Printer Friendly

Compression set vs. compression stress relaxation.

Measurement and specification of elastomer compound response in compression is an important aspect of elastomer seal properties. While compression set is commonly seen as a primary property, many designers prefer to establish compression stress relaxation (CSR) values to more fully characterize a material's sealing capability. Engineers have recently begun to provide data or specify compression set requirements based on a modification of the initial method outlined in ASTM D 395 Standard Test Method for Rubber Property--Compression Set, Method B, Compression Set Under Constant Deflection in Air (ref. 1).

Using the standard test method, samples are compressed at room temperature, placed in a hot air oven for a specified period, and then removed immediately from the fixture and allowed to cool in an uncompressed state on a thermally nonconductive surface to room temperature.

Modifications to this test allow samples to cool to a desired temperature in the compressed state before being removed from fixtures, and allowed to recover to their final thickness. Some samples are removed at room temperature, because it is felt that this is the condition the material will see in a sealing situation. Others have removed samples at an even lower temperature to suggest relative differences in a material's ability to recover, and its potential performance at low temperatures. The criteria for choosing a specific test and test condition, as they pertain to material performance in an intended application, are not always completely obvious and defined.

Another approach used to evaluate sealing capability is CSR. In the compression set test, one measures the amount of the unrecovered strain of a sample after it has been compressed a certain amount, and exposed to a particular aging condition. In CSR, the sample is compressed and aged in a similar manner, but after aging, it is maintained in its compressed condition, where sealing force is measured in a quasi-static condition. This sealing force value is compared to an initial measured value and is reported as % retained sealing force (% RSF).

The quasi-static condition is a situation where a measurement is taken in a manner to reflect as close as possible, a static condition. This is done by using a procedure that limits the polymer's viscous response by minimizing the rate and extent of deformation of the sample. It would be of interest to determine a relationship between compression set and CSR, and to evaluate the strengths and limitations of these tests in defining sealing capability and predicting service life. This paper is intended to provide some data, ideas and opinions in regards to these concerns.

Viscoelastic response

Elastomer compounds are viscoelastic materials. This means that while their equilibrium load and recovery response is elastic, the response changes with time due to its viscous behavior, from its instantaneous response to its equilibrium response. The rate at which this occurs is dependent on the glass transition temperature (Tg) of the material, the compound formulation and its cured crosslinked network.

These factors can affect the relationship between its elastic and viscous response, which will vary as a function of temperature and rate of deformation. When one compresses an elastomer to a given strain at room temperature, the initial load will depend on the rate at which it is compressed, after which the load will decrease over time to an equilibrium elastic load.

Upon removal of the constraining member, the deformation will recover to an equilibrium position, which might not be its original position. Full recovery to its original position might require more time or energy by heating to a higher temperature to overcome the internal viscous drag for specific materials.

By heating the sample to a higher temperature, its energy and molecular motion increase, so through thermal expansion and a decrease in the viscous response, the material is allowed to freely expand and then recover to its final equilibrium position. This position should be the same as the initial position if there was no permanent damage to the morphology and/or change to the crosslinked structure.

These types of evaluations and responses need to be done at strain levels and temperatures where the material is structurally and thermally stable, within an acceptable exposure time frame. This type of response should be reversible and be representative of what is termed physical relaxation. Understanding and measuring this type of response can provide insight into a material's stress-strain properties in applications and how it can affect the test results used to define its sealing capability, using the different test methods and specifications.

Understanding the difference between the part of a material's response that might be a result of its viscous response (physical relaxation), as compared to the portion due its aging response (chemical relaxation), is an important exercise. Chemical relaxation is a term used in CSR literature, which is associated with chemical reactions that involve the making or breaking of crosslinks or polymer chains, or any non-reversible changes to a material's internal structure (ref. 2). This issue is reflected in an elastomer's high initial CSR relaxation and its high compression set values when test samples are removed at lower temperatures.

It can be shown that the physical relaxation is affected by the rate and extent of deformation and has a time dependency that varies with temperature. The physical relaxation occurs more rapidly than the chemical relaxation, but unlike the latter, can provide a certain level of reversible recovery.

