Monte Carlo anlaysis to find Vth mismatch

[Ravinder487]

Hi all,
I've gone through many threads on monte-carlo analysis but I still couldn't get the method to calculate offset of this comparator.Please help me.

 

[Ravinder487]

When I plotted histogram of Vth of one of the transistors,It is showing a standard deviation of 6.5m and mean of 283m.Does this mean Vth mismatch(and hence the offset of comparator) will be around 19.5mV(3*6.5)??

 

[lamoun]

When I am calculating the offset of a circuit I use a slow ramp input voltage close to 0 (differential)
I run a transient analysis via monte carlo and I calculate the input voltage when the output crosses zero.

"Value" of "input voltage" when "cross" of "output" "0" (value and cross are the calculator functions that I use)

Keep in mind that the mean value of the offset will not be right unless you use a slow ramp (until the input is processed by the comparator the input voltage changes a bit)

 

[Ravinder487]

Before comparison output of comparator will be at Vdd.
Sorry I don't understand this

I calculate the input voltage when the output crosses zero.

 

[lamoun]

Before comparison output of comparator will be at Vdd.

Could you post the output and input of the comparator?

Sorry I don't understand this

I mean that in order to calculate the input referred offset you need (by definition) to see what input voltage causes the output to be zero.

 

[Ravinder487]

Could you post the output and input of the comparator?

This output at cross-coupled NAND gate.other output of NAND gate will be inverted version of this.

Later I will post output at intermediate nodes.

 

[Ravinder487]

These are the outputs at darin of M3 and its counterpart.Marked evaluate and precharge phase.

 

[lamoun]

To calculate the offset of a comparator with a reset (precharge - track, what ever) I strobe the output to get only the evaluated data.

For example, this is the output/input of track-and-latch comparator. If I were to check when the output crosses zero, I would get wrong results. So I strobe the output to get only the data I want. (the gray output, superimposed on the real one, is strobed)



Here is the strobing parameters I used above.
I have set the clock period at 1ns -> strobing period 1ns for this circuit.
I have also set the skipstart at .5ns in order to start strobing when the input is already evaluated.

 

[Ravinder487]

Hi Lamoun,
I doubt, your method gives me the resolution of comparator not the offset of comparator,correct me if I'm wrong

 

[lamoun]

With a really slow ramp you could estimate the resolution of the comparator.

But running monte carlo analysis with the equation I mention above gives you the offset of the comparator.

 

[Ravinder487]

That sounds great,I will do the method you suggested and will let you if there are any discrepancies!!

 

[Teddy]

Plotting Vth does not make sense anyway - Vth i susually the parameter to be spread by monte carlo.

 

[Ravinder487]

My idea is to cancel offset of comparator using body biasing of driving transistors in sense amplifier.The method is explained in Pg.No.490 of following attachment