Reply by Ulf Samuelsson February 9, 20052005-02-09
> > If not all functionas are needed, a smaller AVR can easily be used. > > If a small LCD controller is needed an ATmega169 can be used etc. > > > > I think that conservatism is the answer in many cases. > > Note that to replace a Vdd Monitor device, you need a DEFINED reset > drive, from below 1V Vdd. - to get that in a uC, you need one that > drives the Reset pin and has a data sheet defined state - eg the > C8051F30x series state 0.6V is /RST force threshold, whilst AVRs do not > drive the RST pin.
Nope, not neccessary, Connect a pulldown resistor to any I/O pin. When the AVR get brownout, then it is reset. At reset, the I/O pin becomes an input, and the pulldown resistor will activate the reset to the main CPU. When the AVR thinks it is time to pull the main CPU out of reset, it defines the I/O port as an output and drives it high So far, none that I spoken to has complained about this.
> The ideal WDOG_uC would AVR. > ** have a defined Reset Out for < 1V Vcc - Pull down
resistor fixes this...
> ** Have a Brown Out detector that defaulted ON - Yep, fuse option > ** Have an adjustment range on the BOD threshold - The newer AVRs have
64 levels. only three visible
> ** Draw under 1uA, with BOD alive. - Would be
real nice. now it is more like 10 uA.
> > I have not seen one that meets all of the ideals.... >
-- Best Regards Ulf at atmel dot com These comments are intended to be my own opinion and they may, or may not be shared by my employer, Atmel Sweden.
Reply by An Schwob in USA February 9, 20052005-02-09
Steve at fivetrees wrote:
> "Jim Granville" <no.spam@designtools.co.nz> wrote in message > news:4209542a$1@clear.net.nz... > > Note that to replace a Vdd Monitor device, you need a DEFINED
reset
> > drive, from below 1V Vdd. - to get that in a uC, you need one that > > drives the Reset pin and has a data sheet defined state - eg the > > C8051F30x series state 0.6V is /RST force threshold, whilst AVRs do
not
> > drive the RST pin. > > The ideal WDOG_uC would > > ** have a defined Reset Out for < 1V Vcc > > ** Have a Brown Out detector that defaulted ON > > ** Have an adjustment range on the BOD threshold > > ** Draw under 1uA, with BOD alive. > > > > I have not seen one that meets all of the ideals.... > > I don't know the PICs well, but that was also my concern. The one
project I
> was involved in with PICs had brown-out troubles... > > For something as important as reset (and hardware watchdoggery), I
have
> confidence in dumb, essentially analog/mixed devices whose job it is
to
> monitor that the rail is within specs (e.g. 5V +/-5%) and to hold the
CPU in
> reset if not. Nothing else. I don't mind (too much) paying a premium
for
> such a vital function. I don't currently have that same confidence in
e.g.
> PICs. > > The MAX1232 is my old staple part; 50uA operating current (typ.) is
fine for
> my purposes. > > YMMV. > > Steve > http://www.fivetrees.com
Steve and Jim, I did not think about the brownout of the small micro because I was planing to use a battery backup for that one. Basically the small micro would switch to battery when power fails. To monitor the main power, the comparator with its own reference can come in very handy. During active power on, the small micro would act as watchdog / critical data storage and during power off it would be buffered. Large CPUs in new processes have power dowm currents that are enough to operated a small micro while still being active. The LPC900 series offers an EEPROM functionality with byte writes and can be used when power gets turned off to save critical process data from RAM to flash. Then it can go into power down itself, draining the battery probably less than the self discharge. During all your inputs, the ideas have evolved and I can see now that more functionality on the micro really adds value while watchdog/reset just by itself is probably better handled by a low end reset circuit. Thanks everybody for the knowledge you are willing to share! An Schow
Reply by Steve at fivetrees February 8, 20052005-02-08
"Jim Granville" <no.spam@designtools.co.nz> wrote in message
news:4209542a$1@clear.net.nz...
> Note that to replace a Vdd Monitor device, you need a DEFINED reset > drive, from below 1V Vdd. - to get that in a uC, you need one that > drives the Reset pin and has a data sheet defined state - eg the > C8051F30x series state 0.6V is /RST force threshold, whilst AVRs do not > drive the RST pin. > The ideal WDOG_uC would > ** have a defined Reset Out for < 1V Vcc > ** Have a Brown Out detector that defaulted ON > ** Have an adjustment range on the BOD threshold > ** Draw under 1uA, with BOD alive. > > I have not seen one that meets all of the ideals....
I don't know the PICs well, but that was also my concern. The one project I was involved in with PICs had brown-out troubles... For something as important as reset (and hardware watchdoggery), I have confidence in dumb, essentially analog/mixed devices whose job it is to monitor that the rail is within specs (e.g. 5V +/-5%) and to hold the CPU in reset if not. Nothing else. I don't mind (too much) paying a premium for such a vital function. I don't currently have that same confidence in e.g. PICs. The MAX1232 is my old staple part; 50uA operating current (typ.) is fine for my purposes. YMMV. Steve http://www.fivetrees.com
Reply by Jim Granville February 8, 20052005-02-08