Configuration effects

In addition to a material's viscoelastic response, one needs to understand the effect of configuration on a material's load and recovery response. Configuration can be defined as those dimensional values of the test components and constraints that can be defined by a specific number or relationship. These can include % compression, and a sample's shape as reflected by its height, diameter, shape factor or surface to volume ratio.

The shape factor of a sample is the ratio of the area of one compressed face to the area of a sample allowed to expand, as shown in figure 1 (ref. 3). For a cylinder, the area of the compressed face is [pi][D.sup.2]/4; while the area allowed to expand is [pi]Dt, so its shape factor is D/4t. Which of these configurational issues have an effect on the material's test response and its relationship to performance in an application is of interest for design purposes.

Related to the configuration response are other possible constraints, such as friction or contact. These types of constraints can skew the responses in some configurations and can affect initial and aged response as compared to unconstrained samples. One response that is normally accepted as a uniform response over a large range of configurations is that the Stress-Strain response of an elastomer compound in compression should be the same if: 1) The sample is compressed in a zero friction condition; 2) it has a stable sample shape; and 3) it has a "uniform" compressed cross-section. A uniform compressed cross-section is one in which the sample's contact surfaces stay the same over the range of compression and that the cross sectional areas below the compression faces move uniformly and freely with the compression faces and the sides do not bulge out or move inward. This is valid for samples within a certain range of shape factors and % compression that allow contact surfaces to move freely, and where stresses can be developed uniformly between its contact faces.

Samples with shape factors that are too low are unstable and can buckle. In buckling, the stress distribution becomes non-symmetrical within the part, and portions of the sample shift non-uniformly off the center axis, no longer supporting a compressive load. Samples with high shape factors can create problems with friction effects and non-uniform loading, which can result in a non-uniform stress distribution and cause deviations from a uniform stress-strain response. Samples at high % compression can also begin to deviate from this same uniform response, similar to that seen with high shape factor samples.

In cases where the samples have reasonable shape factors and are deformed to reasonable strains (% compression), the stress-strain response of a material, tested at the same tem perature, should be the same. To know the extent to which configurations and conditions show deviations from this, one needs to test a range of samples under different conditions to different strain levels. In doing so, one can begin to define which responses are characteristic of the material and which are due to the extremes or variations in configuration.

Since all of the testing for CSR and compression set begin with an initial deflection and involve both physical and chemical relaxation that could be affected by internal stresses and their distribution, it would be good to know the extent that each test reflects a material's response and what range of configurations might skew this response. In looking at these tests, one can consider them to be a combination of several interrelated tests and conditions. The objective in this exercise is to try to evaluate these responses separately and independently, and then in combination, as done in a CSR or compression set test; to better understand factors that can affect the final values, in relation to the different responses that are occurring in the tests. A compression set test could be thought of as a sequence of the following six tests: 1) An initial instantaneous uni-axial compression; 2) a physical relaxation at room temperature (RT); 3) a differential thermal expansion to an elevated temperature; 4) physical and chemical relaxation at an elevated temperature for a period of time; 5) decompression at an elevated temperature; and 6) a thermal contraction and viscoelastic recovery over a temperature range in an unconstrained state back to room temperature. When the process is complete, the sample's final height is measured and compared to its initial value, to determine a compression set value. This article explores how each of the steps affects the final result and relates the final result to the material and its capability in an application.

Elastomers in compression--physical responses

Instantaneous reponse

To begin the evaluation of the effect of configuration on the compressive response of an elastomer, several different shaped samples were molded with Dyneon FE-5620 fluoroelastomer and post-cured, and then checked for their stress-strain response. A range of sample shapes was molded to evaluate the effect of different configurations and potential constraints, and to note how these responses might also affect their viscoelastic response and tests used to measure their sealing capability. The configurations chosen were those with uniform compressed cross-sections, such as cylinder and washer shapes with different diameters and shape factors, and one with a non-uniform compressed cross-section, a (-214) o-ring. Plied disk samples were also included in this test. This particular range of shapes, shown in figure 2, was chosen because they represent a range of simple shapes that are often used in characterizing a material's sealing response or could reflect the relative size or shape factor of seals used in applications.

An elastomer's response in compression is reflected in changes in the force it develops, as a function of the extent of deflection. This is the most basic response that can be measured. This response can be normalized to the sample height and contact area to develop a stress-strain response for a range of similar shapes, provided the shapes have a uniform compressed cross-section. The contact or internal compressive stress is determined by dividing the load by initial contact area, while the strain is determined by dividing the deflection by the sample's initial height.

