Choosing lowest voltage

[mohd113]

Hello!

I am trying to design a circuit that can have many input voltages but only chooses the lowest one and pass it through. I know how to do it using a mircocontroller but I want to do it without it and only using analogue electronics.

Does anyone have an idea that might help?

Thank you in advance.

 

[ge]

prior to the microcontroller days, we would use comparitor circuits.

 

[crutschow]

You can indeed use comparator circuits but it can get complex for many inputs where you have to decide which one is lowest since each voltage has to be compared with all the other voltages.

 

[ge]

100% agreed crutschow. Speed is the only advantage I see to a comparator circuit.

 

[crutschow]

I think this is a case where an analog solution is overly complex and it should be done with a micro unless speed is of the essence.

 

[D.A.(Tony)Stewart]

I would do analog , with what little you have specified., Trivial.

Lowest man drives the output.

Adjust Supply V & bias R to meet demand current.

 

[erikl]

I want to do it without it and only using analogue electronics.
Does anyone have an idea that might help?

Here's a principle diode circuit which can do it approximately:

 

[crutschow]

I would it analog, with what little you have specified., Trivial.

Lowest man drives the output.

Adjust Supply V & bias R to meet demand current.

View attachment 106888

That's a neat circuit as long as the offset difference between the input and output BJTs is not a problem.

 

[D.A.(Tony)Stewart]

THe offset, I expect is only 1~10 mV with low current matched temperature, so bulk resistance differences yield small offsets. (Re)

Of course the PNP mixers go up 0.65V and the NPN goes down 0.65V and supply voltage must exceed the range required.

 

[crutschow]

THe offset, I expect is only 1~10 mV with low current matched temperature, so bulk resistance differences yield small offsets. (Re)
...................

My simulation with ±5V supplies, all resistors 10kΩ, 2N3906 PNPs and a 2N3904 NPN gave an offset varying from +8mV to -16mV over an input range of ±1V. And that's with idealized transistors and absolutely no temperature difference between the transistors.

The idealized offset can be improved to less than a mV if identical constant current sources are used in place of the resistors.

 

[BradtheRad]

A 'winner-takes-all' circuit was discussed in the thread below. The concept might apply here as well.

https://www.edaboard.com/threads/262870/

 

[crutschow]

A 'winner-takes-all' circuit was discussed in the thread below. The concept might apply here as well.

https://www.edaboard.com/threads/262870/

Good catch. There's a circuit there using precision rectifiers that I had forgotten about. The simulation below shows a two-channel circuit modified to output the minimum of the inputs. It can be expanded indefinitely by adding an opamp and diode for each added channel.

The signal offset is mainly determined by the input offset voltage of the opamp used.

Using a quad opamp package requires only 1 opamp package and 4 diodes per 4 channels.

 

[mohd113]

Thanks for the answers, I wanted to use analog because the speed really matters but I have to ask if I used a higher crystal with the micro could I achieve the same speed taking in consideration the time for fetching and decoding each code line of the program.

 

[KlausST]

Hi,

if micro, then with internal ADC?

If so, then usually the ADC is the bottleneck. Maybe it can convert 15000 times a second. with 6 channels you need about 400us.
The calculation, choosing channel and output the value to a port can be done during one ADC conversion, so it needs no extra time.
(my estimation for a ATMega32)

Klaus

 

[crutschow]

If speed is important than what's the maximum selection time that you can tolerate? You didn't mention that initially and it's apparently a critical design requirement. Also how much signal offset can be allowed?