The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> LC VCO
https://designers-guide.org/forum/YaBB.pl?num=1264229357

Message started by raja.cedt on Jan 22^nd, 2010, 10:49pm

Title: LC VCO
Post by raja.cedt on Jan 22^nd, 2010, 10:49pm

hi,
i am designing a LC vco with 1.2v supply with nmos switching devices and pmos current source, now i juwant to know is there advantage of putting cascode rather than single one? For me there is no problem with swing.

Thanks,
Rajasekhar.

Title: Re: LC VCO
Post by pancho_hideboo on Jan 22^nd, 2010, 10:54pm

You can achieve large output impedance regarding NMOS negative impedance generator.
Also it might be a little useful for isolation from Vdd.

But I don't think cascode is popular.

Title: Re: LC VCO
Post by raja.cedt on Jan 23^rd, 2010, 12:37am

hi pancho_hideboo,
thanks for your reply,some one said casode will reduce phase noise but i didn't understand why it is..because cascode branch don't effect noise. Could you please direct me any reference regading this.

Thanks,
rajasekhar.

Title: Re: LC VCO
Post by loose-electron on Jan 23^rd, 2010, 2:25pm

when you introduce mismatches in the differential stage, the common current source will start to inject some noise.

Common mode rejection is not perfect.

Title: Re: LC VCO
Post by Mayank on Jan 25^th, 2010, 12:48am

Hi Raja,

Jerry once said that Cascode will introduce much more noise than a single transistor current source, his simplest explanation being more the resistance, more it's thermal noise. So, by that logic, cascode should increase the noise pumped into the source node & hence increase phase noise, right Jerry ?

Quote:

For me there is no problem with swing.

If you want to improve you phase noise & you have room to increase your swing, do increase your swing. It should bring down phase noise...unless your vdd is highly noisy and you want better supply insensitivity than usual.

Quote:

some one said casode will reduce phase noise

The only way i see cascode reducing phase noise is by reducing supply sensitivity.

--
Mayank.

Title: Re: LC VCO
Post by macrohan on Jan 25^th, 2010, 11:50pm

I think that cascade can save power compare to the single, if you want obtain the same output swing.

Title: Re: LC VCO
Post by Mayank on Jan 27^th, 2010, 3:42am

Hi,

Quote:

I think that cascade can save power compare to the single, if you want obtain the same output swing.

& How do you think that ? Pls. explain.

--
mayank.

Title: Re: LC VCO
Post by loose-electron on Jan 27^th, 2010, 12:14pm

Mayank -

Current sources create noise - largely because of high impedance. (go model it and run a noise simulation)

A common mode current source injects common mode noise, which is not perfectly removed by differential signal processing.

Mayank - your analysis seems correct to me.

A cascode source is noisier that a simple 1 transistor current source, due to the higher impedance.

Less power with one vs the other? How? the current is the same (or can be controlled to be the same)

jerry

Title: Re: LC VCO
Post by DoYouLinux on Feb 7^th, 2010, 7:55pm

Hi,

It depends on your oscillation frequency. With high enough ones (> 50 GHz), care should be taken with the cascode device size, i.e., its total source capacitance can reduce the effective negative resistance of the whole cross-coupled transistor pair.

Hope this helps as a guideline,

DYL

Title: Re: LC VCO
Post by raja.cedt on Feb 7^th, 2010, 9:10pm

hi DYL,
how cap at the common source node will reduce -ve resistance provided by the diff pair? and generally what i feel is with cascode we can reduce cap the the common source node by simply reducing the cascode size.

thanks,
rajasekhar.

Title: Re: LC VCO
Post by DoYouLinux on Feb 8^th, 2010, 4:36pm

Hi raja,

What I meant is the capacitance at the source of the cascode transistor. Imagine that as the bottom transistor is trying to push a current to the tank node, this capacitance will spoil the drain current from the bottom transistor, reducing the efficiency in producing negative conductance.