The objective here is to see initially how the difference in these shapes and the level of strain can affect deviations from a material's intrinsic stress-strain response. It is also of interest to see if, or how, these configurations affect a material's sealing response based on compression set and CSR tests.

The dimensions of the samples that were molded for the different evaluations are listed in table 1. Some samples were not used in certain tests and only selected data are shown, which best illustrate specific points. However, all of the data generated during the testing were used in developing the conclusions that were drawn.

The first evaluation was to determine the instantaneous stress-strain response of different samples. This test was performed both with and without lubricant. The objective here was to see how much a lubricated surface could affect the stress-strain response over the range of sample sizes and shape factors. Deviations from a constant stress-strain response would provide information on situations that could create non-uniform stress conditions. The results showed that the stress-strain response with lubricant is similar over the range of samples tested, shown in figure 3. The two samples that show higher stress levels, in comparison, are the ones evaluated without lubricant. They also show that the magnitude of the stress increase is related to the corresponding increase in diameter and shape factor. Lubricated plied samples were also evaluated and showed slightly higher stresses than the solid lubricated samples, but not as high as the un-lubricated samples, as shown in figure 4. The higher stress levels for these samples may be due to a smaller actual or "effective" compressed area. When samples are die-cut out of sheets, the edges of the sample tend to be tapered and not vertical, and when stacked and loaded, may not be aligned exactly or transmit loads uniformly at their edge. This results in a response that would be more typical of a solid sample with a smaller diameter. If the difference between the assumed and "effective" dimensions could affect compression set or CSR results, then it might explain possible test result differences between pried disks and molded samples.

[FIGURES 3-4 OMITTED]

The deviation in the response seen with the non-lubricated samples relative to uniformly lubricated samples is a result of a frictional constraint, which is configuration dependant and, as such, does not reflect an intrinsic material property. If the relaxation response is a function of the level of stress developed at a given strain, then the type of samples (e.g., plied disks) or the lack of use of a lubricant could affect the final test results. When lubricant is used within a reasonable range of sample sizes with solid uniform samples, the load response is more uniform and its instantaneous stress-strain response at room temperature can be considered an intrinsic material property.

Along with the stress-strain response, one can also look at samples in the context of their load-deflection response. This response is a reflection of both the material and its configuration. Samples of the same diameter will show the same load at the same level of strain (% compression), but as the shape factor increases, so does the stiffness (slope of the load-deflection curve). The stiffness developed by these materials is analogous to the spring constant used to define the response of springs. These responses are shown in figure 5.

[FIGURE 5 OMITTED]

Figure 6 shows a comparison between an o-ring and a theoretical load-deflection response for a washer shape with the same ID, OD and thickness, assuming the uniform stress-strain response that was noted earlier. The response for the washer is fairly linear up to about 25% deflection, after which the load increases at a faster rate. This response can be due to polymer chain orientation, strain crystallization or friction effects that can begin to occur at higher strains. This can be considered part of a configurational response. When the load-deflection curve of an o-ring is compared to this response, it shows more curvature. The non-linear response for the o-ring is due to the fact that as the o-ring is compressed, more of the top and bottom surfaces come into contact with the compression faces. Because of this, the actual % compression over the contact faces will vary non-uniformly between 0 and 25% deflection. Figure 7 shows how much this varies, assuming no lateral deformation, which does occur in the real situation. As a result of this non-uniform compression, the internal stress levels across the cross-section of the o-ring will also be non-uniform. This difference might affect its relaxation response as compared to samples with uniform compressed cross-sections.

[FIGURES 6-7 OMITTED]

Physical relaxation response at room temperature

Comparing instantaneous load-deflection responses for an elastomer is relevant if one were only concerned with the initial compressive response of a material. But, once the sample is compressed to a specific strain level, the load will decrease as a function of time, based on the physical relaxation of the material and any interfacial movement. To evaluate this response, relaxation tests were performed at room temperature with lubricant on all of the samples noted previously. A comparison of 13 mm diameter disks is shown in figure 8. The sample with the lowest shape factor showed slightly less relaxation. This was also seen with the 25.4 mm (1") diameter samples.

