Table 1 lists the experimental methods used in dynamic mechanical testing. Of the experiments summarised below thermal scan mode (method 5) is the most commonly used technique, especially in analytical laboratories. Here the typical use is to look for differences in materials’ batches, thermal history, different grades, etc. The stepped isotherm experiment of method 6 is mainly used in studies involving detailed mechanical property determination for either structural or vibrational analysis. Such tests are only normally carried out in R&D environments. Methods 1, 2 and 3 can be regarded as "quick" experiments, taking at the most 30 minutes and more typically a few minutes. Method 4 is application specific.

It is the thermal scan method (5), usually at a constant frequency of 1 Hz, that is the most commonly used experiment. All relevant dynamic mechanical data are obtained from a single thermal scan taking only 1 to 2 hours. The frequency is held at a constant setting (or possibly several discrete settings) and the temperature is scanned at a constant rate, usually from low to high This technique is suited to small specimens. Accuracy will be better for smaller, thinner samples that achieve thermal equilibrium quickly and generally for slower heating rates for the same reason. In comparative studies, temperature errors can be tolerated, since any lag will be the same for similar samples.

In the stepped isotherm method (6) a whole series of frequencies is scanned at a constant temperature. The temperature will then be set to a new value and the scan repeated. This will be carried out until the measurements have bracketed the range of interest for the sample under test. This method generates significantly more data than any other experiments. It can be regarded as the classical experiment for viscoelastic property determination. Such results are usually more accurate, but experiments may take several days.

Simple materials, such as PMMA and PC which are amorphous, will exhibit little strain dependent behaviour, provided that glassy strains are kept below about 0.5%. Certain filled materials, especially carbon filled rubber will, however, demonstrate significant mechanical property changes, not only as a function of strain and temperature, but indeed as a function of recent sample history. Specific measures have been developed for dealing with such samples. As a general rule, an unknown material should be evaluated over a range of strain both above and below the glass transition temperature. It is most likely that little or no dependence will be found, but this should be confirmed before final test conditions are decided.

The most important property that a dynamic mechanical analyser measures is stiffness. For a good instrument, with uniform, accurately machined samples that are within the instrument’s measurement range, accuracy of better than ± 5% should be attained. This is best tested using 3point-bending geometry and steel bar samples. The first test should use a sample whose stiffness falls at or slightly above the instrument’s lower stiffness limit. After calibrating the instrument a value of 200GPa should be obtained for mild steel (it is better to use ordinary steel, as opposed to stainless steel, as the modulus of the latter can change markedly with varying alloying additions – steel is more consistent). Secondly a thicker steel bar should be used, at or slightly in excess of the instrument’s upper stiffness limit. If the same value is obtained all is well. If a lower value is obtained, the difference is probably explained by the compliance of the instrument itself during measuring, which accounts for some of the measured displacement, yielding a lower apparent modulus value for the sample.

All other modes of geometry (excepting shear) will exhibit errors due to imperfect clamping of the sample. This can only be rectified by using clamps that are massively stiff in comparison to the sample. This is not practicable in dynamic testing instruments due to mass considerations. Generally speaking tension suffers the least error, compression the worst, with clamped bending somewhere in between. The error will depend on sample stiffness and any correction must take this into account during a test which typically involves a modulus change of up to x1000 times.

Any polymeric sample will take a finite time to reach thermal equilibrium and this is frequently commensurate with the timescale of the experiment. The most accurate measurements are made isothermally, as a soak period can be arranged, in which time the sample reaches thermal equilibrium. Poorly designed ovens can exacerbate the problem.

If accurate temperatures are required for measured transitions a simple test is to make measurements on the same samples at different thermal scanning rates, say 1, 2, 3° Cmin^-1. If all transition temperatures, in heating experiments, are coincident then the sample is in thermal equilibrium in all tests. If higher transition temperatures are measured for the faster rates however, then the sample is lagging behind the measured temperature. For cooling experiments lower values will be seen at the faster cooling rates if the sample is lagging.

Some materials are more easily tested than others. The ones that represent the largest challenge are those whose modulus changes the most at the glass transition temperature, Tg. Therefore amorphous PMMA or PC make good test samples. Both of these polymers show a sharp Tg and a rubbery modulus of approximately 1MPa. Such samples afford the opportunity to test both the temperature accuracy and the instrument’s lower stiffness limit. Both glassy and rubbery moduli results from tension and bending experiments should be comparable (3 point-bending is not a good mode to use through the glass transition). Shear measurements (G’) made above Tg should be approximately one third of the E’_rubbery value obtained.

Generally tension mode is the first choice, unless sample sizes are too large for the instrument’s upper stiffness limit. One class of materials that may cause problems in tension is the polyolefines. As these have a predominately linear structure, they have a propensity to creep excessively, especially near the melting point, Tm. If excessive elongation occurs during the course of the experiment, clamped single-cantilever bending geometry is a better choice of sample geometry.

Natural rubber presents a formidable challenge to any DMA instrument. This is certainly one of the most difficult samples to run in a tensile experiment. The tan d peak at the glass transition is particularly narrow and the rubbery modulus can be very low. The best procedure is to load a sample (from a rubber glove, etc) in tension at room temperature and then cool to -100° C. At this point the sample will be glassy and the pretension should be applied. Now the dynamic measurements can commence and these should be made as fast as possible (at least one point per 10 seconds). The heating rate should not exceed 3° C/min and would be slower if more than one frequency was being measured.

A smooth curve should be obtained for tan d through the glass transition and the pretension should reduce rapidly to avoid any over-extension of the sample once it becomes rubbery. If these results are obtained the instrument is performing well.