Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.arch.embedded > #28645 > unrolled thread

H8S2676 faulty chips

Back to article view | Back to comp.arch.embedded

Contents

  H8S2676 faulty chips xyman8@gmail.com - 2019-05-29 01:23 -0700
    Re: H8S2676 faulty chips Dimiter_Popoff <dp@tgi-sci.com> - 2019-05-29 19:22 +0300
      Re: H8S2676 faulty chips xyman8@gmail.com - 2019-05-30 04:18 -0700
        Re: H8S2676 faulty chips Mike Perkins <spam@spam.com> - 2019-05-30 19:39 +0100
          Re: H8S2676 faulty chips Chris <xxx.syseng.yyy@gfsys.co.uk> - 2019-05-31 13:02 +0100
            Re: H8S2676 faulty chips Theo <theom+news@chiark.greenend.org.uk> - 2019-05-31 17:18 +0100
              Re: H8S2676 faulty chips xyman8@gmail.com - 2019-06-07 01:50 -0700
                Re: H8S2676 faulty chips Theo <theom+news@chiark.greenend.org.uk> - 2019-06-07 12:56 +0100
                  Re: H8S2676 faulty chips xyman8@gmail.com - 2019-06-08 15:50 -0700
                    Re: H8S2676 faulty chips Paul Rubin <no.email@nospam.invalid> - 2019-06-08 17:44 -0700
                    Re: H8S2676 faulty chips Rick C <gnuarm.deletethisbit@gmail.com> - 2019-06-08 20:43 -0700
                      Re: H8S2676 faulty chips xyman8@gmail.com - 2019-06-08 22:40 -0700
                        Re: H8S2676 faulty chips Paul Rubin <no.email@nospam.invalid> - 2019-06-08 23:03 -0700
                          Re: H8S2676 faulty chips xyman8@gmail.com - 2019-06-08 23:08 -0700
                        Re: H8S2676 faulty chips Rick C <gnuarm.deletethisbit@gmail.com> - 2019-06-09 12:46 -0700
                          Re: H8S2676 faulty chips Paul Rubin <no.email@nospam.invalid> - 2019-06-09 12:53 -0700
                            Re: H8S2676 faulty chips Rick C <gnuarm.deletethisbit@gmail.com> - 2019-06-09 13:27 -0700
                              Re: H8S2676 faulty chips Theo <theom+news@chiark.greenend.org.uk> - 2019-06-11 14:05 +0100
                                Re: H8S2676 faulty chips xyman8@gmail.com - 2019-06-12 01:11 -0700
                            Re: H8S2676 faulty chips Stef <stef33d@yahooI-N-V-A-L-I-D.com.invalid> - 2019-06-11 11:10 +0200

#28645 — H8S2676 faulty chips

We use the Hitachi H8S2676 chip in one of our products and a while ago Renesas told us the chip was going out of production.  Someone here forgot to do a lifetime buy from them in time but we recently did a lifetime buy of several thousand from a third party company.  In the first production run of over 100 boards, every single one failed the manufacturing test.  It seems at least one of the problems is when it tries to access internal memory from external memory it gets zeros - or something like that (don't have the exact details).  We suspect the chips are counterfeit or seconds or something.  

Has anyone heard of this kind of problem?  

Does anyone know a reliable source of H8S2676 chips?

[toc] | [next] | [standalone]

#28646

On 5/29/2019 11:23, xyman8@gmail.com wrote:
> We use the Hitachi H8S2676 chip in one of our products and a while ago Renesas told us the chip was going out of production.  Someone here forgot to do a lifetime buy from them in time but we recently did a lifetime buy of several thousand from a third party company.  In the first production run of over 100 boards, every single one failed the manufacturing test.  It seems at least one of the problems is when it tries to access internal memory from external memory it gets zeros - or something like that (don't have the exact details).  We suspect the chips are counterfeit or seconds or something.
> 
> Has anyone heard of this kind of problem?
> 
> Does anyone know a reliable source of H8S2676 chips?
> 

Can't be of any help but thanks for sharing the info.

Can you perhaps name the third party company, this might
help other people in the future.

