30.07.2008
Load cells (LCs) with single-crystal mechanical springs (MSs) and sputtered-on strain gauge (SGs), due to the ideal elastic properties of the MSs, promise a low time-dependent behaviour and a high reproducibility of the measuring signals. These sensors would thus be optimally suited for use in mechatronic systems in order to digitally compensate for further influence factors such as temperature dependences and non-linearities. Investigations carried out on an MS made of silicon (Si) have confirmed the - by a factor of 100 lower - time-dependent behaviour compared to metallic MS. Presently, Si LCs are being investigated as to their linearity, hysteresis and the reproducibility of the measuring signals. First comparative measurements on metallic and Si load cells have confirmed the better properties of the Si LCs.
Interferometric investigations of the time-dependent deformation of a Si MS after load changes have shown that the relative mechanical aftereffects of the mechanical spring made of single-crystal silicon are smaller than 2·10-5 and thus lower by a factor of 100 than in the case of metallic MSs. In a next step, thin-film SGs were applied to the Si MSs using the method of sputter deposition (see Fig. 1). Contrary to glued SG foils, the direct bond created through sputtering significantly reduces the SG creep and increases reproducibility.
Figure 1: Silicon load cell with sputtered-on thin-film strain gauges
Currently, the properties of Si LCs are being investigated as to their linearity, hysteresis and reproducibility. For this purpose, measurement series at 20°C and increasing and decreasing load steps from 0 kg to 1 kg are carried out on Si LCs and, for comparative purposes, on metallic LCs of equal geometry. The analysis of the measurement results is done according to the international OIML recommendation R 60 for the testing and classification of LCs in legal metrology.
In Figure 2, the load cell error EWZ (blue line) and the reproducibility error ERep (red marks) are plotted for three metallic and one Si LC. The error limits of the respectively highest achievable accuracy class are represented by black lines.
Figure 2: Load cell error EWZ (blue line) and reproducibility error ERep (red marks) according to the OIML Recommendation R 60 as a function of the load L for various load cells; the black lines represent the error limits of the achieved accuracy class.
The aluminium LC with glued SGs achieves accuracy class C with 1000 intervals. A limiting criterion is the load cell error which is caused by the hysteresis of the LC signal. Theoretically, it is possible to achieve a higher accuracy class by compensation of hysteresis effects.
The steel LC with glued SGs achieves accuracy class C with 2500 intervals. Contrary to the aluminium LC, not the load cell error, but the reproducibility error is the limiting criterion. This means that, even by means of digital compensation, it would not be possible to achieve a higher accuracy class.
Compared to the steel LC with glued SGs, the steel LC with sputtered-on SGs shows a low reproducibility error and achieves accuracy class C with 8000 intervals. Similar to the aluminium LC, the load cell error is the limiting criterion; a higher accuracy class can be achieved by compensation of hysteresis effects.
The Si LC with sputtered-on SGs achieves accuracy class C with 10,000 intervals and thus the highest class among all investigated LCs. Here, the limiting criterion is the load cell error. Contrary to the metallic LCs, however, this error is not caused by hysteresis, but by non-linearities. A higher accuracy class can be achieved by compensation of the non-linearity.
Digital compensation of non-linearities is easier to achieve than the compensation of hysteresis effects since models with load-dependent history are necessary for hysteresis compensation. Due to the dominating hysteresis influence in metallic LCs, a compensation of non-linearity is not sensible. Digital compensation of non-linearity by using a linear model allows the LC to comply with the error limits of accuracy class B with 30,000 intervals. In a next step, the investigations will be extended to a temperature range from -10°C to 40°C.
Contact person:
Sascha Mäuselein, FB 1.1, AG 1.12, Email: sascha.maeuselein@ptb.de