AD8227 output voltage limitations @ 0..100mV CM input

Document created by analog-archivist Employee on Feb 23, 2016
Version 1Show Document
  • View in full screen mode

I am considering using your AD8227 instrumentation amplifier to replace an
existing part. The existing part is unable to drive sufficient output voltage
at the common mode input voltage I have. I have a common mode input voltage
approaching 0V. I can see from the datasheet that the output voltage will drive
up to 2.8V with a 5V single supply rail, VREF connected to 0V. Can you tell me
what the output voltage capability will be limited too for this configuration
over the device temperature range? Are the plots in the datasheet (specifically
Figure 10) typical or maximum? If typical what variation would be anticipated? 
This is for a critical application which I need to be sure the output voltage
will not be limited over the temperature range of the device.

I am primarily concerned with common mode input voltages above 0V (up to 100mV
or so). My system design uses a gain of 20 providing output of 2 – 4V. However
I see the AD8227 cannot drive this output with such a small common mode
voltage. I am able to reduce the gain and re-scale my circuit but I want to
know to what extent I have to reduce the gain to guarantee operation over
temperature. I want to keep the gain as high as possible but ensuring expected

The supply voltage is 5V single supply. VREF is tied to 0V.  

The datasheet appears to show that with my setup and a gain of 5 I can expect
the output to be capable of driving to 2.8V. How does this vary with
temperature? How much margin do I need to add for tolerance/ through life? 

The application is low side monitoring a 100mR shunt resistor for a high
current discrete output (solid state relay). The AD8227 drives an overcurrent
shutdown circuit which should shut down if 1.5A flows through the shunt (150mV
input voltage – common mode 75V). My concern is what gain I should select and
therefore the scaling of the trip points to ensure the shutdown circuit will
operate over the full temperature range.

This is a commercial aircraft flight control application where we use 20
amplifiers per box.

Thank you for your calculations. I just wanted to check the calculations for
the AD8226 were truly applicable to the AD8227 as the AD8227 datasheet states:

"Most instrumentation amplifiers have a very limited output voltage swing when
the common-mode voltage is near the upper or lower limit of the part’s input
range. The AD8227 has very little of this limitation. See Figure 9 through
Figure 16 for the input common-mode range vs. output voltage of the part."

Comparing figures 10 and 13 between the datasheets appears to show two very
different responses. I was just wondering why this might be if both devices are
limited by the input stage?

If I were to reduce the gain to 5 the 1V out would be ok but this gives me
another problem because the output is dual purpose. The output is used to trip
a shutdown circuit at Vin = 150mV but also trigger a current flowing monitor
threshold at Vin = 10mV. With a gain of 5 this would only provide 50mV on the
output to indicate current flowing. The gain error of the device is not
specified below 0.2V. Can you tell me what the gain error would be or how to
work it out? Also what would the maximum output offset voltage be when the Vin
= 0? How would these vary with temperature? I have a load of approximately 3K
to ground (which could be increased if performance is improved).  

Thank you for the proposed devices. Unfortunately we have committed the design
to PCB using an MSOP8 package therefore we are trying to find a pin for pin
compatible device in the industrial temperature range. Do you know of any that
would be a drop in replacement for the AD8227 with improved functionality?


What gain do you need and at what voltage are you going to drive the reference?
The good news is that there are a few inamps that will work for applications
with common-mode voltages near the negative supply. The bad news is that the
plots shown on data sheets are not very conservative. These are from a single
part, tested at the time of release, at ambient temperature. So, they reflect
typical performance at best. On top of that, they represent the saturation
boundary for the amplifier, which means that there will be a gain error
(non-linearity) before reaching the limits.

So, how to determine whether or not this will work? Of course, this will
require you to test it in your particular application. But some guidance that
will make finding the appropriate choice easier :

a) The reason AD8227 goes below the supply is because the input transistors are
bipolar PNP devices, as can be seen in Figure 58. This means that nodes 3 and 4
will be ~0.6V above the voltage at the input. If the preamplifiers could only
swing, say 0.1V above the supply, this means you can have both inputs 0.5V
below the supply and all the transistors will be happy. This does not mean that
the subtractor will be, since it may see a little less voltage. Hence, you can
only go 0.3V below the supply or so.
b) The base emitter voltage on Q1 and Q2 is not well controlled, and it drops
as temperature increases by roughly 2mV/C. A 100C increase in temperature above
ambient (worst-case) will bring this voltage down by 200mV, thereby limiting
the input swing to merely 0.1V below the supply. But this may still be enough
for you.
c) By the same token, at lower temperatures, this part will perform better in
this regard.

