Capacitive Liquid Level Sensing

From LabAutopedia

Jump to: navigation, search
Invited-icon.jpgA LabAutopedia invited article

Authored by: Caliper Life Sciences

(Originally published as Caliper Life Sciences Application Note 701; reprinted with permission)



Liquid level detection is a common option for liquid handling robots. Automated pipettors use conductive tips or fixed metal cannulas with capacitance measuring electronics to find the surface of the working liquid. The tip is sent down toward the surface of the liquid, the robot is stopped at the moment a change in capacitance is measured. The system now knows the height of the liquid surface in the consumable. This information allows the system to decide how far into the liquid to drive to aspirate the requested amount. The absolute amount found can be returned for record keeping purposes. Empty wells can be flagged. The only caveat is the liquid must be conductive. Non-polar liquids, such as oil, cannot be found through capacitive liquid detection. This article will explore changes to the sensitivity (Positive Threshold), reproducibility, and accuracy of finding liquid in a 384-well plate.



  • Sciclone ALH 3000 equipped with a liquid level sensing Z-8 (Caliper Life Sciences)
  • Calibrated single channel pipette (Ranin P100)
  • Deionized water (NERL CAT# 9801)
  • Dimethyl sulfoxide (JT Baker, Baker Analyzed)
  • 10% Trition-X 100 (Sigma T-8787)
  • TE Buffer pH 8.0 (Ambion CAT# 9858)
  • Clear polystyrene 384-well plate (Greiner 7811-01)
  • Conductive 100-μL disposable tips (Caliper Life Science p/n 111625)
  • Nalge-Nunc 300-mL reservoir (Nalge 1200-1300)

One of the settings in the configuration of the Liquid Level Detection for the Z-8 is the Positive Threshold, which is the target change in pipette channel capacitance (in picofarads) to be used for liquid level detection (see Liquid_handling_automated_performance_monitoring for technical details on capacitive sensing). The value is found in the Setup for the Sciclone (Figure 1). The values were varied to cover values of 0.05 to 10 picoFarad (pF). A smaller value will make the tip more sensitive to capacitive changes and thus to the liquid surface. The Nalge reservoir was filled with 200 mL DI water. A cover was fashioned from aluminum foil to limit evaporation. An area of the cover was removed to allow access to the “A1” location. This area was large enough to prevent interaction with the conductive tip, but still limit evaporation. The large volume and surface area, along with the cover was used to minimize any volumes changes during the experiment. Each tip was sent to the “A1” location of the reservoir. The volume found was calculated from the dV/dH from the Sciclone consumable library. The value was written to an Excel workbook and data from each Z-8 tip compiled. Each Positive Threshold value was repeated 10 times to generate reproducibility data.

Caplier LLS Figure 1.jpg
Figure 1. Liquid Level Detection Configuration

False Positive Detections

The default threshold value (0.35 pF) was used to determine how many false positives would be returned from an empty 384-well plate. Each Z-8 tip was sent to an empty Greiner polystyrene 384-well plate 200 times. The volume found was returned to an Excel workbook. To demonstrate false positive detection, the experiment was performed with a slightly more sensitive setting (0.33 pF).

Reproducibility Hitting a Constant Liquid Surface

The same 300-mL Nalge reservoir was used to determine the reproducibility in each Z-8 tip. The consumable was filled with 150 to 200 mL of water. The reservoir was covered except for the area over the “A1” location. Each tip was sent to the “A1” location of the reservoir 100 times. The volume found was calculated using the dV/dH from the consumable library and sent to an Excel workbook.

Accuracy within a 384-well Plate

A calibrated, manual pipette (Ranin P100) was used to deliver various volumes of test liquids to wells of a Greiner Clear Polystyrene plate. The transparent plate allowed visual inspection of the liquids tested. Three replicate volumes were placed in adjacent wells. The Sciclone was instructed to find the liquid in each of the three wells using all 8 cannula. This was repeated 3 times (total of 9 points per cannula) for each volume (100 mL, 50 mL, and 20 mL). Limiting the number of replicates reduces the effects of evaporation on the data. In each case, the dV/dH was used to calculate the volume found. This value was sent to an Excel workbook. Data was collected for de-ionized water, dimethyl sulfoxide, 10% Trition-X 100, and TE buffer.


The Positive Threshold value has a direct impact on the liquid sensitivity of the tips. As the value of positive threshold is lowered, the sensitivity increases (Figure 2). While this improves the accuracy, it does increase the chances of finding liquid when none is present. The more sensitive to liquid, the faster the threshold is met as the tip is submerged. The robot is stopped loser to the surface, thus giving a more accurate Z-axis measurement of the surface inside the well. Eventually, sensitivity can be heightened to the point where the tip can sense the top plane of the consumable (< 0.1). This is not desirable when looking for the surface of the liquid inside the consumable. Increasing the positive threshold makes the tip less sensitive. Once the value is greater than 1 the accuracy suffers. More of the tip is required to be submerged before the capacitance change is large enough to trigger a stop command. Thus a balancing of returning few false positives and accuracy must be met. Figure 3 and Figure 4 demonstrate a 0.02-pF change in positive threshold. Note that cannula 1 reports many false positive values when set to 0.33 pF, while it reports none at 0.35 pF. There is very little difference in the reported volume for these two positive threshold values however the electronics in cannula 1 are slightly more sensitive, reporting more false positives. A threshold value that returns fewer or no false positive is favored.

