Microcontroller with 12-bit a/d subsystem?

[John Hudak]

I am looking for a micro with a 12-bit a/d subsystem on it. I've
searched quite a few mfg, but seems 10-bits is the maximum. Anybody
have any suggestions for microcontrollers that have a 12-bit a/d?
Thanks
John

 

[Spehro Pefhany]

I am looking for a micro with a 12-bit a/d subsystem on it. I've
searched quite a few mfg, but seems 10-bits is the maximum. Anybody
have any suggestions for microcontrollers that have a 12-bit a/d?
Thanks
John

You didn't look very hard. AD, BB/TI, TI. Silabs, Microchip and
others..

Best regards,
Spehro Pefhany

 

[Tim Wescott]

You didn't look very hard. AD, BB/TI, TI. Silabs, Microchip and
others..

Few, if any, of those will actually be good to 12 bits. A/D converter
bit count in microcontrollers has become a big marketing game, and
accuracy has gone by the way side.

I'd sure be interested in a microcontroller that had an honest 12 bits to
it, though.

--
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html

 

[Vladimir Vassilevsky]

Few, if any, of those will actually be good to 12 bits. A/D converter
bit count in microcontrollers has become a big marketing game, and
accuracy has gone by the way side.

Right. The accuracy of the 12bit ADC of TI 28xx is nothing but a
disappointment.

I'd sure be interested in a microcontroller that had an honest 12 bits to
it, though.

Look at the ADUC series from AD. It is a x51 core with a decent ADC.
There also used to be the ADSP218x with the delta-sigma audio codec.



Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[John Devereux]

Right. The accuracy of the 12bit ADC of TI 28xx is nothing but a
disappointment.



Look at the ADUC series from AD. It is a x51 core with a decent
ADC. There also used to be the ADSP218x with the delta-sigma audio
codec.

They do the ARM7 core as well (ADUC7000 series). The ADCs seem as good
as any other 12 bit type. They are 1us so quite fast too - OP may be able to
reduce any transition noise with some averaging.

 

[Spehro Pefhany]

Few, if any, of those will actually be good to 12 bits. A/D converter
bit count in microcontrollers has become a big marketing game, and
accuracy has gone by the way side.

I'd sure be interested in a microcontroller that had an honest 12 bits to
it, though.

TI has "16 bit" sigma-delta ADCs, and others have 24-ish nominal bits.
also in sigma-delta. As always, ENOB depends on the sampling, and
probably other factors.. BB claims very low noise (sub 100nV) but
their stuff is pretty pricey. Usually for control applications, a high
resolution is better, all other things being equal, so a 16-bit
converter with 12-bit accuracy (and maybe the LSB is mostly noise)
will be better than a really solid 12-bit converter.

Best regards,
Spehro Pefhany

 

[Tim Wescott]

Right. The accuracy of the 12bit ADC of TI 28xx is nothing but a
disappointment.



Look at the ADUC series from AD. It is a x51 core with a decent ADC.
There also used to be the ADSP218x with the delta-sigma audio codec.

Unfortunately, whenever I put the words '8051' and 'microcontroller'
together, the phrase 'lousy excuse for' starts floating around in my head...

Be that as it may, you're probably right. I take that line of ADCs the
way ADI markets them -- they're not microcontrollers with ADCs tacked
on, they're ADCs with microcontrollers tacked on. This is not a bad
thing, and I should take a close look at the ones with the ARM cores.

--

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

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

[Tim Wescott]

TI has "16 bit" sigma-delta ADCs, and others have 24-ish nominal bits.
also in sigma-delta. As always, ENOB depends on the sampling, and
probably other factors.. BB claims very low noise (sub 100nV) but
their stuff is pretty pricey. Usually for control applications, a high
resolution is better, all other things being equal, so a 16-bit
converter with 12-bit accuracy (and maybe the LSB is mostly noise)
will be better than a really solid 12-bit converter.

Come to think of it, I've used the TI one -- but I neither carefully
studied the data sheet nor did I characterize my processor in circuit,
because my application wasn't that critical. I still suspect that for
SAR circuits the last few bits of the 12 are there for chest-beating.

You're correct about the resolution vs. accuracy issue -- getting those
last few bits, even if you can't trust them for more than a few seconds,
can make a big difference in a control system's behavior.

--

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

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

[Nico Coesel]

I am looking for a micro with a 12-bit a/d subsystem on it. I've
searched quite a few mfg, but seems 10-bits is the maximum. Anybody
have any suggestions for microcontrollers that have a 12-bit a/d?
Thanks
John

The MSP430 series come to mind. Dunno if they are 12 bit though.
However, their analog behaviour is very good according to some
people...

 

[Anthony Fremont]

Vladimir Vassilevsky wrote:
Unfortunately, whenever I put the words '8051' and 'microcontroller'
together, the phrase 'lousy excuse for' starts floating around in my
head...
LOL

Be that as it may, you're probably right. I take that line of ADCs
the way ADI markets them -- they're not microcontrollers with ADCs tacked
on, they're ADCs with microcontrollers tacked on. This is not a bad
thing, and I should take a close look at the ones with the ARM cores.

