ISO124 Gain error

Hello cafain,

Unfortunately we only have a typical specification for that parameter.

Best regards,

Jose

Hello Jose,

I understand.

Thank you,

Hello !
Please explain me, how to estimate gain error, based upon datasheet, if ambient temperature is 60 degrees. What includes a gain error?

Parametrs:
Power supply: + -15V

Rload 30k

At an ambient temperature of 25 degrees:

Input voltage ISO124: 5.0006V DC

Output voltage ISO124: 5.0138V DC

Gain=5.0138V/5.0006V=1.00264

_______________________________________

At an ambient temperature of 60 degrees:

Input voltage ISO124: 5.0002V DC

Output voltage ISO124: 5.0228V DC

Gain=5.0228V/5.0002V=1.00572

_______________________________________

Gain error = (1-(1.00264/1.00572))*100=0.306%

I am confused by this gain error,  when device heated. Is it OK? If so, how this gain error to correlate to datasheet?

Thank you.

P.S. Sorry for my English.

Hello Alexander Temchuk,

The process to find the gain error is a little more involved than the calculations showed in your post.

In order to find the gain error of the device you need to slowly sweep the input from -10 V to 10 V (for power supply set at +/- 15 V and Rload = 2 kohm) and take several measurements (for example in 0.5 V steps) of input and output. This sweep will reveal that the transfer function of the device has three sources of error:

1)      Offset error

2)      Gain error

3)      Non-linearity

The gain error can be found by obtaining the slope of the best-fit straight line through your collected data, then subtracting that slope from 1 (since the ideal gain for the device is 1 V/V) and finally multiplying by 100 (so that the error is expressed in percent).

The typical gain error drift is obtained by dividing the obtained gain error by the specified temperature range of the device (-25 deg. C to 85 deg. C). Therefore, the typical deviation one can expect is 1100 ppm (0.11%) throughout the specified temperature range.

For narrower temperature ranges (such as the one indicated in your post) one cannot predict the exact expected gain drift because the gain error versus temperature behavior of the device might not be monotonic.

In summary, a typical unit is expected to have a gain error in the +/-0.16% range throughout the specified temperature range.

Hope this helps.

Best regards,

Jose

Hello Jose Duenas!

In my previous post I incorrect described the problem about gain error. I was embarrassed not by the gain error vs temp. only, but the overal gain error. However, I spent several experiments based on your recommendations.

First of all I find overall gain of the device, at an ambient temp. 25 and 60 degrees. I took measurements of the input and output signal in 0.5V steps (Vs=+/-15, Rload 2kOhm). Then, I find the best-fit straight line through my collected data, and slope of this line. Of Course, I find some offset voltage (near zero), and exclude one from gain characteristic.  Then i subtract slope of corrected gain characteristic from 1 and multiplying by 100. The results is:

__________________________

for 25°:

Offset voltage = 8,5 mV

Max nonlinearity = 0.008%

Gain error (without nonlinearity and offset) = 0.08%

for 60°:

Offset voltage = 13.22 mV

Max nonlinearity = 0.01%

Gain error (without nonlinearity and offset)  = 0.22%

__________________________

Analyzing these results we can approve that in case of overheating of 35°(temperature rise from 25° to 60°), increment of the gain error is 0.14% (0.22%-0.08%). But, the typical deviation gain error drift from datasheet is 10ppm/C°, ie 350 ppm (0.035%). Also, in datasheet shows maximal gain error - +/-0.5%, which contain Gain vs Temp (1100ppm - 0.11% max) and Nonlinearity (0.01% max). So, the sum of gain vs temp. and nonlinearity  not equal maximal gain error. Of course, we can observe, that Gain vs Temp is only typical value. I just try to determine, is it possible to accurately predict gain error vs temp drift ? Can overal Gain Error include anything else except nonlinearity and Gain error vs temp, if we do not take into account offset voltage vs temp. (This fact we can  exclud  by calibrating the main device, e.g.). 

What can I realy wait of ISO124 throughout the specified temperature range?

A more detailed calculation, done in "Mathcad" software:

Thank you!

Hello Alexander Temchuk,

Thank you for sending the additional measurements and calculations. There are a few points that I need to clarify from your post:

1)      There is a slight mistake in the way you calculate the gain error. The correct derivation (although the answer will be the same) is to divide the 5.004 by 5 and then subtract 1 from this quotient.

2)      You mentioned: “Also, in datasheet shows maximal gain error - +/-0.5%, which contain Gain vs Temp (1100ppm - 0.11% max) and Nonlinearity (0.01% max).”
Please bear in mind that the 0.5 % maximum gain error specification applies only to 25 degree Celsius and is for the slope of the best fit line, in other words, the slope of your Mathcad operation “line(x25, y25)”. It does NOT include the effect of temperature, offset or non-linearity.

I can see that you found a 0.14 % increase in the gain error of the device when going from 25 degrees C to 60 degrees C. This change clearly falls outside the typical gain error drift specification.

With your measurements you have found two points in the Gain vs. Temperature curve of the specific device that you used. I expect the device to behave similarly if you repeat the experiments at 25 deg. C and 60 deg. C; however, there is no guarantee that another device will behave the same as the one you just characterized. The only way to accurately predict the gain vs. temperature behavior is by doing what you just did.

Note also that there is another compounding factor: long term drift. This subject is described in the following blog post:

http://e2e.ti.com/blogs_/b/precisionhub/archive/2014/02/20/ic-long-term-stability-the-only-constant-is-change

Best regards,

Jose

Hello Jose Duenas!

Thank you for your detailed response! But I still have some questions.

I researched 18 amplifiers (ISO124) used in my measuring transducers. And found that ~40% devices is with similar deviations.
If this change of gain falls outside the typical gain error drift specification, does this mean that the device (ISO124) is defective (may be counterfeit)?

Thank you.

Hello Jose!

I am sending pictures of 15 amps. 3 of them are varnished, that is why titles on the cases is not visible. But this titles are same as the others.

Supplier: COMPEL, LLC, Moscow. (http://www.compel.ru/)

Receipt date: 26.08.2013 

Files:

http://www.filedropper.com/iso124photo

Thank.

Hello, Jose !

Thank you for conclusion of marketing department .

What about isolation rating in my devices. Required isolation rate is 500Vrms. Also It should be taken into account varnished pcb, and minimal distance between traces running under the ISO124 units of 1mm. I believe that these factors should be sufficient to provide isolation of 500Vrms. Future tests will answer this question.

My last question about ISO124 gain error.

We have determined, that changes of gain error (when going from 25 degrees C to 60 degrees C) clearly falls outside the typical gain error drift specification. Typical gain error drift specification is 10ppm/°С (in practice deviation of some units at 3-4 times more than the typical specification). But this is only typical drift, not max. And a maximum drift in datasheet is unknown. Therefore we can say that, in terms of datasheet, ISO124 with gain error drift more than the typical at  3-4 times, works correctly (we are talking about the obligations of the manufacturer). Isn't it ?

Thank you!