AREF bypass capacitance on ATMega2560?

[Joerg]

Nico Coesel wrote:

[...]

So far, in over 25 years, I had only one screw-up with Atmel. One of

their 8051 series uC would not reliably run at full spec'd clock speed

and I had to back off to 75%. They did fess up and apologize though,

something only very few and good companies (such as LTC) do when caught

with a bug.



--

Regards, Joerg



http://www.analogconsultants.com/

Atmel does not explain the USB pins of their Mega CPUs very well.

It can be worse. Many connector datasheets have no (!) amperage rating
for their contacts. That is kind of an important parameter.

 

[John Devereux]

Folks,

What's the usual capacitance? Any stability issues there? I was planning
on using a 1uF X7R ceramic cap on the AREF pin of an ATMega2560, in
order to be able to use its internal bandgap reference. I saw people
using 0.1uF and 0.47uF. The datasheet is silent about stuff like that,
as usual.

MCU manufacturers are so generally so clueless with respect to the
analog aspects... I would just experiment with a devkit if at all
worried.

 

[Klaus Kragelund]

Indeed! Didn't see it. I really hate such netlist style schematics,

better to draw it right at the pin.

Sometimes. Other times you just need to proceed without too much hassle

Cheers

Klaus

 

[Spehro Pefhany]

Copying machines should never ever second-guess what is on a document.
That is IMO a very bad design and I would never buy a machine like that.

It's important that they send the correct information to the NSA over
their Ethernet port.

 

[John Devereux]

It's important that they send the correct information to the NSA over
their Ethernet port.

You know, funny, that does not even sound all that paranoid now.

 

[Klaus Kragelund]

It's important that they send the correct information to the NSA over

their Ethernet port.

I think they may have some back-door installed in the copying machines.

On another note, some machines must by law have a money bill detector. If a bill is detected, they insert some pixels on the print to identify the machine

Regards

Klaus

 

[Joerg]

I think they may have some back-door installed in the copying
machines.

On another note, some machines must by law have a money bill
detector. If a bill is detected, they insert some pixels on the print
to identify the machine

Then a text appears on the little LCD "Dude, you're screwed!" and the
flashing of blue lights can be seen down at the street.

Even restrooms have such features:

http://www.funnycorner.net/funny-pictures/4347/183.jpg

 

[Lasse Langwadt Christensen]

I think they may have some back-door installed in the copying machines.



On another note, some machines must by law have a money bill detector. If a bill is detected, they insert some pixels on the print to identify the machine

I believe all (color?) printers insert micro print so print out can be
identified regard less of what you print

-Lasse

 

[Spehro Pefhany]

I think they may have some back-door installed in the copying machines.

On another note, some machines must by law have a money bill detector. If a bill is detected, they insert some pixels on the print to identify the machine

Regards

Klaus

Many color printers put an ID dot code on the output page regardless
of the source.

http://w2.eff.org/Privacy/printers/docucolor/

http://www.washingtonpost.com/wp-dyn/content/article/2005/10/18/AR2005101801663.html

Photoshop (and Paint Shop Pro) have a banknote detection algorithm:
http://fstoppers.com/photoshop-wont-let-you-work-with-images-of-currency

 

[Nico Coesel]

Nico Coesel wrote:

[...]

I hope Atmel didn't lie to you. Atmel likes to be creative with their
specs... I had to put Atmel on my blacklist.

So far, in over 25 years, I had only one screw-up with Atmel. One of
their 8051 series uC would not reliably run at full spec'd clock speed
and I had to back off to 75%. They did fess up and apologize though,
something only very few and good companies (such as LTC) do when caught
with a bug.

I learned the hard way that none of Atmel's parts specced to run at
1.8V will work reliably at 1.8V.

Yikes, that does not sound good. What did they say about it?

Not much. Tech support didn't respond after I concluded the device
isn't suitable for running on 1.8V. I guess they knew they lost a
customer. I solved the problem by adding an extra 2V regulator in the
next production runs.

 

[Joerg]

Nothing special.


Do not. Use Vcc as reference.
Internal reference is inaccurate.

Yeah, sure looks ghastly. I provided my own reference from another board
and filtered it with 0.1uF.

It doesn't really matter.

However, not everyone knows that because they don't say much about the
innards of the chip. It's my first ATMega case, or maybe the 2nd.

 

[Tim Williams]

However, not everyone knows that because they don't say much about the
innards of the chip. It's my first ATMega case, or maybe the 2nd.

Finally gave up and bit the uC bullet? ;-)

Tim

 

[Klaus Kragelund]

Yeah, sure looks ghastly. I provided my own reference from another board

and filtered it with 0.1uF.









