High precision, good accuracy Current Source Schematic

yungman

There are plenty of op-amp that work down to 1.8V single supply!!! That's not the problem. The uncertainty for me is the gate source turn on voltage is about 3V, that eat up your supply voltage. Look into it in more detail and see whether you can make it work for 5V. Also a hybrid darlington using a p-JFET and a PNP might help as the gate source voltage is in opposite direction and it might even help. You have to do the calculation, I am only giving a suggestion. Try look for low threshold P-MOSFET.

Maybe it's not a problem. If you can find a P-MOS with operating gate source voltage of -3V, you still have 2V on Rs to play with. Say if the op-amp output is at 0V for max current, the source will be at +3V, so operating voltage for constant current output at the drain can be from 0 to about +3V. Maybe it's not a problem.

jim hardy

i think i used a n mosfet, could just swap opamp's +&- inputs for p installed other way..

here's an interesting transistor..
http://www.onsemi.com/pub_link/Collateral/ENN7508-D.PDF
see curve Ic versus Vce, pages aren't numbered it's about middle of second page.

will those 3 volt opamps drive as much as a milliamp?

never mind, just looked at one.
http://www.ti.com/lit/ds/symlink/lmv711-n.pdf

old jim

maxim

This forum is fantastic, and you're really lovable persons!

So, being inspired by your good advices, I put together the latest project.

It uses a low power Differential Amplifier (AD8276B) supplied by a Voltage Reference (ADR3425), a low bias current Op Amp (AD8603) for a feedback loop, then an external current source BJT (T1, to be defined).



Please, have a quick look to that and let me know if makes sense.

Later on I will write here my considerations and calculations for the component sizing, based on datasheet and app notes from Analog Devices.

Of course, any idea, suggestions or criticisms are welcome!

maxim

maxim

About the schematic above, here I have put few comments, hopefully updated by volunteers:

The AD3425 serves as voltage reference to provide a stabilized 2.5V (V_REF)

The AD8276B is used here as difference amplifier, where V_REF is applied to its non inverting input (+) to control the amount of output current, I_O. The inverting input (-) is grounded. The output of the AD8267B is used to drive the transistor (T1 to be defined) since a current of a few hundreds of mA is required.

The AD8603 op amp is used in the feedback loop of the circuit; it has low bias current (< 1 pA), low offset voltage (< 50 μV), and low temperature drift (< 4.5 μV/°C ) .

So, since the AD8276B has R_F1/R_G1 = R_F2/R_G2 = 1, the output current Io depends only from the ratio V_OUT/R₁ = V_REF/R₁. The accuracy of R₁ is critical, so it should have 0.1% tolerance or less.
Considering V_REF = 2.5V and I_O = 100mA, R₁ value is 25 Ω.
R_HALL is ~ 2 Ω

Minimal conditions have to be met, fixing Vs = 5V:

1. V_HALL = I_O × R_HALL must be within the AD8603 op amp input range (−0.3 V , 5.2 V)

2. V_OUT = I_O × (R_HALL + R₁) must be within the AD8276 SENSE pin voltage range (−10.2 V, 7 V)

3. I_O × (R_HALL + R₁) + 2(−Vs) − 0.2 V to 2(+Vs) − 3 V must be within the AD8276 output voltage range (−7.5 V, 7 V)

First 2 conditions are OK. I am puzzled about the last one.

jim hardy

lets see here

my simple thinking goes thus:

AD8276 will hold his inputs equal

which dictates that since his inverting input sees Vout/2



and his noninverting sees (2.5+Vhall)/2

equating those yields

Vhall = Vout - 2.5
so your circuit will hold 2.5 volts across R1 giving the constant current you seek.


Now a sanity check to make sure feedback is proper direction:
Assume Rhall drops a teeny bit
Vhall accordingly decreases a teeny bit, so (Vout - Vhall) got larger
however the decreased voltage at noninverting input lowers pin6 voltage, driving Vout in proper direction to restore balance at inputs.

Looks Great so far, said the blind man !

At 100 ma i see Vout = 0.1 * (2 + 25) = 2.7 volts, {corrected - jh}
and if hall sensor varied between zero and 0.5 5 ohms Vout need only range between 2.50 and 3.0 volts

That amplifier doesn't have a lot of output current capability but datasheet says it can drive to within 0.1 volt of supply with 10 k load so you might get away with a TIP120 darlington driver , 3 Volts + 1.5 for Vbe leaves 0.4 volt headroom.
Yungman is way better than me at selecting a transistor.

Using a PNP darlington emitter side up and swapping Mr 8276's pins 2&3 would more nearly center his pin 6 voltage at mid-supply. Perhaps with just a tad of emitter resistance for loop stability....
You guys will find a better driver i'm sure, there are even IC's made specifically for boosting opamp drive capability.

Your third concern i think is from their warnings to assure headroom,
which looks okay from my simplistic approach above.
Inputs and Vout are near mid-supply and pin 6 has ~ 1/2 volt headroom even with a darlington's two eb junctions. PNP darligton driver (TIP125?) would give about 1.3 volt ..



