Design a current bias circuit

[gogomi]

Hi,

I would like to desing a current bias circuit for LC-VCO and charge pump.
Could anyone give me some suggestions on reference or circuit topology.

Actually, I've tried to build a supply-independent bias circuit but it's hard
to get the current I need and it fails to be supply-independent.

Thank you for your help

gogomickey

 

[steer]

Start with voltage reference, then convert voltage into a current.

 

[Colbhaidh]

See very simple circuit as a basis for I-Bias

 

[steer]

See very simple circuit as a basis for I-Bias

Be aware that this simple bandgap reference is relatively noisy. Might even become the dominant noise source depending on your LC oscillator Q and its bias sensitivity.

 

[gogomi]

Thank you guys! I don't know if I should built a bias reference circuit for
every block, which needs one, or just built one bias circuit for all the blocks.

Would you also please let me know what kind of bias circuit is considering
a better or low-noise topology. Thank you again!

gogomickey

 

[steer]

Would you also please let me know what kind of bias circuit is considering
a better or low-noise topology.

First look how sensitive your system to the oscillator noise. For example, your PLL loop might take care about phase noise below 10MHz. Then you can use a simple BG bias like above as its noise is usually a LF one.
Second, does your spec really requires a cutting edge performance?
Third, how sensitive your LC oscillator to the bias noise?

Answering on these three questions help you to deside.
In my experience for many LC-based applications a simple resistor to VDD often does an excellent job.
If not, Vth over R bias is simple and easy to make low-noise.
Bandgap requires some experience and complications to get its noise low, but is it really necessary in your case?

 

[gogomi]

First look how sensitive your system to the oscillator noise. For example, your PLL loop might take care about phase noise below 10MHz. Then you can use a simple BG bias like above as its noise is usually a LF one.
Second, does your spec really requires a cutting edge performance?
Third, how sensitive your LC oscillator to the bias noise?

Answering on these three questions help you to deside.
In my experience for many LC-based applications a simple resistor to VDD often does an excellent job.
If not, Vth over R bias is simple and easy to make low-noise.
Bandgap requires some experience and complications to get its noise low, but is it really necessary in your case?

I think one part of my question is that I am not sure what kind of bias circuit is
good enough for a regular LC-VCO or PLL. I don't really want to touch the BJTs
of BG ref since I don't have much time for the whole design. I think I will stick
to Vth over R bias circuit. Actually, I've done some simulations on it but didn't
get the results as expected. Anyway, I'll study it over again.

Thank you for your help!

 

[cetc1525]

I think Colbhaidh's circuit is good.

 

[rfsystem]

If your intend is to stabilize a LC-VCO it is best to stabilize the internal oscillation amplitude. The amplitude has the biggest impact on PN. So compare the amplitude with a bandgap and drive the bias current to fulfill this target.

 

[raymond33]

Sorry I don't think so Golbaidh's circuit is wrong.
I give you an example PNP1 emitter is connected to the Base that is wrong , the right answer is Base connected to the ground.Second point Bandgap's PMOS,the upper one has got PMOS gate connection is wrong.I hope this will help

 

[steer]

Raymond, you've a sharp eye. This is a non-working circuit indeed.

 

[leo_o2]

It is better to keep the voltage on both sides balanced.

 

[Colbhaidh]

Sorry, absolutely right. The base should be grounded on the PNP's. Also I used Circuitmaker to quickly draw the schematic so all the PMOS substrates should be connected to Vcc and not as shown. Need to ditch Circuitmaker some time !

 

[markty]

which part of Bandgap reference is critical for low noise application?
I built a BG reference once, the simulation results is good but the real performance is not. The output voltage varied every time, like 1mv. The temp coefficient is not a constant.