[FIGURE 8 OMITTED]

Evaluations such as these are useful to help define a minimum time to wait before checking the initial sealing force in a CSR test. One would like the time interval to be long enough so that the initial test time will have allowed for sufficient relaxation to occur, so uniform consistent readings would be determined regardless of general variations in test timing, but short enough to allow a reasonable test time. In figure 8, it shows that at 70 minutes the samples had relaxed to 95% of their final values for the 120 minute relaxation test.

When all of the samples were compared as a function of % relaxation vs. shape factor, there appeared to be some correlation with shape factor, but it was not too definitive, as shown in figure 9. How much difference slight changes in positioning, lubricant and configuration made could not be defined more specifically in this study. Overall, the samples appeared to relax about 25% over the 120 minute test, showing a slight relationship with shape factor.

[FIGURE 9 OMITTED]

Physical relaxation response over a temperature range

Although physical relaxation is generally done at room temperature, relaxation as a reflection of the material's viscoelastic response can take place over a range of temperatures, particularly at high temperatures. This occurs in applications where elastomer seals are assembled at room temperature and then exposed to elevated temperatures for a period of time, short enough to not produce any chemical relaxation or aging, but long enough to reach the elevated temperature and undergo physical relaxation. As the temperature increases at constant compression, the differential expansion between the elastomer and the constraining material creates higher stress on the samples. This additional driving force, along with the increased thermal molecular motion in the polymer system, forces it to relax even more at elevated temperature. If the temperature is not too high with respect to the stability of the polymer system, and the time is long enough to allow relaxation, but short enough to prevent any chemical relaxation, one can measure the extent and reversibility of this relaxation corresponding to changes in measured dimensions and loads. This type of response is shown in figure 10. It shows the sample's load and strain response as it: 1) Is compressed at RT; 2) heated to 150[degrees]C; 3) cooled to RT; 4) uncompressed at RT; 5) heated to 150[degrees]C; and 6) cooled to RT. It shows the combinations of physical relaxation and thermal expansion effects over a range of compressions and thermal exposure that bring the sample back to its original condition with its initial dimensions at no load. Within this test sequence, samples can be measured at discrete intervals to provide CSR and compression set values. The larger points on the blue line are the defined conditions or measured values used to determine CSR % RSF, while the large points on the red line are used to define compression set.

[FIGURE 10 OMITTED]

Chemical relaxation--configuration comparisons

This same relaxation test could be run at higher temperatures for longer periods of time to provide an aged response that results in chemical relaxation and permanent set. By performing both of these tests, one can delineate between the physical and chemical relaxation. Of interest here was to see how these two types of relaxation would affect the compression set values, based on specific test procedures. In addition to this, the test was intended to see if, or how much, the test configuration (sample shape and size) might affect these results. In a previous study, the results showed that the CSR % RSF was strongly related to a sample's shape factor and to a less extent on its diameter (ref. 4). The samples with higher shape factors and larger diameters showed poorer % RSF.

In this study, a similar range of samples was measured for CSR response, but was evaluated at 150[degrees]C. Some of the samples were also removed from the CSR test at defined time intervals to allow them to be measured for both initial and permanent set. The samples were checked for sealing force at room temperature (RT) and then removed from the jigs. They were then allowed to recover for about 30 minutes at RT, before being measured for thickness. These values were used to determine the value defined as initial set. The samples were then heated to 150[degrees]C for one hour in an uncompressed state and cooled back to RT. The thicknesses of the samples were measured again to determine their permanent set. The interest in this test was to determine the relationship between CSR % RSF and initial and permanent set.

The aged results over 168, 336 and 504 hours.showed that there was a general relationship between CSR and initial set, but not an exact correlation, with most of the values clustered in a certain range shown in figure 11. The results show that % RSF decreases as compression set increases, but for a range of samples that showed a more uniform response, there was not the same direct correlation between CSR % RSF and compression set (where each compression set value corresponded to a specific CSR % RSF value). Some of the samples with high shape factors did show higher compression sets, as well as lower CSR % RSF. This corresponded with the previous study (ref. 4). Since 150[degrees]C was not a severe aging condition for this material, increasing the temperature and time intervals for the test might provide a better response separation. Figure 12 shows a comparison of one of the samples for its initial and permanent set. With this sample and the others, there is a noticeable difference between the two values, but they all had a similar parallel response, with permanent set values less than 15% after 504 hours.

