Rail-to-rail op-amp suggestion?

[[email protected]]

Not being an analog guy, I'm having trouble choosing an op-amp from the
vast array available. I just want to use it as a voltage follower to
handle a slight impedance mismatch between a sensor and an ADC.

5V supply with output swing from at least 0.25 to 4.5V.
Low-bandwidth application (DC to 300Hz)
Would like a dual-gate 8-pin SOIC or (T)SSOP if possible.

I realize this question sounds stupid, but is an LM358 suitable for
this application?

 

[[email protected]]

Not being an analog guy, I'm having trouble choosing an op-amp from the
vast array available. I just want to use it as a voltage follower to
handle a slight impedance mismatch between a sensor and an ADC.

5V supply with output swing from at least 0.25 to 4.5V.
Low-bandwidth application (DC to 300Hz)
Would like a dual-gate 8-pin SOIC or (T)SSOP if possible.

I realize this question sounds stupid, but is an LM358 suitable for
this application?

No. The input common mode voltage range of the LM358 only extends up to
1.5V below the rail - 3.5V with a 5V rail. A 4.5V output voltage
doesn't seem too realistic either.

You might want to look at the Motorola (now Freescale?) MC33201. It's
particular virtue is a realtively low output impedance, which is handy
when you are driving nasty capacitative ADC inputs.

 

[PeteS]

No. The input common mode voltage range of the LM358 only extends up to
1.5V below the rail - 3.5V with a 5V rail. A 4.5V output voltage
doesn't seem too realistic either.

You might want to look at the Motorola (now Freescale?) MC33201. It's
particular virtue is a realtively low output impedance, which is handy
when you are driving nasty capacitative ADC inputs.

An amp I used successfully for driving nasty capacitive loads with rail
to rail in and out (within 100mV of the rail for both), rated at 2.5V
to 30V supply (typical 5V) is the LM2861. I was using this as a unity
gain buffer, which is your application.

http://www.national.com/pf/LM/LM8261.html

Cheers

PeteS

 

[PeteS]

An amp I used successfully for driving nasty capacitive loads with rail
to rail in and out (within 100mV of the rail for both), rated at 2.5V
to 30V supply (typical 5V) is the LM2861. I was using this as a unity
gain buffer, which is your application.

http://www.national.com/pf/LM/LM8261.html

Cheers

PeteS

For the OP - unity gain is usually the worst case for capacitive
loading, as it typical (as Bill alluded) on an ADC input. There's an
excellent app note about it from National (marketing their designs, of
course, but hey, that's business)

http://www.national.com/appbriefs/files/AppBrief108.pdf

Cheers

PeteS

 

[Phil Hobbs]

An amp I used successfully for driving nasty capacitive loads with rail
to rail in and out (within 100mV of the rail for both), rated at 2.5V
to 30V supply (typical 5V) is the LM2861. I was using this as a unity
gain buffer, which is your application.

http://www.national.com/pf/LM/LM8261.html

Interesting. There was one thing in the datasheet that made my brain
hurt a bit though--maybe I've been doing device work too long, but I'm
puzzled by this passage from Page 15:

The output stage Figure 1 is comprised of complementary
NPN and PNP common-emitter stages to permit voltage
swing to within a VCE(SAT) of either supply rail. Q9 supplies
the sourcing and Q10 supplies the sinking current load.
Output current limiting is achieved by limiting the VCE of Q9
and Q10; using this approach to current limiting, alleviates
the draw back to the conventional scheme which requires
one VBE reduction in output swing.

Q9 and Q10 are the output transistors. But how does limiting VCE
perform current limiting? Maybe they meant VBE? I can think of ways to
do that, but VCE?

How do you IC guys do current limiting in a bipolar RRO amplifier?

Cheers,

Phil Hobbs

 

[Spehro Pefhany]

Not being an analog guy, I'm having trouble choosing an op-amp from the
vast array available. I just want to use it as a voltage follower to
handle a slight impedance mismatch between a sensor and an ADC.

5V supply with output swing from at least 0.25 to 4.5V.
Low-bandwidth application (DC to 300Hz)
Would like a dual-gate 8-pin SOIC or (T)SSOP if possible.

I realize this question sounds stupid, but is an LM358 suitable for
this application?

No, for reasons Bill described. You might be able to get the output
swing with a pullup, but the input stage can't handle signals close to
the +ve (ObJohnLarkin: 'positive') rail. So, forget it.

Last time I had this requirement for a moderate quantity design, I
picked a Microchip MCP602x series part for its low max Vos** (and a
number of other factors), but there are many, many parts which will
meet your stated requirements.

** Note that with RRIO opamps the Vos can behave a bit strangely as
you transition between the two input stages (Vos shifts suddenly by
scores of uV with CM input voltage (and the temperature sensitivity
changes in the case of the MCP602x series)). If your ADC is ~10 bits
you can virtually always ignore this, because it's quite a bit less
than the quantization error (~5mV in this case if Vref = 5.0V/10bits).


Best regards,
Spehro Pefhany

 

[James Beck]

Not being an analog guy, I'm having trouble choosing an op-amp from the
vast array available. I just want to use it as a voltage follower to
handle a slight impedance mismatch between a sensor and an ADC.

5V supply with output swing from at least 0.25 to 4.5V.
Low-bandwidth application (DC to 300Hz)
Would like a dual-gate 8-pin SOIC or (T)SSOP if possible.

I realize this question sounds stupid, but is an LM358 suitable for
this application?

