Hmmm.. it’s a magnetic sensing component, you seem to have it clamped in a huge block of ferrous material (the vice). Does it read different away from the vice ?
The rubber pads on the vise jaws likely provide enough distance that the steel would have little influence. I suppose it is possible that the steel has become quite strongly magnetized, but that would almost certainly produce an offset error rather than a gain error.
It's likely a mislabeled acs704 or 706, both have a 133mv/A sensitivity. Check the resistance between pins 6 and 7, on the 704/706 they're shorted together. The 704/706 are discontinued so are probably dirt cheap.
Assuming the IC is authentic, the device in datasheet quite clever. It has a hybrid of electrical chopper-stabilize & Hall offset auto null in-one. So the opamp offset auto-nulls at the same time as the magnetism canceled out. Not only that but works like this over temperature range. That makes the zero signal output = 0.5 Vdd. Failure of this chopper stab would cause the output to be offset from 0.5 Vdd and drift much more than spec. I reckon you can test for defective chopper by watching the zero-current signal drift with temperatura and chsnge of external field magnetic.
Is there a part specified at 135 mV/A? Possibly, the chip's label is wrong. Is there possibly a shunt so that not all of the current goes through the chip? Is it reasonable that the current is not divided equally between the two pin pairs? This might make the magnetic field's distribution problematic.
@@IMSAIGuy in UK dot is the separator between integer and decimal parts ;-) But other countries in Europe use coma as separator. World is strange: billion = milliard etc. ;-)
2.1kV... the discussion was about the separator but there I put it right for you and I give shit loads of shit about a shit load of shit...actualy ;)@@zyghom
Beware of those boards !.... Several years ago I bought from AliExpress plain ACS712-05 chips and found them OK BUT...a few months ago I bought 3 times from different vendors this kind of board with ACS712-05. They were OK about the scaling factor but the outputs were so noisy that, to use them, I had to heavily filter the output for lowering that noise (and bandwith as well). The datasheet says noise = 21 mVpp max for a 2 kHz bandwith... I had more than 100 mVpp noise with a Cfilter = 47 nF, that means an equivalent noise of more than +/-0.5A reading... The chip is built around a chopper amp and I guess that the ones that I bought were rejects from the factory. These boards are therefore only usable for very slow signals (& DC of course)
The IC output is RATED as HALF the supply voltage with no current flowing, so the accuracy and stability of its supply would affect its output. This is not unusual in this sort of ICs.
One of the graphs in the datasheet says it should be pretty much bang on 2.500 V for no current (at 5V bias, 25°C) and going below freezing should still kick it off less than 10mV I can't even conceive how the user could mess with that (short of wrong supply, but you said you set it to exactly 5V so ¯\_(ツ)_/¯ )
These sensor modules are rather popular, meaning it is likely that they are highly faked. The part series is very wide... 712, 714, 724, on and on. Still available on Mouser, last I checked.
You obviously have some nonmagnetic causes going on, which, according to the datasheet, causes the quiescent output voltage to deviate, which needs to be factored in the calculation using the ratiometric calculations, accuracy graph etc. Or it may just be rubbish.
I had good success with ADS1261. Used +/-5V for excitation and +/-2.5V for the ADC analog section, but you can just use clean +5V for excitation and the ADC, except that the you will get double the noise/signal ratio. I run it at 500Hz with circular STM32 DMA and it can go much faster.
@@IMSAIGuy I would vote for out of spec because its deviation is very accurately the same in both directions and is very linear when you double the current, which would hardly be the case with a poorly made copy. I am actually amazed how well it tracked the current changes - the AMOUNT of change per ampere is not a big deal when you can always adjust your circuit for that, either in hardware or in code. Out of spec would also be consistent with what we often get from AliExpress as one guy accurately put it: factory floor sweeping specials, meaning masses of out of spec rejects are sold cheaply BECAUSE they couldn't be sold as such otherwise.
Without an internal voltage reference, the 1,5% accuracy become the smallest problem ^^ So it would be better if it e.g. requires 5V makes a 3.3v reference or something and measures this against…
This chip has ratiometric output from supply voltage. In many cases it is better than using a internal voltage reference. If the ADC uses the same supply voltage as reference as the sensor then the dependence on the reference voltage disappear. If it had an internal reference then errors in both the internal reference and the reference for the ADC would add to the sensor error.
@@larslindgren3846 okay yeah correct but i would never use a supply voltage as my reference voltage not even in rapid prototyping it’s 2-5 more lines of code if internal in uC MSP432 in my case.
