Many thanks to "bill knight", Chris (kris?) and Al, and others for their kind responses to my inqury about Flash Corruption. I am currently juicing my Vcc from 3.3V to 3.6V, so as to stay within spec, having an 8MHz crystal, even though I do not believe this is a root cause of my problem. However, if I am experiencing Power supply noise, say something additive(or subtractive), then an extra 0.3V will help my immunity, to keep Vcc happily high enough for the flash to function during field upgrades. In reponse to some of your("your"-pluaral) points: My Vcc bypassing consists of 4-220uF tantalum, plus a number of 1nF and 1uF tantalums. One of you indicated that you could "count on one hand" the number of flash failures you have experienced. I know this was meant to encourage me, but frankly I found that news to be devastating. I was rather hoping, you know, for zero failures. We will never know, for sure, in your case, because some of your products get overclocked at 16MHz, if I read correctly. But this is still a troubling data point. Finally, there is something in my system which might be playing a role here. I have a 1 Watt RF transmitter, about 900MHz, and the antenna is sitting just inches from my MSP430F149. What might be the chances of THAT dithering my Vcc, or maybe even my flash? I am still wondering, actually, how I might instrument that thing to monitor Vcc. But until then, your insights are welcome. Regards, Sam
Could a 1Watt RF Field Cause Flash Corruption?
Started by ●December 19, 2003
Reply by ●December 19, 20032003-12-19
Samuel Waters wrote: > Many thanks to "bill knight", Chris (kris?) and Al, and others for > their kind responses to my inqury about Flash Corruption. > > I am currently juicing my Vcc from 3.3V to 3.6V, so as to stay within > spec, having an 8MHz crystal, even though I do not believe this is a > root cause of my problem. > > However, if I am experiencing Power supply noise, say something > additive(or subtractive), then an extra 0.3V will help my immunity, > to keep Vcc happily high enough for the flash to function during > field upgrades. > > In reponse to some of your("your"-pluaral) points: > > My Vcc bypassing consists of 4-220uF tantalum, plus a number of 1nF > and 1uF tantalums. I agree with several others, you have too much bypassing here. > > One of you indicated that you could "count on one hand" the number of > flash failures you have experienced. I know this was meant to > encourage me, but frankly I found that news to be devastating. I was > rather hoping, you know, for zero failures. We will never know, for > sure, in your case, because some of your products get overclocked at > 16MHz, if I read correctly. But this is still a troubling data point. You must take this in context. I have been testing MSP430 systems almost to destruction for 3 years now. many systems run continuously. The failures of flash that I have seen occurred during a flash write cycle excercise, designed to determine the realistic number of flash writes you might expect. Nothing is defect free, so my code attempts re-writes, logs anything that takes more than 2 attempts, maintains failed bit/byte logs and does re-attempt these at later times. but it treats the flash like RAM, it continuously writes data to flash. When the defined flash file is filled it erases segments, and starts again. failures have been due to aging. But that aging requires typically many thousands of write cycles. IIRC only 2 failed below the spec. ALL at 3V3, tests conducted from my deep freeze to 70C. I wouldn't think that cause for concern. Any reasonably robust filing system can handle that. I have thousands of boards out there using the flash as a filing system, none have failed in the field. > > Finally, there is something in my system which might be playing a > role here. > > I have a 1 Watt RF transmitter, about 900MHz, and the antenna is > sitting just inches from my MSP430F149. What might be the chances of > THAT dithering my Vcc, or maybe even my flash? I am still wondering, > actually, how I might instrument that thing to monitor Vcc. But > until then, your insights are welcome. I run RF with many of my designs and haven't found it a problem, however the highest power unscreened system I have is 200mW. Mostly I screen the RF and micro. Certainly add rf decoupling in the power supply lines Cheers Al > > Regards, > Sam > > > > > . > > > > > > >
Reply by ●December 19, 20032003-12-19
