Single-Source PIC, AVR & Alternatives

[steve]

Some of what you say is true, but the apps where the wide voltage range
of the AVR is useful is only a subset of embedded apps. It may be that
*all* of *your* apps fall in that category, but that still does not
mean that that is a large market compared to the rest.

When you talk about the SAM7 not beating the ARM other than under
"strict conditions", you may not understand the SAM7. With the PLL on,
it is still lower power than the ATmega128 at the same clock speed.
The information I have from Atmel shows the current for the "clock
divisor" with the PLL on at under 300 uA or about three times the
current for the rest of the circuit running at just under 1 MHz. With
the PLL off, the chip current can get below 40 uA at very low clock
speeds. Meanwhile, the spread sheet shows the chip drawing 6 mA at
12.5 MHz with PIO, I2C and SPI all running.

The sleep current of the AVR will be lower than the SAM7, but that will
only be useful if your sleep ratio is higher than 99% of the time.
Also what you say about the "wake up" time of the ARM is not
universally valid. For most apps you never have to put the CPU into
"sleep", you just slow down the clock with the divisor. It all depends
on how deep you need to "sleep".

No, the ARMs of any flavor do not cover the same range as the low power
8 bit processors. But there is a lot of overlap and the ARM is clearly
the better in those cases. Part of the selection needs to consider
what your last design used (for reuse) and next design will require
(again for reuse). For many of us, the ARM is the right answer based
on the need to provide more processing power in a small package with
low power. Some still need the wide Vin range and ultra low sleep
current.

The parts I think will really kick some 8 bit butt is the next
generation of CM3 devices from LM. I don't have any hard data yet, but
once they target the low power modes I think the CM3 has a lot of
potential to do low power "right"! I just hope they don't give up 5
volt tolerance.

The new SAM7L series is claimed to be ultra low power and have LCD
drivers, which is the combo I need, but Atmel is also coming out with
Xmega AVR's series shortly (from the rumor mill), who knows what
features they will have.

 

[rickman]

The new SAM7L series is claimed to be ultra low power and have LCD
drivers, which is the combo I need, but Atmel is also coming out with
Xmega AVR's series shortly (from the rumor mill), who knows what
features they will have.

Yes, but the term "ultra low power" has little real meaning just like
"pico power". The pico power parts have low static current draw, but
that is only an issue in apps that have to remain in sleep mode for
more than 99% of the time. Otherwise the SAM7 parts were designed for
very low active power and were made in a newer process than the AVRs,
so they get better power consumption. I don't think the Xmega parts
are going to do any better.

Who knows what is going to be out tomorrow? I am trying to get info
from LM on the next generation of CM3 parts. Our stuff has to be low
power, but the uA of sleep current mean little. It is the active
current that counts to us. By this time next year I fully expect the
CM3 offerings to be dominating the new product announcements.

 

[Jim Granville]

Yes, but the term "ultra low power" has little real meaning just like
"pico power". The pico power parts have low static current draw, but
that is only an issue in apps that have to remain in sleep mode for
more than 99% of the time. Otherwise the SAM7 parts were designed for
very low active power and were made in a newer process than the AVRs,
so they get better power consumption. I don't think the Xmega parts
are going to do any better.

Who knows what is going to be out tomorrow? I am trying to get info
from LM on the next generation of CM3 parts. Our stuff has to be low
power, but the uA of sleep current mean little. It is the active
current that counts to us. By this time next year I fully expect the
CM3 offerings to be dominating the new product announcements.

Banking for 'new process' to deliver hoped for lower powers may be
optimistic. {vaporware can promise anything, it's real silicon that
counts )

New process has not done this on FPGA, and if one looks at the newest
200Mhz process devices from Atmel, you see these comments :


" The AT91SAM9263 consumes about 700 uA of static current on VDDCORE at
25'C. This static current rises at up to 7 mA if the temperature
increases to 85'C. On VDDBU, the current does not exceed 3 uA @25'C, but
can rise at up to 20 uA @85'C.
A software-controllable switch to VDDCORE guarantees zero power
consumption on the battery when the system is on.
For dynamic power consumption, the AT91SAM9263 consumes a maximum of
70 mA on VDDCORE at maximum conditions (1.2V, 25°C, processor running
full-performance algorithm)."

Depends on your targets, but 700uA-7mA static are not leaps down in
Icc in most designer's yardsticks (well, maybe if you come from a PC
background ?!
It does not matter HOW fancy your core is, or how fancy your clock
gating, this static Icc is always there, until you remove the power
entirely.
[This is why dual processor solutions are appearing, and why fewer are
expecting one chip to be the final solution ]


Marketing is still driven by MHz first, power 2nd (tho power is
certainly now on the radar), and simple maths and physics are
still there...

Maybe this will be the solution, for those who still think in uA ?
http://www.eetimes.com/news/semi/sh...5M0GXMQSNDLRCKHSCJUNN2JVN?articleID=196602004

-jg

 

[Ulf Samuelsson]

The new SAM7L series is claimed to be ultra low power and have LCD

drivers, which is the combo I need, but Atmel is also coming out with
Xmega AVR's series shortly (from the rumor mill), who knows what
features they will have.

I guess that, if the rumour mill is true, that would be me

 

[Ulf Samuelsson]

Yes, but the term "ultra low power" has little real meaning just like
"pico power". The pico power parts have low static current draw, but
that is only an issue in apps that have to remain in sleep mode for
more than 99% of the time. Otherwise the SAM7 parts were designed for
very low active power and were made in a newer process than the AVRs,
so they get better power consumption. I don't think the Xmega parts
are going to do any better.

The Picopower has a power curve that starts at close to zero and increases
in almost a linear fashion.
The main advantage of Picopower is when you are in sleep mode
or are running at low clock frequencies.
When running at high frequency, the main benefit is from the new
cell library, which is not present in the ATmega128, that Rickman
compares with.

 

[steve]

I guess that, if the rumour mill is true, that would be me

No, I haven't seen any comments from you on xmega, unless you are using
different user names on different sites

 

[Jim Granville]

No, I haven't seen any comments from you on xmega, unless you are using
different user names on different sites

I think Ulf was answering (strictly) the question
"Who knows what features they will have ?"
- hence the smiley...

a better question would be "who is prepared to say what features
they will have?" - if the rumour is true, of course.

Seems there is always something new in the pipeline....


-jg

 

[Neil]

This has been a really interesting thread to read, with all the opinions
and all.

But what I'm curious about is which companies have you had good
experiences with over the years, and which ones have left you feeling
like you'll never be dumb enough to design one of their parts into a
product ever again?

Thanks in advance.

Microchip has been good with old parts. Find some old numbers and hit
them into Digi-key. Plus a lot of there parts migrate upward. a 28pin
part is always pin compatible with other parts. Handy if you need more
features or power later.

 

[Ulf Samuelsson]

I think Ulf was answering (strictly) the question
"Who knows what features they will have ?"
- hence the smiley...

a better question would be "who is prepared to say what features
they will have?" - if the rumour is true, of course.

I guess, if the rumour mill is true, that would not be me ;-)

 

[rickman]

The Picopower has a power curve that starts at close to zero and increases
in almost a linear fashion.
The main advantage of Picopower is when you are in sleep mode
or are running at low clock frequencies.
When running at high frequency, the main benefit is from the new
cell library, which is not present in the ATmega128, that Rickman
compares with.

I looked at the data sheet for the pico power part you recommended, the
ATmega324P. It does not seem to be significantly better than the
ATmega128. It draws 7.4 mA at 8 MHz which is 37 mW at the rated 5 volt
power. It draws 2.1 mA at 4 MHz and 0.4 mA at 1 MHz. This is
essentially the same (or worse) as the SAM7S parts, no? The price
listed in the press release is not any better than SAM7 prices.

As much as you wave your arms, I have not seen any information that
shows that either the SAM7 or the AVR devices will do a good job of
competing against the CM3 devices coming out in 2007. I understand
that with limited resources it may not be the best idea for Atmel to
develop a CM3 device line at this time. But I think it is pretty clear
that the CM3 is the way that ARM designs will be headed in the near
future and that the days of the ARM7 are numbered.

 

[Joerg]

I looked at the data sheet for the pico power part you recommended, the
ATmega324P. It does not seem to be significantly better than the
ATmega128. It draws 7.4 mA at 8 MHz which is 37 mW at the rated 5 volt
power. It draws 2.1 mA at 4 MHz and 0.4 mA at 1 MHz. This is
essentially the same (or worse) as the SAM7S parts, no? The price
listed in the press release is not any better than SAM7 prices.

As much as you wave your arms, I have not seen any information that
shows that either the SAM7 or the AVR devices will do a good job of
competing against the CM3 devices coming out in 2007. I understand
that with limited resources it may not be the best idea for Atmel to
develop a CM3 device line at this time. But I think it is pretty clear
that the CM3 is the way that ARM designs will be headed in the near
future and that the days of the ARM7 are numbered.

" ... and that the days of the ARM7 are numbered". That's exactly why
some things such as the 8051 are so successful in the marketplace. Large
parts of the industry (including pretty much all of my clients) cannot
consider any components that are likely to fall from grace within a few
years.

 

[przemek klosowski]

I don't know, but I suppose many of us have applications that very low
sleep current, which ARM's don't have, or very fast wake up times, which
ARM's don't have, or wide operating voltages (e.g., 1.8 to 5 volts for
AVR's) which ARM's don't have, or require integrated LCD controllers to
drive simple LCD glass, which ARM's don't have. PICs, AVR's,MSP430s, H8's
,8051s, or Elans all have the above features in a wide variety of flavors
and tiny sizes.