@@steinmar2 If you for some reason (e.g. other sensors that has absolute voltage output) uses a reference voltage other than supply voltage in the ADC then you can compensate for the supply voltage simply by measuring it with an other ADC chanel and calculate the ratio. From a theoretical point of view it is better to measure the ratio of two signals than an absolute voltage. Many precision measurements are made ratiometric e.g. strain gauges, temperature, differential transformer sensors and other measurement bridges. Ratios can theoretically be self calibrating while a voltage reference needs traceable calibration to an primary voltage standard.
@@larslindgren3846 One important caution with use of the supply rail as a reference, even in ratiometric circuits, is to be sure the supply is adequately clean and stable at the relevant physical locations in the circuit and at the relevant times. The latter is important if you want to calculate the ratio of two signals. Sampling both as closely spaced in time as possible is generally best. Collect both samples, _then_ do any required processing. If you do something like reading multiple times to get an average, interleave the readings of the two channels, don't do all the samples for one then move on to the other. Don't forget that "calibration" in code can result in missing counts. Offset correction is usually easy and clean. Gain (scale) correction often isn't. The resistance of the wires used to connect to a bridge device can introduce significant error in some instances. This can often be fixed with four-wire schemes and some good instrumentation amps carefully applied.
@@d614gakadoug9 Good points but at 1.5% accuracy I don't think they mater most of the time. In a high accuracy situation all relevant sources of errors need to be considered. A separate clean reference for the sensor and the ADC is a good idea. The point is that you don't want to have two separate voltage references for the sensor bridge and the ADC since errors in the voltage references then directly translates to measurement errors.
Hmmm.. it’s a magnetic sensing component, you seem to have it clamped in a huge block of ferrous material (the vice).
Does it read different away from the vice ?
The rubber pads on the vise jaws likely provide enough distance that the steel would have little influence. I suppose it is possible that the steel has become quite strongly magnetized, but that would almost certainly produce an offset error rather than a gain error.
The more modern chips that use GMR technology have wider bandwidth and lower noise. Check for example ACS70331.
ACS724 also does stray magnetic field cancellation much better than ACS712.
I used the same board for monitoring battery charging. I calculated the coefficient while using dedicated USB PS. Input voltage to DAQ. It's cheap.
Have you checked the output with an oscilloscope, in case it is oscillating?
Did you tried to measure directly at chip legs? Maybe there you get values from spec.
It's likely a mislabeled acs704 or 706, both have a 133mv/A sensitivity. Check the resistance between pins 6 and 7, on the 704/706 they're shorted together. The 704/706 are discontinued so are probably dirt cheap.
now that's the first explanation I can believe.
Assuming the IC is authentic, the device in datasheet quite clever. It has a hybrid of electrical chopper-stabilize & Hall offset auto null in-one. So the opamp offset auto-nulls at the same time as the magnetism canceled out. Not only that but works like this over temperature range. That makes the zero signal output = 0.5 Vdd.
Failure of this chopper stab would cause the output to be offset from 0.5 Vdd and drift much more than spec.
I reckon you can test for defective chopper by watching the zero-current signal drift with temperatura and chsnge of external
field magnetic.
Interested! The voltage difference with datasheet, it maybe temperature issue.
Out of spec waste parts bought cheaply from OEM.
My ACS712 measures .190volts per amp. The board looks exactly like yours.
Chinese printed 185mv instead of 136mv on spec sheet? Wouldn't be my first time seeing that.
Could be quite a useful part.
Is there a part specified at 135 mV/A? Possibly, the chip's label is wrong. Is there possibly a shunt so that not all of the current goes through the chip? Is it reasonable that the current is not divided equally between the two pin pairs? This might make the magnetic field's distribution problematic.
@00:53 - I am not sure how you read this in USA but in Europe we read it "two point 1" rather than "one point two" ;-)
well it does have a 1 and a 2 and a point. I need to practice the order. anyway in Europe it would be 2,1 right? is comma still a point out loud?