We have a more recent inamp that is slightly less sensitive to temperature
variations, the AD8237 (assuming the other performance parameters still make it
viable option). Because AD8237 uses a discrete-time level shifting, it does not
have the same limitations as the traditional three-opamp architecture.

Thank you for the additional information; now it's clear why you have a problem.

As you have probably noticed, the diamond plot shifts upwards at high
temperatures. And like mentioned before, this is really due to the input
transistors having a lower base-emitter voltage, which also shifts the input
voltage range for each input.

I think it would be helpful to look at the AD8226 data sheet, page 20. AD8226
has a very similar input stage as the AD8227, so this information under the
“Input Voltage Range” section also applies to AD8227. From this page, you can
see that the maximum gain can be determined from equation (1). Equations (2)
and (3) must be evaluated as well, but in this case, it is the first equation
the one that imposes the limitation. In this case, Vcm=Vin/2 and Vdiff=Vin.
Solving for G we get that

G < 1+ (2*V_-LIMIT/Vin)

What this equation does is taking into account the base-emitter junction of the
input transistors and the maximum swing of the preamplifiers given the input
conditions and gain. These boundaries have been calculated and given in Table
8. Like mentioned before, the worst-case scenario is at high temperature (you
can see the drop in the negative limit value by 200mV in the table going from
25C to 125C). Solving the equation above for Vin (max) = 0.15V this results in
G<1.67. However, because AD8227’s subtractor has a gain of 5, we need to
multiply this result by this factor to get a maximum gain of 8.3.

However, from the information you provided, I assume your maximum Vin is 0.2V
(because you expect a maximum output voltage of 4 with G=20), in which case,
the gain is limited to be less than 7.5. This does not take into account much
margin, gain accuracy, gain drift and offsets. So, then I would further reduce
the gain to a value between 6 or 7; but at this point I would rather save the
trouble of having an additional resistor and use the default gain of 5, which
yields better gain accuracy and gain tempco. Of course, this creates the
problem that the output will only reach 1V, and will require further
amplification with a second gain stage, such as an opamp configured for a gain
of 4.

Have you considered using a current-sense amplifier? These will work best and
many come with pre-set gain of 20. Some options include AD8211, AD8215, AD8207,

If your application is power sensitive, another suggestion is AD8237. This
inamp won’t have the common-mode range limitation and is also specified up to
125C. The only question is if it has enough bandwidth for your application, but
assuming it does, it can solve the problem with enough gain in a single

Even though both devices are limited at the input stage, AD8227's second stage
has a gain of 5, while the AD8226's only has a gain of 1. This means that when
operated at gain of 5, the AD8226 will be more limited than AD8227. This is why
gain "G" was used for AD8226 to perform calculations and for the same
conditions on AD8227 at a gain of "G+5".  Anyway, this is why the figures look
so different.

Well, you have a difficult problem. If you are constrained by your current PCB
design, there are not a lot of options. If you want a drop-in replacement, then
this is it, AD8227 is your best choice.

With regards to the output swing and gain accuracy. Please keep in mind that
even though these are rail to rail amplifiers, they do not swing all the way to
ground. This is especially true for bipolar devices, such as AD8227. Their
nominal VCE(sat) is around 0.1V (at room temp), and it varies with temperature.
Asking for 50mV from ground is simply not realistic. This is why we don’t
specify much below 0.2V; when the output devices start to enter into
saturation, the gain becomes very non-linear. So the answer is yes, they will
swing below 0.1V and maybe down to 50mV, but I think that is pretty much it.
So, the gain error at such levels can be as large as 100% (or what I mean is
that if you had 0V at the input, the output would still remain at 50mV if so).
Usually, the solution for this is to place to bias the reference pin above
ground, to get away from this area of operation, but a lot of people forget
about that.

If the current solution does not work, you have two options: (1) you should
consider performing your own characterization/screening of the parts to find
the ones that work in an intermediate gain to balance all your design
requirements or (2) you need to consider redesign. You need to judge in your
case which one is the least costly.