Hi I designed an SH3-7709 (80MHz) based industrial controller about 2 years ago, which has been used in a number of applications. This has worked fine after initial debugging and board validation etc. and has been in production now for over 18 months. Recently the contract assembly house was changed to get better quality and value, as the manufacturing volumes are beginning to grow. I have just received the first of the production boards from this new supplier and I have a problem! The CPU (HD6417709F80) doesn't function after power-on-reset; i.e the 10MHz crystal feeding the internal oscillator doesn't start at switch-on. Only when I inject a manual reset into the cpu that it starts to oscillate! I have checked all the component specifications etc and I don't see any difference between the old board and the newly built one. The reset circuit (using MAX811/5v and MAX6361/3v3 in tandem) generates a clean POR after switch-on. The load capacitors for the crystal are correct value (after all they function after a manual reset!). All the other components for the PLL 1 & 2, and all the necessary pull-ups and mode switches for the cpu are correct and as before. There are only two unrelated and minor changes with the board artwork; the board manufacturing spec is the same (4layers with ground and power planes, 1oz copper, gold-flash etc). The finished board from the new supplier looks great; it is very clean and well assembled. Compared to this I still have my very first hand-assembled prototype board with a bird's nest of cuts and links which still works fine. After the power-on, and me providing a manual reset the board and all the software works fine as per the designed functionality. I am totally stumped at the moment! What has gone wrong? I need this board to work in the machines at power-on without fail. I am beginning to doubt my sanity or that something has changed in the 7709 that nobody has told me about. Latest data from Renesas now refers to 7709S and the hardware manual I downloaded today still gives identical info as the 7709/80MHz regarding the oscillator/PLLs and reset conditions etc. Can you please help? Or do you know someone who can? Thanks in anticipation, Khalid
SH3-7709 (80MHz) board - Power-on Reset Problem
Started by ●January 16, 2004
Reply by ●January 16, 20042004-01-16
"Khalid" <alvidesign@aol.com> wrote in message news:f80840ad.0401161027.2df5b707@posting.google.com...> The CPU (HD6417709F80) doesn't function after power-on-reset; i.e the > 10MHz crystal feeding the internal oscillator doesn't start at > switch-on. Only when I inject a manual reset into the cpu that it > starts to oscillate!What is the reset period of each reset controller? Could it be that your power supply is coming up more slowly than you expect? Are both rails (5V and 3.3V) coming up together? How fast? I've used various Maxim reset controllers, but not these ones - does your circuit guarantee that both rails are within tolerance (from the POV of the CPU) before the reset period starts? Steve http://www.fivetrees.com http://www.sfdesign.co.uk
Reply by ●January 16, 20042004-01-16
Steve at fivetrees wrote:> "Khalid" <alvidesign@aol.com> wrote in message > news:f80840ad.0401161027.2df5b707@posting.google.com... > >>The CPU (HD6417709F80) doesn't function after power-on-reset; i.e the >>10MHz crystal feeding the internal oscillator doesn't start at >>switch-on. Only when I inject a manual reset into the cpu that it >>starts to oscillate! > > > What is the reset period of each reset controller? Could it be that your > power supply is coming up more slowly than you expect? Are both rails (5V > and 3.3V) coming up together? How fast? > > I've used various Maxim reset controllers, but not these ones - does your > circuit guarantee that both rails are within tolerance (from the POV of the > CPU) before the reset period starts?I'm not familiar with the CPU in question, but I did have a similar problem a couple of years ago. I was using a Cypress clock generator and the part would not start if the power ramped up slowly. I had used the parts with no problem for a couple years before and this hit just as we were gearing up to do a big, important build. It turned out the Cypress had done a die shrink on the part and had broken it. After some spirited discussion, they did the right thing and got me good parts and modified their production test to include slow power rampup.
Reply by ●January 17, 20042004-01-17
Hello Steve, Thanks for your reply. Following your suggestions, I have looked at the reset timings more carefully and compared it with the cpu data book. Both the 5V and the 3V3 rails come from the same switchmode package. As the power is switched on, 5V rises and is stable at 5.04V after 300uS (microSecond). The 3V3 rail is much slower to rise and is stable at 3.28V after 1.6mS (milliSecond). The reset signal from the 5V monitor feeds into the 3V3 monitor which then drives the reset pin of the cpu. While there may be a 0.5V glitch on the reset line when the ON switch is thrown, this line is kept low (0V) by the Maxim6361 device for a minimum of 150mS (millisecond) AFTER the slower 3V3 rail has stabilised. Typically this reset period is closer to 190mS before the reset circuit drives it high to bring the cpu out of reset. This compares fine with the cpu data which recommends a minimum reset timing of 10mS to allow for the cpu internal clock oscillations to stabilise. So, I can't see a problem with reset timings itself. It is the intrnal cpu clock which fails to start after switch on; when I monitor the clock output line (which is used elsewhere on the board, this just goes high and stays there after power-on, wtihout any hint of oscillations. It is only when I short the manual reset input pin of the Maxim monitor that the oscillations start and the cpu comes to life. I would very much appreciate your further thoughts! Best regards Khalid "Steve at fivetrees" <steve@NOSPAMTAfivetrees.com> wrote in message news:<4008384f$0$61069$65c69314@mercury.nildram.net>...> "Khalid" <alvidesign@aol.com> wrote in message > news:f80840ad.0401161027.2df5b707@posting.google.com... > > The CPU (HD6417709F80) doesn't function after power-on-reset; i.e the > > 10MHz crystal feeding the internal oscillator doesn't start at > > switch-on. Only when I inject a manual reset into the cpu that it > > starts to oscillate! > > What is the reset period of each reset controller? Could it be that your > power supply is coming up more slowly than you expect? Are both rails (5V > and 3.3V) coming up together? How fast? > > I've used various Maxim reset controllers, but not these ones - does your > circuit guarantee that both rails are within tolerance (from the POV of the > CPU) before the reset period starts? > > Steve > http://www.fivetrees.com > http://www.sfdesign.co.uk
