is dated 2014 and we have two more reels of the 3X2510-1 lot)  We have other reels in stock of the ATTINY87’s from after the Microchip switch, we are going to try those in production.

thanks for the feedback.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If your AVR fresh from the reel and fresh on board only answers ID request when you feed it external clock, the clock fuse is set wrong from factory, it should be factory set as Internal RC Oscillator.

In this case, you must contact immediately Microchip engineering support (Atmel), for them to have a very bad day of internal discussions and finger pointing sections... 

Wagner Lipnharski
Orlando Florida USA

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

matt6ft9 wrote:
1402C

Things have changed over the years on detail markings.  Back when I was your age that would be week 2 of 2014 and chip rev C.

 

Did you happen to get some kind of overage/surplus that was set up for a particular customer?  [hmmm--if so, then the whole reel would have external clock etc.]

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

-It looks like we have purchased the ATTINY87’s through Avnet, Digikey, mouser, and Newark.  The reel in production didn’t have any distributor markings.  So, hopefully they were not floor sweepings or dumpster dive parts. 

 

-‘good point about the “the whole reel would have external clock etc.”

 

-So far, the tech has replaced about 15 of the ATTINY87’s with the same 3X2510-1 lot number.  Every part she has soldered on by hand has programmed and functioned correctly.  Which points me to something we are doing wrong in house.  One of the other engineers here suggested looking at the reflow profile, he thought we might overcooking them.  We have a 7 stage reflow oven.  The profile is set at a maximum of 260C, which seems reasonable.  The engineer in charge of the reflow oven stated even though the setting is 260C, the surface of the board get up to only 245C.  (According to his thermocouple logger)

 

-The next step seems to be to contact Microchip, but that will probably take weeks to send failed parts to them and have them evaluated.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

matt6ft9 wrote:
One of the other engineers here suggested looking at the reflow profile, he thought we might overcooking them.

I thought of process issues, but kind of dismissed that since you said this isn't the first back of the app.

 

But I suppose that if you haven't done a batch for a while then the process could change?

 

Any problems possible during placement, such as lack of static protection?  But it would seem unlikely that static zaps would produce the same symptoms over and over.

 

5V OK? 

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

matt6ft9 wrote:
-The markings on the chips are: ...

This is before the Microchip buy out.

The tiny87 in issue may have bottom markings also.

First was the reel re-mark then quite some time later came the package remark which is :

https://www.microchip.com/mymicrochip/NotificationDetails.aspx?pcn=GBNG-15KQFZ896

open first attachment, go to pages 8 thru 10 for 300mil wide SOIC.

 

via https://www.digikey.com/product-detail/en/microchip-technology/ATTINY87-SUR/ATTINY87-SURTR-ND/2508038

and https://www.microchip.com/pcn

 

"Dare to be naïve." - Buckminster Fuller

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You can do a test, just remove 5 random boards from the production before reflow, take out the AVR and reflow without it, then solder it manually.

Check against other 10 to 15 produced regularly via reflow.

 

It means, you WILL solder by hand the exactly same chip that would reflow, and see if the reflow chips will present some problems.

 

You can verify the reflow soldering with microscope and check continuity between pins waveforms with a Huntron tracker unit, comparing with a working board.

 

Also, remember, those Attinys are not made by Atmel anylonger... rsrs

 

Wagner Lipnharski
Orlando Florida USA

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Check the specs of your regulator - especially in regards to the ESR of the output capacitor. It might be oscillating.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You can also build/purchase a parallel programmer and try to check the chip using that tool, it is more straight forwarded than ISP one.

If parallel doesn't talk to the "bad" chip removed from the board, then you may have a "killing" or "waking dead" situation, from the unpack from the reel to the post reflow steps.

Of course, if chips work after manually replacement, then you don't have a board situation, like power supply, oscillation, etc.  It is a chip problem.

There are few possibilities for ISP fail to communicate, mostly related to fuses, this is why a parallel programming would probably answer some questions for you.

Being a SOIC device, you will need an adapter to connect to the parallel programmer, or just take patience and solder wires to the chip and to programmer, or an adapter from SOIC-20 to DIP and protoboard.

I think the situation worth the parallel programming pain.

 

Wagner Lipnharski
Orlando Florida USA

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

So Wagner is it really true that goannas are falling off trees frozen in Florida?

John Samperi

Ampertronics Pty. Ltd.

www.ampertronics.com.au

* Electronic Design * Custom Products * Contract Assembly

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Is anything else on the board sharing the ISP lines?..could have a problem with such components making things edgy.  From your description, this doesn't seem the culprit, but anything is game now.

Could it just be a bad connection with your ISP test pads, due to soldermask or some residual coating?   We once were stymied by some very thin clearcoat getting into some 1mm connectors where it couldn't be seen ...other than not working!

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

- theusch:  The ATTINY87’s are almost a daily runner for us.  (we used 195 of the ATTINY87’s on January 5, 45 on January 8, etc.)  As far as I can tell, no changes have been made to the production process for at least 4 weeks.  But, there is always that possibility of the reflow oven being set wrong, static damage, or other calamity.  Mistakes can definitely be made out there.  It is difficult to get good information from production.

 

-gchapman:  These are 2014 vintage MCU’s, so there are no markings on the bottom of the IC’s.

 

-Kartman:  I don’t see any troublesome oscillations/ringing when probing with the o-scope.  During programming, the MCU is powered from the MK-II, so the on PCB 5V linear regulator does not come into play. 

 

-avrcandies:  no other devices are on the ISP lines.  We use spring loaded pogo pins that rotate when pressed on the 6 pin ISP header.  We have been using these for as long as I have been here.  (11 years)  Usually, they clear out any residue.  In my first photo above, some solder flux residue can be seen to the right of the ISP header.  This is from the wave solder process for the 5mm LED’s.

 

Wagnerlip:  I will see if I can setup the parallel programming of the removed ATTINY87’s.  Last I knew, the STK500 could do that, but it has been many years.