Turbidity Calibration Standards Evaluated from a Different Perspective

Formazin was established as the first calibration standard for turbidimeters in the 1950's.  Machine performance and Environmental Protection Agency (EPA) approval for turbidimeters was structured around formazin as the calibration standard.  The EPA method 180.1 also outlined design parameters for turbidimeters used for testing surface source drinking water.  The design parameters include a white light source and photodetector(s) positioned at 90° to the light source.  The nephlometric design was to optimize the detection of sub-micron particulate.  Refer to Brumberger et al, Light Scattering is a Function of Light Wave Length and Particle Size.  That is, the characteristics of a given particle depend on its refractive index, shape, and size.  Sub-micron particles scatter short wavelengths light (white light) at optimally 90°.

Today the scenario is unchanged except for additional EPA approved calibration standards.  Besides "scratch" formazin, there is formazin concentrate (4000 NTU – Nephelometric Turbidity Units), stabilized formazin and submicron polymer suspensions.

The submicron polymer suspensions are cross-linked copolymer microspheres suspended in ultrapure water.  Size distribution is 0.02 to 0.203 microns and mean diameter is 0.121 microns.  They are stable down to 0.10 NTU for at least 1 year.  No preparation is required; vigorous mixing or agitation will degrade the accuracy of the polymer standards.  They are non-toxic, which means they can be used in any setting without fear of contamination to the user, equipment, or the environment.  The EPA approved the polymer suspensions in 1984 as a calibration standard for turbidimeters under the name AMCO AEPA. Today, the improved standards are known as AMCO CLEAR.

The polymer suspensions are unique among the approved standards in several obvious ways:

  • non-toxic
  • stable
  • ready to use
  • accurate +/- 1% of stated value lot to lot
  • submicron in particle size distribution
  • values available from 0.10 NTU to 10,000 NTU
  • size, shape, and particle size distribution is always the same, regardless of lot
  • no special storage requirements
  • no special disposal requirements
  • can be frozen solid for four hours before degradation of suspension

Turbidity reporting is done on finished water (effluent), not incoming water (affluent), because filtered water is the final product.  The remaining particulate in finished water after the filtration process is mainly submicron in size, unless there is a filter breakthrough.

The polymer suspension is a better standard for the calibration of surface source water analyzing turbidimeters.  The primary reason is its particle size distribution, 0.02 micron (m) to 0.2m and a mean size of .121m.  As mentioned the EPA dictates a turbidity design configuration to maximize submicron particle detection less than 1.0m.  It is therefore advantageous to calibrate a nephelometric turbidimeter with a standard that most closely matches the size of the particulate it is analyzing.  Plus the standard is more precise and the size distribution is consistent.

 

Formazin, by contrast, is not consistent in size distribution.  From lot to lot sizes range from less than 0.1 to 10 microns.  The majority of sizes are above 1.0 m.  This can be demonstrated by withdrawing a sample of the clear liquid off of a 4000 NTU concentrate once the large particles have settled out.  The turbidity is that of the formulation water.  Eighteen years of testing at the APS Analytical Standards labs, only once, has this not been the case.

It has been argued that since real world water samples have a wide distribution of particle shapes and sizes; the perfect turbidity standard should be of the same matrix.  Perhaps true if the filtered final water still consisted of that composition, however, this is not the case.  The large particles have been removed.  Remember that turbidity reporting is done on finished water.

 

See Figure 1 of the three particle sizes.  Figure 1(A) most closely resembles the remaining particulate in finished treated water.  The size 1/10th the wavelength of white light less than 60nm = 0.06m.  White light wavelength is 400 to 600 nm = 0.4 to 0.6microns.  Again to emphasize the fact, the EPA protocol of nephelometric turbidimeter design optimizes detection of submicron particulate that scatters light in a 90° direction.  Formazin is represented by Figure 1(C), 6000nm = 6.0m.  Formazin is outside the box; too large in size by several factors to equate to the particles that are analyzed.  Does Formazin represent real world samples?




Formazin can be reproduced +/- 1% batch to batch.  This is true under ideal conditions; which involves quality chemicals, precise volumetric glassware, ultra pure water and excellent laboratory technique.  The formulation process is tedious and timely.  The final diluted standards are time sensitive and it is commonly recommended not to prepare standards below 2.0 NTU.  The EPA requires turbidity values not to exceed 0.3 NTU for surface source drinking water.

The reality is that the majority of water plants instead of formulating scratch formazin purchase the stock concentrate from several commercial sources.  Accuracy of the concentrate does not exceed +/- 5% of stated value; plus having the limitations of scratch formazin.  There is also an alternative stabilized formazin standard.  However it also has the limitations of formazin regarding particle size, toxicity, and re-suspension after a period of non-use.