I have to talk my way through a circuit before attacking its equations.

A most elegant looking solution, my friends.


old jim

jim hardy

hope i didnt say something socially wrong - we "Aspies" can do that unwittingly.

fixed my voltage blunder above..

maxim

Back after a lot of calculations... maybe I am wrong but I've the impression that swapping pins 2 & 3 nothing changes other than the polarity of Vref:

Vout = [V(+IN) – V(-IN)] x gain

So if you tie Vref to pin 2 instead of pin 3, Vout will be a negative voltage (assuming negative supply).


Going back to the original configuration, now I need help to refine calculations.

Let say I want to drive 300mA but keep the same conditions, I would have to increase Vref or R1.
If I change Vref to Vref = Io * R1 = 0.3*25 = 7.5V, this is outside the range of AD8276 on a 5V supply.

The other option is to change R1. So for example, I could change R1 to 10 Ohms, that means Vref = 1V gives 100mA and 3V gives 300mA, and I think it meets all the input range limitations...

Then a TIP120 has a Vbe(on) of 2.5V... Why not use a regular NPN BJT like the 2N1711?
It has a Vbe(sat) of 1.3V (max) with IC = 150 mA & IB = 15 mA

What do you think about?

jim hardy

wow, is my face red. i took an old shortcut and it bit me - adding an inverting stage behind an opamp often lets you just swap input pins, and i assumed, without thinking it through, that just swapping 2&3 would do the job..
:Rats. i'm old enough to know better.
I extend my apology.
Eternal vigilance ...

Soo,, much chagrined - This time i did my algebra.
What i missed before is that to swap that internal opamp's input pins , i'd have to swap pins 1&5 in addition to swapping 2&3.
That'd leaves us as follows:
Pin 2 = +2.5 volts, Vref
Pin 3 = 0v
Pin 1 = Vhall + I_o*R1
Pin 5 = Vhall
NOW internal opamp sees at his Inverting input: (Vpin5 + Vpin2)/2
or, Inverting = (Vhall + Vref) /2
and at his Noninverting input: (Vpin1 + Vpin3)/2
or, Non-inverting =(Vhall + I_o*R1 +0)/2
Equating those yields
(Vhall + Vref) /2 = (Vhall + I_o*R1 +0)/2 ;

(Vhall + Vref ) = ( Vhall + I_o*R1)
or,
Vref = I_o*R1 ;;
I_o = Vref/R1 which is what we want.



Now sanity check for direction of feedback:
with a PNP driver and both pairs of those pins swapped

IF I_o increases by Δi,
Non-Inverting input goes up by (Rhall + R1) * Δi
but Inverting input goes up by just Rhall * Δi
inputs now see difference R1*Δi , with Non-inverting more positive

Since Non-inverting input is now greater than Noninverting, AD8276's pin 6 moves positive
which moves PNP's base positive, nearer emitter shutting it off, restoring balance.
So a PNP could work if you need one.
I remember waking up in middle of night thinking i should check on this, but for some reason didn't get it done.

You are right. I didnt finish my thinkling and caused you extra work.

_____________________________________________________________

Now to your question -

Indeed everything is supplied by +5, your ADR3426 would be hard pressed to deliver 7.5.

Your circuit as you drew it will keep I = Vref/R1
so as you said
Agreed.
The internal opamp's inputs need to be kept 1.5V below supply, ie 3.5 volts.
So Pin 5 must stay below 7 Volts, okay there.
And (pin 1 + pin3)/2 must stay below 7 volts, okay again because Rhall is only 2 ohms and i think you're planning on less than an amp though it.
AD8276 datasheet page 14
And you can't go negative with just a +5 supply.

The 15ma Ib worries me. That's a heavy load for your AD9276.
Note his datasheet says , page 4 "output characteristics", short circiut current limit is 10ma. Graphs of output voltage page 9 figs 19-21 have 1kΩ as lowest load. I dont think he's intended to deliver much current.

But with minimum hfe of 100 at 150 ma it should work...

But.... Just in case;;
The TIP125 datasheet

http://www.fairchildsemi.com/ds/TI/TIP125.pdf
fig 2 page 3
shows Vce of ~0.75 volt
and Vbe of ~1.25 volt
when Ic = 250*Ib.
That would place Mr AD8276 in the enviable position of sinking very modest current(~ a milliamp) with output voltage well above mid-supply, about 4.25 volts.
With the 2N1711 he could have to source 15ma at (Vhall + VRef +Vbe) = 0.2 + 2.5 + 1.3 = 4.0 volts.
That leaves only one volt room to +supply. That's right at his limit, see fig 21 of AD8276 datasheet.

We should get a cross-check from Yungman on this point.

That's my thoughts.
Let me know how it looks to you.

Thanks for catching my error in previous post. If this one looks cross-threaded anywhere please challenge me on it. You know i'd much rather admit my mistakes than have them cause you aggravation.

Thank you , kind sir !

maxim

Hi Jim,

No worries about possibly error, it's the same for me.
Is always a pleasure to talk to nice people like you!

