[4200A -CVU] Prolem with CV measurement at 1Khz

Hello my friends,
I 'm doing test on wafer with CV measurement. As requirement testing, the result should be with in 3 sigma ( below 1%).
But actually, Almost result is upper 1%, this problem only happen at frequency 1Khz. Frequency upper than 1Khz almost result is below 1% ( within 3-sigma)

Below is prameter config and result.
Have anyone testing like that or have exprerience about this testing, Please give me advise, thanks so much
 

Config	preSoak Voltage	Hold Time (s)	Conidtion Test	Average(pF)	Standard Deviation	3Sigma
- DC Bias : 0.3v - AC Drive Signal: 30mV - Interval: 20 - Mode: Sampling - Freq: 1kHz	-0.3	10	1	1.424814	0.040005262	8.42%
		30	2	1.401139	0.033187129	7.11%
		60	3	1.4367645	0.041729935	8.71%
	-1.5	10	4	1.8695225	0.066384102	10.65%
		30	5	1.886752	0.04348203	6.91%
		60	6	1.882425	0.042995434	6.85%
	1.5	10	7	1.896749	0.047060789	7.44%
		30	8	1.9012545	0.042947112	6.78%
		60	9	1.907959	0.053639368	8.43%
	3	10	10	1.4861455	0.063937909	12.91%
		30	11	1.4844695	0.049281225	9.96%
		60	12	1.5566555	0.057454766	11.07%

Can you explain the 3-sigma value?  How compute it?

In your config, what sort of speed settings?  Use Quiet if looking for small standard deviations.

It looks like your DUT value is a few single digit pF.
At 1KHz, it will be very big Z and very small current relative to using higher test freq.
In our datasheet for CVU, we do not publish value for pF at 1KHz.

 

Hi Andrea,
Thanks for your response.
In speed setting, I'm using "Normal" but look like not good. I changed to custom and using Delay Factory, Filter and A/D Apeture time with config as below
:CVU:SPEED 3,1,20,0.05 --> Look better and standard deviations is small 

But our system is measurement pad on wafer and using probe tip to touch on pad of wafer to measure ( see picture as attached). During loading/unload wafer, The value of Capacitor which measure on wafer was unstable ( refer excel file)
Could you give me some adviese for this case?