However, not everyone knows that because they don't say much about the

innards of the chip. It's my first ATMega case, or maybe the 2nd.

Its just a bypassing capacitor for the reference of the ADC (and what other analog uses the REF)

I have never seen a microcontroller datasheet that spelled out that it should be a certain value. It corresponds to bypassingm your average bandgap IC...

Cheers

Klaus

 

[Joerg]

Its just a bypassing capacitor for the reference of the ADC (and what
other analog uses the REF)

I have never seen a microcontroller datasheet that spelled out that
it should be a certain value. It corresponds to bypassingm your
average bandgap IC...

Yeah, but if you bypass, for example, the simple TL431 with the wrong
value of capacitance it can really sing the blues. Becomes unstable.
It's best to know if there is any risk of that. Other companies such as
Analog Devices do spell out the ref bypass values in their datasheets.

 

[Joerg]

Finally gave up and bit the uC bullet? ;-)

Sometimes you need it. A while ago I even had a switcher design that
would have been totally impossible to do without a uC. But it does raise
eyebrows if I request timers and port pins and MIPS, probably because of
my analog background. "YOU want some of the uC resources? What for?"

 

[Tim Williams]

Its just a bypassing capacitor for the reference of the ADC (and what
other analog uses the REF)

I have never seen a microcontroller datasheet that spelled out that it
should be a certain value. It corresponds to bypassingm your average
bandgap IC...

No, I don't think it's that. If it's configured for internal reference,
yes, but otherwise, it's just a bypass to keep sampling and whatever
clean. As a result, one would assume it should be pretty simple (just
bypassing for external noise and low impedance), unless using the internal
reference, in which case, some stability concern might be warranted (give
or take how crappy the reference is... which, as has been mentioned, is
rather awful to begin with!).

Tim

 

[Spehro Pefhany]

No, I don't think it's that. If it's configured for internal reference,
yes, but otherwise, it's just a bypass to keep sampling and whatever
clean. As a result, one would assume it should be pretty simple (just
bypassing for external noise and low impedance), unless using the internal
reference, in which case, some stability concern might be warranted (give
or take how crappy the reference is... which, as has been mentioned, is
rather awful to begin with!).

Tim

Here's a similar requirement on a Microchip part:

The LDO voltage regulator requires an external bypass
capacitor for stability. The VUSB pin is required to have
an external bypass capacitor. It is recommended that
the capacitor be a ceramic cap between 0.22 to 0.47 uF.
On power-up, the external capacitor will look like a
large load on the LDO voltage regulator. To prevent
erroneous operation, the device is held in Reset while
a constant current source charges the external
capacitor. After the cap is fully charged, the device is
released from Reset. For more information, refer to...



Best regards,
Spehro Pefhany

 

[Spehro Pefhany]

Yeah, but if you bypass, for example, the simple TL431 with the wrong
value of capacitance it can really sing the blues. Becomes unstable.
It's best to know if there is any risk of that. Other companies such as
Analog Devices do spell out the ref bypass values in their datasheets.

I've seen errata too.. I forget the wimpy term they used..
"fluctuations" or something like that, when they really meant
"oscillate like a banshee".


Best regards,
Spehro Pefhany

 

[Tim Williams]

Here's a similar requirement on a Microchip part:

The LDO voltage regulator requires an external bypass
capacitor for stability. The VUSB pin is required to have
an external bypass capacitor. It is recommended that
the capacitor be a ceramic cap between 0.22 to 0.47 uF.

<snip>

Considering the simplest absolute on-chip tolerances are factors of 2 to 5
(e.g., current limit on an average LDO), I wouldn't feel comfortable at
any point in that ridiculously narrow range!

The TL431 is, of course, an upside-down LDO, and must be compensated
accordingly. At least it's dominant-pole compensated, so it goes as
easily as any op-amp (keeping in mind the open collector output).

Tim

 

[Spehro Pefhany]

<snip>

Considering the simplest absolute on-chip tolerances are factors of 2 to 5
(e.g., current limit on an average LDO), I wouldn't feel comfortable at
any point in that ridiculously narrow range!

Presumably they've already worked those tolerances into the
requirements. It's a valid point, though, and why testing a sample or
two is not a very good way to get a reliable product. The next batch
of chips may be quite a bit different even if they don't fool around
deliberately with process parameters, die shrinking etc.

The TL431 is, of course, an upside-down LDO, and must be compensated
accordingly. At least it's dominant-pole compensated, so it goes as
easily as any op-amp (keeping in mind the open collector output).

Tim

There's a nice figure here (bottom left page 8) illustrating the edges
of the (potential) instability zones:

http://www.diodes.com/datasheets/TL431_432.pdf

Again, "presumably" most typical devices would be stable well onshore
the islands of doom.