I have not been burned yet but I have heard plenty of
stories about chips bought not via the large, known
distributors. There looks to be quite a busyness living
on selling you "whatever you need, we'll just have to
imprint the logo on a fitting package".

OTOH we bought some fast HV diodes via ebay twice (not sure if it
was the same seller both times), were OK.

Dimiter


======================================================
Dimiter Popoff, TGI             http://www.tgi-sci.com
======================================================
http://www.flickr.com/photos/didi_tgi/

[toc] | [prev] | [next] | [standalone]

#28647

On Thursday, May 30, 2019 at 4:22:45 AM UTC+12, Dimiter wrote:
> On 5/29/2019 11:23, xy...@gmail.com wrote:
> > We use the Hitachi H8S2676 chip in one of our products and a while ago Renesas told us the chip was going out of production.  Someone here forgot to do a lifetime buy from them in time but we recently did a lifetime buy of several thousand from a third party company.  In the first production run of over 100 boards, every single one failed the manufacturing test.  It seems at least one of the problems is when it tries to access internal memory from external memory it gets zeros - or something like that (don't have the exact details).  We suspect the chips are counterfeit or seconds or something.
> > 
> > Has anyone heard of this kind of problem?
> > 
> > Does anyone know a reliable source of H8S2676 chips?
> > 
> 
> Can't be of any help but thanks for sharing the info.
> 
> Can you perhaps name the third party company, this might
> help other people in the future.
> 
> I have not been burned yet but I have heard plenty of
> stories about chips bought not via the large, known
> distributors. There looks to be quite a busyness living
> on selling you "whatever you need, we'll just have to
> imprint the logo on a fitting package".
> 
> OTOH we bought some fast HV diodes via ebay twice (not sure if it
> was the same seller both times), were OK.
> 
> Dimiter

Thanks for the reply.  I can't name the company as we're still not exactly sure of what's going on.  I would have thought that anyone with the technology to manufacture these kind of chips would have the ability to do it correctly but it seems they don't.

[toc] | [prev] | [next] | [standalone]

#28648

On 30/05/2019 12:18, xyman8@gmail.com wrote:
> On Thursday, May 30, 2019 at 4:22:45 AM UTC+12, Dimiter wrote:
>> On 5/29/2019 11:23, xy...@gmail.com wrote:
>>> We use the Hitachi H8S2676 chip in one of our products and a
>>> while ago Renesas told us the chip was going out of production.
>>> Someone here forgot to do a lifetime buy from them in time but we
>>> recently did a lifetime buy of several thousand from a third
>>> party company.  In the first production run of over 100 boards,
>>> every single one failed the manufacturing test.  It seems at
>>> least one of the problems is when it tries to access internal
>>> memory from external memory it gets zeros - or something like
>>> that (don't have the exact details).  We suspect the chips are
>>> counterfeit or seconds or something.
>>> 
>>> Has anyone heard of this kind of problem?
>>> 
>>> Does anyone know a reliable source of H8S2676 chips?
>>> 
>> 
>> Can't be of any help but thanks for sharing the info.
>> 
>> Can you perhaps name the third party company, this might help other
>> people in the future.
>> 
>> I have not been burned yet but I have heard plenty of stories about
>> chips bought not via the large, known distributors. There looks to
>> be quite a busyness living on selling you "whatever you need, we'll
>> just have to imprint the logo on a fitting package".
>> 
>> OTOH we bought some fast HV diodes via ebay twice (not sure if it 
>> was the same seller both times), were OK.
>> 
>> Dimiter
> 
> Thanks for the reply.  I can't name the company as we're still not
> exactly sure of what's going on.  I would have thought that anyone
> with the technology to manufacture these kind of chips would have the
> ability to do it correctly but it seems they don't.

It is possible the chips are genuine but a different variant, destined 
for a large manufacturer. It is also possible that access to the 
internal memory is a feature that is turned on by setting a bit not 
normally accessible.

I presume the company you purchased through is a long time served 
supplier? Can they not assist and confirm their source is legitimate?