Ulf Samuelsson wrote:
>>Our design needs to have a timed reset (shorter!) to get up to speed >>faster. > > >>External supervisory circuits are more expensive and a lot less >>versatile than low end micros... >>With all these benefits and still so many reset devices on the market, >>I must have missed some important disadvantages of such a >>configuration. > > ... > >>What else am I missing? >> >>Your feedback is appreciated, Schwob >> > > As other people have noted, just a pure rest function is maybe not > costeffective.
... Or even technically correct/possible. <snip>
> If not all functionas are needed, a smaller AVR can easily be used. > If a small LCD controller is needed an ATmega169 can be used etc. > > I think that conservatism is the answer in many cases.
Note that to replace a Vdd Monitor device, you need a DEFINED reset drive, from below 1V Vdd. - to get that in a uC, you need one that drives the Reset pin and has a data sheet defined state - eg the C8051F30x series state 0.6V is /RST force threshold, whilst AVRs do not drive the RST pin. The ideal WDOG_uC would ** have a defined Reset Out for < 1V Vcc ** Have a Brown Out detector that defaulted ON ** Have an adjustment range on the BOD threshold ** Draw under 1uA, with BOD alive. I have not seen one that meets all of the ideals.... -jg
Reply by Jim Granville February 8, 20052005-02-08
An Schwob in USA wrote:
> Our design needs to have a timed reset (shorter!) to get up to speed > faster. We use a 32-bit CPU with external crystals and the required > time to keep the device in reset is min. 10 ms. If I use a > RC-combination to generate the reset pulse and include the factory > tolerance my typical reset low time will be much longer than 10 ms. > > External supervisory circuits are more expensive and a lot less > versatile than low end micros. If the micro has an internal oscillator > and a defined startup time, it is very easy to control the reset > timing. It is also very easy use the device as watchdog with > programable actions. This watchdog could even try to establish > communication with the larger micro before it generates a reset. > The LPC903 or the LPC904 with ADC seem to be perfect. > > With all these benefits and still so many reset devices on the market, > I must have missed some important disadvantages of such a > configuration. > > For one the supervisor circuit uses less power than a micro but the > micro is still less than 100 uAs active. > What else am I missing?
I'd look closely at the Power On reset, or Vpor, on the uC, if you want to use it as a WDOG/RST. The SiLabs C8051F305 etc family might be a better solution, as they have a more robust and precise BOD and /RST pin-drive handling. -jg
Reply by Ulf Samuelsson February 8, 20052005-02-08
> Our design needs to have a timed reset (shorter!) to get up to speed > faster.
> External supervisory circuits are more expensive and a lot less > versatile than low end micros... > With all these benefits and still so many reset devices on the market, > I must have missed some important disadvantages of such a > configuration.
...
> What else am I missing? > > Your feedback is appreciated, Schwob >
As other people have noted, just a pure rest function is maybe not costeffective. there is lot of other things you can do. In the Comet board, an AT90CAN128 is a supervisory controller. It will do: UART expansion CAN expansion Reset the AT91RM9200 Can program the boot order. * External /Internal Dataflash over SPI * Internal Dataflash over SPI * I2C (using internal AVR Flash down.loaded over CAN) * UART * USB Brownout Detection Port Expansion Real Time Clock - little more power than a discrete solution Temperature Measurement Touch Screen Control (ADC) Key Debouncer LED Drive If not all functionas are needed, a smaller AVR can easily be used. If a small LCD controller is needed an ATmega169 can be used etc. I think that conservatism is the answer in many cases. Don't know the volume you have but there are many AVRs which should go under a dollar in volume. Best Regards, Ulf Samuelsson ulf@a-t-m-e-l.com This message is intended to be my own personal view and it may or may not be shared by my employer Atmel Nordic AB
Reply by Raivo Leini February 8, 20052005-02-08
"Dan" <dan@dontspammecauseidontlikit.com> wrote in message
news:qt8g01162f79pmgfbn3s164qt0rrkgluba@4ax.com...
> On 7 Feb 2005 16:37:53 -0800, "An Schwob in USA" <schwobus@aol.com> > wrote: > > > >For one the supervisor circuit uses less power than a micro but the > >micro is still less than 100 uAs active. > >What else am I missing? > > You also have to go to the extra step of programming the micro. The > reset chip is simply soldered into place and it works. > > Dan
right, but you can get from micro much more, like I2C slave RTC,ADC, battery powered RAM, EEPROM , software protection key and Reset device for more powerful (ARM,MIPS) system. r.
> >
Reply by Peter Jakacki February 8, 20052005-02-08
That's funny, I just answered a similar question this morning on LPC2000 
groups. Someone had mentioned they used a TPS3306-20-Q1 but when I 
checked the TI 1K price it was over a $1. For years I have used little 
PIC12C509s for reset and ISP control etc, what flash micros don't need 
some easy kind of ISP control? The newer devices such as the LPC900s or 
PIC12Fs are even better and cheaper. To save double-talk I'll just 
repost my previous message.