Reply by ●January 17, 20042004-01-17
Dear Jim Many thanks for your reply. Please see my reply to Steve regarding power supply timings. I should take heart from your experience! While I am still looking for a hardware solution, I am coming round to the idea that I may well have bad 7709 cpu ICs fitted; persuading Hitachi/Renesas of my problem is not going to be easy - so wish me luck! Khalid Jim Stewart <jstewart@jkmicro.com> wrote in message news:<H_GdnUuhveukppXdRVn-vg@omsoft.com>...> Steve at fivetrees wrote: > > "Khalid" <alvidesign@aol.com> wrote in message > > news:f80840ad.0401161027.2df5b707@posting.google.com... > > > >>The CPU (HD6417709F80) doesn't function after power-on-reset; i.e the > >>10MHz crystal feeding the internal oscillator doesn't start at > >>switch-on. Only when I inject a manual reset into the cpu that it > >>starts to oscillate! > > > > > > What is the reset period of each reset controller? Could it be that your > > power supply is coming up more slowly than you expect? Are both rails (5V > > and 3.3V) coming up together? How fast? > > > > I've used various Maxim reset controllers, but not these ones - does your > > circuit guarantee that both rails are within tolerance (from the POV of the > > CPU) before the reset period starts? > > I'm not familiar with the CPU in question, but I did > have a similar problem a couple of years ago. I was using > a Cypress clock generator and the part would not start if > the power ramped up slowly. I had used the parts with no > problem for a couple years before and this hit just as we > were gearing up to do a big, important build. > > It turned out the Cypress had done a die shrink on > the part and had broken it. After some spirited discussion, > they did the right thing and got me good parts and modified > their production test to include slow power rampup.
Reply by ●January 17, 20042004-01-17
Dear Khalid, Khalid wrote:> Both the 5V and the 3V3 rails come from the same switchmode package. > As the power is switched on, 5V rises and is stable at 5.04V after > 300uS (microSecond). The 3V3 rail is much slower to rise and is stable > at 3.28V after 1.6mS (milliSecond). > > The reset signal from the 5V monitor feeds into the 3V3 monitor which > then drives the reset pin of the cpu. While there may be a 0.5V glitch > on the reset line when the ON switch is thrown, this line is kept low > (0V) by the Maxim6361 device for a minimum of 150mS (millisecond) > AFTER the slower 3V3 rail has stabilised. Typically this reset period > is closer to 190mS before the reset circuit drives it high to bring > the cpu out of reset.I had similar problems (with a different CPU) and it turns out that simple crystal oscillators circuits need up to 1 second after power-up to oscillate properly. Interestingly, there was an oscillation visible on the scope, but obvioulsy not stable, so the processor did not start properly. For a test I suggest you use a longer RESET time (500 to 1000 ms) and see if that was the problem. HTH Wolfgang -- From-address is Spam trap Use: wolfgang (dot) mahringer (at) sbg (dot) at
Reply by ●January 19, 20042004-01-19
Hello Wolfgang Thank you for your suggestion, which I did consider and tested for. However, the actual problems turns out to be more mundane and completely unexpected. It turns out that on this board vrsion a ground connection is missing! I finally homed in on this yesterday afternoon, as I searched for a solution in despair. This was an error I had spotted two years ago when I was checking the artwork for the first production boards and had it corrected, so it never went into production with the error. But this time when I had some minor (and totally unrelated) changes made to the artwork for the current set of boards, the pcb designer, who it turns out had not updated his own archive with the corrected artwork, re-introduced it!!! This missing ground commits a couple of mode pins on the 7709 cpu to 0V. Since the basic operating mode for the processor was indeterminate, it never booted up properly, hence all the aggravation I had last week. The solution was a quarter-inch long fine wire link. Today, the artwork engineer has had his wrists well and truly slapped! So there is a happy ending after all. Many thanks to everyone who took the time and trouble to reply and make suggestions; this kept me going. Anyone with similar tales of woe? How can one prevent such a thing happening again? Best regards Khalid ps: by the way this is what we use these boards for: http://www.yorkdetection.com Wolfgang Mahringer <yeti201@gmx.at> wrote in message news:<ICcOb.1$xc4.219260@news.salzburg-online.at>...> Dear Khalid, > > Khalid wrote: > > Both the 5V and the 3V3 rails come from the same switchmode package. > > As the power is switched on, 5V rises and is stable at 5.04V after > > 300uS (microSecond). The 3V3 rail is much slower to rise and is stable > > at 3.28V after 1.6mS (milliSecond). > > > > The reset signal from the 5V monitor feeds into the 3V3 monitor which > > then drives the reset pin of the cpu. While there may be a 0.5V glitch > > on the reset line when the ON switch is thrown, this line is kept low > > (0V) by the Maxim6361 device for a minimum of 150mS (millisecond) > > AFTER the slower 3V3 rail has stabilised. Typically this reset period > > is closer to 190mS before the reset circuit drives it high to bring > > the cpu out of reset. > > I had similar problems (with a different CPU) and it turns out that > simple crystal oscillators circuits need up to 1 second after power-up > to oscillate properly. > Interestingly, there was an oscillation visible on the scope, but > obvioulsy not stable, so the processor did not start properly. > > For a test I suggest you use a longer RESET time (500 to 1000 ms) and > see if that was the problem. > > HTH > Wolfgang