BJT vs FET current Mirroring

Hello,
I'm kind of new to all this IC design and have a couple questions about a circuit im trying to design. Im trying to create a current mirror thats powered off a 2.5V rail. I'm not sure whether i should use a BJT or a MOSFET configuration and was wondering whats the benefits to each. My specific needs are:

1) low power consumption
2) package size
3) large current range (varied by changing the bias resistor).

Thanks!

Some additional background:

I originally had a circuit where i was sampling the output voltage whichwas proportional to my variable resistor i was using. However, i realized that i had better characteristics if I made the output voltage proportional to the inverse of variable resistor. So i designed three circuits which could do this, a bjt current mirror, a mos current mirror, and an opamp with negative feedback. Im trying to make a comparison between the three circuits to see which gives me the largest range in voltage output, the smallest power dissappation and the smallest package size.

I would have thought the choice would have been determined by the process used for other functions, unless that is the only function of the chip.

To compare bipolar vs MOSFET current mirrors you would need to define what the specification is e.g. current range, voltage range, mirror accuracy under certain conditions, speed etc. Then for each technology there are various designs which have certain benefits e.g. cascode mirrors for improved current source resistance or bipolar ones which eliminate base current error.

Keith.

hi keith,

thanks for your response. To clarify, the only function is the function of the chip. I'm actually not interested in the accuracy of the mirror, i'm more interested in the range of current i can create.

For example, if i were able to change the biasing resistor on my mirror, i would ideally like to exploit the full range of Vgs (ie: 0-Vdd), which in turn would allow me to get my maximium range of currents. However, MOSFETs are limited by their threshold voltage making it impossible to keep the MOSFET is saturation and having Vgs less then Vt.

not sure how much that cleared things up haha. I'm not very good at explaining things.

As a side note, i also wouldnt mind learning the benefits in general of FETs vs BJTs as current mirrors in general.

You can operate MOS current mirrors in subthreshold;
in fact they're liable to be better, over a wider range of output
voltage, if you do - running a smaller device "hotter" will
raise the "knee" voltage where your mirror becomes
accurate.

BJTs have a wasted base current term, on the order of 2% at
hFE=100 (1/100 the current into each of the bases). You can
buffer the base current so long as you have 2*Vbe headroom
on the reference side, or drive the rack's base with some amp
and use one mirror output as feedback. This could be effective
down to maybe 1V. The nice thing about BJTs is that your
headroom is pretty fixed across a wide current range.
The bad thing is, many integrated BJTs (once you get away
from old-school, engineered analog bipolar processes)
have a fairly narrow peak-hFE plateau. You'd like decades
of roughly-peak hFe. You may not see this on a device
optimized for frequency at lower voltage, because the higher
dopings kill low current density efficiency.

Some processes have lower-VT MOSFETs which will let you
cascode without much headroom penalty. More likely in
RF and analog CMOS flows.