Testing of the ACS712 breakout board consisted of a number of stages. Initially, dummy loads where connected to produce currents of known magnitude with the output of the ACS712 Vout pin monitored with a digital multimeter. I have a DIY electronic dummy load which made applying known currents easy. If this is not available, simply use a suitable load resistor (and varying the applied voltage and or resistor value to develop the desired current value).
The ACS712 produces a voltage proportional to measured current which is nominally 185mV/A ± 5mV/A. The voltage produced by the ACS712 is centered on Vcc/2 to enable bi-directional current sensing i.e., "zero" current flow is indicated by a voltage of 2.5V with 5V supply to the ACS712. The ACS712 (x05B version) has a nominal measurement range of ± 5A which means a nominal voltage output (Vout on pin 7) of 1.575V to 3.425V.
However, it was found with the eBay breakout board version that observed voltage output from the ACS712 was only 135mV/A. The output voltage in reference to input current was measured on two discrete items of the ACS712 breakout boards that were purchased, in order to check if the low sensitivity was isolated to a particular board. However, the same results showing significantly decreased sensitivity compared to that stipulated in the datasheet were found. Hence, a potential reason why the breakout board versions (which include breadboard ready headers, LED power indicator and free shipping) cost approximately one-third the cost of just the ACS712 chip from a "western" source (Mouser, Digikey, RS Components etc). No doubt you "get what you pay for".
The effect of varying the supply voltage of the ACS712 (Vcc) was examined, since the "zero" current point is centered at Vcc/2 to enable bi-directional current sensing. Hence, drift and or variation in ACS712 Vcc will produce a measurement error. The ratiometric output from the supply voltage resulted in an observed 135mV/A at various levels of Vcc (ACS712 has an absolute maximum Vcc of 8V with recommended 4.5 - 5.5V). However, as expected, the output "zero" current point was highly dependent on supply voltage Vcc (although if Vcc did not vary the "zero" current point was also stable).
Despite the negative results obtained by measuring ACS712 output with a multimeter (i.e. the low sensitivity of the breakout board compared to the datasheet), the ACS712 was still interfaced to the PIC18F248 ADC to check the results. The multimeter used has limited resolution and perhaps was not capable of adequately monitoring the expected relatively small voltage changes.
The same results were observed using the PIC18F248 ADC to measure ACS712 output versus applied current (i.e., only 135mV/A compared to the datasheet value of 185mV/A ± 5mV/A.
The ACS712 is potentially advantageous for current measurements as the component has a 2.1 kVRMS minimum isolation voltage with a simple direct output voltage proportional to bi-directional current flow/magnitude.
However, the limited sensitivity (185mV/A ± 5mV/A) restricts the potential usefulness compared to other similar IC's (see MAX471 current sensor). Particularly, in the case of the ebay breakout board version which in the case of the items tested did not meet the datasheet specification i.e., possibly production rejects which explains the cheap pricing (there is a photograph of the ACS721 IC on the breakout board in the Photographs Section). Additional hardware/circuitry could be used to scale the output voltage range of the ACS712 to better utilise the measurement range of the PIC18F248 ADC, however, this negates the potential low cost/simple use advantages of the ACS712.
As an update note, the Allegro ACS723 05AB (SOIC 8-pin package $4.50/each from RS-Online at time of writing) is now available. This current sensor has the same advantages of the ACS721 (isolation etc) but a sensitivity of 400mV/A (for the 05AB version, ± 5A).