I've been severely bitten by the ARM bug. Been playing with one of these
recently:
http://www.olimex.com/dev/lpc-2378stk.html
It makes for a way cool toy.

 

[John Hudak]

You didn't look very hard. AD, BB/TI, TI. Silabs, Microchip and
others..

Best regards,
Spehro Pefhany

Actually I did. I have used a few as well..MicroChip, Renseas, Atmel.
Both performance and accuracy is pitiful....

 

[Spehro Pefhany]

Actually I did. I have used a few as well..MicroChip, Renseas, Atmel.
Both performance and accuracy is pitiful....

You didn't say you needed 12 real bits rather than marketing bits. ;-)

I have not done much with on-chip SAR converters lately, but 10-bit
ones are so stable and noise-free that I suspect that the 12-bit ones
are not too bad, even if they are not any better. That's assuming the
layout is good, the supply voltage is reasonably high and clean, there
is as good ground plane, there are not large untamed currents flowing
around, and all the other usual analogish stuff. I have no doubt that
others can get worse performance out of them. ;-)

Best regards,
Spehro Pefhany

 

[Hal Murray]

You're correct about the resolution vs. accuracy issue -- getting those
last few bits, even if you can't trust them for more than a few seconds,
can make a big difference in a control system's behavior.

I'm not sure what your "few seconds" comment refers to.

I generally divide ADCs into two types.

The old style are rated in terms of linearity of low bits.
They work for things like measuring temperature or voltage that
changes slowly.

The new style are intended for dynamic signals - audio or radar.
They are rated in ENOB - Eveective Number of Bits, usually
the data sheet has an FFT picture.

 

[John Devereux]

Unfortunately, whenever I put the words '8051' and 'microcontroller'
together, the phrase 'lousy excuse for' starts floating around in my
head...

Be that as it may, you're probably right. I take that line of ADCs
the way ADI markets them -- they're not microcontrollers with ADCs
tacked on, they're ADCs with microcontrollers tacked on. This is not
a bad thing, and I should take a close look at the ones with the ARM
cores.

The peripherals sometimes seem a bit basic compared to other
"microcontrollers with ADCs tacked on" - although this also means they
are relatively simple to understand.

There are also some oddball design choices, such as the embedded
programmable logic array, or the DDS synthesiser with "100 ohm line
driver". Presumably a customer-specific part released to the masses.

My impression is that this family is not as active as lines from other
manufacturers. And the parts seem slightly more expensive; I guess you
are paying for the analog performance.

I've just noticed a new one with dual 16 bit delta-sigma ADCs!

<http://www.analog.com/en/prod/0,2877,ADUC7032%2D8L,00.html>

And it operates directly from 12V! See what I mean about oddball?

 

[Richard Henry]

Few, if any, of those will actually be good to 12 bits. A/D converter
bit count in microcontrollers has become a big marketing game, and
accuracy has gone by the way side.

I'd sure be interested in a microcontroller that had an honest 12 bits to
it, though.

What are your specific problems with those devices (just so I'll know
in the future)?

 

[john jardine]

[...]

I've been severely bitten by the ARM bug. Been playing with one of these
recently:
http://www.olimex.com/dev/lpc-2378stk.html
It makes for a way cool toy.

Was looking at an ARM development board on Sunday. Are they straighforward
to program in assembly?.
I've now had 10 years of PICs and am heartily fed up with their obtuse,
infernal internals.
On average I get through maybe 4 jobs a year that use a PIC and each time
round I have to relearn the bit settings, the register defaults, the
interrupt sequencing, the pin allocating, the ... . It's a grade#1 pain.
Maybe it's because I'm not really interested in programming and see it as
just a means to an end, yet I still find it very easy to read Z80 assembly
listings from 15 years back and could still program one from scratch.
I reckon the PICs are doing my head in. There must be something better out
there. Perchance the ARM?.

 

[Anthony Fremont]

[...]

I've been severely bitten by the ARM bug. Been playing with one of
these recently:
http://www.olimex.com/dev/lpc-2378stk.html
It makes for a way cool toy.

Was looking at an ARM development board on Sunday. Are they
straighforward to program in assembly?.

I really like the instruction set, much more powerful than the typical micro
IMO.

I've now had 10 years of PICs and am heartily fed up with their
obtuse, infernal internals.

;-) I wanted to do some work in C and learn something about ARM, so I
finally got around to getting a dev board. The dev board in the link I
posted is my second one. I started with a LPC2106 board to get my feet wet
and then bought the second board a couple of weeks later. I've been having
a ball with this one, what with all the nifty peripherals that I haven't
played with before.

