AD8428 Comon mode issue(?)

[DivadLarsen]

Hello,

I have an issue with the AD8428 instrumentation amplifier, that looks like common mode issue, but I might be mistaking.

Basically I drive a custom laser diode using an LT3092 current source (basically a Voltage-to-current source in my application), and use the AD8428 to measure voltage across a 0.1R resistor in series with the laser. (See attached schematic)
The current measurement is needed in a 3.3V micro controller, hence the AD8428 ref-pin is grounded.
Since the laser Anode is inherently grounded, all driving is done from a -12V rail, which works well; Set-voltage and actual laser current has been verified using a multimeter.



However, when I measure the output of the AD8428 there's a non-constant error, depending on current through the sense resistor.
Below are some measurements of the set current vs the current measured using the AD8428:

mA set mA measured
1 0.869
2 0.1859
5 4.844
10 9.826
12 11.815

With a fixed gain of 2000, this translates into an offset error going from 26mV to 37mV from 1 to 12mA.
As the laser voltage increases with driving current it seems logical that the error increasing with current could be a common mode issue.

Now, I get that due to the high gain of 66dB, CMR is only ~74dB (CMRR=140dB), but that should still only be around 1mV error with roughly 5V across the laser.

Am I missing something, is my math off or am I just stright up an idiot?

NB: The AD8428 datasheet suggests 10M resistors to GND from inputs, but since that's used to create a GND-point for input offset currents, I figure that shouldn't be much of a problem with a constant connection through the LT3092.
(Also I did a quick test, which seems to verify this - but honestly I don't really trust ratsnests on high-gain, high impedance inputs ).

 

[KlausST]

Hi,

Your circuit really needs improvement.

It really makes no esns to use a 0.1 Ohms resistor as shunt, which guves a very tiny output voltage .... then amplify thus tiny voltage with a gain of 2000.

On the other hand there is a 150 Ohms resistor in series .... with the same current as the shunt ... giving a factor of 1500 times the output voltage than the shunt.....
Thus the voltage across the 150 Ohms resistor is in the same range as the expected output voltage.

--> A simple difference amplifier with a gain of 4/3 can do the job.

And if you place the 150 Ohms to the supply ... then you decrease the common mode voltage travel.

Klaus

 

[DivadLarsen]

Hello Klaus,

thank you for your response.

Using a differential amplifier is actually a good idea if I can find sufficiently large resistors with close tolerances.
The choice of AD8428 was mainly due to low thermal drift, but that should be somewhat matchable with good resistors and a decent opamp.

With regard to the 150R resistor, it's used as a stabilizing measure for the LT3092 (which is not inherently stable with the lasers as the only load), and hence is subject to change value.
I could still split it into two parts, though; one for measuring and another for stability optimization. Assuming I can find close tolerances and low drift resistors in those kind of sizes/values - probably 50->100R-ish.

With regard to moving the sense resistor to supply (by which I assume you means GND), is not an option since the laser Anode is physically grounded, making the schematic node inside the laser itself. (Hence the wording "inherently grounded", but I might be off on my vocabulary, as english is not my native language(?))

Except for my circuit being shitty, do you have any idea as to why I'm seeing the offset issue?

Regards.

 

[FvM]

Expectable current measurement error due to AD8428 input offset is up to 1 mA (A Grade) respectively 250 uA (B Grade), in any case much higher than observed 130 uA. What are you complaining about?

Did you check the offset with zero laser current?

There's apparently an additional scaling error, how accurate is the 0.1 ohm resistor?

 

[DivadLarsen]

Hello FvM,

thanks for your reply!
I see your point on input offset, I completely missed that - do you happen to know if this scales with input voltage (diff or common mode)?
But assuming it's the input offset, that means whenever someone uses this amp (A grade), they shouldn't expect better performance than +/-0.2V on its outputs? I know it has a high gain, but that seems a bit muh to me...

Unfortunately, I'm embarrased to say that I didn't get an offset measurement with no current. I will try to do that as soon as possible...

The 0.1R is only a 1% 75ppm/C component. However I did also observe the issue with a 0.1% 15ppm/C resistor, although I do know have any data on how badly that (mis)scaled with current.
But that would only account for a constant error, right? Not the scaling issue...?

 

[DivadLarsen]

Yes, I could do a software correlation, but the issue may be likely to change over time/devices (depending on what's causing it), so for creating multiple devices a software solution might not be appropriate

Btw, please ignore my previous comment on expected performance (the "mumble, mumble +/-0.2V... mumble"-thing) - I do realized that is only valid for this very specific type of application / input scheme... My bad, I was a little hasty...

 

[FvM]

I presume 0.2V additional output offset doesn't overload the ADC range. How about a 2-point correction (y = ax + b) of offset and scaling error performed in software?

- - - Updated - - -

I should have said individual adjustment per channel rather than software correction.

Alternatively, if you can life with +/- 1% scaling error caused by the shunt, you could provide a means to shut off the current source and perform autozero adjustment in operation.