I want to take your friendly inputs seriously, now. Please help me understand a little better. The product is up-and-running for a long time before flash upgrades occur. When questioning the bypass strategy, were you thinking of a power-on delay dealy proble, or were you thinking of some inherent characteristic of the devices which might amplify their noise? Thanks, Sam ========================== --- In msp430@msp4..., onestone <onestone@b...> wrote: > > Samuel Waters wrote: > > Many thanks to "bill knight", Chris (kris?) and Al, and others for > > their kind responses to my inqury about Flash Corruption. > > > > I am currently juicing my Vcc from 3.3V to 3.6V, so as to stay within > > spec, having an 8MHz crystal, even though I do not believe this is a > > root cause of my problem. > > > > However, if I am experiencing Power supply noise, say something > > additive(or subtractive), then an extra 0.3V will help my immunity, > > to keep Vcc happily high enough for the flash to function during > > field upgrades. > > > > In reponse to some of your("your"-pluaral) points: > > > > My Vcc bypassing consists of 4-220uF tantalum, plus a number of 1nF > > and 1uF tantalums. > > I agree with several others, you have too much bypassing here. > > > > > One of you indicated that you could "count on one hand" the number of > > flash failures you have experienced. I know this was meant to > > encourage me, but frankly I found that news to be devastating. I was > > rather hoping, you know, for zero failures. We will never know, for > > sure, in your case, because some of your products get overclocked at > > 16MHz, if I read correctly. But this is still a troubling data point. > > You must take this in context. I have been testing MSP430 systems almost > to destruction for 3 years now. many systems run continuously. The > failures of flash that I have seen occurred during a flash write cycle > excercise, designed to determine the realistic number of flash writes > you might expect. Nothing is defect free, so my code attempts re- writes, > logs anything that takes more than 2 attempts, maintains failed bit/byte > logs and does re-attempt these at later times. but it treats the flash > like RAM, it continuously writes data to flash. When the defined flash > file is filled it erases segments, and starts again. failures have been > due to aging. But that aging requires typically many thousands of write > cycles. IIRC only 2 failed below the spec. ALL at 3V3, tests conducted > from my deep freeze to 70C. I wouldn't think that cause for concern. Any > reasonably robust filing system can handle that. I have thousands of > boards out there using the flash as a filing system, none have failed in > the field. > > > > > Finally, there is something in my system which might be playing a > > role here. > > > > I have a 1 Watt RF transmitter, about 900MHz, and the antenna is > > sitting just inches from my MSP430F149. What might be the chances of > > THAT dithering my Vcc, or maybe even my flash? I am still wondering, > > actually, how I might instrument that thing to monitor Vcc. But > > until then, your insights are welcome. > > I run RF with many of my designs and haven't found it a problem, however > the highest power unscreened system I have is 200mW. Mostly I screen the > RF and micro. Certainly add rf decoupling in the power supply lines > > Cheers > > Al > > > > > Regards, > > Sam > > > > > > > > > > . > > > > > > > > > > > > > >
Reply by ●December 19, 20032003-12-19
Hi Samuel,
> I have a 1 Watt RF transmitter, about 900MHz, and
the antenna is
> sitting just inches from my MSP430F149. What might be the chances of
> THAT dithering my Vcc, or maybe even my flash? I am still wondering,
> actually, how I might instrument that thing to monitor Vcc. But
> until then, your insights are welcome.
1 Watt will produce already a significant field strength.
Bear in mind that any so-so solder joints or even many non-linear elements
can
easily rectify RF into DC. Being all digital (?) the introduced "DC
shifts"
shouldn't affect it too
much, but because of low power many ports on the MSP430 are very high
impedance,
watch out there. The most sensitive are eg. LFXTAL1 - any spikes or some
noise will
typically show up as for example "a real time clock running faster on
occasion", because
of the extra clocking of these pulses into the oscillator.
If any non-burst noise finds its way into the clock, it would overclock -
and could *possibly*
on occasion overclock the Flash Timing Generator. That __could__ be a source
of
Flash corruption.
If this is easily changed, it is paramount in strong RF field that (apart
from the usual)
1. Vdd/GND supply lines : If it's pair of wires, run them
"twisted" to
your board to
combat EM field pick up a bit.
2. Any bus-ish ribbon stuff (for example) ideally is shielded, or at
least has alternating
GND wires, and twisted.