[FIGURES 11-12 OMITTED]

Chemical relaxation--compression set procedure comparisons

In performing this test, 13 mm D x 6.3 mm H x 0.5 SF disks were evaluated for both CSR and compression set. Different sets of the same samples were run in parallel for both CSR and compression set. The samples used were based on FE-5620 FKM and were heat aged at 200[degrees]C and measured at intervals of 24, 70 and 168 hrs. This temperature was used to try to generate more chemical relaxation in these shorter time intervals. The objective of this work was to see how much initial and permanent compression set values varied with the different procedures and how they compared to the aged CSR response. Three different compression set procedures were used, where the temperature at which the samples were removed from the constraining test fixtures was varied. They were removed: 1) At the elevated temperature as is required by ASTM D395 Method B; 2) after they had cooled to room temperature; and 3) upon cooling to 0[degrees]C. In the first two procedures, the sample heights were measured at room temperature, while the third was initially kept at 0[degrees]C and measured, and then allowed to warm to room temperature, where it was measured again. These values were used to define initial set of the samples for each of the conditions. These samples were then heated to 200[degrees]C for one hour and cooled back to RT to measure their permanent set. The comparative sealing force values for CSR were only measured at room temperature.

The results of this testing show that the samples removed at the lower temperatures gave higher compression set values, as shown in figure 13. It also showed that after the samples had been heated back up to the elevated temperatures, all of the samples showed the same permanent set. These permanent set values corresponded to the initial set values of the samples taken out of the fixtures at the elevated temperature. This reflects on the fact that at lower temperatures, the polymers are more viscous and cannot easily recover to an equilibrium position over the time frame of interest. Once heated to elevated temperature, though, all of the samples recovered to the same height as the samples taken out of the fixture at the elevated temperature. This shows that the difference in the responses for each of the procedures is a result of the viscous response, and not a result of material degradation or chemical relaxation. When the initial set values are compared to the CSR % RSF in figure 14, they show a near parallel linear response for the different removal temperatures, except for one value in the low temperature test.

[FIGURES 13-14 OMITTED]

As a result of this testing, it seems there may be no significant difference in the quality or importance of the compression set numbers determined by each of the compression set procedures, except that one test might be easier to perform than the others. Since the responses are uniform and parallel for the different procedures, one only needs to define acceptable values for a particular procedure to provide an indication of the relative material response. If only comparative values are of interest, then any value or procedure might be used to define limits for acceptability. However, if one wanted to use a compression set value to predict the performance of a material in an application, there might be some question as to what value and procedure should be used.

In comparison, although sealing force was measured at room temperature, it could be measured at any temperature. A recent paper suggested that fluid leakage past a seal is possible when the fluid pressure exceeds the contact pressure exerted by the seal (ref. 5). If this is the case, then knowing the sealing force at a specific temperature, after a defined environmental exposure, should allow one to predict when this condition might occur. This is a more direct way to match the potential performance in the application to a measured material property.

Chemical relaxation--aging mechanism--diffusion limited oxidation

In this study, as well as in a previous one (ref. 4), the data seemed to show that the FE-5620 FKM samples having the same shape factor had about the same rate of loss of sealing force. It also appeared that the dimensional recovery after aging was uniform over the width of the compressed surfaces. This suggests that the degradation of this material is uniform in nature and may result from making or breaking bonds uniformly throughout the polymer matrix. In some materials, this has been shown not to be the case, and degradation may vary over the exposed cross-section as a result of diffusion-limited oxidation (ref. 6). In these cases, the CSR % RSF response might be expected to vary as a function of surface area to volume ratios of the sample. To evaluate this, samples of FKM and HNBR were molded that had the same shape factor, but different diameters and resulting surface to volume ratios. The samples were then heat aged and measured for sealing force and their retention as a function of time. The HNBR compound was not specially formulated to optimize its performance, but only to evaluate this possible effect, if it occurred. The compound was made orange in color, since with a previous orange HNBR compound it had been observed to turn black at the interfaces exposed closest to the air. The previous sample also showed that its dimensional recovery was non-uniform across the compressed face, increasing progressively more toward the center of the sample. The samples were evaluated at 150[degrees]C.

