Why does my current mirror not work

[ericgibbs]

I was thinking about the next one:
NXP|PBSS4021SN|TRANSISTOR,NPN/NPN,20V,SO8 | Farnell Nederland

But probably I'll need one with a heatsink in the end.

hi,
Current rating looks OK, the wattage is lowish, have you checked the dissipation in the transistor using LTspice at 2amps, the transistor will be linear mode.

EDIT:
Using a 2N3055 model. sim result, note 0.11R in left emitter.

 

[turbo_henky]

I first wanted to know if the current mirror would work. When that's working I'm going to look the the dissipation, which is also depending on the voltage over the peltier elements. When increasing the current through a peltier element, the voltage increases also, so less voltage will be over the transistor.

But I will try to make a circuit with 0.1R emitter resistors. The advantage of the Emitter resistors I can change the multiply factor of the current mirror.

 

[ericgibbs]

I first wanted to know if the current mirror would work. When that's working I'm going to look the the dissipation, which is also depending on the voltage over the peltier elements. When increasing the current through a peltier element, the voltage increases also, so less voltage will be over the transistor.

But I will try to make a circuit with 0.1R emitter resistors. The advantage of the Emitter resistors I can change the multiply factor of the current mirror.

hi,
By trimming emitter value you can of course improve the tracking.
As you say, a heat sink will be required.
Let us know how the mirror works in the final application.

 

[kinarfi]

I understand the theory as explained on the link. My simulation is working as expected. Why should I post my simulation file?
My real circuit does not work.

So other can play with it and experiment and be helped as well as help, why not share it?

 

[audioguru]

How did you make the two transistors at exactly the same temperature? Some people glue them together.

Many years ago Texas Instruments made current mirror ICs in a TO-92 case. They were available in many fixed small currents.
Now a few semiconductor manufacturers make them but in tiny surface-mount packages.

 

[ericgibbs]

How did you make the two transistors at exactly the same temperature? Some people glue them together.

Many years ago Texas Instruments made current mirror ICs in a TO-92 case. They were available in many fixed small currents.
Now a few semiconductor manufacturers make them but in tiny surface-mount packages.

hi,
He did use a dual package on one trial.

I also build the circuit with a PBSS4021SN, which has 2 transistors in  one package. This configuration does also not work. The current through  R3 is increasing when decreasing R3

 

[carbonzit]

My textbook (Malvino's Electronic Principles) gives this as the generic current mirror:

**broken link removed**

Is this still the preferred configuration?

The second diagram implies that it doesn't matter whether you use the same supply or separate supplies; is this true?

He makes the point later about current mirrors using discrete components:

For a complementary mirror to work properly, the curves of the compensating diodes must match the transconductance curves of the transistors. In discrete circuits, this is not easy to do because of diode and transistor tolerances; so you have to settle for less than perfect mirror action. But diode bias really shines when it comes to linear integrated circuits. Since the compensating diodes and emitter diodes are on the same chip, they have almost identical characteristics; this means almost perfect mirror action over a large temperature range.

 

[Roff]

I think too much is being made of gain (beta) mismatches. IMHO, the most sensitive parameter is emitter area mismatch, which exhibits itself as Vbe mismatch. If you want to play with this, add an area multiplier term after the transistor's name. For example, "2N3904 M=1.1" will make the transistor 1.1 times larger. In the Philips model that I have, this causes a Vbe mismatch of about 2.7mV at Ic=10mA. This in turn causes a mirror current mismatch of 10%, which we should expect, given the 10% area mismatch.
Now take two 2N3904s with identical areas, and change the beta on only one of them to 100 (nominal is 300 in the model). The mirror current mismatch simulates at about 5%.

 

[turbo_henky]

How did you make the two transistors at exactly the same temperature? Some people glue them together.

Many years ago Texas Instruments made current mirror ICs in a TO-92 case. They were available in many fixed small currents.
Now a few semiconductor manufacturers make them but in tiny surface-mount packages.

I did a test with the two BC548 transistors. When pressing the two transistors together the current in leg two drops a little. Since the voltage over the transistor in the mirror leg is higher more power is dissipated so also the temperature is rising. With a rising temperature I see a rising current.
So it is important for the transistors to have the same temperature.

 

[Roff]

I did a test with the two BC548 transistors. When pressing the two transistors together the current in leg two drops a little. Since the voltage over the transistor in the mirror leg is higher more power is dissipated so also the temperature is rising. With a rising temperature I see a rising current.
So it is important for the transistors to have the same temperature.

That's true. It is also important that the Vbe's be matched within 1 or 2 millivolts at the approximate current that will be passing through one leg of your mirror.
To match transistors, use the circuit on the reference leg of your mirror, and measure Vce. You will have to let the temperature stabilize before making the measurement.

 

[turbo_henky]

I found the next item on internet:
Voltage Controlled Current Source - Current Servo

I build the first leg with a FET (Same as Q1) with the Gate and Drain connected to each other and the resistor value of R3 from the Source to the GND. I took the voltage over this resistor as input voltage of the voltage controlled current source.

See attached file.

The real circuit is working great, but in the simulation is the two currents are not equal!

 

[turbo_henky]

I did some thinking and came to the conclusion that I don't need the FET in the first LEG.
I have attached the simulation file

 

[Roff]

R1 and R3 can be MUCH larger. Try R1=1Meg, R3=10k. The output current will be the same. This is because the op amp's input current is miniscule, in this case, 200nA max.

 

[ericgibbs]

R1 and R3 can be MUCH larger. Try R1=1Meg, R3=10k. The output current will be the same. This is because the op amp's input current is miniscule, in this case, 200nA max.

hi Roff,
I 'think' what the OP is trying to do, is to tap into an existing 2Amp current in a Peltier circuit in order to drive another Peltier at the same current by using a 'mirror'.?

 

[Roff]

hi Roff,
I 'think' what the OP is trying to do, is to tap into an existing 2Amp current in a Peltier circuit in order to drive another Peltier at the same current by using a 'mirror'.?

I missed that.

 

[RCinFLA]

The best devices for current sources are ones with large early voltage, Va, (constant current vs. Vce slope is small and parallel at various Ib->Ic drive levels) and good saturation characteristics (active region knee at low Vce voltage).

I doesn't take much connection resistance difference in one of the emitter legs to screw up the expected ratio. If you can stand putting in emitter resistors they will wash out variations in devices and temp. Two devices on same die with same wafer lattice orientation is best, two die in same package (most of time they are taken from near adjacent spots on wafer) is next, two parts epoxy'd together is least accurate at keeping temps the same between devices.

In simulation, check mirrored device collector current versus Vce voltage. A really good device can keep the mirror accuracy down to 0.2 to 0.4 volt Vce.