On average I get through maybe 4 jobs a year that use a PIC and each
time round I have to relearn the bit settings, the register defaults,
the interrupt sequencing, the pin allocating, the ... . It's a
grade#1 pain. Maybe it's because I'm not really interested in
programming and see it as just a means to an end, yet I still find
it very easy to read Z80 assembly listings from 15 years back and
could still program one from scratch.

I don't think the ARM architecture is going to make you suddenly enjoy
programming, but I think you will find the instruction set allot easier on
your sanity. My biggest problem with PIC chips is the W reg, it makes life
so tedious.

Something I noticed pretty quickly about the ARM world is that while the
instruction set is the same between manufacturers, the way the built in
peripherals work can vary quite a bit. I originally had planned to start
using Atmel SAM7 parts, but due to a shortage in the dev boards, I ended up
going with the LPC(NXP/Phillips) part. I'm kinda glad I did as I like the
GPIO port design in the NXP parts better......I think. ;-)

I reckon the PICs are doing my head in. There must be something
better out there. Perchance the ARM?.

This may be somewhat of an apples/oranges issue. IMO the PIC will always
have its place because of its ruggedness and reliability not to mention the
miniscule amounts of power it uses. OTOH, the ARMs are getting really cheap
and they are just sooo frigging fast it's amazing. I'm used to PIC
instructions taking .5 to 1 uS to execute, now I'm burning the place down
with instruction cycle times < 15nS. Seems so odd to call a subroutine to
delay for 1uS, but I'm not complaining mind you. ;-) OTOH, welcome to the
world of wait states when running from flash. :-( I guess you can't have
everything.

 

[Roger Hamlett]

Actually I did. I have used a few as well..MicroChip, Renseas, Atmel.
Both performance and accuracy is pitiful....

Funnily enough, the 'old' PIC14000, performed very well, on slow
changing signals, and could easily give 14bit accuracy. The current
units can be 'OK', provided you put the processor to sleep for the
conversion, but the way most people use them, you are lucky to get
8bit accuracy.... :-(
It really does depend mssively though on the nature of the signal
involved, and what else is going on round the system. Getting a
genuine 12bits, requires a reference that is better than many, and
significant care in layout. I suspect that most micro designers, are
thinking in terms that with the noise round the processor itself, you
are better off using a small I2C interfaced ADC, that is placed away
from this enviroment, into a 'cleaner' enviroment, and so their
designs are targetted at simplicity, rather than quality. It shows...

Best Wishes

 

[Richard Henry]

[...]

I've been severely bitten by the ARM bug. Been playing with one of these
recently:
http://www.olimex.com/dev/lpc-2378stk.html
It makes for a way cool toy.

Was looking at an ARM development board on Sunday. Are they straighforward
to program in assembly?.
I've now had 10 years of PICs and am heartily fed up with their obtuse,
infernal internals.
On average I get through maybe 4 jobs a year that use a PIC and each time
round I have to relearn the bit settings, the register defaults, the
interrupt sequencing, the pin allocating, the ... . It's a grade#1 pain.
Maybe it's because I'm not really interested in programming and see it as
just a means to an end, yet I still find it very easy to read Z80 assembly
listings from 15 years back and could still program one from scratch.
I reckon the PICs are doing my head in. There must be something better out
there. Perchance the ARM?.

I agree. The PIC assembly code is horrible. I recommend a low-level
PIC-compatible C compiler.

 

[Nobody]

Was looking at an ARM development board on Sunday. Are they straighforward
to program in assembly?.

The 32-bit ARM instruction set has to be the "cleanest" instruction
set in existence. The 16-bit version which is widely used on uC versions
is still a lot neater than typical 8/16-bit microprocessor instruction
sets.

However, that's just the core instruction set. Once you start dealing with
all the "bundled" extras (MMU, FPU, SIMD, I/O, Java, etc), things start
getting more involved.

It's also worth bearing in mind that the architecture was originally
conceived as the CPU for desktop systems rather than for twiddling I/O
pins, so the overall system is vastly more complex than a typical uC.

OTOH, there's likely to be less of a benefit to using assembler when you
have 16x32-bit registers and 128/256/512KiB flash and 32/64KiB of RAM with
a flat address space.

You might want to take a look at the datasheets for the AVR32 family to
get an idea:

http://www.atmel.com/dyn/products/datasheets.asp?family_id=682

I've now had 10 years of PICs and am heartily fed up with their obtuse,
infernal internals.
On average I get through maybe 4 jobs a year that use a PIC and each time
round I have to relearn the bit settings, the register defaults, the
interrupt sequencing, the pin allocating, the ... . It's a grade#1 pain.
Maybe it's because I'm not really interested in programming and see it as
just a means to an end, yet I still find it very easy to read Z80 assembly
listings from 15 years back and could still program one from scratch.
I reckon the PICs are doing my head in. There must be something better out
there. Perchance the ARM?.

If you're looking for overall system simplicity, you probably don't want a
32-bit RISC chip which is capable of being used as the main CPU for
everything from mobile phones to TCP/IP routers to Unix workstations.