Maker Pro
Maker Pro

Splitting voltage

huttojb

May 3, 2015
52
Joined
May 3, 2015
Messages
52
hey all

Don't know if this is possible, I'm looking if it's possible to split a voltage into 2 outputs to increase the sensitivity. I'll try and explain on what I'm thinking.

I have a 12v voltage which I want to split to a high side and low side for ADC.

So therefore some examples;

Input voltage of 12v.

Output A 5v
Output B 5v

Input voltage of 9.5v

Output A 5v
Output B 2.5v

Input voltage 6v

Output A 5v
Output B 0v

Input voltage 3v

Output A 2.5v
Output B 0v

Some complications, the 12v input is a vehicle battery so this could potentially vary between 11.5 to 14.5v

Is this even possible, and how? Thinking Op amps may have something to do with it??

Jason
 

AnalogKid

Jun 10, 2015
2,884
Joined
Jun 10, 2015
Messages
2,884
Everything you describe can be done with 3-terminal regulators. These function as electronically controlled variable resistors to maintain a constant output voltage when the input voltage or output current vary. An example would be the LM7805, which can take anything from +8V to +25V input and produce a steady +5V output at any current from 0 to 1 A. Similar fixed output regulators are available for 6 V and 9 V outputs. Or you can use an LM317, which has an output that is adjustable from approx 1.3 V to 10 V with a 12 V input. You are correct, all of this can be done with opamps and power transistors. But the regulator chips are way easier.

If you want a negative output from a positive voltage source, things get a bit more complicated but not much. An example of this would be needing +5V and -5V to power an opamp.

ak
 

BobK

Jan 5, 2010
7,682
Joined
Jan 5, 2010
Messages
7,682
I do not understand your examples at all. How would you reconstruct the actual voltage from the split voltages? It does not look like any simple relation would do this.

If all you want to do is read a voltage of up to 14.5V with a 5V ADC, simply use a voltage divider that will reduce the max voltage to 5V. I would set the top voltage at 15 and therefore use a ration of 2:1 for the two resistors. The lower resistor would then have a max of 5V across it and you would simply multiply the ADC reading by 3 to get the actual voltage. A voltage divider does not lose any accuracy (assuming you use resistors with appropriate precision) If you want more accuracy, you need a more accurate ADC.

In an automotive environment you should also use a Zener diode to suppress any peaks above 15V before the voltage divider, since the electrical system is very noisy and can have large spikes. I am not an expert on that, but others here are so maybe they will weigh in with a good protection circuit.

Bob
 

huttojb

May 3, 2015
52
Joined
May 3, 2015
Messages
52
Analog kid,

My output voltage will vary to situation so a regulator won't work.

Bobk. I agree with your theory and this is my initial option. But the single ADC 5v range will have low resolution. This is why I'm looking to split it into 2 channels.

The 2 ADCvoltages will be imputed to a microprocessor which will be joined back together. Therefore increasing resolution.

Sorry, when I get into my laptop I'll send more detailed specification.

Jason.
 

BobK

Jan 5, 2010
7,682
Joined
Jan 5, 2010
Messages
7,682
What is the resolution of the ADC? If it is 12 bits, you would have to have very precise electronics to do the splitting and make it come out higher. And at most, you are getting 1 more bit of resolution.

Simply taking more samples and averaging them will also increase the resolution.

I would also question why you would need to know the voltage of a vehicle electrical system to such precision. You do realize that is constantly changing depending on demand, engine speed, etc., right?

Bob
 

huttojb

May 3, 2015
52
Joined
May 3, 2015
Messages
52
Hello Bob.

Sorry, I was writing the post whilst sitting in a Finance meeting, you can imagine how fun it was (only another 2 days of it!!!!) lol.

Ok, now I have more time to compose the post, I can explain more.

I'm monitoring the resistance from the fuel sender, I will put the fuel sender into a potential divider. With my initial thoughts I will have a source of 5V so therefore, I only have a voltage range of 0v to (probably) approx 4.7v. Voltage will be in poportion to fuel levels.

I am monitoring the potential divider output into a 10 bit ADC.

I'm concerned that the range of the input voltage to the ADC will not give a great deal of variance between actual differences. So my thoughts....

If I use Vehicle Battery as the source (which I already monitor so I know Max value) I can get an output of say 8volts. Obviously I can not send 8v into my ADC so I split this into a most significant and least significant bits.

This will give me more range on the input voltage giving me more resolution ( I think ).

So for example;

If vehicle batt is 13v and I have 6.5v coming from my potential divider. This shows me I have 1/2 tank of fuel.

I would want 5v on pin A and 0v on pin B.

Example 2

Same battery voltage, but out from pd is 9.7v. Identifying I have 3/4 tank of fuel (being 0.75 of max value)

I would want 5v on pin A (showing the bottom half of tank is full) and 2.5v on pin B identifying top half of tank is half full.

Etc...

I hope this makes more sense. This is all currently at the theory stage so I'm just bouncing ideas around to get the best solution.

Jason.
 

AnalogKid

Jun 10, 2015
2,884
Joined
Jun 10, 2015
Messages
2,884
Your theory is clear, but will not work in the real world. The worlds most perfect 10 bit A/D converter has a resolution of 1 part in 1024, or almost exactly 0.1% accuracy. If your voltage dividers use 0.000001% accurate resistors, and you take the two readings and combine the results, the result will have an error tolerance of 0.2% and there nothing you can do to improve that. If you use 1% tolerance resistors in two scaling dividers, the result will have a maximum possible error of over 2%. Note that this is 100% worse than a single 1% tolerance divider driving the A/D converter. And none of this includes the errors accrued in the subtraction circuit that "splits" the input into two different voltages.

Full range input voltage > the best voltage divider you can afford > single range A/D converter > best possible result.

ak
 

huttojb

May 3, 2015
52
Joined
May 3, 2015
Messages
52
Your theory is clear, but will not work in the real world. The worlds most perfect 10 bit A/D converter has a resolution of 1 part in 1024, or almost exactly 0.1% accuracy. If your voltage dividers use 0.000001% accurate resistors, and you take the two readings and combine the results, the result will have an error tolerance of 0.2% and there nothing you can do to improve that. If you use 1% tolerance resistors in two scaling dividers, the result will have a maximum possible error of over 2%. Note that this is 100% worse than a single 1% tolerance divider driving the A/D converter. And none of this includes the errors accrued in the subtraction circuit that "splits" the input into two different voltages.

Full range input voltage > the best voltage divider you can afford > single range A/D converter > best possible result.

ak

Thanks for your response. And I "think" I follow your explanation. Maybe we are not talking about the same concerns.

If we scale the full tank between 0 and 100, 0 being empty and 100 being full. If I use 5v I need a 0.05v change to increase the fuel tank. This is what I'm concerned with.

My proposal to use 2 lines will allow a 0.1v change to effect the fuel level. I think it will be difficult to monitor a 0.05v change accurately.

Maybe I just need to try to see how it reacts.

Jason
 

huttojb

May 3, 2015
52
Joined
May 3, 2015
Messages
52
Just doing some more thinking on this.

10 bit ADC will break the range in 5/1024 = 4.8mV approx so therefore I can easily monitor a change of 50mV. So maybe I'm overthinking this.

Jason
 

AnalogKid

Jun 10, 2015
2,884
Joined
Jun 10, 2015
Messages
2,884
Now you're getting there. Engineers working with data conversion talk about ENOB - effective number of bits. As above, in round numbers:

0.1% is effectively 10 bits.
0.2% = 9 bits
1% = 7 bits
2% = 6 bits

So a voltage divider made with 1% resistors turns your 10 bit A/D into a 7 bit A/D in terms of its absolute accuracy. ENOB a way of seeing where your error budget is being spent, and where to attack first if you are trying to improve the system performance. You have more than enough resolution to do what you want without the splitter/subtractor.

ak
 

GPG

Sep 18, 2015
452
Joined
Sep 18, 2015
Messages
452
Last edited:

AnalogKid

Jun 10, 2015
2,884
Joined
Jun 10, 2015
Messages
2,884
Welcome to high school physics. Today's topic is the cosine of gravity...
 

huttojb

May 3, 2015
52
Joined
May 3, 2015
Messages
52
AK / GPG

As this is monitoring the fuel sender, I don't need a quick response and I have to take into consideration liquid movement within the tank my reading will, be in processing world, very long.

I would imagine I'll take a reading every 1000mS and average out probably 10 reading before outputting giving me a 10s update pole.

Thanks for everyone's advice.

Jason.
 
Top