Current sources for load referred to ground

atferrari

Well-Known Member
Most Helpful Member
After reading about constant current sources, I defined the four cases shown in the pdf. Current should go up to 250 mA.

Implementing B and C with the attached circuit (and its counterpart) is straightforward, but I am not sure how to proceed for A and D. I cannot have Rsense referred to ground anymore.

How could I go away with no fancy ICs?

Gracias for any help.
 

Attachments

Wade's is very temperature sensitive.

How stable is the Supply voltage? If not very stable, then you should use a voltage reference (like a TL431) connected between the supply and the non-inverting input of the opamp. You can get by running the opamp on the same supply (no neg supply needed). I would use a straight PNP Darlington emitter follower or a PFET; not the compound NPN-PNP inverting connection which is likely to cause instability because it is inside the opamp's feedback loop.

Here is how I would do a ground-reference Load: Simulation is for load resistances varying from 1 to 15 Ohms at four different supply voltages: 9,10,11 and 12V

 
Last edited:
Hola Mike

No chances to run LTSpice where I am right now.

With the TL431 you ensure a stable value at the noninverting input but V+ could still change wildly, right?

I would like to change current from 0 to 250 mA at will.

Last night I started to consider using a current mirror. Is that too daring?
 
I made a +-50 mA current source, compliance around 10V using an LT1010 as a buffer. Other stuff had to drive a capacitative load and was unstable, so I used the info in the application note.
 
A current mirror built out of discrete NPNs, PNPs, and diodes will have a temperature drift problem. Using a TL431 in place of the Zener would greatly improve the Power Supply Rejection Ratio (PSRR), and also improve the temperature related drift. Here is a way of making it adjustable:

 
Hey Mike. Happy Thanksgiving!

Can you throw some capacitance on the output and see what happens? e.g. 1 uF

Seems like the topic of the month is alternators, doesn't it?
 
Here is the transient response. V1 introduces a 200mV step. The settling time and over-shoot of V(g) is the critical node to watch.

 
Last edited:
The thing I was doing is making a 4-terminal current to voltage converter. When I was done it had 4 ranges with +-10V representing +-100 mA. it was capable of +-10 v biasing. It had a Voc mode. It did fine on the AC performance and reasonably well. It had like a 40 pA offset without any nulling. I could not complete the nulling circuit and that wasn't the interest. No manual nulling was possible in the design. I added +-50 mA of suppression, but that meant that the bias was limited to +-5V. Usually we were interested in +-1.5V max anyway. I used 400 meg resistors so the sense leads could be disconnected.

It was a front end to a DSP lock-in amplifier to measure the spectral response of a solar cell. Monochromatic light was chopped at about 40 Hz and the solar cell was connected to this I-V converter. During bench tests it was fine, but when I attached a calibration 1 cm*2 cell, it went unstable.
 
Sounds interesting, but I don't see how that's particularly related to the title of this thread.
 
crutschow: The suppression section was indeed a current source referred to ground, but it was bipolar. I also had stability issues that I fixed.
 
Gracias Mosaic

Been there...
 
I simulated the bipolar current mirror in LTSpice and Proteus ISIS using 2n3906 as a current source. . LTspice sims it fine...it does indeed work. ISIS has issues...things like Vce-SAT in excess of spec going on, end result being currents aren't mirrored at all.

EDIT: Multisim Blue seemed to do a better job than both other sims w.r.t. Vce -sat . I t showed an 8% variation in current when using different colour LEDs and Vf as the loads.
 
Last edited:
Today I will start testing the topology I posted in the OP, for the four cases. It is to have them ready built at the bench to avoid assembling one every time I need to test something.

Gracias for replying.
 
Interestingly, I was trying to establish a constant current for some 7 LEDS (3.2Vf) from a Li Ion battery...3.6 to 4.1V with emitter degradation and a base V ref with NPN's works but requires as many resistors as transistors (7) plus a Vref. About 16 parts for constant current.
Resistors (22 ohm) alone cause current to vary from about 40mA down to 18mA over the batteries useful voltage range. 7 parts, current with 55% tolerance
Using the current mirror bipolar approach (8 transistors) a single 82 ohm derives the mirror currents at 40mA/4.1V which drops to 34mA at 3.6V from the LTspice sim. 9 parts delivering a current with 15% tolerance.

Seems the current mirror will do the trick in a tight PCB that uses discrete thru hole parts.
 
Cookies are required to use this site. You must accept them to continue using the site. Learn more…