REPOST from http://groups.yahoo.com/group/lpc2000/

The problem with reset chips is that they are overly expensive (why?)
and customized. I find it cheaper to employ an 8-pin micro
(PIC12F,LPC900)complete with power-on timer/brown-out detect etc. I can
configure these for my reset threshold, reset time etc. These micros are
typically stand-alone, not requiring any external circuitry such as
crystal etc to operate.

PLUS!, there is always the bonus of having it monitor the coms line and
being able to reset or ISP the main processor as well as being able to
operate as an independant configurable watchdog.

Summary:
Configurable intelligent watchdog/reset/monitor *costs less* than a
fixed reset chip.

my2cents
*Peter*


-- An Schwob in USA wrote:

> Our design needs to have a timed reset (shorter!) to get up to speed > faster. We use a 32-bit CPU with external crystals and the required > time to keep the device in reset is min. 10 ms. If I use a > RC-combination to generate the reset pulse and include the factory > tolerance my typical reset low time will be much longer than 10 ms. > > External supervisory circuits are more expensive and a lot less > versatile than low end micros. If the micro has an internal oscillator > and a defined startup time, it is very easy to control the reset > timing. It is also very easy use the device as watchdog with > programable actions. This watchdog could even try to establish > communication with the larger micro before it generates a reset. > The LPC903 or the LPC904 with ADC seem to be perfect. > > With all these benefits and still so many reset devices on the market, > I must have missed some important disadvantages of such a > configuration. > > For one the supervisor circuit uses less power than a micro but the > micro is still less than 100 uAs active. > What else am I missing? > > Your feedback is appreciated, Schwob
Reply by An Schwob in USA February 8, 20052005-02-08
Hello Jack,

great ideas. Watchdog and Reset are probably not worth the while but
adding some sanity check functionality is really excellent.
In fact I have seen hot plug functionality with CAN in the past as well
but the really cheap micros don't offer this yet. The HC12 is a little
too expensive for what I am looking at, the ARM7 device is competing in
price with many of the HC12s. Looked at PIC but the 10-series is just a
toy, the PIC12F is somewhat similar in functionality to the LPC904 but
does not offer the same serial features or the same great internal
oscillator. The smallest AVRs seem more expensive.

In fact your approach could be best implemented with the LPC916,
multiple serial interfaces, 2k flash and an ADC and I got quotes
definitely below $1. Could offload the CPU quite a bit for serial
communication and some low end analog inputs.

Thank you for your great feedback!