Check out AD 'analog ARM's from the ADuC7xxx series:

http://www.analog.com/en/subCat/0,2879,762__0__0_,00.html

$5 40MHz 8kB SRAM/62kB Flash, 10 ADC, 4 DAC, analog and digital
inputs 5V-tolerant, PWM/timers/I2C/serial, 50 uA sleep, 50 mA running.
They have a nice ADUC7020 Miniboard eval kit for $30, which is in a 40-pin
DIL form factor---looks ideal for prototyping:

http://www.analog.com/UploadedFiles/Associated_Docs/153397299aduc7020_adap_a.pdf

There's a lot of choice in the ARM space, and as a result nice
leapfrogging---the customer wins.

Disclaimer: I have no connection to ARM or AD---just read about their
new stuff.

 

[Tim Wescott]

Microchip has been good with old parts. Find some old numbers and hit
them into Digi-key. Plus a lot of there parts migrate upward. a 28pin
part is always pin compatible with other parts. Handy if you need more
features or power later.

I can bitch about their instruction set all day, but that pin
compatibility is sure sweet.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

[Neil]

I can bitch about their instruction set all day, but that pin
compatibility is sure sweet.

The Hi-tech compiler generates Very reasonable sized Code. I just
ignore it. It is still 8 bits of course.

 

[steve]

Check out AD 'analog ARM's from the ADuC7xxx series:

http://www.analog.com/en/subCat/0,2879,762__0__0_,00.html

$5 40MHz 8kB SRAM/62kB Flash, 10 ADC, 4 DAC, analog and digital
inputs 5V-tolerant, PWM/timers/I2C/serial, 50 uA sleep, 50 mA running.
They have a nice ADUC7020 Miniboard eval kit for $30, which is in a 40-pin
DIL form factor---looks ideal for prototyping:

http://www.analog.com/UploadedFiles/Associated_Docs/153397299aduc7020_adap_a.pdf

There's a lot of choice in the ARM space, and as a result nice
leapfrogging---the customer wins.

Disclaimer: I have no connection to ARM or AD---just read about their
new stuff.

I have used that chip, at full speed the power/mips ratio is great, but
at 1Mhz, the typical power requirement is 40X greater the MSP430/AVR,
at 300Khz, 100X greater, and that is with no A/D's or DAC running

 

[rickman]

I have used that chip, at full speed the power/mips ratio is great, but
at 1Mhz, the typical power requirement is 40X greater the MSP430/AVR,
at 300Khz, 100X greater, and that is with no A/D's or DAC running

I have looked at the ADUC ARM stuff and it is a nice chip if you need
good analog. But it has a 7 mA quiescent power draw. I was just at a
seminar for these parts and they confirmed that this is true, but did
not give an explanation for it.

Like you say, at full power it is ok, but if you are trying to save
power by running a slow clock it is not a good choice.

 

[Joel Kolstad]

I have looked at the ADUC ARM stuff and it is a nice chip if you need
good analog. But it has a 7 mA quiescent power draw. I was just at a
seminar for these parts and they confirmed that this is true, but did
not give an explanation for it.

7mA has got to be a design error (I mean, it's not like a Pentium with 65nm
transistors that leak like sieves...), so I'd expect they'd come out with a
better version ASAP.

I'm sure someone's butt has been roundly kicked...

 

[rickman]

7mA has got to be a design error (I mean, it's not like a Pentium with 65nm
transistors that leak like sieves...), so I'd expect they'd come out with a
better version ASAP.

I'm sure someone's butt has been roundly kicked...

I don't think it is that simple. These devices have been out for well
over a year and I don't think they have any intention of respinning
them. I would speculate that the precision analog has something to do
with the high quiescent draw, but I am not a chip design expert, so I
can't really say.

 

[steve]

7mA has got to be a design error

don't know, but I was following the ADUC ARM's since the first
preliminary datasheets were released (a few years ago), the low speed
power draw was much lower back then, but on the final datasheet it
jumped up to 7mA typical, 10mA max

 

[Jim Granville]

don't know, but I was following the ADUC ARM's since the first
preliminary datasheets were released (a few years ago), the low speed
power draw was much lower back then, but on the final datasheet it
jumped up to 7mA typical, 10mA max

Some of Maxims MAXQ's are similar.

There are two main current hogs: FLASH memory, and Analog cells.
Some devices leave the FLASH charge pumps/sense amps enabled,
which gives faster speeds, but has a static Icc cost.

Most wake up Flash as they need it, but that has a time-cost,
so your MIPS are lower than they could be.

Analog cells can be powered on/off, sometimes that can be simply overlooked.

Or, they might find that the impedance of the Vcc switch
adds too much impedance/noise to meet the important Analog
targets, so they do a metal layer change and short it out.

-jg