The results in figure 15 show that the FKM samples had a similar % RSF over the time frame, regardless of the sample diameter, while the HNBR samples showed that the samples with the higher air-exposed surface to volume ratio showed poorer CSR % RSF. Pictures of the HNBR samples in figure 16 show a difference in the color of the internal compressed contact area of HNBR closest to the air and that, as the diameter of the sample increases, the ratio of the black area as a percent of the total volume decreases. It was also seen that, as the outside surface changed color and hardened, there was a non-uniform recovery of the sample, with the interior compressed contact area recovering closer to its original position.

[FIGURE 15 OMITTED]

In trying to define a test procedure for predicting performance in an application, it is important to know and understand when a measured response is a material response, or when it may be affected by the configuration, and when the specific configuration effects can vary with different materials. If one used large buttons to predict the performance of seals with a much smaller exposed cross-sectional width, then the testing might over-estimate the performance of this HNBR compound in an application.

Conclusion

A number of elastomer principles was discussed in this article and a significant amount of test data was generated to provide some background for characterizing elastomers in compression for sealing applications. The following ideas were presented and evaluated:

* The load-deflection responses that are seen with elastomers in compression are a reflection of both its elastic and viscous behavior.

* The instantaneous stress-strain response in compression is the same for a given material over a reasonable range of sample sizes and strains when performed in a lubricated frictionless condition at the same rate of strain.

* Deviations from a uniform stress-strain response can be found at extremes in configurations, which can include high and low shape factors, high strains and conditions of higher friction or non-uniform deformation.

* The elastomer tested showed similar physical relaxation responses over the full range of sample configurations.

* There is a direct relationship between CSR and compression set. As compression set increases, the CSR % RSF decreases. Compression set values will vary with the specific test method used. Samples removed from test fixtures at lower temperature will produce higher compression set values, but these values are a reflection of the test and not a result of any difference in material degradation.

* Compression set values are only useful as a relative gauge of capability based on a specific test used. It has been suggested that sealing force and its retention at a given temperature can be a more direct predictor of sealing performance and service life.

* The FKM samples tested show a uniform aging response, with the % RSF of samples of the same shape factor being similar.

* The HNBR samples showed a non-uniform degradation mode, with air-exposed surfaces being prone to diffusion-limited oxidation. This shows that the actual sealing response of HNBR is a function of both its material response and the test configuration.

References

(1.) ASTM D 395, Standard Test Method for Rubber Property--Compression Set, Method B, Compression Set Under Constant Deflection in Air.

(2.) ASTM D 6147, Test Method for Vulcanized Rubber and Thermoplastic Elastomer--Determination of Force Decay (Stress Relaxation) in Compression.

(3.) R. Brink, D. Czernik and L. Horve, Handbook of Fluid Sealing, McGraw-Hill, New York, 1993, p. 26.

(4.) P. Tuckner, "The effect of configuration on sealing force measurement and compression stress relaxation response," SAE technical paper 2003-01-0946 (2003).

(5.) M. Otsuka, T. Okamura, N. Suetsugu, T. Ohta and S. Ono, "A new concept of static rubber gasket for sealing rough surface," SAE technical paper 2003-01-0485 (2003).

(6.) K. Gillen, M. Keenan and J. Wise, "New method for predicting lifetime of seals from compression stress relaxation experiments," Die Angewande Makromolekulare Chemie 261/262 (1998), 82-92 (Nr. 4,619).

by Paul F. Tuckner, Dyneon LLC
COPYRIGHT 2006 Lippincott & Peto, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2006, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Tuckner, Paul F.
Publication:Rubber World
Date:Jul 1, 2006
Words:5360
Previous Article:Characterization and integrity testing utilizing corona beam technology.
Next Article:New machinery suppliers.


Related Articles
Long term sealing ability of butyl o-rings.
Silicone making autos quieter than ever.
Stress relaxation tester. (Suppliers Showcase: Instruments).
Continuous compressive stress relaxation of elastomers used in engine sealing applications.
TPE masterbatch can be combined with SBC.
Factors influencing low temperature performance of EPDM compounds.
Rubber testing.
Polyacrylate elastomers.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters