Perhaps it's time to whip out Python and write a simulator for this problem.

"I'll be back."

 

I have finished my Python calculator...

It's designed to be extensible. In future I may add features such as standard deviation or temperature coefficients or other characteristics to the engine...

Only one mode, the "FB" mode, is supported.

You will need Python to use this. You can use Python directly on most Mac OS X and Linux computers, but you will need to install it on Windows.

The file is attached. It is licensed under a GPL version 2 license.

Here is some sample output (copyright notices truncated):

$ python sim.py Mode="FB" R1="115k 0.1%" R2="59k 0.1%" Vref="4.096V 0.1%"
R1: 114.885K <= 115.0K <= 115.115K
R2: 58.941K <= 59.0K <= 59.059K
Vref: 4.0919V <= 4.096V <= 4.1001V
Vout: 12.0533V <= 12.07855V <= 12.10379V
Range: +/-0.209%
$ python sim.py Mode="FB" R1="10k 0.1%" R2="10k 0.1%" Vref="4.096V 0.1%"
R1: 9.99K <= 10.0K <= 10.01K
R2: 9.99K <= 10.0K <= 10.01K
Vref: 4.0919V <= 4.096V <= 4.1001V
Vout: 8.17645V <= 8.19119V <= 8.20594V
Range: +/-0.180%

 

I am somewhat confused when you talk about a two resistor divider in which the output voltage is greater than the input voltage Vref.

Can you supply a schematic of the resistor divider you speak of so we can see what your circuit looks like?

hgmjr

 

It's part of a feedback mechanism. It is simulating the operational amplifier as well. This scales the percentage appropriately and is more useful for me. If people want, I'll add a normal divider.

 

The output accuracy of an ideal feedback loop, comparing a feedback signal (in this case voltage divider), and a reference voltage, is to my mind what comes out of the summer junction. The nuances you may want to review are:

- The point of out voltage measurement, and how different is that to the voltage you really want (eg. if it is a voltage going into a device and there is a current flowing then there will be distribution voltage drops).

- whether the signal is dc or has a frequency component.

- whether there is any bleed in the summer circuit, such that it impacts on either the reference or feedback signals (no issue if it just part of the subsequent loop gain stage).

- The magnitude of noise in both the feedback and the ref signals. This would be influenced by correlation and loop delay and I think forward loop gain.

Maybe someone can point to a good reference on all this - but obviously we are starting to get down to influences that are typically considered negligible in normal designs if discussing <0.1% accuracies.

Do we have any idea if this widgetr is being made in numbers more than 1 - as I would have thought going into SD's was appropriate mainly for large populations of widgets and a red herring if only 5 widgets are to be made.

Ciao, Tim

 

I am pretty sure that if a resistor is rated 1%, it means that the resistance is guaranteed to be within 1% of the nominal value.

Wow, amazing how a simple question can generate so much math! Equations are blooming like water lilies!

As for DC error sources in the feedback loop, I think it's really just:
1. voltage divider error
2. reference voltage error
3. comparator offset error
4. comparator input bias current error (probably negligible if divider is reasonable)

As for AC sources, I think those could be all over the map and may relate to feedback loop over- or under-damping and responsiveness to quick pulse currents, and distribution drops as you mentioned.