On Sat, Jun 5, 2010 at 6:08 PM, YES NOPE9 wrote: > If I did the calculations correctly, it seems > that an Allegro A1323 linear hall effect sensor > ( HES ) (1) should work. =A0This part has an output of > 2.5 mv / G. =A0I am planning to measure a range > of 1 amp to 200 amps. =A0Should I make the > HES part of a magnetic ring around the > wire ? The ring would consist of two "C" sections > married to each other sandwiching the HES. > Would steel be a suitable material > given I am measuring DC to 120 Hz ? =A0Does the > steel have to be laminated ? =A0Or would a ferrite > ring be necessary ? > > After single point calibration , I would like to > see an accuracy of .1% . > > (1) ( costs $2 to $.60 depending on quantity. ) Hall-effect sensors exhibit significant drift over temperature and some of this drift has a temperature hysteresis (i.e., you can't compensate it out because it depends not only on the present value of the temperature but on the recent history of the temperature). For this reason alone, I highly doubt that you will be able to get anywhere near 0.1% over a range of 1 A to 200A. For comparison, I have a 100A clamp-on current probe, hall-effect based, which cost about $400 US and it is only accurate to about 3% over typical indoor temperature fluctuations. Also, while hall sensors themselves are very linear, I believe that they almost always have a built-in iron core to increase sensitivity. This causes nonlinearities and magnetic hysteresis. If you add an additional external iron piece, that will make these problems even worse, although it will likely improve your temperature sensitivity due to the higher signal level per Amp of wire current. Also, the external iron ring makes the sensor much less sensitive to relative positioning of the wire and sensor, as long as the wire runs through the ring. As someone else mentioned, higher-end current sensors typically perform a compensation feedback loop - they have an extra multi-turn coil around the iron ring and a circuit drives a current through this coil to force the output of the hall sensor to zero. By keeping the magnetic field in the iron ring near zero, hysteresis and non-linearities are compensated out. The measurement output is then proportional to the current in the multi-turn coil, not the hall sensor output. This technique is used in the A63xx series of current probes from Tektronix. These were several kilodollars new and still fetch at least $300 when in good shape on the used market. Granted, they can measure to over 1MHz BW as well. Most high accuracy current measurement systems use current shunts with Kelvin terminals. These are calibrated resistors of very low value (usually a few milliohms) and made of materials which have a very low temperature coefficient of resistance. The ultimate limiting factors for accuracy with shunts are inductive effects (inductive reactance may be significant compared with resistance), heating and voltage drop due to shunt resistance (i.e., too low a resistance and you can't measure tiny currents, too high a resistance and the shunt melts at high currents, or at least gets hot enough to vary in resistance by a small amount), measurement amplifier limitations such as 1/f noise and offset voltage, and parasitic thermocouple effects due to temperature gradients in the shunt and the connecting wires. In general, even at a max frequency of 120Hz, anything better than 1% accuracy over a 1:100 dynamic range of current is going to be expensive. Now, if you mean that you need 0.1% of full scale (in other words, you can tolerate 0.2 Amps error even when the current is around 1 Amp), things will get somewhat easier but still not in the price range you are suggesting. Sean -- = http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist