Influence Of Disturbance On Measurement Precision Using Ad Plug

Transcript

Influence of Disturbance on Measurement Precision Using AD Plug-in Boards Vladimír Haasz, František Pištínek Abstract A disturbing field in a PC case is one of the causes which influence the actual number of effective bits of the AD plug-in board used. Three sorts of tests were executed to examine this influence - measurement of an internal magnetic field inside PCs, measurement of the number of effective bits of the tested AD board in the defined magnetic field, and the measurement of an actual influence of a disturbing field in the PCs mentioned above on the number of effective bits. Finished experiments showed, that a disturbing field can be a significant cause of a decrease in the number of effective bits in measuring systems based on PC plug-in boards. It concerns, above all, 16-bit AD boards and 12-bit AD boards using a higher gain. Keywords: AD plug-in board, number of effective bits, disturbance, disturbing field. 1. Introduction For the measurement of dynamic signals, in both laboratory and industrial applications, systems using various types of AD-modules with standard interfaces are increasingly used instead of stand-alone instruments (transient memories). However, it is very difficult to estimate real measurement errors for such systems. Undefined (or insufficiently defined) parameters for the dynamic quality of AD-modules, and a dependence on internal EMC conditions complicates their usage in many cases (see [2], [3]). For this reason, reference methods should be proposed for the determination of dynamic parameters of AD-modules with standard interfaces regarding real EMC conditions. To analyse the influence of a disturbing field inside the PC case on the increase in the real number of effective bits of AD plug-in board, three types of tests were executed: - measurement of an internal magnetic field inside the PC (in defined points), - measurement of a decrease in the number of effective bits of the tested plug-in board in a defined magnetic field - measurement of an influence of an internal disturbing field on the decrease in the number of effective bits using the same PC and the same plug-in board mentioned above. 2. Measurement of an internal magnetic field inside a PC case A special plug-in board (full size) made from insulated material was used to define primary 8 points (1 - 8) for disturbing field measurement. The measuring points were placed 2 cm from board margins in an equidistant distance (see Fig. 1.). This plug-in board was placed in all ISA slots of the PC used, and the intensity of a magnetic field was measured (the component perpendicular to the board). The close-field probe with an amplifier was used for this purpose (HP 11941A for frequency to 30 MHz and HP 11940A for the frequency range above 30 MHz). The output voltage was measured using a digitising oscilloscope HP 54645D (termination 50 Ω must be used for this purpose). The intensity HD of a disturbing field was calculated using the probe constant and the frequency fD of the fundamental component of the measured voltage waveform. The PCs with only a graphical board inside, and also the PCs with further plug-in boards were examined. The 486DX2 and Pentium 133 MHz were tested. The first measurements showed, that there are two significant sources of disturbance. The motherboard, which radiates the component with fundamental frequency 33 MHz above all, and some plug-in boards, which radiate a disturbing magnetic field with other fundamental frequencies. The value of the intensity HD of the disturbing field is practically independent of the used slot in the near empty PC (with graphical board in the most distance slot from the tested one). From all tested plug-in boards (I/O boards, AD boards, graphical boards, network boards etc.) only network boards, and a graphical board radiate a significant disturbing field. The tested boards radiated on frequency fD = 49 kHz or fD = 60 kHz (network boards, where DC/DC converter is a main source of disturbance), or fD = 250 kHz (a graphical board). The mentioned values of frequency depend on the type of boards, of course, but these values determined approximately a 2 range of frequencies of internal disturbing fields. A disturbing voltage induced in the input circuit is proportional to the product of the frequency fD and the intensity of a magnetic field HD (in the ideal case). The shielding and further influences partially distort this dependency, but nevertheless this product can be a characteristic parameter of the disturbing field. The results measured in the near empty PC, and also in the neighbouring slots of boards, which are a significant source of the disturbing field, are published in Table 1. From the measured results, it follows, that the spatial distribution in the area of the neighbouring slots is very non-homogenous. To make this measurement more conclusive, the measuring points 9 - 13 were added in the area, where the input circuit is mostly placed in AD boards (see Fig. 1). The results mentioned below validate the presumption, that the disturbing signal with frequency 33 MHz does not disturb most of the multifunction plug-in boards, because their cut off frequency is many times lower (e.g. for the used plug-in board AT-MIO-16E2 it is 1 MHz),. 3. Measurement in the reference AC magnetic field outside of PC 3.1. Method of measurement used The Helmholz coils with a diameter of d = 50 cm were used for generation of the defined AC magnetic field. It warrants a homogenous magnetic field in a central space (with a diameter of d´= 20 cm and with non-homogenity lower then 5 %) of coils. The tested multifunction plug-in board was placed in the ISA extension box (external slots without shielding of used AD plug-in boards) out of a PC case. The used ISA extension box was connected to the PC with a 1 m longshielded cable. The cable contains both data and supply wires. This box with a tested plug-in board was inserted into the Helmholz coils so, that the anolog part of tested board including the AD converter was placed in the space of the homogenous field - see Fig. 2. When the extension box is inserted in the area of the Helmholz coils, the distribution of the magnetic field is deformed, of course. However, it can be neglected for this purpose (the measurement validates the changes smaller then 5 % in the area of a tested board). The generated field with the comparable value of fD.HD to the value published in Table 1 were applied for the following tests. 3 Two methods can be used to determine the decrease in number of effective bits due to a disturbing field. It is possible to presuppose, that for the low signal frequency the results using curve fitting method (or FFT methods - see [1]) and the results using a measuring of output should be comparable. At first, both methods mentioned above were tested to verify this presumption. In the first case, the curve fitting method and the input testing signal with frequency 1 kHz was used for the determination of the decrease in the number of effective bits due to the defined magnetic field. In the second case, the method using a measuring of the output noise was examined. The input of a tested AD plug-in board was short-circuited using a resistor with the resistance equal to the output resistance of a generator used for the tests mentioned above. Also a grounding was realised in the same way. Using this method, the calculated number of effective bits nef´ corresponds to an ideal AD channel which influences only the disturbance and an internal noise. The number of effective bits can be calculated from the signal to noise ratio S = 6.02 nef/ + 1.76 N (1) In the case mentioned above, where no input signal is connected and the value S/N is related to the input range UR (for bipolar input is the corresponding RMS value U R / 2 2 ), the quantisation noise must be also taken into account. The value S/N can be calculated for this purpose using the formula UR S = 20 log N 2 2 U q2 + U N2 (2) where UR is the range of ADC used, UN is the RMS value of the measured output noise, and Uq is the RMS value of the quantisation noise. It follows from equation (1) and (2), that the number of effective bits can be calculated using the formula nef/ = UR 1 1.76 20 log − 6.02 2 2 U q2 + U N2 6.02 (3) The results are published in Table 2. They show that both methods are usable in this case, the first one also takes into account the non-linearity of the whole AD channel, the second one takes into account only the influence of a disturbance and an internal noise. 