Anyone Familiar With This Circuit Logic in Analog Way?

[Mechatrommer]

i can do it with mcu, but it seems to be overkill and unstable (PWM output need taming to make it a stable voltage, (DAC)), and probably slower. so i'm seeking analog way of doing this, not sure the correct term for this kind of circuit to google.



basically the circuit will output the lowest voltage between its input voltages. i'm thinking of some transistors arrangement, but i cant think of a proper way.

spec:
- both input impedance should be high somewhere greater than 100Kohm (1Mohm preferably) because Vi1 and Vi2 come from high resistive voltage divider impedance of around 10Kohm.
- BW: dont care, as long as it can produce stable voltage. somewhere above 10KHz will be acceptable
- output capacity is to be used as Vref for other circuit, so output impedance doesnt necesarily to be significantly low since it wont drive any significant load.
- a slight voltage drop (diode) is acceptable since it can be compensated on the inputs.

[Signal32]

How about comparator which drives two MOSFET's, one MOSFET is on when comparator output is high and other MOSFET is on when comparator output is low.
MOSFET source connected together, here is where you will find your voltage.
Something like attached ( appropriate transistor resistors not included + you may also have to swap which voltage goes to which MOSFET drain)

[bson]

A comparator and an analog 2:1 mux.  Buffer the inputs.

[Mechatrommer]

thanks Signal32. can i change the n-mosfets with npn bjts? which one better? i heard mosfet is less stable compared to bjt in linear mode, no?
bson, i dont have 2 channel analog multiplexer. i'm trying the discrete solution.

[Signal32]

BJT's will not work here at all because of multiple reasons. You're trying to forward a high-impedance voltage. There will be a voltage drop across the junction of the BJT. Also you'll be controlling the base with a different voltage which will mess up forwarding the original voltage.

[dmills]

100K resistor from a supply more positive then your maximum input to the base of a NPN transistor, 22K emitter resistor.
A diode from the base to each of your inputs, output is the emitter, job done?

The base will be ~0.6 volts more positive then the lower one of your inputs (Due to the diode), so the emitter will be at a voltage equal to your lower input.

If you want to buffer, replace the diodes with two PNP transistors, and lower all the resistors by a factor of 10 (but do respect the reverse Vbe limit of the PNP parts).

Regards, Dan.

[Mechatrommer]

job done?

if you meant this, i think the base voltage will follow the larger input...

[rfeecs]

Maybe something like this?

[Mechatrommer]

Maybe something like this?

it seems to do the job in simulation. but requiring 2 opamps. thanks.

[dmills]

if you meant this, i think the base voltage will follow the larger input...

Diodes the other way around, output from the other end of the 22K, bottom of the 22K to negative rail.
Or replace diodes with PNP transistors, emitter to the base of the NPN, collector to negative rail input to base of PNP.

Pretty sure either will get it done.

Regards, Dan.

[StillTrying]

Yes, 2 op amps and 2 diodes seem to be quite good, output will only go down to 0.8V though.

Posted in full in case any LT'er wants to continue with it.

[rfeecs]

Yes, 2 op amps and 2 diodes seem to be quite good, output will only go down to 0.8V though.

Posted in full in case any LT'er wants to continue with it.

Those little glitches at the transitions might start to be a problem up at 10kHz.

Here's another possible version of Dan's circuit:

[StillTrying]

rfeecs/Dan's 2 transistor version.

Emitter resistors 1K, diodes pull-up resistor 10K, freq 10 KHz. Input Z probably varies 150K-400K, output Z
8 to 5K ?

[Mechatrommer]

Diodes the other way around, output from the other end of the 22K, bottom of the 22K to negative rail.
Or replace diodes with PNP transistors, emitter to the base of the NPN, collector to negative rail input to base of PNP.

understood and it worked in sim, thanks.

[dmills]

The three transistor variant has much higher input Z if that matters to you.

Regards, Dan.