Microchip has some parts made just for this type of application. I use
one of their MCP6271 for a little signal conditioning on one of our
products. Just go to microchip.com and take a look. Should be several
that they list as rail to rail input and output.

Jim

 

[[email protected]]

Hi Bill (et al - this is a global reply):

No. The input common mode voltage range of the LM358 only extends up to
1.5V below the rail - 3.5V with a 5V rail. A 4.5V output voltage
doesn't seem too realistic either.

The main reason I was looking at the 358 is that I bought twelve reels
of them a while ago, and still have six and a half left. It's the only
dual op-amp in an 8-pin package that I have in my home lab, in fact. I
have a regulated +12V rail in my application, and for that matter a
rather wiggly +/-10V available at very low current. I had some crazy
plan to power the 358 off +12, and bias the input up off the 0V rail,
but I kind of stopped there.

I've downloaded the datasheets for the parts suggested by everybody,
thanks for the input. I happen to have a friendly Microchip rep keen to
get her parts into my work, so it will PROBABLY wind up being the uChip
part....

 

[[email protected]]

** Note that with RRIO opamps the Vos can behave a bit strangely as
you transition between the two input stages (Vos shifts suddenly by

I think this factor can be ignored in my application. The sensor value
is never going to be static - it wanders constantly. The software
filters the input; it maintans a running average of sixteen samples in
a ring buffer. If "this" sample differs from the average by more than a
threshold, a second ring buffer is populated with a copy of the current
ring buffer, and the new value is put in the second buffer. If, after
sixteen further samples, the average of buffer 2 is further away from
the original average than 1/16th the difference between the first
maverick and the original average (i.e. if a trend is detected), the
second buffer is copied into the first buffer and it goes back to
"quiet" mode. I'm not quite sure what to call this sort of filter, but
it works very well in the application.

The software that needs the sensor reading only sees the filtered
average.

 

[Spehro Pefhany]

I think this factor can be ignored in my application. The sensor value
is never going to be static - it wanders constantly. The software
filters the input; it maintans a running average of sixteen samples in
a ring buffer. If "this" sample differs from the average by more than a
threshold, a second ring buffer is populated with a copy of the current
ring buffer, and the new value is put in the second buffer. If, after
sixteen further samples, the average of buffer 2 is further away from
the original average than 1/16th the difference between the first
maverick and the original average (i.e. if a trend is detected), the
second buffer is copied into the first buffer and it goes back to
"quiet" mode. I'm not quite sure what to call this sort of filter, but
it works very well in the application.

The software that needs the sensor reading only sees the filtered
average.

I think I didn't explain this fully... it's a DC effect.

"Suddenly" refers to the slope of Vos vs. Vcm (offset voltage vs.
common mode voltage) not Vos vs. time (or imagine you were increasing
common mode voltage at certain rate). But it's well under a mV (100 or
200uV, IIRC) in size, so unless you have a lot of bits or are
dithering with noise or something like that to get more effective bits
it will be well hidden.


Best regards,
Spehro Pefhany

 

[Spehro Pefhany]

Hi Bill (et al - this is a global reply):



The main reason I was looking at the 358 is that I bought twelve reels
of them a while ago, and still have six and a half left. It's the only
dual op-amp in an 8-pin package that I have in my home lab, in fact. I
have a regulated +12V rail in my application, and for that matter a
rather wiggly +/-10V available at very low current. I had some crazy
plan to power the 358 off +12, and bias the input up off the 0V rail,
but I kind of stopped there.

Well, you didn't say anything about having 12V available (and we
assumed you didn't). In that case, you probably can use the LM358, but
you'd have to clamp the output so it doesn't cause problems with the
ADC. The LM358 also has a rather large maximum Vos (that means a DC
offset error! perhaps several bits), but aside from that it works
almost down to the - rail. You could get a -0.6V with a diode and
resistor and a +12V from your regulated supply if it was necessary to
get down to the rail. If your ADC can tolerate a couple of K you could
use the second amplifier as a clamp using a diode and the 5V supply as
a reference. If you only need to go to 50mV or so then you wouldn't
need the - supply, ground would do.

I've downloaded the datasheets for the parts suggested by everybody,
thanks for the input. I happen to have a friendly Microchip rep keen to
get her parts into my work, so it will PROBABLY wind up being the uChip
part....

The CMOS parts from Microchip are nice, and not insanely expensive,
but they cost a lot more than an LM358 bought in bulk (probably under
a dime). You do get 500uV Vos, almost no bias current, and few
external parts for the money. OTOH, there are some applications where
the bias current is actually an advantage (detect an open connection).


Best regards,
Spehro Pefhany

 

[Fred Bartoli]

I think I didn't explain this fully... it's a DC effect.

"Suddenly" refers to the slope of Vos vs. Vcm (offset voltage vs.
common mode voltage) not Vos vs. time (or imagine you were increasing
common mode voltage at certain rate). But it's well under a mV (100 or
200uV, IIRC) in size, so unless you have a lot of bits or are
dithering with noise or something like that to get more effective bits
it will be well hidden.

Not all the RRI opamps exhibit this effect.
See the National's range of CMOS opamps.

For example:
http://cache.national.com/ds/LM/LMC6484.pdf

 

[Spehro Pefhany]

Not all the RRI opamps exhibit this effect.
See the National's range of CMOS opamps.

For example:
http://cache.national.com/ds/LM/LMC6484.pdf

Yes, DS011714-60 and DS011714-61. Much better.


Best regards,
Spehro Pefhany