Low power , low voltage Current Source

[erikl]

... it is worth raising doubt for that the circuit will achieve 20% current accuracy.

Hi Leo,
raising doubt with good cause always is legitimate ;-)

It is related to k5 that is process dependent. k5 effect is not described in the paper.

k5 simply is the W/L ratio of mosfet M5. Its effect is described in equ. (9), and considered in the following equations through the defined ratios p, t, & l (p. 3805 of the paper).

The essential idea behind this current stabilization process relies on the (admittedly) process-dependent n/p mobility ratio, which can be compensated by appropriate choice of the k- resp. W/L ratios. Of course the resulting design is not transferable to other processes, probably not even to other processes with the same structure size.

For additional info, you might want to communicate with the authors of the paper directly.

 

[raducrv]

so nu suggestion for the start-up circuit?

 

[FvM]

so nu suggestion for the start-up circuit?

Don't you have them in your text book? E.g. Razavi Fig. 11.37 ?

 

[raducrv]

This is a suggestion of start up circuit (M9 M10 M11) but we have to find something that works at even lower supply voltage . I don't have any book

---------- Post added at 21:31 ---------- Previous post was at 21:28 ----------

M7 and M8 for mirroring the current (Id6)

 

[FvM]

I was referring to this circuit



P.S.: Or a different one:

 

[leo_o2]

Thanks for discussion.
k5=u.Cox.(W/L).
So besides (W/L), u.Cox also will lead to variation.

Hi Leo,
raising doubt with good cause always is legitimate ;-)


k5 simply is the W/L ratio of mosfet M5. Its effect is described in equ. (9), and considered in the following equations through the defined ratios p, t, & l (p. 3805 of the paper).

The essential idea behind this current stabilization process relies on the (admittedly) process-dependent n/p mobility ratio, which can be compensated by appropriate choice of the k- resp. W/L ratios. Of course the resulting design is not transferable to other processes, probably not even to other processes with the same structure size.

For additional info, you might want to communicate with the authors of the paper directly.

---------- Post added at 02:16 ---------- Previous post was at 02:10 ----------

In your schematic, M10 might be PMOS. When VDD=0, the node "U" will be discharged thru M11's body diode.
So when VDD is ramping, M10 will be on and it will inject start-up current. M11 will charge the node "U" to high, then M10 will be turned off. M11 could be design with large L and small W. It will have longer start time.

This is a suggestion of start up circuit (M9 M10 M11) but we have to find something that works at even lower supply voltage . I don't have any book

---------- Post added at 21:31 ---------- Previous post was at 21:28 ----------

M7 and M8 for mirroring the current (Id6)

 

[raducrv]

ok , i have problems in Ltspice , transient simulation!!
Simulation time is raising very slow from 23,07us !!!
Is this related to the start-up circuit's transistor dimmension(L,W) ???
Here is the schematic and N_u1 Nad P_1u lib file

 

[raducrv]

The problem is that in tran mode the circuit will instant go up to 23us and then will increase very slow the simulation and it would take probably a few hours until 100us

 

[leo_o2]

Did you observe any oscillation in waveform? Usually oscillation leads to very slow simulation. Pls check if your start-up circuit is unstable. Pls try to reduce gain of start-up circuit.

 

[raducrv]

here is the waveform fot V(u) VDD and Iref(Id(M6))


---------- Post added at 12:01 ---------- Previous post was at 12:00 ----------

and as you can see the sim stops

 

[ic98]

lots of docs on ieeexplore

 

[FvM]

I think it's a LTSpice bug respectively an unsuitable default setting of transient simulation parameters, sometimes even occuring with passive circuits. But it mostly has to do with extreme time constant ratios. I expect that it can be avoided by adding a compensation capacitor to the feedback path.

 

[raducrv]

how do i pick the value?

 

[leo_o2]

V(u) is a negative voltage? It is strange. Is it possible your MOSFET model is incorrect?

 

[raducrv]

yes leo , that is very strange .i don't know what else to do , i give up

 

[erikl]

Re: Low power, low voltage Current Source

...i don't know what else to do , i give up

Not yet: At least your M8 & M9 should be nMOSFETs (N_1u) :idea: :!:

You're not far from what you want: 2.6µA (with my process models), and the start up works fine:
Just play a little bit with your W/L ratios; I'd think you also could reduce both W and L by a factor of 10.

Good luck! erikl

 

[leo_o2]

Pls try other simulator (like hspice) with correct device model. Negative voltage is unreasonable in your simulation.

yes leo , that is very strange .i don't know what else to do , i give up

 

[raducrv]

erikl it works fine now, that was the problem , whith M8 and M9 . Thanks a lot you all.:lol:
Another question : I want to do a monte carlo simulation , does ltspice have **** one? I have declare the W and L as .param to use them as parameters in monte carlo simulation .what do i have to do

---------- Post added at 10:50 ---------- Previous post was at 09:52 ----------

a curious thing is that even with nMosFet for M8 and M9 it still doesen't work fine until i set the temperature -40 27 89
I don't know why , but for 90 degrees it won't work

 

[erikl]

Try something like this:

 

[raducrv]

i tried but the same problem with the simulation for 90 degrees temp , anyway , i don't think that is a major problem , because as i wrote it works fine for 89 degrees. Cand you help me with the other post about the monte carlo simulation