-- 
Mike Perkins
Video Solutions Ltd
www.videosolutions.ltd.uk

[toc] | [prev] | [next] | [standalone]

#28649

On 05/30/19 19:39, Mike Perkins wrote:
> On 30/05/2019 12:18, xyman8@gmail.com wrote:
>> On Thursday, May 30, 2019 at 4:22:45 AM UTC+12, Dimiter wrote:
>>> On 5/29/2019 11:23, xy...@gmail.com wrote:
>>>> We use the Hitachi H8S2676 chip in one of our products and a
>>>> while ago Renesas told us the chip was going out of production.
>>>> Someone here forgot to do a lifetime buy from them in time but we
>>>> recently did a lifetime buy of several thousand from a third
>>>> party company. In the first production run of over 100 boards,
>>>> every single one failed the manufacturing test. It seems at
>>>> least one of the problems is when it tries to access internal
>>>> memory from external memory it gets zeros - or something like
>>>> that (don't have the exact details). We suspect the chips are
>>>> counterfeit or seconds or something.
>>>>
>>>> Has anyone heard of this kind of problem?
>>>>
>>>> Does anyone know a reliable source of H8S2676 chips?
>>>>
>>>
>>> Can't be of any help but thanks for sharing the info.
>>>
>>> Can you perhaps name the third party company, this might help other
>>> people in the future.
>>>
>>> I have not been burned yet but I have heard plenty of stories about
>>> chips bought not via the large, known distributors. There looks to
>>> be quite a busyness living on selling you "whatever you need, we'll
>>> just have to imprint the logo on a fitting package".
>>>
>>> OTOH we bought some fast HV diodes via ebay twice (not sure if it was
>>> the same seller both times), were OK.
>>>
>>> Dimiter
>>
>> Thanks for the reply. I can't name the company as we're still not
>> exactly sure of what's going on. I would have thought that anyone
>> with the technology to manufacture these kind of chips would have the
>> ability to do it correctly but it seems they don't.
>
> It is possible the chips are genuine but a different variant, destined
> for a large manufacturer. It is also possible that access to the
> internal memory is a feature that is turned on by setting a bit not
> normally accessible.
>
> I presume the company you purchased through is a long time served
> supplier? Can they not assist and confirm their source is legitimate?
>

Seems to me that if you are buying from any unproven vendor, you would
get samples and verify that they were good before a bulk buy ?. With so
much fraudulent stuff coming from places like China, we all need to be
more vigilant...

[toc] | [prev] | [next] | [standalone]

#28650

Chris <xxx.syseng.yyy@gfsys.co.uk> wrote:
> Seems to me that if you are buying from any unproven vendor, you would
> get samples and verify that they were good before a bulk buy ?. With so
> much fraudulent stuff coming from places like China, we all need to be
> more vigilant...

Common practice at dodgy vendors (of lithium batteries in this case, but
probably applies to semiconductors too) is to supply small orders from
genuine stock, and counterfeits for volume orders.  That means the samples
work fine, and also means any bad feedback on marketplaces like Alibaba is
drowned out by good feedback from small customers.

Theo

[toc] | [prev] | [next] | [standalone]

#28661

Are there are semiconductor manufacturers that could make 6000 or so of these chips if we paid them enough money?

[toc] | [prev] | [next] | [standalone]

#28662

xyman8@gmail.com wrote:
> Are there are semiconductor manufacturers that could make 6000 or so of these chips if we paid them enough money?

Theoretically: yes
Practically: no

To do it, you'd need the design files from Renesas.  That includes any
proprietary blobs inserted by the foundry.  In practice you'd need to use
the same foundry they used, because the design will be targeted to a
particular process.

Next you need a packaging plan.  You need someone to take your silicon
dice, and bond it into a package.  That may or may not be the same as the
existing package - depends on the capability they have.

Then you need a testing strategy.


It's 'only' half the work of designing a new chip from scratch.

Instead I'd be looking at redesigning the product - that's probably a lot
easier.

Theo

[toc] | [prev] | [next] | [standalone]

#28663

Thanks.
I'm told that we've been unable to determine where the chips were made.  Some people here are speculating that the chips were actually made by Renesas but were rejected during manufacture.  Does this seem believable?  