Hope this clarifies :)

DYL

Title: Re: LC VCO
Post by rfcooltools.com on Feb 9^th, 2010, 1:29pm

Raja, you are probably better off not using a current source at all. Instead use a resistor tied to a voltage reference and adjust the reference based on current draw from a replica circuit. This will improve phase noise by reducing the conversion of of even harmonics of noise. In addition the resistor will reduce the common capacitance conversion to differential when the tank is imbalanced (which is always the case to some degree) which should improve your tuning range.
J

Title: Re: LC VCO
Post by love_analog on Apr 23^rd, 2010, 10:27am

If you don't use a current source, the Q of the tank is degraded. Typically the advantage of lower upconverted noise is not sufficient to overcome the Q degradation so most products out there do have the current source

you may put a resistor or even better an inductor incread of the current source. In the latter case your current will vary all over the map. I don't know how well a resistor performs as compared to using a current source. Would welcome ideas on that. Most papers I see use a current source but you can use a resistor servoed by an amplitude control loop to set the amplitude.

Title: Re: LC VCO
Post by Mayank on Jun 9^th, 2010, 10:26pm

Quote:

This will improve phase noise by reducing the conversion of of even harmonics of noise.

I hope this doesnt refer to the 2f upconversion of noise coz i dont see how replacing tail current source with resistor will remove 2f upconversion.

If you are talkin about MOS transistor's inherent non-linearity in ID equation, then it's contribution to even harmonics of Fosc in Phase Noise Plot is minimal as compared to 2f upconversion of flicker.

Plz if you can clarify.

--
Mayank

Title: Re: LC VCO
Post by rfcooltools.com on Jun 10^th, 2010, 11:06pm

Mayank,
Actually the problem is subtle
A low impedance at the common source acts to turn the rectified voltage at this node to a current which will ultimately flow easier into the tank.
What makes up the voltage at this node? it is the accumulation of bias current and the noise associated with it. This just a different way of looking at a well derived phenomenon (well at least my viewpoint).

Current mirrors have Cgd and Cdb which have some value call it Cdrain. If a resistor can be used to bias the VCO where the value of this resistor is higher than the impedance of the Cdrain at the second harmonic frequency then there should be a net improvement due to this effect. And the effect is compounded when you consider higher harmonics where the capacitors impedance keeps reducing and a resistor remains constant. In addition to this the Cgd of the current source will act to lower the impedance and act to upconvert 1/f to 2f especially if the capacitor of gate bypassing cap is not sufficiently large.

http://rfcooltools.com

Title: Re: LC VCO
Post by Mayank on Jun 11^th, 2010, 12:23am

Quote:

A low impedance at the common source acts to turn the rectified voltage at this node to a current which will ultimately flow easier into the tank.
What makes up the voltage at this node? it is the accumulation of bias current and the noise associated with it. This just a different way of looking at a well derived phenomenon (well at least my viewpoint).

If we look at it in a more canonical way, considering noise injection is in terms of current which is actually the case. -- i_n² noise current from the Tail current source (both flicker & thermal , generated due to it's transcoductance gm)

Reason to consider noise injetion in terms of Current not Voltage --
If the node under consideration has an infinite impedance(as is the case at gate of a MOS), we would have considered it as noise voltage.
If we are investigating noise at a node with a finite impedance, then we should always consider noise sources as current & then drop it across appropriate node impedance & convert it to noise voltage.

That being said,
If we now consider noise injection in terms of current,
Higher the cap from common source node to GND,
OR Higher the Inductance from drain of tail current source to Common Source of diff pair,
Lesser is the propagation of Noise current into the TANK,
& so is the conversion of Noise current to Noise voltage,
& so lesser is the upconversion.

Meaning,
more the parallel Cdrain (or series L), lesser the effect of noise from tail current source.

as i have understood & is quoted in many texts.

--
Mayank