To demonstrate the relationship of machine design on formazin and the polymer calibration standards, three different lots of stock formazin data (1,2) and instrument specific polymer standards were tested in four different turbidimeters.  The difference in this analysis is that the machines are calibrated with both types of calibration standards instead of just formazin and compared against each other.

Each machine employs a different optical and photo detector design:

HF Micro 100

  • true EPA method, 180.1 design
  • light source - white light  400-600 nm wavelength
  • 90° photo detectors

Hach 2100 AN

  • EPA method design 0.00 NTU to 39.9 NTU; 40.0 NTU to 10,000 NTU in the ratio mode
  • light source - white light  400-600 nm wavelength
  • ratio mode - four photo detectors; one at 90°, one at backscatter position, one at
  • forward scatter, and one at 180° position - (transmitted light)

Hach 2100 AN ISO

  • same design as the 2100 AN but uses a light emitting diode; 860nm wavelength

McVan 160 Analite

  • light source - fiber optic IR, 870 nm wavelength
  • 90° light scatter

Analyzing the test results demonstrates several key points (refer to the 2 data sheets).

The different formazin lots do not stay within the 1% variance that is claimed by the manufacturer.  The importance of the variance relates to the premise that it is reproducible by any end user.

When APS does comparative studies of formazin to the polymer calibration standards, as in this case, our chemists use ultra pure water, class A volumetric glassware and the dilution technique is redundant.  All standards are formulated at the same time and the machine is calibrated immediately to minimize variance.

From 1995 through 1997, APS Analytical Standards had a contractual agreement with a major turbidimeter manufacturer to supply them the polymer calibration standards for their turbidimeters, as an alternative standard to formazin.  Several quality control procedures were established to verify the equality of the polymer standard with the performance of their formazin.  Several lots of the polymer calibration standards were formulated with each new lot pre-approved before shipping product.  The Quality Control (QC) procedure required APS to calibrate three machines of the same model using their stock formazin as a starting point diluted down to the appropriate calibration values.  The quality of the dilution water and the Standard Operating Procedure (SOP) were directed by them.  During a period of time, lots of APS calibration standards failed QC specifications.  APS determined that their machines were being calibrated with different lots of formazin stock than the machines they provided us.  The companies agreed to use the same lots of formazin and combine three lots to make one stock concentrate to calibrate all instruments.  Once this procedure was in place, all QC issues disappeared.

They asked APS to make all of our standards based on this procedure instead of our procedure of comparing our standards lot to lot using turbidimeters and spectrophotometers and to two years of retention samples, using formazin as a back up check.  Unfortunately if APS were to follow this procedure we would lose our +/- 1% variance from lot to lot for our other customers.  We could not do this because the accuracy of the APS calibration standard would be chasing the variance of formazin.

Evaluating the data sheets for the HF Micro 100 and the McVan 160 probe, the worst case variance is 6.8% per NTU value.  Discarding the outliers (2) the average variance is 1.56%.  The machines do not change into ratio mode above 40.0 NTU.  The polymer calibration standards are instrument specific due to the wavelength of the light sources, which are extremely different; HF 400-600nm and McVan 870 nm.  The light source wavelength for the McVan is almost twice that of the HF.  The impact of the difference is realized in what the two machines see.  Imagine two wire mesh screens; one sized 0.4m and the other size at 0.82m, which one is going to trap smaller particles?  Remember the EPA turbidimeter design criteria for filtered drinking water wavelength?  The white light HF machine with its shorter wavelength, 400-600nm, will strike more small particles than the McVan machine.  Visualize ping-pong balls verses basketballs.

The two Hach machines data sheets are the most complex to decipher.  First, only the Hach 2100 AN instrument specific polymer calibration standards were used in the testing of both machines. At the 20 NTU reading, overall variance is 1.45 %.  At the 200 NTU value, variance is 3.7%.  At the 1000 NTU value, the variance is 10.43%.  Lastly, the variance is 4.38% at the 4000 NTU calibration point.  The percent error is large for both machines at the 1000 NTU and 4000 NTU polymer standard, why?

One, the polymer standards are specific for the 2100AN machine.  Two, the machines are in the ratio mode at the 200 NTU, 1000 NTU, and 4000 NTU calibration points.  Thus, multiple detectors at different angles other than 90° are being used, and transmitted light is also measured.  These additional detectors are not seeing as much of the polymer suspension as with the 90° photo detector.  Three, the ISO machine uses an infrared light source, 860 nm, as opposed to a white light source, 400-600 nm.  Four, when calibrating the AN machine with the polymer suspension, the formazin standards read high in the ratio mode.  The additional detectors are seeing the formazin therefore, inflating their turbidity readings.  Also, more polymer suspension is needed to read matching formazin values at 200 NTU, 1000 NTU, and 4000 NTU.  This is demonstrated in the ISO machine where the polymer suspension standards are not instrument specific.  Once the ISO machine is calibrated with the non-instrument specific standards the calibration points are undervalued.  This is shown by low formazin readings.