So, I would actually use a TIP125 transistor as final stage, it convinces me.
Please, help me to fix the polarization network at its best and to choose the proper power supply / Vref for the whole circuit.

Your contribution is greatly appreciated.

jim hardy

Thanks for your kind words.

Well i always seem to have "just one last correction" ..

"""That would place Mr AD8276 in the enviable position of sinking very modest current(~ a milliamp) with output voltage well above mid-supply, about 4.25 volts."""

SHOULD have read
"" That would place Mr AD8276 in the enviable position of sinking very modest current(~ a milliamp) with output voltage well slightly above mid-supply, about 4.25 3.75 volts.""

--------------------------
Are you saying you want that range of adjustment ?

Soounds easy enough. You already figured R1.
A potentiometer across output of the reference and a voltage follower amp to buffer it would be straightforward approach. AD makes a dual version of AD8603 , the AD8607 so it'd be same package count.

If you like i'll try to sketch something tonite.
How much room do you have for power supply?
What physical arrangement do you envision ? Will you make a circuit board or is this a one-of-a-kind hand wired?

What range of current adjustment would you like to have?


old jim

maxim

They're deserved.



--------------------------

You know, the sensor need 100mA of polarization current. Up to 300mA is for working condition, so...

So far, so good!

I do not want to ask you to waste your time with me, although I'd appreciate it a lot!

Let say I have the option to choose the proper supply. Started from 5V because it was obvious (other divices provide 5V lines), we can change this value freely. Only the Vref voltage I wish were provided by Mr. ADR34xx for stability.

Only the sensor will be stored inside a probe for high magnetic fields. I think I would keep its wires c.a. 20-40cm long.

Yes, this is the next step. I have to make a PCB as small as possible :)

I guess 100-300mA is enough.

You're my friend now !

All the best, jim!

jim hardy

Sorry for delay, just dont get to everything anymore.
Beware of that "smartfortress virus" it's out again, turns off your antivirus and tells you you are infected , and blocks most things you can do to get rid of it. Cant even start up in safe mode if it gets far.
Microsoft antivirus recognized it yesterday, a couple hours late for me.
If you get it google smartfortress and it'll take you to a microsoft link that kills it.
I had to run their killer program and then 'restore' to a few days ago because of damage it did.

Anyhow ---
.
Okay - your scheme is so good it doesn't need much change i think.

All I did was three things:

1. make Vref variable to give 0-300ma adjustment range with 10 ohm R1.
To that end,
Added Radj a ten turn pot with locking dial. That lets you set current by a ten turn dial with good resolution.
http://search.digikey.com/us/en/prod...02L-ND/1088592
http://search.digikey.com/us/en/prod...US2011_Catlink

Added a buffer voltage follower amp A2 to Radj. 500 and 2500 ohm divider on output make his output 0-3 volts for 0-2.5 in. If you use instead ADR3430, omit those resistors.
A2 can be another AD8603, or half an AD8607 with other half replacing your 8603
http://www.analog.com/en/all-operati...s/product.html

The buffer keeps the resistance of your potentiometer from appearimg in series with Mr8276's internal 40K resistors.

2. Added provision for "Criss-Cross" jumpers at inputs of Mr AD8276.
That lets you use either NPN or PNP driver.
Install horizontal jumpers for NPN like your 2N1711 or one of Yungman's Mosfets
Install diagonal jumpers for PNP like my TIP125.

3. Showed three dots instead of drive transistor.
Make two sets of holes on you board, one set for TIP and another for your 2N1711.
That way you can use either NPN or PNP or MOSFET driver, emitter up for PNP and emitter down for NPN.


http://s232.photobucket.com/albums/e...rMaxim1001.jpg




NOW while typing i realized that you may not need infinite adjustment.
If not, consider this:
Omit buffer and Radj, go back to 2.5 volt Vref.
Make your R1 measuring resistor be three 25 ohm resistors and a switch that lets you select one, or two or three of them in parallel. That'll give you 100, 200 or 300 ma.

And in keeping with Detecive Columbo's "One Last Thing";;

Yungman's sketch in post 10 had a capacitor to stabilize things. Always a good idea.
That additional stage behind Mr8276 increases his open loop gain , but only if we use PNP driver.
Sooo we should add a zero ohm resistor (jumper) between top of driver transistor
and Vsupply , to be replaced with just a couple ohms to reduce loop gain in case PNP exhibits instability.
Can add that to sketch.

What's your thoughts so far?

jim hardy

i see size is a matter. That ten turn pot is big.

BTW while you're experimenting, try that LM317 circuit from post 14

1.25V/.1amp = 12.5 ohms

maxim

So happy to be an "old" Linux user who uses windows only for playing chess...

That's ok for me.

Simply so smart!

Yes, I see.


http://s232.photobucket.com/albums/e...rMaxim1001.jpg




This circuit is a wonderful solution!
I will keep you informed about progress.

Right. I will try also the LM317' solution: easy and compact.