Very low input voltage voltage controlled oscillator

[VeeJayy]

I have designed a ring vco with the input voltage ranging from 0.7 to 1.8V, but i need it to work below threshold voltage too. The increase in frequency with respect to the voltage has to be linear.

Regards,
Vidj

 

[dick_freebird]

Ring oscillators aren't going to be naturally frequency-
linear in any case, and less so as you slide across the
low end where current is exponential w/ VIN (worse
than square-law at the big end). You probably need
to make the bias scheme more elaborate, with current
being the function of control voltage which makes the
linear-frequency result (may not be linear current, as
the travel-range of the stages also moves with VIN).

I'd start with the supply and control voltage sims and
determine what the biasing "right answer" is, first. Then
worry about making that bias generator.

 

[VeeJayy]

1. Biasing "right answer" as in? How will i know what the right value is?

2. And are you saying that there has to be a different biasing circuit altogether? If so, using Cmos isn't it difficult to make the transistor "work" below 0.7 V.

3. I am trying to find a solution to this from the past week or two but have come up with no relevant papers for this. What would this biasing circuit be called as ? If anything pops into your mind which can be a possible solution to this, please do state the same.

 

[dick_freebird]

First you need to find the current to frequency transfer
function. This is going to vary with supply (presuming the
stage number is high enough that all of the inverters are
operating at full logic swing - but this also may change
when stage drive gets very weak). You say you want
frequency linear with voltage, presuming you mean the
control voltage that requires a transform from Vctrl to
"right answer current".

Now I'd be inclined to challenge this requirement as a
VCO generally is part of a control loop that can seek
the "right answer" itself and doesn't need true dead
linear operation, just monotonic and not changing its
gain hugely (such that loop compensation is affected
or fails stability at one end of the other).

I might suggest, if your Vctrl range has to get close
to Vss, that you use a closed loop buffer to impose Vctrl
on a resistor, and take the current that it took to to
that off the drain of the buffer's follower output device.
That's a voltage to current converter that's fairly linear
and as well controlled as your resistor tempco & "make"
tolerances.

You might sum into that current, some function of
supply (following or opposing) to take out Vdd influence.
Similarly perhaps a PTAT current to fight temperature.

But again, maybe this is the PLL phase detector's job
and the problem isn't a problem at all. Check that....

 

[LvW]

I have designed a ring vco with the input voltage ranging from 0.7 to 1.8V, but i need it to work below threshold voltage too. The increase in frequency with respect to the voltage has to be linear.View attachment 135293
Regards,
Vidj

Because you want to work "below threshold" - does this mean that you want to design "translinear" circuit operation (log domain)?

 

[VeeJayy]

1. So, it is important for me to find the right voltage to current converter which is linear so that the frequency which depends on this current, linearly relates to the current if i'm not wrong?

2. "if your Vctrl range has to get close
to Vss, that you use a closed loop buffer to impose Vctrl
on a resistor, and take the current that it took to to
that off the drain of the buffer's follower output device.
That's a voltage to current converter that's fairly linear
and as well controlled as your resistor tempco & "make"
tolerances."

Is this the voltage to current converter example that you are talking about? I'm sorry, but it sounds very complex to me since i have just started IC designing since a month or so. Can you please break it down a wee bit more?

3. I wont be designing the complete PLL, just the VCO.

4. I have seen some concepts like level shifting, rail to rail(for low voltage working), body effect of cmos (for linearity). Am i in the right path? My professor said that i should probably look into some digital circuitry which would make my life easier. What do you suggest?

 

[VeeJayy]

Yes, assuming translinear means the current should vary linearly with the voltage rather than exponentially which is usually the case below threshold voltage.

 

[LvW]

Yes, assuming translinear means the current should vary linearly with the voltage rather than exponentially which is usually the case below threshold voltage.

FET operation below the threshold voltages is always based on an exponential I-V relation. This allows translinear operation (translinear: linear transfer properties becasue of the linear relationship between current and transconductance.)

 

[VeeJayy]

FET operation below the threshold voltages is always based on an exponential I-V relation. This allows translinear operation (translinear: linear transfer properties becasue of the linear relationship between current and transconductance.)

So am i right in asking "How can translinear operation at subthreshold input voltages help in obtaining high frequencies in the VCO" then? I have to obtain Mhz operation at 0 input volts too.

requirements : Mhz frequency range for 0 to 1.8V since i am using a 1.8V Pmos and Nmos in Cadence.

 

[FvM]

You could supplement another transistor to form a current mirror with M5, getting a linear I/f characteristic. Add a V/I converter if necessary.

 

[VeeJayy]

You could supplement another transistor to form a current mirror with M5, getting a linear I/f characteristic. Add a V/I converter if necessary.

Wouldn't it be the same circuit if i mirror a current from one nmos to the other since the Vgs of both will be the same? What would the current mirror do in this case?

 

[FvM]

The current mirror performs a linearization so that you get a linear frequency characteristic which is apparently required.

Vgs is in fact the same, and will always have a non-linear relation to VCO frequency. For linear V to f control, you still need a linear V/I converter.

 

[VeeJayy]

And the output current that is obtained from this converter will be fed to the current source and sink in my circuit(or may be as a current source itself)? rather than having a voltage source control the circuit. Am i right in stating this?

 

[VeeJayy]

I needed to know a few things before proceeding in the project.

1. Do the current sources and sinks need to be in saturation for the proper functioning? As in my circuit the inverters seem to be in saturation and the sources and sinks seem to be in the active region.

2. Does the current have a linear function with the frequency and if it does i need a linear voltage to current converter to get a linear voltage to frequency oscillator at the end?

3. The output current from this V to I converter will be fed to the drain of the "biasing circuit" of the circuit that i have?