Is that a flaw in the polymer standard?  No, because in the ratio mode the machines are  "tuned" to measure large particles and to compensate for color.  Neither of which is a parameter in the analysis of finished drinking water.

The last test results (data sheet 3) demonstrate the variance of the generic EPA formulated polymer calibration standard in six different design parameter machines.  The term generic is defined as the standard to be used to calibrate any turbidimeter that meets the EPA design parameters.

A criticism of the polymer calibration standards is that the turbidity values are established by comparing point to point against formazin, down to 0.1 NTU.  Discard the outliers and factor this into the variance from 0.1 to 1000 NTU, then deduct 5% for the expected accuracy of formazin.  The compared deviation is 3.37%!

Sorting all the data to come to a variance of 3.37% for the generic EPA formulated polymer standards to formazin over an NTU span of 0.1 NTU to 1000 NTU mutes the criticism of the polymer calibration standards.

Obviously, machine design can make radical differences in readings but they are outside of the EPA design parameters.  Reverse the standard comparison.  Let the polymer calibration standards be the gauge.  Consider the benefits:

  1. The polymer concentrate is formulated in batches that could be a 10 - 20 year supply.  Batch to batch particle size variance +/- .001%.
  2. Retention samples that could last indefinitely.

Published in a turbidimeter manufacturer catalog is an announcement of a new turbidimeter that is several times more effective at detecting submicron particulate in filtered water.  It states that current machines are not adequate to analyze the new ultra low turbidity requirements of the EPA.  However, they offer formazin as the calibration standard for the machine.  In a position paper they claim the polymer calibration standards have a very narrow particle distribution and particle shapes may be outside the constraints of many real world samples.  This contradicts their latest claims of their new machine and the use of formazin as a calibration standard.

Why is formazin acceptable to calibrate a machine that is designed to detect submicron particulate, while the polymer calibration standard is criticized as not sufficient to calibrate EPA design parameter machines?  The AMCO Clear standards have been in use for 18 years with thousands and thousands of standards used to calibrate turbidimeters.  In addition, Syracuse University under the sponsorship of AwwaRF conducted a one-year study of the performance of the calibration standards and turbidimeters.  All of the EPA approved calibration standards were evaluated including AMCO Clear in a variety of bench top, online, and portable turbidimeters, some not meeting the EPA design criteria.  After calibration the machines tested “real world” water samples form various water plants around the country.

Two interesting revelations came out of that study.  First, on Page 76 of the study it states  “In every case the least squares fitted value of the slope and intercept fall within the 95 % confidence limits of the other three calibration methods.  The calibration method does not seem to have a significant effect on the agreement or lack of agreement between instrument-modes.”  Second, on page 75 it says “The coefficients of determination for the Group B instrument-modes are all less than 0.8, and for the Group A instrument-modes all but one value are greater than 0.9.  These results are consistent with the earlier observation that there are obvious and significant difference between the measurements made with the Group A and Group B instruments.”  The Group A machines are all from one manufacturer who claims AMCO Clear Standards do not work in their machines.  Group B machines are from a mix of manufacturers.

In summation, if the intent of the application is analyzed, it is clear which is a better standard.  The explanation of why the polymer calibration standards are instrument specific reveals a deviance of machine design as opposed to a shortcoming of the standard.

Factoring in the design variance, the generic polymer calibration standards on average are well within the tolerance of formazin (+/-5%).

The final consideration is safety, the polymer calibration standards are environmentally safe; formazin in any form is toxic.