3. Of course, the PCB layout itself. Liberal use on "ground planes"
isn't
just the beyond and
all they are commonly presumed to be. Ground planes can be just as
"hot" as any other parts.
Maybe DC probe a few important lines on MSP430 (close-in) and key the TX on
and off,
and see if you notice any DC level change, that's a good thing to do.
This is of course a bit late, at post production stage, but if it is
critical Flash-wise, look up
how to do a PCB layout for a proper guard ring and local GND plane on XTAL
oscillator.
It will give more piece of mind on the next field PCB.
Hope that helps a bit,
-- Kris
www.microbit.com.au
PS : CCR on MSP430 actually stands for Credence Clearwater Revival :-)
Reply by ●December 20, 20032003-12-20
I'm mostly concerned with power on delay, and excessive start up
currents. early Flash members of the MSP430 family are particularly
sensitive to slow starting power supplies. Just because you spot the
flash corruption when you upgrade does not necessarily mean that it
occurred then, only that it went undetected until then.
Al
Samuel Waters wrote:
> I want to take your friendly inputs seriously, now. Please help me
> understand a little better.
>
> The product is up-and-running for a long time before flash upgrades
> occur.
>
> When questioning the bypass strategy, were you thinking of a power-on
> delay dealy proble, or were you thinking of some inherent
> characteristic of the devices which might amplify their noise?
>
> Thanks,
> Sam
>
>
> ==========================>
> --- In msp430@msp4..., onestone <onestone@b...> wrote:
>
>>Samuel Waters wrote:
>>
>>>Many thanks to "bill knight", Chris (kris?) and Al, and
others
>
> for
>
>>>their kind responses to my inqury about Flash Corruption.
>>>
>>>I am currently juicing my Vcc from 3.3V to 3.6V, so as to stay
>
> within
>
>>>spec, having an 8MHz crystal, even though I do not believe this
>
> is a
>
>>>root cause of my problem.
>>>
>>>However, if I am experiencing Power supply noise, say something
>>>additive(or subtractive), then an extra 0.3V will help my
>
> immunity,
>
>>>to keep Vcc happily high enough for the flash to function during
>>>field upgrades.
>>>
>>>In reponse to some of your("your"-pluaral) points:
>>>
>>>My Vcc bypassing consists of 4-220uF tantalum, plus a number of
>
> 1nF
>
>>>and 1uF tantalums.
>>
>>I agree with several others, you have too much bypassing here.
>>
>>
>>>One of you indicated that you could "count on one hand"
the
>
> number of
>
>>>flash failures you have experienced. I know this was meant to
>>>encourage me, but frankly I found that news to be devastating. I
>
> was
>
>>>rather hoping, you know, for zero failures. We will never know,
>
> for
>
>>>sure, in your case, because some of your products get overclocked
>
> at
>
>>>16MHz, if I read correctly. But this is still a troubling data
>
> point.
>
>>You must take this in context. I have been testing MSP430 systems
>
> almost
>
>>to destruction for 3 years now. many systems run continuously. The
>>failures of flash that I have seen occurred during a flash write
>
> cycle
>
>>excercise, designed to determine the realistic number of flash
>
> writes
>
>>you might expect. Nothing is defect free, so my code attempts re-
>
> writes,
>
>>logs anything that takes more than 2 attempts, maintains failed
>
> bit/byte
>
>>logs and does re-attempt these at later times. but it treats the
>
> flash
>
>>like RAM, it continuously writes data to flash. When the defined
>
> flash
>
>>file is filled it erases segments, and starts again. failures have
>
> been
>
>>due to aging. But that aging requires typically many thousands of
>
> write
>
>> cycles. IIRC only 2 failed below the spec. ALL at 3V3, tests
>
> conducted
>
>>from my deep freeze to 70C. I wouldn't think that cause for
>
> concern. Any
>
>>reasonably robust filing system can handle that. I have thousands
>
> of
>
>>boards out there using the flash as a filing system, none have
>
> failed in
>
>>the field.
>>
>>
>>>Finally, there is something in my system which might be playing a
>>>role here.