4 Differences form 0.5 effective bit for a low gain to 1 effective bit for a high gain correspond to it. Ideal EMC conditions are very difficult to achieve. It is, in principle, possible to insert a tested AD module in a shielded chamber and to supply them using an ideal DC voltage source. However, the technical realisation is very complicated. Therefore, the dependence of the decrease in the number of effective bits on the disturbance must be determined as the difference of two measurements - with and without the defined disturbance. The decrease in the number of effective bits due to the defined disturbance can then be determined as 2 U q2 + U D1 10 ∆n = log 2 6.02 U q2 + U D2 / ef (4) where UD1 is the RMS value of the disturbing component in the output signal without defined disturbance (the state near ideal EMC conditions - only an internal noise applies), and UD2 is the RMS value of the disturbing component in the output signal, when the defined disturbance influences the AD board. 3.2. Testing of an influence of a disturbing field on the decrease in the number of effective bits for several types of AD boards 11 types of boards were tested. They were marked from A to K for their identification. The influence of a disturbing magnetic field dominates all types of boards with low voltage (high gain) ranges (A, B, C, D, G; K). In the case of the boards with low gain rages only (C, E, H) the other source of disturbance dominates (see below). The measured decreases in the number of effective bits due to the defined disturbing field are shown in the Fig. 3 for the gain G = 10 (a) and the gain G = 100 (b). The influence of an input connection on the decrease in the number of effective bits was also investigated. The results are shown in Table 3 for the AD board AT-MIO-16E2. Both basic connections (single ended and differential) were used (Fig. 4a, b). In the third case the GND pin of the analog input of the tested plug-in board was grounded using the separate wire (Fig. 4c). 5 It was also investigated if an influence of a disturbing field depends on the channel used. The results measured for the same AD boars are published also in Table 3 (only channels with the lowest and the highest disturbation). The results of the both tests mentioned above showed, that differences of a disturbance sensitivity using both different channels and different types of input connection are not significant. The measurement made for several frequencies of a disturbing field and for the product fD.HD in the range of 20 to 150 kHz.A/m confirmed the premise, that the output noise arisen due to a disturbing field is approximately proportional dependent on the product of the frequency and the intensity of a magnetic field fD.HD (see Fig. 5), and a frequency dependence is not significant in a low frequency range for the board without shielding. The differences in the frequency range of 10 to 30 kHz are lower then 0.4 bit in all the cases mentioned above, and a disturbance effect slowly decreases with frequency by several boards. It can be caused due to eddy currents induced in large grounding areas of the printed circuit board, which are used here. The dependence of the number of effective bits on the frequency fD of the disturbing field is shown in Fig. 6. This dependence is more complicated in the case of the AD board AT-AI-16XE-10, where the whole input part is shielded. The shielding case is made from a ferromagnetic metal plate. The measured results showed, that in this case the shielding effect decrease with frequency for the constant product of the frequency and the intensity of a magnetic field fD.HD (see Fig. 7). According to the results analysis, it can be caused by a change of the permeability of the shielding case due to the change of the intensity of a magnetic field. The detailed analysis of this problem is more complicated, because also a penetration depth and a rise of eddy currents in the shielding case are dependend on frequency. There are also, however, other sources of disturbances, which can become significant in some cases. Quite different results were obtained e.g. for low-cost AD boards with low gain ranges only. The relatively high disturbing component (especially for gain G = 8) in the output signal is practically independent of the disturbing field for these boards. Hence it follows that the disturbing field is not the major source of the disturbing component in the output signal in this case (see Table 4). 