In a production run of 100 boards here, all boards failed.  We bought 6000 of these chips.  I find it hard to believe that Renesas would have as many as 6000 failures or that they would allow these chips to escape their factory.  The reason I'm interested is that if the chips were made by a non Renesas company, it's possible they have a solution and secondly, how did they get the design files?

[toc] | [prev] | [next] | [standalone]

#28664

xyman8@gmail.com writes:
> In a production run of 100 boards here, all boards failed.  We bought
> 6000 of these chips.  I find it hard to believe that Renesas would
> have as many as 6000 failures

Maybe you could contact Renesas and offer to pay them to examine some of
those chips and see if they are legit.

[toc] | [prev] | [next] | [standalone]

#28665

On Saturday, June 8, 2019 at 6:50:05 PM UTC-4, xym...@gmail.com wrote:
> Thanks.
> I'm told that we've been unable to determine where the chips were made.  Some people here are speculating that the chips were actually made by Renesas but were rejected during manufacture.  Does this seem believable?  
> 
> In a production run of 100 boards here, all boards failed.  We bought 6000 of these chips.  I find it hard to believe that Renesas would have as many as 6000 failures or that they would allow these chips to escape their factory.  The reason I'm interested is that if the chips were made by a non Renesas company, it's possible they have a solution and secondly, how did they get the design files?

Uhhh, design files?  They could have beach sand inside for all you know, right?  It is not uncommon for counterfeit chips to have totally bogus chips inside the package so they x-ray looking like something is in there.  

What happened with the company you bought them from?  Have they told you where they got them? 

-- 

  Rick C.

  - Get 1,000 miles of free Supercharging
  - Tesla referral code - https://ts.la/richard11209

[toc] | [prev] | [next] | [standalone]

#28666

On Sunday, June 9, 2019 at 3:43:17 PM UTC+12, Rick C wrote:
> On Saturday, June 8, 2019 at 6:50:05 PM UTC-4, xym...@gmail.com wrote:
> > Thanks.
> > I'm told that we've been unable to determine where the chips were made.  Some people here are speculating that the chips were actually made by Renesas but were rejected during manufacture.  Does this seem believable?  
> > 
> > In a production run of 100 boards here, all boards failed.  We bought 6000 of these chips.  I find it hard to believe that Renesas would have as many as 6000 failures or that they would allow these chips to escape their factory.  The reason I'm interested is that if the chips were made by a non Renesas company, it's possible they have a solution and secondly, how did they get the design files?
> 
> Uhhh, design files?  They could have beach sand inside for all you know, right?  It is not uncommon for counterfeit chips to have totally bogus chips inside the package so they x-ray looking like something is in there.  
> 
> What happened with the company you bought them from?  Have they told you where they got them? 
> 

Nope, they don't have beach sand.  They're functional to a point.  They were either made by Renesas or someone with the design files and the equipment to make them.  However I guess the chance that they were made by a legitimate manufacturer must be pretty slim as we wouldn't get 100% failure rate if they were legitimate.

[toc] | [prev] | [next] | [standalone]

#28667

xyman8@gmail.com writes:
> However I guess the chance that they were made by a legitimate
> manufacturer must be pretty slim as we wouldn't get 100% failure rate
> if they were legitimate.

Have you swapped a chip from an old, good board into one of your new
failing boards?  Did it work?

[toc] | [prev] | [next] | [standalone]

#28668

On Sunday, June 9, 2019 at 6:03:26 PM UTC+12, Paul Rubin wrote:
> xy...@gmail.com writes:
> > However I guess the chance that they were made by a legitimate
> > manufacturer must be pretty slim as we wouldn't get 100% failure rate
> > if they were legitimate.
> 
> Have you swapped a chip from an old, good board into one of your new
> failing boards?  Did it work?

I wasn't involved in the investigation but the guys who did it are fairly capable and they've established the new chips are only partially functional  - but I will check.