 Caplier LLS Figure 2.jpg  Caplier LLS Figure 3.jpg
Figure 2. Changes in Liquid Level Detection Positive Threshold values. N=10 Figure 3. Detection of Liquid in an empty plate. N = 200. Positive Threshold = 0.35 pF.
Caplier LLS Figure 4.jpg Caplier LLS Figure 5.jpg
Figure 4. Detection of Liquid in an empty plate. N = 200. Positive Threshold = 0.33 pF. Figure 5. Reproducibility of the Z-8 finding the surface of a liquid. CVs = 1%. N=100

Reproducibility is affected by many factors. The first is the Positive Threshold value. As stated above, a value that is too sensitive will be problematic because the tip may find the consumable as easily as it does the liquid surface. The next factor is the meniscus of the working liquid. Liquids have different surface tensions and interact with each plate material differently. Calculating the liquid volume in the well relies on a “change in volume with change in height” or dV/dH equation. This assumes a perfect square cylinder in the case of a 384-well flat bottom plate.

Some liquids will be close to this, others will have deep bowls for the meniscus. This bowl will alter the height where the surface is found. It also means that if the positional accuracy and precision in the X and Y axis is not good, the tip may find the upper edge of the bowl sometimes and the bottom of the bowl other times. If the bowl is really deep, this could mean a few μL difference in reported volume. Evaporation plays a large role when working with aqueous solutions. Grove et al. show the magnitude of evaporation for several liquids.[1] A covered 300-mL open reservoir was used to demonstrate reproducibility with limited evaporation and meniscus effects. The 8 cannula were sent to the “A1” location of the reservoir 100 times. Each cannula had a CV of less than 1%. All 8 averaged together have a CV less than 1% (Figure 5).

Caplier LLS Figure 6.jpg Caplier LLS Figure 7.jpg
Figure 6. Accuracy Calibration using Artel’s MVS Calibration Kit Figure 7. Calibration of DMSO in polystyrene 384-well plate.

The greatest effect on accuracy within the 384-well plate was the shape of the meniscus. Dimethyl sulfoxide (DMSO) has a pronounced meniscus. The deep bowl shape is easily seen in the data (Figure 6). The liquid surface was found deeper in the well resulting in an apparent low bias in accuracy. While DMSO is easily found with liquid level detection, using detection to catalog the volume would give values less than the actual volume. The bias can be removed through a calibration of true volume versus found volume in the well (Figure 7). Using the calibration curve, an accurate volume in each well would be calculated and returned to the catalog. Volumes less than 20 mL would be difficult to determine because of the deep meniscus. The TE buffer had a flat meniscus. Thus the returned value for the volume was very accurate. Water was similar to TE buffer, but the Triton-X solution gave results that were not as precise as the other liquids tested. This could be the result of the way the surfactant interacts with the tip. Great care was taken to prevent bubble formation during initial manual addition of the liquid to the wells. Bubbles would create a raised section on the meniscus causing the returned volume to be greater than a well with no bubbles.


Liquid level detection is a powerful tool for the liquid handling robot. It empowers the robot with the ability to find liquid surfaces and act on the information. Once the surface is found, the robot can travel just far enough into the liquid to perform the required liquid handling operation, eliminating the need to program a constant relative move into the plate. The height of the liquid surface from the bottom of the well, in conjunction with the dV/dH values for the consumable can be used to catalog the volume in the well. A balance between sensitivity to liquid and insensitivity to the environment is required to achieve optimum results. The positive threshold can be set to find the liquid of interest as accurately and precisely as possible while still knowing if the consumable is indeed empty. The default setting of 0.35 pF was found to be the best positive threshold value. Sensitivity does not vary between the liquids tested in this experiment, but the shape of the meniscus inside the well directly impacts what the robot perceives as the amount of liquid present. This perception is only important if the volume found is being catalogued or the amount to be withdrawn exceeds the perceived volume found. Water based liquids have less meniscus than DMSO solutions therefore accuracy of the volume in the well is more accurately measured. Some logistics in programming the Sciclone could be used to correct for the bias caused by the meniscus shape. Surfactants, while water based, are a challenge because they tend to have small bubbles present on the surface. Care must be taken in the previous liquid handling steps to insure bubbles do not exist if liquid level detection accuracy is important.

Related articles

Click [+] for other articles on  The Market Place for Lab Automation & Screening  Automated Liquid Handling Systems Microplate Transfer Systems
Click [+] for other articles on 
The Market Place for Lab Automation & Screening  Valves and Pumps
Click [+] for other articles on 
The Market Place for Lab Automation & Screening  Microplate Dispensers