Need an independent supply current reference circuit

hello

i need a supply independent current reference circuit.

Vdd range 1.3-3.3V Iref=2uA

I have found this ciuruit in Maloberti's book. The thing is it is not accurate in the corners . What can you advice me? Of course the startup circuit will be added.

What resistor will be most stable over temperature? I'm using AMS 0.35um

Maybe you can suggest another circuit?


regards

2ua current source

for a stable over this much temp and process variation...u may have to come up with some other config/equations...some times blend of <RES>+<MOS>+<BJT/HBT> may give better result....u may try it....but 2uA is too low for BJT....may be base current will be near that value....u may have to scale down later....

if i can find any docs....will let u know...

gd luck...
sankudey

Re: current reference

sankudey said:

for a stable over this much temp and process variation...u may have to come up with some other config/equations...some times blend of <RES>+<MOS>+<BJT/HBT> may give better result....u may try it....but 2uA is too low for BJT....may be base current will be near that value....u may have to scale down later....

if i can find any docs....will let u know...

gd luck...
sankudey

Click to expand...

i am restricted by low current unfortunately.
BTW: what solution is better

1uA current reference and than multiplying that to higher values by current mirrors (max. 20times)

or 10uA and multiplying two times and dividing by two for another branch?

low current is the main specification

thanks and regards

current reference

You can use a low voltage bandgap to generate the stable reference current which is independent with vdd, temp and mos, but this current will be effected by resistor's corner.

Re: current reference

I think going for 40uA/80uA reference and then dividing by two would be better....

BTW: this depends on the topology....as mentioned earlier that ...if some topology incorporates BJT to have a better and stable reference...then u have to do the above....but thing depends on the topology......multiplication/divition u may think of later.....first....in my opinion....u have to make a stable current reference with corner variations within the limited tolerance (in %)...than scaling up/down may not be that much troublesome....


gd luck...
sankudey

Re: current reference

Hi Jutek,

What specifically falls apart at the corners, and which corner?

Here is a delta-Vgs circuit I am drawing for the Openschemes op-amp that has worked pretty well. If you need a Vgs/R you can build one in a similar manner, but it looked like your original circuit was also a delta-vgs.

One warning - when using on-chip resistors, you will always have variation proportional to this resistor. Since this circuit generates dVgs/R, it has a linear relation to R. In my experience, this is not a big variation in comparison to temp and input voltage, and can be further cancelled by bringing this same resistor dependence into your circuit later. (dVgs/R)*R=>dVgs regardless of process.


On to the circuit...

MP39 is a generic peaking mos current source, to bias up the delta-vgs core with a 2uA always-on current. Shut it off by opening R9's path to ground with an NMOS if you need 0-current shutdown. By using an always-on current, this circuit is guaranteed to start at about 1.5v.

MN20/MN34 are a 4:1 delta-Vgs. Upon startup, these are imbalanced since both branches of the dVgs core are driven with the same 2uA current from the mos peaking source. When they are imbalanced, drain of MN34 rises up until the delta-Vgs voltage is supplied by MN32. Eventually, MN32 supplies enough current to R10 that the circuit rises into balance.

The current through MN32 is therefore delta-Vgs/50k.

MN33/MN34 are split so I can high-swing the PMOS mirrors by creating a cascode voltage with MP44. I think I set the drains of MP41/MP28 at about 200mV. Vdsat for the rest of the mosfets are 150mV except MN32, who can do anything he likes since he's a gain stage.


These sizes are lambda sizes, not microns! If I recall correctly, lambda here is 0.25um. Let me know if it's any use!


For use only in non-commercial and educational applications, copyright 2006 Openschemes.com, your free schematic warehouse.. Or something like that.

Re: current reference

Hi,
I was going thru some docs....these are uploaded here.....u may go thru these....










u may combine goods of each ...or good compilation of BJT/MOS/RES.....hope would help u....


gd lk....

sankudey

Re: current reference

you can use a bandgap...or in your circuit a improve will come from using 2 resistors in series - a ntype and a ptype...by doing this you'll get some stability over temperature...not much though...

Re: current reference

mircea said:

you can use a bandgap...or in your circuit a improve will come from using 2 resistors in series - a ntype and a ptype...by doing this you'll get some stability over temperature...not much though...

Click to expand...

I agree with mireca....its a good idea and gives good result.......
some times in technoilogy files....resistors with both +ve trmp. coeff. and -ve temp. coeff. are available.....may be salicided/non-salicided.....u may check it ..... if find it...now can go for smooth cancellation by chhosing the ratio of the resistor value to their temp coeff......

gd luck...
sankudey

current reference

you need to know what the current is for. a temperature-independent current is not often useful. to bias an amp, you want a delta-Vgs current. To bias an oscillator, you want a current proportional to your threshold voltage. I suppose you could use a temp-independent current here, but why not use a current proportional to your threshold over temp, then it all washes out and frequency is stable over temp.

mixing two kinds of resistors means you will have independent variation in each so you either have to trim both portions, or just trim one and suffer the mismatch in tempco. it's rare when this trick works in production.

sankudey, mircea: can you explain why you want a perfectly temp-independent current? what is it for?

current reference

its PSRR is poor

Re: current reference

Hi, electronrancher,
What's the merit of this circuit compared to the traditioal constant gm current reference?Thanks!
regards!

Added after 3 hours 44 minutes:

hi,all
Anyone has a current reference that does not use resistor for it has a quite large variation at all process corners.