[toc] | [prev] | [next] | [standalone]

#28669

On Sunday, June 9, 2019 at 1:40:31 AM UTC-4, xym...@gmail.com wrote:
> On Sunday, June 9, 2019 at 3:43:17 PM UTC+12, Rick C wrote:
> > On Saturday, June 8, 2019 at 6:50:05 PM UTC-4, xym...@gmail.com wrote:
> > > Thanks.
> > > I'm told that we've been unable to determine where the chips were made.  Some people here are speculating that the chips were actually made by Renesas but were rejected during manufacture.  Does this seem believable?  
> > > 
> > > In a production run of 100 boards here, all boards failed.  We bought 6000 of these chips.  I find it hard to believe that Renesas would have as many as 6000 failures or that they would allow these chips to escape their factory.  The reason I'm interested is that if the chips were made by a non Renesas company, it's possible they have a solution and secondly, how did they get the design files?
> > 
> > Uhhh, design files?  They could have beach sand inside for all you know, right?  It is not uncommon for counterfeit chips to have totally bogus chips inside the package so they x-ray looking like something is in there.  
> > 
> > What happened with the company you bought them from?  Have they told you where they got them? 
> > 
> 
> Nope, they don't have beach sand.  They're functional to a point.  They were either made by Renesas or someone with the design files and the equipment to make them.  However I guess the chance that they were made by a legitimate manufacturer must be pretty slim as we wouldn't get 100% failure rate if they were legitimate.

What do they do right and what do they do wrong? 

If they are largely functional but are failing some small detail it is possible these are clone chips, a fresh design, not a literal duplicate of the original device.  

-- 

  Rick C.

  + Get 1,000 miles of free Supercharging
  + Tesla referral code - https://ts.la/richard11209

[toc] | [prev] | [next] | [standalone]

#28670

Rick C <gnuarm.deletethisbit@gmail.com> writes:
> If they are largely functional but are failing some small detail it is
> possible these are clone chips, a fresh design, not a literal
> duplicate of the original device.

It's weird: are you saying some illicit chipmaker did an unauthorized
redesign of this somewhat obscure and not exactly simple part, in order
to sell them under counterfeit labels, and left out functionality?  Does
that really happen?  Now that there is a market for working units of the
discontinued chip, could the illicit maker issue an erratum and spin
another run?  There is at least one possible customer...

[toc] | [prev] | [next] | [standalone]

#28671

On Sunday, June 9, 2019 at 3:53:58 PM UTC-4, Paul Rubin wrote:
> Rick C <gnuarm.deletethisbit@gmail.com> writes:
> > If they are largely functional but are failing some small detail it is
> > possible these are clone chips, a fresh design, not a literal
> > duplicate of the original device.
> 
> It's weird: are you saying some illicit chipmaker did an unauthorized
> redesign of this somewhat obscure and not exactly simple part, in order
> to sell them under counterfeit labels, and left out functionality?  Does
> that really happen?  Now that there is a market for working units of the
> discontinued chip, could the illicit maker issue an erratum and spin
> another run?  There is at least one possible customer...

You say "obscure", but no one makes low volume MCUs.  Most likely you are buying a chip that is not so obscure other than perhaps the combination of speed, package, temperature, etc that you chose.  Once they have made a die, they are going to sell it into each part number they can.  It might not even be a new chip design.  It's not impossible to reprogram the microcode on an existing chip to emulate another instruction set.  Then they only need to deal with the pin out issues.  

There was a serial port to USB converter chip that was cloned by reprogramming a different MCU chip.  It pretty much worked the same, but not quite.  The company making the original chip reverse engineered the clone enough to figure out how to reprogram it again, bricking the cable when used with their driver.  That didn't set well with Microsoft since they were distributing the driver with their software. 

I don't know what you have, but it sounds like you don't either.  I think you need more detail from your tech guys to figure out if you have a clone or a factory reject.  I would say turning a chip over to the factory would be useful, but they may not be interested in helping since they no longer make that chip. 

This is exactly why it's not a great idea to buy parts from third parties without some sort of testing and guarantee.  I was going to buy some EOL FPGAs that were at a good price at a reputable B list seller.  They said they had tested them and would guarantee the purchase price, but my assembly house said they wouldn't be responsible for the rework if they were faulty.  So I paid the higher price to get them from Arrow who still has stock more than five years after EOL. 

-- 

  Rick C.

