AS8500 - Data Sheet austriamicrosystems
Revision 1.2, 08-Junel-06 www.austriamicrosystems.com Page 35 of
41
9 General application hints
Since the AS8500 is optimised for low voltage applications extreme care should be taken that the signal is not disturbed by influences like bad ground
reference, external noise pick-up, thermal EMFs generated at the transition of different materials or ground loops. The influence of these error
sources can be quite high and they may completely shadow the excellent properties of the device if not handled properly. The following sections are
supposed to supply additional informations to the design engineer how to get around some of these problems.
9.1 Ground connection, analog common
The analog common terminal where all voltages are referring to is RSHL. All ground lines of the external circuitry of VBAT, ETS and ETR as well as
the voltage sense line of the low ohmic current sensing resistor should be connected to each other in a star like ground point. It is recommended that
this point is as close as possible situated to the low side sense terminal of the current sensing resistor. It should also be connected to the VSS and
VSSD terminal, but the return line of both must leave this point separately. Also the power decoupling capacitors should be connected to the analog
common.
To give an example of the magnitude of possible errors consider that the ground return of the power supply is not connected properly and 5 mm of a
copper track 35µm thick and 0.1 mm wide are within the measuring circuit with a current flow of 5 mA. This will result in an offset of 120 µV which is
more than 500 times higher than the typical offset of the ASSP. In addition the current fluctuations will act as an extra noise voltage which is also way
above that of the device itself.
9.2 Thermal EMF
another major source of error for low level measurements are thermal voltages (electromotive force, thermal EMF) or Seebeck voltages which are
principally produced by any junction of two dissimilar materials. On PC-boards pairs of dissimilar materials may consist of the copper tracks and the
solder, the leads of different components or different materials used in the construction within the components. Any temperature difference between
two connection produces a voltage which is superimposed to the measuring voltage.
A number of strategies are known to detect or minimise their influence on the measuring result:
- in cases were a current has to be measured directly or a current is to be used to activate a resistive sensor (like
Ohm-meter or temperature measurement with RTDs, NTC or PTC) a switch in the circuit could be used to interrupt or invert the current thus
producing a current change dI. In the difference of the two voltage states dU the EMFs as well as the Offset voltages of the amplifier are fully
eliminated. For resistance measurements this method is known as ‘true Ohm’ measurement.
- in applications were this is not possible and the problematic device (i.e. the input resistor of an amplifier) can be located it may help to place a
dummy device of the same type in the circuit as close and thermally connected as good as possible to compensate the influence of the first
one.
- Since the thermal EMFs are proportional to the temperature difference it is important to maintain a homogeneous temperature distribution in
the vicinity of the sensitive area. This is possible by keeping this area as small as possible, by avoiding any heat sources nearby or by
increasing the heat conductivity of the substrate, i.e. wide and thick copper tracks, multilayer board or even metal substrate.
- The best solution of all however is to avoid the thermal EMFs by using only components which are matched to the copper world which means
that their thermo-electrical power against copper is zero. This is specially important for current measurements in the range of 10- 1000A. In this
case the resistance value has to be very low (down to 100µOhms) to limit the measuring power and avoid an overheating of the sensing
resistor. On the other hand the voltages to be detected are extremely low if a high resolution is required. If for instance a current of 10 mA has
to be measured with a 100µOhm resistor, the resolution of the measuring system must be better than 1µV and the error voltages due to
thermal EMFs must be below this limit. Quite often people are trying to use the well known Konstantan (CuNi44) for current sensing resistors.
This is a bad choice since the thermal EMF versus copper is very high.
With –40µ V/deg already a temperature difference of
2.5 K is enough to produce an error which is 100 times lager than the required resolution. Or vice versa a temperature fluctuation of only 1/100
K produces a ‘thermal noise’ which is equivalent to the required resolution.
With such materials and high currents of 10A and above the other thermoelectric effect, the so called Peltier-effect, can also play an important
role. Under current flow this effect generates heat in one junction and destroys the same amount of heat in the other junction. The amount of
heat is proportional to the current and its direction. The result is a temperature difference which in turn generates a thermal EMF proportional to
it. Finally this means that such a resistor produces its own error voltage and it is never possible to measure better than 1-2% with such badly
matched materials. The precision resistance materials Manganin, Zeranin and Isaohm are perfectly matched to the copper world and resistors
made from these materials can achieve the high quality that is necessary for low
level measurements and high resolution.
9.3 Noise considerations
for every low level measuring system it is essential to know the origin of noise and to accept the limitations given by it. Three major sources of noise
have to be considered. The input voltage noise and the input current noise of the amplifier and the thermal noise (Johnson noise) of resistors in the
external circuitry around the amplifier. Due to the fact that these three sources are not correlated they can be added in the well known square root
equation.
In most applications the input resistor or input divider is low ohmic (i.e. below 10 kOhms) which mean that the noise voltage produced by the input
current noise is negligible compared to the input voltage noise. The input noise density (En) of the AS8500 is with only 35 nV/sqr(Hz) extremely low.
This could be achieved with a special internal analog and digital chopper circuitry which eliminates the CMOS typical 1/f-noise completely. Even
though the overall noise will be dominated by the input amplifier as long as the external resistors are below 10 kOhm.