Generic 2100 AM 2100 ISO 43900 18900 Mirco 100 *2100 p Variance
0.02 .033 0.029 0.001 0 0.03 0.11  
0.1 0.155 0.141 0.124 0.1 0.11 0.22 26%
0.2 0.293 0.275 0.27 0.22 0.23 0.37 29%
0.4 0.482 0.462 0.472 0.4 0.4 0.57 17%
0.5 0.625 0.598 0.63 0.55 0.52 0.75 17%
0.6 0.715 0.692 0.706 0.63 0.61 0.9 12%
0.8 0.933 0.905 0.932 0.85 0.82 1.06 11%
1 1.16 1.12 1.163 1.07 1.02 1.29 11%
2 2.29 2.21 2.32 2.1 2.08 2.46 10%
3 3.38 3.24 3.42 3.2 3.08 3.58 8.70%
4 4.38 4.23 4.46 4.3 4.05 4.59 7%
5 5.37 5.18 5.48 5.2 5.04 5.61 5%
6 6.34 6.13 6.47 6.2 6.01 6.5 4%
8 8.36 8.08 8.52 8.3 8.02 8.49 3%
10 10.4 10 10.63 10.4 10 10.8 3%
20 20.9 20.3 21.4 21 19.7 20.9 3%
30 31.4 30.5 32 31 30.3 Total % Variance of 0.1 – 40 = 9.69%
40 40.6 39.3 41.4 41 40.1
50 50.9 49.6 52 52 49.9 47.3 1.76%
60 61.5 60.6 69.6 62 60.2 56.5 4.63%
80 79.5 79.2 90.3 81 79.9 71 2.40%
100 102 101 105.2 106 100 86.3 2.84%
200 204 207 231 - 201 158 5.38%
300 327 315 352 - 300 219 7.80%
400 416 430 458 - 402 271 6.60%
500 516 546 565 - 495 310 6.10%
800 863 957 924 - 800 445 10.75%
900 966 1084 1028 - 896 Total % Variance of 50 – 1000 = 6.45%
1000 1091 1237 1149 - 1000

*Data not used in calculation
Combined Total Variance = 8.07%

Hach 2100 AN / ISO

 
calibrated lot # A1019
 
 
 
 
 
calibrated lot # A1082
 
 
 
 
 
calibrated lot # A1176
 
 
 
 
 
calibrated APS stds
 
  lot # A1019 lot # A1082 lot # A1176 APS stds variance
20 20.1 19.7 20.3 20.1 0.25%
200 201 199 203 208 1.38%
1000 1000 984 1018 1098 2.50%
4000 3997 3953 4041 4796 4.92%
           
20 20.3 20.1 20.7 20.4 1.88%
200 202 201 204 212 2.38%
1000 1025 1001 1035 1106 4.18%
4000 4043 4000 4089 4853 6.16%
           
20 19.7 19.7 20 19.6 1.20%
200 199 196 200 205 0.00%
1000 987 970 1000 1090 1.15%
4000 3954 3918 4000 4956 3.93%
           
20 19.8 19.5 20.1 20.1 0.60%
200 171 170 173 201 10.60%
1000 653 644 661 1000 26%
4000 3464 3402 3516 3972 10%

Total Variance = 4.90%

 

Hach 2100N / AN

 
calibrated lot # A1019
 
 
 
 
 
calibrated lot # A1082
 
 
 
 
 
calibrated lot # A1176
 
 
 
 
 
calibrated APS stds
 
  lot # A1019 lot # A1082 lot # A1176 APS stds variance
20 20.1 19.7 20.4 20.9 1.38%
200 201 199 203 208 1.25%
1000 1002 992 1017 994 0.13%
4000 4002 4048 4054 4048 0.24%
           
20 20.4 20 20.6 21.3 2.80%
200 202 200 204 211 2.13%
1000 1021 1000 1035 1000 1.38%
4000 4065 4000 40124 4101 1.81%
           
20 19.9 19.6 19.9 21 0.50%
200 198 196 200 207 0.13%
1000 984 969 1001 987 1.50%
4000 3955 3893 4000 4000 1%
           
20 19.2 18.9 19.5 20.1 3%
200 231 230 235 200 12%
1000 1622 1597 1655 1000 46.60%
4000 4407 4375 4425 4001 7.60%

Total Variance = 5.22%

 

MCVAN 160 PROBE

calibrated lot # A1019
 
 
 
calibrated lot # A1082
 
 
 
calibrated lot # A1176
 
 
 
calibrated APS stds
 
 
  lot # A1019 lot # A1082 lot # A1176 APS stds variance
10 10.03 9.97 10.25 10.03 0.70%
100 100.3 98.8 100.7 102 0.45%
           
10 10.14 10 10.54 10.06 1.90%
100 100.7 100 102.4 105 2.00%
           
10 9.8 9.67 10.1 9.75 1.20%
100 98 97 100.1 103 0.27%
           
10 10 9.82 10.22 10 0.10%
100 95.6 94.8 96.6 100 3.30%

Total Variance = 1.32%

MCVAN 160 PROBE

calibrated lot # A1019
 
 
 
calibrated lot # A1082
 
 
 
calibrated lot # A1176
 
 
 
calibrated APS stds
 
 
  lot # A1019 lot # A1082 lot # A1176 APS stds variance
10 10. 9.88 11.4 9.82 2.80%
100 1000 987 1007 1001 0.13%
           
10 10.2 10 11.6 10.9 6.80%
100 1015 1000 1024 1015 0.14%
           
10 9.56 10.6 10 10.4 1.40%
100 980 979 1000 980 0.25%
           
10 10.6 10.3 11.8 10 3%
100 1002 987 1012 1000 0.03%

Total Variance = 1.8%