>>>
>>>I have a 1 Watt RF transmitter, about 900MHz, and the antenna is
>>>sitting just inches from my MSP430F149. What might be the
>
> chances of
>
>>>THAT dithering my Vcc, or maybe even my flash? I am still
>
> wondering,
>
>>>actually, how I might instrument that thing to monitor Vcc. But
>>>until then, your insights are welcome.
>>
>>I run RF with many of my designs and haven't found it a problem,
>
> however
>
>>the highest power unscreened system I have is 200mW. Mostly I
>
> screen the
>
>>RF and micro. Certainly add rf decoupling in the power supply lines
>>
>>Cheers
>>
>>Al
>>
>>
>>>Regards,
>>>Sam
>>>
>>>
>>>
>>>
>>>.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>
>
>
> .
>
>
>
>
>
>
>
Reply by ●December 22, 20032003-12-22
RF radiation is not good to have around ANY electronic circuits. I spent literally months on a contract troubleshooting a problem with 900 MHz walkie talkies interfering with an 18-bit ADC. I think Kris is right on. A 1 watt field is really quite strong and can easily be demodulated in many circuits. >If any non-burst noise finds its way into the clock, it would overclock - >and could *possibly* >on occasion overclock the Flash Timing Generator. That __could__ be a >source >of >Flash corruption. I would not doubt that for a second at that power level. I would say the biggest thing you could do at first is: separate the transmitter from the circuit as far as possible and try to shield the circuit from the transmitting field. Separating the antenna is most effective since the radiation drops proportional to the square of the distance in an open area. If there are metal enclosures to worry about, you could get standing waves even though things are somewhat shielded. >1. Vdd/GND supply lines : If it's pair of wires, run them "twisted" to >your board to > combat EM field pick up a bit. Also consider adding ferrite beads to any lengths of wire. Any wire, even circuit board traces will form antennas and pick up the RF. If the length of the wire happens to be an odd multiple of quarter wavelengths, it is now a tuned radiator which means the RF signal is efficiently delivered to the termination point of the wire. Be careful when using any ferrite beads, which are inductors, on clocked lines. The extra inductance can distort waveforms and cause problems. In-line inductance and capacitance to ground is what will help attenuate RF interference. Be sure the capacitors you use are good at the frequency of the radiation. >Maybe DC probe a few important lines on MSP430 (close-in) and key the TX on >and off, >and see if you notice any DC level change, that's a good thing to do. Remember though, as you probe the circuit you are adding some capacitance which could affect what you are measuring. I can remember my problems going away as long as the scope probe was attached. Then I knew it was time to add some capacitance at a certain point. I hope you do not need to do what I had to do on my project. I had to isolate the ADC chip itself inside a metal can with each line to the ADC going through a T-network filter. It worked but it taught me one thing I will never forget: KEEP THE RF AWAY FROM THE CIRCUIT! Sorry I had to yell, but those were some of the toughest days of my contracting career, and I will never forget it. Lou _________________________________________________________________ Check your PC for viruses with the FREE McAfee online computer scan. http://clinic.mcafee.com/clinic/ibuy/campaign.asp?cid963
Reply by ●December 22, 20032003-12-22
Hi Lou, It's great to see some fellow RF fellas on here :-) I think you covered the remaining issues quite thoroughly enough. Maybe one more comment for other people that might use RF near MCU stuff. > I would not doubt that for a second at that power level. I would say the > biggest thing you could do at first is: separate the transmitter from the > circuit as far as possible and try to shield the circuit from the > transmitting field. Separating the antenna is most effective since the > radiation drops proportional to the square of the distance in an open area. > If there are metal enclosures to worry about, you could get standing waves > even though things are somewhat shielded. Just to enlighten further here, keeping the antenna away is of course the sure fire way. In that case it is very important to have good matching onto transmission line, and good matching onto the antenna. Standing waves will create RF return currents on the braid of coax, and they'll radiate and/or conduct "back in" - so the problem could partially remain. If there is still doubt, use a so-called attenuator pad (say 3 dB) - that will improve VSWR dramatically. -- Kris