6 A reason fir this, can be a bad filtering of the supply voltage. It corresponds also to the fact that the disturbing effect (the decrease in the number of effective bits) is in two cases higher for the measurement in the ISA extended box. The plug-in board is in this case supplied by the PC using about a 1 meter long connection, and it can increase the output impedance of the power voltage source, so to make worse the quality of the supply voltage for the plug-in board. The problem concerning the disturbance propagating over supply lines will be investigate separate in the near future. 4. Measurement of an influence of internal disturbing field inside PC case on the decrease in the number of effective bits To test a disturbing effect inside PCs, a tested AD board was placed in the same slots close to the network interface boards, where the disturbing field was measured in par. 2. The measured results are published in Table 5. Let´s presuppose, that only a disturbing field in the space of the input circuit can significantly influence the number of effective bits. The area of measuring points 1, 9, and 2 corresponds to the space of the input circuit of tested AD boards. The measured values „left“ of fD.HD e.g. for network board A are in these points in the range of 18 to 29 kHz.A/m (see Table 1). The corresponding decrease of the number of effective bits should be about 2 bits, but the measured values were for tested boards several times smaller (see Table 5). It shows, that an estimation of the disturbing effect is very difficult in the case of small local sources of a disturbing field. It is also necessary to mention here that a difference between the disturbing effect in the modelled and in the real internal disturbing field can be quite different for another AD board thanks to a different layout used. 5. Conclusion The measurements made in the defined disturbing field validated most of the tested boards, that a disturbing field can be a significant cause of a decrease in the number of effective bits. The results further validated the premise that this decrease corresponds to the product of frequency and intensity of a disturbing field, but the dependence on the frequency of the disturbing field is not significant for boards without shielding. Only a 7 slow decrease in a disturbing effect with frequency was found in some cases. It is probably caused due to eddy currents induced in a large grounding area of a printed circuit board, which is used. The EMC conditions concerning disturbing fields can be easily defined for these types of AD boards. It was also validated that this dependence is more complicated in the case of high resolution AD boards, which use a shielding case made from a ferromagnetic metal plate. It will be necessary to determine the dependence of the shielding effect on the frequency and the intensity of a disturbing field. The new Helmholz coils with a frequency range to 300 kHz were design and realised for this purpose, and the next measurement in this frequency range will follow. A great difference between the decrease in the number of effective bits estimated from measured parameters of a disturbing field radiated from a board in the neighbouring slot and the real decrease is also very interesting. It can be probably caused due to the very non-homogenous field radiated by the used network boards. It shows, that an estimation of the disturbing effect is very difficult in the case of small local sources of a disturbing field. Therefore, it is the best solution to measure the output noise for a concrete arrangement of the system. However, it is possible to warrant the lower disturbing effect than corresponds to the results of measurements in a homogenous field in any cases. However, the conclusions mentioned above are not valid in the case, when the main source of an output noise is not a disturbing field, but e.g. a bad filtering of supply voltages etc. The influence of a disturbance propagating over supply lines should be investigated in detail in the near future. It will make it possible (together with the results mentioned above) to define the EMC conditions inside PCs, for which the dynamic parameters of tested AD boards will be warranted. Acknowledgement The results of bilateral Czech-German project „Analog-to-digital converters and A/D modules by electromagnetic influence“ supported by the Czech Ministry of Education and the first theoretical results of the project No. 101/98/1508 „New methods 8 for testing of systems for dynamic measurement“ which is supported by the Grant Agency of the Czech Republic were applied in this paper. References [1] Roztocil, J., Mascio, F., Pokorny, M.: Practical Aspects of A/D Plug-in Boards Dynamic Testing. XIV IMEKO World Congress, Tampere 1997, pp. 274-276 [2] Haasz, V., Pištínek A.: Influence of a Disturbing EM Field on the Real Number of Effective Bits of PC Plug-in Boards. XIV IMEKO World Congress - International Workshop on ADC Modelling, Tampere 1997, pp. 183-186 [3] Pokorný, M., Roztoèil, J.: Influence of Magnetic Field on Measurements Using A/D Plug-in Boards. IMTC´97, Ottawa 1997, pp. 302-306 9 1 10 9 11 2 12 3 4 5 13 6 7 8 Fig. 1. Helmholz coils Plug-in board PC HD Fig. 2. 