An Schwob

Jack Klein wrote:
> On 7 Feb 2005 16:37:53 -0800, "An Schwob in USA" <schwobus@aol.com> > wrote in comp.arch.embedded: > > > Our design needs to have a timed reset (shorter!) to get up to
speed
> > faster. We use a 32-bit CPU with external crystals and the required > > time to keep the device in reset is min. 10 ms. If I use a > > RC-combination to generate the reset pulse and include the factory > > tolerance my typical reset low time will be much longer than 10 ms. > > > > External supervisory circuits are more expensive and a lot less > > versatile than low end micros. If the micro has an internal
oscillator
> > and a defined startup time, it is very easy to control the reset > > timing. It is also very easy use the device as watchdog with > > programable actions. This watchdog could even try to establish > > communication with the larger micro before it generates a reset. > > The LPC903 or the LPC904 with ADC seem to be perfect. > > > > With all these benefits and still so many reset devices on the
market,
> > I must have missed some important disadvantages of such a > > configuration. > > > > For one the supervisor circuit uses less power than a micro but the > > micro is still less than 100 uAs active. > > What else am I missing? > > > > Your feedback is appreciated, Schwob > > I've used an approach like this for ten years, but I wouldn't use it > just for reset and watchdog. You still need a power supply monitor, > and perhaps an external reset circuit anyway, in any safety critical > or high reliability. > > It can be extremely useful to put a secondary microcontroller as an > intelligent slave on a board. I used an 8051 deritave with a 486 a > long time ago, and our current architecture uses a 16-bit HS12 from > Motorola as a slave/IO coprocessor to an ARM 9. > > But in addition to the programmable watchdog function, the
coprocessor
> handles relatively low speed tasks to spare the main processor > overhead, such as running an A/D converter, debouncing switch and > mechanical sensor inputs, controlling LED display outputs. > > On our newest product, the hand controllers are plug-and-play over a > low speed CAN interface, The coprocessor sends out a broadcast
status
> request four times a second, assembles all the replies from however > many are connected (there are multiple ports), and sends them to the > main processor in a single 115k baud packet using the serial protocol > we developed. > > It is also extremely handy for the board assembly house to > troubleshoot boards, and locking the board -- and the whole system -- > into a fail safe state if something goes wrong. > > The typical 32-bit processor/controller requires a large number of > devices and board traces to be working to do anything at all. The > secondary micro uses no external memory and really needs nothing but > power and its oscillator. > > -- > Jack Klein > Home: http://JK-Technology.Com > FAQs for > comp.lang.c http://www.eskimo.com/~scs/C-faq/top.html > comp.lang.c++ http://www.parashift.com/c++-faq-lite/ > alt.comp.lang.learn.c-c++ > http://www.contrib.andrew.cmu.edu/~ajo/docs/FAQ-acllc.html
Reply by Jack Klein February 7, 20052005-02-07
On 7 Feb 2005 16:37:53 -0800, "An Schwob in USA" <schwobus@aol.com>
wrote in comp.arch.embedded:

> Our design needs to have a timed reset (shorter!) to get up to speed > faster. We use a 32-bit CPU with external crystals and the required > time to keep the device in reset is min. 10 ms. If I use a > RC-combination to generate the reset pulse and include the factory > tolerance my typical reset low time will be much longer than 10 ms. > > External supervisory circuits are more expensive and a lot less > versatile than low end micros. If the micro has an internal oscillator > and a defined startup time, it is very easy to control the reset > timing. It is also very easy use the device as watchdog with > programable actions. This watchdog could even try to establish > communication with the larger micro before it generates a reset. > The LPC903 or the LPC904 with ADC seem to be perfect. > > With all these benefits and still so many reset devices on the market, > I must have missed some important disadvantages of such a > configuration. > > For one the supervisor circuit uses less power than a micro but the > micro is still less than 100 uAs active. > What else am I missing? > > Your feedback is appreciated, Schwob
I've used an approach like this for ten years, but I wouldn't use it just for reset and watchdog. You still need a power supply monitor, and perhaps an external reset circuit anyway, in any safety critical or high reliability. It can be extremely useful to put a secondary microcontroller as an intelligent slave on a board. I used an 8051 deritave with a 486 a long time ago, and our current architecture uses a 16-bit HS12 from Motorola as a slave/IO coprocessor to an ARM 9. But in addition to the programmable watchdog function, the coprocessor handles relatively low speed tasks to spare the main processor overhead, such as running an A/D converter, debouncing switch and mechanical sensor inputs, controlling LED display outputs. On our newest product, the hand controllers are plug-and-play over a low speed CAN interface, The coprocessor sends out a broadcast status request four times a second, assembles all the replies from however many are connected (there are multiple ports), and sends them to the main processor in a single 115k baud packet using the serial protocol we developed. It is also extremely handy for the board assembly house to troubleshoot boards, and locking the board -- and the whole system -- into a fail safe state if something goes wrong. The typical 32-bit processor/controller requires a large number of devices and board traces to be working to do anything at all. The secondary micro uses no external memory and really needs nothing but power and its oscillator. -- Jack Klein Home: http://JK-Technology.Com FAQs for comp.lang.c http://www.eskimo.com/~scs/C-faq/top.html comp.lang.c++ http://www.parashift.com/c++-faq-lite/ alt.comp.lang.learn.c-c++ http://www.contrib.andrew.cmu.edu/~ajo/docs/FAQ-acllc.html