  -- Get 1,000 miles of free Supercharging
  -- Tesla referral code - https://ts.la/richard11209

[toc] | [prev] | [next] | [standalone]

#28673

Rick C <gnuarm.deletethisbit@gmail.com> wrote:
> You say "obscure", but no one makes low volume MCUs.  Most likely you are
> buying a chip that is not so obscure other than perhaps the combination of
> speed, package, temperature, etc that you chose.  Once they have made a
> die, they are going to sell it into each part number they can.  It might
> not even be a new chip design.  It's not impossible to reprogram the
> microcode on an existing chip to emulate another instruction set.  Then
> they only need to deal with the pin out issues.

I doubt this is a clone - I'd have thought this family was an unlikely
target for cloning (they don't seem high margin parts).

But you might be right in that it could be one part of the family remarked
as another.  For instance, MCU families like the Cypress PSoC are all the
same die internally, but with different features enabled by firmware.  An
unscrupulous supplier could remark the cheap chips as the expensive chips -
either without any changes (so you'd only notice if you try and use the
third UART, or whatever), or with hacked firmware to enable extra features.

But, as mentioned upthread, it would be worth checking whether there have
been any production changes to the genuine chip.  It could be the stock you
bought was a v1 chip while the manufacturer was selling v2 chips until EOL.
They might behave functionally the same, until you deviate slightly into
undefined behaviour that differs between the different versions.

Theo

[toc] | [prev] | [next] | [standalone]

#28675

The engineers here think the faulty chips were made by Renesas but rejected during manufacturing.  We found short a short term supply and we're going ahead with re-designing the board.  Thanks for all the comments.

[toc] | [prev] | [next] | [standalone]

#28672

On 2019-06-09 Paul Rubin wrote in comp.arch.embedded:
> Rick C <gnuarm.deletethisbit@gmail.com> writes:
>> If they are largely functional but are failing some small detail it is
>> possible these are clone chips, a fresh design, not a literal
>> duplicate of the original device.
>
> It's weird: are you saying some illicit chipmaker did an unauthorized
> redesign of this somewhat obscure and not exactly simple part, in order
> to sell them under counterfeit labels, and left out functionality?  Does
> that really happen?  Now that there is a market for working units of the
> discontinued chip, could the illicit maker issue an erratum and spin
> another run?  There is at least one possible customer...

Maybe not even illicit or unauthorized (but you could probably check
that). I remember a case with an 68000 processor (I may have mentioned
that here before, but it seems apropriate in this topic)

We used a CPU card with a 68000 on it a long time ago. These boards were
in a rack, connecting to other cards via an 8-bit bus. This design was
in use for many years and boards were equiped with Motorola and TI
versions of the 68000 without issues. But then we created a new daughter 
card for the 8-bit bus. This card worked fine when a cpu card with a
Motorola 68000 was used, but failed when a TI 68000 was used.

I think that we can be pretty sure that the TI version was not
unauthorized, but I never did check. ;-)
So what made the TI versions fail?

I do not remember the exact details, but the cause was in 16-bit access
over the 8-bit bus, which depended on two consecutive accesses. The bus
control signals were derived from the 68000 bus signals by a PAL. In
this PAL, one of the 68000 signals was not included in the generation
of the 8-bit bus signals. But due to the timing of the other signals,
the timing of the bus signals was correct with the Motorola version of
the 68000. With the TI version however, a double access was generated
on both 8-bit halves of the 16-bit access, resulting in invalid
transfers.

After studying the datasheets, we concluded that both processors 
generated signals within specification. And the error was in our
generation of the 8-bit bus control signals. The design was probably
originally tested with the Motorola processor and afterwards only
a functional test was performed with the TI version. Which worked
because all cards at that time only used 8-bit access.

A simple change in the PAL code fixed this problem and then both
Motorola and TI 68000's worked as intended.

So what I am trying to say is that there is a possibility that your
new processors are in fact functioning correctly, but that an error
/omission in your board causes the difference in functionality.
Could be something as simple(?) as changed timing due to a new
manufacturing process. Something worth to check, in my opinion.


-- 
Stef    (remove caps, dashes and .invalid from e-mail address to reply by mail)

The lines are all busy (busied out, that is -- why let them in to begin with?).

[toc] | [prev] | [standalone]

Back to top | Article view | comp.arch.embedded

csiph-web