10 decrease in the number of eff. bits 0 -1 D1 (diff) -2 D2 (SE) B1 G -3 A K (G=16) -4 0 50 100 f.H (kHz.A/m) 150 a) G = 10 decrease in the number of eff. bits Fig. 3a) 0 D1 (diff) -1 D2 (SE) B1 -2 G A -3 -4 -5 -6 0 50 100 f.H (kHz.A/m) b) G = 100 Fig. 3b) 11 150 1 1 2 50 Ω 1 2 50 Ω 2 50 Ω 16 Rp Rp 8 8 1´ 2´ 1´ 2´ Rp Rp COM GND a) 8´ 8´ GND GND b) c) Fig. 4 Output noise (mV) 70 D1 B1 A 60 50 40 30 20 10 0 0 50 100 Disturbing field (kHz.A/m) Fig. 5 12 150 Number of effectve bits D1 8.5 B1 A (G=500) 8 7.5 7 6.5 6 10 15 20 f (kHz) 25 30 Number of effective bits Fig. 6 20 kHz.A/m 12 50 kHz.A/m 11 150 kHz.A/m 10 9 8 7 10 15 20 f (kHz) Fig. 7 13 25 30 Captions of figures. Fig. 1. Measuring points on a special PC plug-in board Fig. 2. The tested AD board in an extension box placed in the Helmholz coils Fig. 3. The measured decreases in the number of effective bits due to the defined disturbing field Fig. 4. The input connections of a tested AD board Fig. 5. The dependence of the output noise on the product of frequency and intensity of a magnetic field fD.HD for the gain G = 100 Fig. 6. The dependence of the number of effective bits on the frequency of a disturbing field for fD.HD = 100 kHz.A/m and G = 100 (A: G = 500) Fig. 7. The dependence of the number of effective bits on the frequency of a disturbing field for the board AT-AI-16XE-10 14 Tables: Table 1. The disturbing magnetic field in a PC behind the network interface boards (A, B), and the graphical board (G), the values of fD.HD (kHz.A/m) for the most significant components Meas. point 1 2 3 4 5 6 7 8 9 1 0 Empty PC 7 5 5 5 5 6 1 7 6 3 1 2 1 1 5 1 2 6 6 3 Board A left 2 9 1 6 5 8 Board A 1 1 right 5 4 Board B left 1 8 1 Board B 5 9 right Board G left 1 6 0 Board G 1 1 right 8 1 5 4 4 4 5 4 3 3 4 4 3 34 3 8 8 7 4 7 1 6 5 0 5 1 0 2 0 8 3 6 0 1 5 1 3 7 7 1 4 3 6 1 1 8 - 1 4 - 7 5 1 9 m fD 3 ax. (kHz) 5 13 3300 0 5 34 60 2 6 2 38 60 6 2 87 49 6 3 47 49 5 - 19 250 2 2 8 3 4 2 46 1 6 2 2 1 2 0 8 6 4 1 1 3 3 3 6 5 9 - 1 - - - - 50 250 1 Table 2. The comparison of the measured decrease in the number of effective bits using both methods (fD.HD = 100 kHz.A/m, 12-bit AD board) Gain ↓ 1 2 5 10 20 50 100 curve fitting method without dist. with dist. field field 11.4 10.9 11.5 11.3 11.4 9.5 11.3 9.0 11.0 7.8 10.1 6.4 8.5 5.4 15 measurement of output noise without dist. with dist. field field 12.0 11.5 12.0 11.8 12.0 11.0 12.0 9.9 11.7 8.8 10.8 7.4 10.0 6.5 Table 3. The influence of the input connection and the influence in the channel used on the decrease of the number of effective bits ∆nef due to the defined disturbing magnetic field (fD.HD = 100 kHz.A/m, fD = 19 kHz) Influence of the input connection Gain ↓ 1 10 100 SE 0.4 0.4 4.2 DIF 0.4 0.9 3.9 DIF, GND 0.2 0.8 3.8 Influence of the channel used min. ∆nef max. ∆nef 0.3 0.4 0.9 1.0 3.6 3.9 Table 4. The influence of a disturbing magnetic field on the number of effective bits and on the disturbing component in the output voltage Meas. value of Set gain nef in „near and tested empty“ PC (curve fitting method) G =8 UD / UR (mV/V) UR - range of ADC (V), UD - RMS value of disturb. component (mV) fDHD = 100 fDHD = fDHD ≈ 0 (kHz.A/m) 100 (kHz.A/m) 12 0.02 0.11 11.9 1.26 1.47 fDHD ≈ 0 fDHD ≈ 0 E H 11.9 13.3 12 12.2 I E H 10.5 - 10.2 10.3 9.4 10.2 10.1 9.3 4.8 4.29 8.59 4.8 5.07 9.01 I - 9.7 9.7 6.6 6.8 boards G =1 Meas. value of n´ef in ISA ext. box, (calculation from disturbing component in the output signal according eq. (1)) 16 Table 5. The comparison of the results of the test inside a PC case with and without the network board (board A - left, G = 100), and in the similar homogenous field. AD board G D1 B1 A PC without network board 13.2 11.3 10.9 11.6 PC with network board A 12.5 11.0 10.9 11.5 17 Ref. dist. field 20 kHz.A/m 10.3 9.5 8.6 10.8