How do I set the adjustable reference voltage for TSM102A Dual Op-amp - Dual Comp & Voltage Ref IC.

[ChrisP58]

Yes, I would ideally like to operate the IC at the highest Vka available through Vcc.
Leaving good head room for IC tolerance.

If you start with a high voltage needing interface with Arduino.
Say 100v Max.
Divide to the highest reliable Vka, to amplify accuracy at 5v level.
.4% @ 20v = .04% accuracy after voltage division to 2v.

The final voltage on input of Arduino would still be of Max 5v.

I could be wrong with this basic logic, let me know.

Thanks

Wrong.
You do not increase accuracy when you go through a voltage divider. At best you maintain accuracy if you have perfectly matched resistors in the divider. But in most cases your initial accuracy is further degraded by the precision of the resistors in your divider.

 

[JimB]

Divide to the highest reliable Vka, to amplify accuracy at 5v level.
.4% @ 20v = .04% accuracy after voltage division to 2v.
This uses 2v example just to highlight the improved accuracy.

I could be wrong with this basic logic, let me know.

I am otherwise a bit busy at the moment, but yes you are wrong.

A 0.4% error remains a 0.4% error no matter how you multiply or divide it.

In reality it will get worse due to the errors in divider resistors and gain setting resistors in the amplifier.

JimB

 

[Gazza_AU]

I am otherwise a bit busy at the moment, but yes you are wrong.

A 0.4% error remains a 0.4% error no matter how you multiply or divide it.

In reality it will get worse due to the errors in divider resistors and gain setting resistors in the amplifier.

JimB

My thought process already assumes the need and use of voltage divider.
The question relates to a preexisting high voltage to interface.
My only basis of improved accuracy is the higher reference voltage with the same accuracy.
But this has turned out to be wrong.

Not sure what confused me, but it is clearly wrong.
Oh well

 

[JimB]

OK, I am back at the keys with a bit more time available.

In the "standard" version of the TSM102 the reference voltage has an error of +/- 1%.
ie it can have any value between 2.475V and 2.525V. That is +/- 25mV.

If we design and build a "perfect" voltage divider (R1 and R2) as per the expression in the datasheet, to give an output voltage of 25volts when the Vref is 2.5volts, then the output from our voltage reference can be anywhere between 24.75 and 25.25 volts. That is +/- 250mV.

In both cases, the error is +/-1%, but the absolute value of the error has increased from 25mV to 250mV, ie by a factor of 10, just like the output increased by a factor of 10.

JimB

 

[JimB]

Not sure what confused me, but it is clearly wrong.

I get days like that sometimes.

My only basis of improved accuracy is the higher reference voltage with the same accuracy.

I guess it all depends what you mean by accuracy.
Without raking through the datasheet again, it may be possible to include an adjustment potentiometer into the R1/R2 chain and adjust the (say) 25.15v volts that we get from some randomly selected TSM102, and make it EXACTLY 25.000 Volts, assuming that we have a voltmeter which is good enough to measure 25.000 volts accurately.
With time and temperature variations, it is likely that our 25.000v will drift to (say) 25.017v due to changes in the Vref, and the R1, R2 resistors.

Whether such a variation is acceptable all depends on what you are trying to do.

JimB

 

[Gazza_AU]

guess it all depends what you mean by accuracy.
Without raking through the datasheet again, it may be possible to include an adjustment potentiometer into the R1/R2 chain and adjust the (say) 25.15v volts that we get from some randomly selected TSM102, and make it EXACTLY 25.000 Volts, assuming that we have a voltmeter which is good enough to measure 25.000 volts accurately.
With time and temperature variations, it is likely that our 25.000v will drift to (say) 25.017v due to changes in the Vref, and the R1, R2 resistors.

Whether such a variation is acceptable all depends on what you are trying to do.

I want to accurately represent "the average voltage of any one cell" in a string of 17 cells.
The area of interest 3v to 3.55v, adjusted and feed to Arduino.
My first approach was to create 1/17th voltage divider. This will remain the base line.
But this is only academic, given Arduino could extract the same information via algorithm.

MAX6126A21+
5v 0.02% Initial Accuracy
3 PPM / C
$5.99

This could do the op-amp bias ref and Arduino Aref pin.
I would like to find a more affordable ref if possible, but it is an option.

https://au.mouser.com/ProductDetail/Maxim-Integrated/MAX6126A21+/?qs=sGAEpiMZZMuBck1X%2b7j9fB2T4ftaG3guOyx5Ymmp6ys=
DataSheet: https://www.mouser.com/ds/2/256/MAX6126-88935.pdf

Or;
MAX6250ACSA+
5v 0.02% Initial Accuracy
2 PPM / C
$9.17

https://au.mouser.com/ProductDetail/Maxim-Integrated/MAX6250ACSA+/?qs=sGAEpiMZZMuBck1X%2b7j9fB2T4ftaG3guusM4a1A2rpc=
DataSheet: https://www.mouser.com/ds/2/256/MAX6225-MAX6250-87956.pdf

 

[Mikebits]

I want to accurately represent "the average voltage of any one cell" in a string of 17 cells.
The area of interest 3v to 3.55v, adjusted and feed to Arduino.

I am sorry, but I am confused. I thought you wanted a ref voltage around 23 volts. Based on this I am now unsure?
I say this because I was actually trying to find a solution for you, but now I am unsure what actual ref voltage you need.

Also, even though you ignored my previous question, I asked how would you test the accuracy of your circuit. This is a very valid question and one you should consider. A HP 8.5 digit meter cost is very expensive, so if you were doing this project to build a product, you would have to verify that your product meets your specifications. I was not trying to be a wise guy by asking this question.
I think you would have received better responses if you had stated your reason for needing such accuracy (which I am sorry to say, unrealistic).

One more thing. You are looking to make a unbelievably accurate reference and you wish to do so very inexpensively. Well I can tell ya, this is not going to happen. To achieve the accuracy you proposed in your earlier post would be anything but cheap. In fact to scale your voltage would require resistors in the neighborhood of hundreds of dollars.

 

[Gazza_AU]

I am sorry, but I am confused. I thought you wanted a ref voltage around 23 volts. Based on this I am now unsure?
I say this because I was actually trying to find a solution for you, but now I am unsure what actual ref voltage you need.

Also, even though you ignored my previous question, I asked how would you test the accuracy of your circuit. This is a very valid question and one you should consider. A HP 8.5 digit meter cost is very expensive, so if you were doing this project to build a product, you would have to verify that your product meets your specifications. I was not trying to be a wise guy by asking this question.
I think you would have received better responses if you had stated your reason for needing such accuracy (which I am sorry to say, unrealistic).

One more thing. You are looking to make a unbelievably accurate reference and you wish to do so very inexpensively. Well I can tell ya, this is not going to happen. To achieve the accuracy you proposed in your earlier post would be anything but cheap. In fact to scale your voltage would require resistors in the neighborhood of hundreds of dollars.

You have misunderstood pretty much everything in this reply.
This was expected, and no problems.

I explored some factors that effect accuracy in order to make the best of the few dollars I have for components.
I do like pulling ideas apart, to understand more about how they work.

As a very successful electronic engineer once said to me, we just put chips together.
Not a degree in electronics, but an idea what to buy.
Components have changed in over 30 years, and asking for someones knowledge on Surface Mount technology seems valid.
The 23v reasoning has already been discussed and I have gratefully accepted my mistake and your replies, no worries.

Not here to argue, nor demand.
Sorry, but I don't want to argue for the truth on ever issue.
Moving on.

Anybody that has an interest to follow this post is welcome to ask me questions.