@@IMSAIGuy in UK dot is the separator between integer and decimal parts ;-) But other countries in Europe use coma as separator. World is strange: billion = milliard etc. ;-)
2.1Kv is two point one Kilo Volts. It's a mistake we all make every now and then lol ! no comma just a dot will do us Brits....cheers@@IMSAIGuy
@@andymouse wrong: kilo is small letter, Volt is capital letter. you guys don't give a s...t about anything I see ;-)
2.1kV... the discussion was about the separator but there I put it right for you and I give shit loads of shit about a shit load of shit...actualy ;)@@zyghom
Beware of those boards !.... Several years ago I bought from AliExpress plain ACS712-05 chips and found them OK BUT...a few months ago I bought 3 times from different vendors this kind of board with ACS712-05. They were OK about the scaling factor but the outputs were so noisy that, to use them, I had to heavily filter the output for lowering that noise (and bandwith as well). The datasheet says noise = 21 mVpp max for a 2 kHz bandwith... I had more than 100 mVpp noise with a Cfilter = 47 nF, that means an equivalent noise of more than +/-0.5A reading... The chip is built around a chopper amp and I guess that the ones that I bought were rejects from the factory. These boards are therefore only usable for very slow signals (& DC of course)
If the chopper is defective in the IC it'll be "good enough quality for Ebay"
Did I see that the output changed as you changed the VCC? That would be a bummer.
The IC output is RATED as HALF the supply voltage with no current flowing, so the accuracy and stability of its supply would affect its output.
This is not unusual in this sort of ICs.
Are we dealing again with a capacitor that varies with voltage. have a look at the capacitor between pin 5 & 6.
Could you do a vid on how the ad524ad works in a overcurrent circuit
One of the graphs in the datasheet says it should be pretty much bang on 2.500 V for no current (at 5V bias, 25°C) and going below freezing should still kick it off less than 10mV
I can't even conceive how the user could mess with that (short of wrong supply, but you said you set it to exactly 5V so ¯\_(ツ)_/¯ )
The "0.5x Vdd" is laser-trimmed in the chip factory. The laser zaps stop automatically when meets spec.
These sensor modules are rather popular, meaning it is likely that they are highly faked. The part series is very wide... 712, 714, 724, on and on. Still available on Mouser, last I checked.
You obviously have some nonmagnetic causes going on, which, according to the datasheet, causes the quiescent output voltage to deviate, which needs to be factored in the calculation using the ratiometric calculations, accuracy graph etc. Or it may just be rubbish.
Quiescent Vout depends on the supply voltage, it's stated in the datasheet.
I am interested in a better chip for load cells, a competitor to HX711 that talks SPI ? 😊
MCP3561 24-bit A to D converter. We are using this in one of our test bench.
I had good success with ADS1261. Used +/-5V for excitation and +/-2.5V for the ADC analog section, but you can just use clean +5V for excitation and the ADC, except that the you will get double the noise/signal ratio. I run it at 500Hz with circular STM32 DMA and it can go much faster.
I am wondering if it is an out of spec or a poorly made copy, or an out of spec poorly made copy.
I vote for the latter
@@IMSAIGuy I would vote for out of spec because its deviation is very accurately the same in both directions and is very linear when you double the current, which would hardly be the case with a poorly made copy.
I am actually amazed how well it tracked the current changes - the AMOUNT of change per ampere is not a big deal when you can always adjust your circuit for that, either in hardware or in code.
Out of spec would also be consistent with what we often get from AliExpress as one guy accurately put it: factory floor sweeping specials, meaning masses of out of spec rejects are sold cheaply BECAUSE they couldn't be sold as such otherwise.
I found ones for up to 20 amps. Giving me ideas for a psu.
Without an internal voltage reference, the 1,5% accuracy become the smallest problem ^^
So it would be better if it e.g. requires 5V makes a 3.3v reference or something and measures this against…
This chip has ratiometric output from supply voltage. In many cases it is better than using a internal voltage reference. If the ADC uses the same supply voltage as reference as the sensor then the dependence on the reference voltage disappear. If it had an internal reference then errors in both the internal reference and the reference for the ADC would add to the sensor error.
@@larslindgren3846 okay yeah correct but i would never use a supply voltage as my reference voltage not even in rapid prototyping it’s 2-5 more lines of code if internal in uC MSP432 in my case.
@@steinmar2 If you for some reason (e.g. other sensors that has absolute voltage output) uses a reference voltage other than supply voltage in the ADC then you can compensate for the supply voltage simply by measuring it with an other ADC chanel and calculate the ratio. From a theoretical point of view it is better to measure the ratio of two signals than an absolute voltage. Many precision measurements are made ratiometric e.g. strain gauges, temperature, differential transformer sensors and other measurement bridges. Ratios can theoretically be self calibrating while a voltage reference needs traceable calibration to an primary voltage standard.
@@larslindgren3846
One important caution with use of the supply rail as a reference, even in ratiometric circuits, is to be sure the supply is adequately clean and stable at the relevant physical locations in the circuit and at the relevant times. The latter is important if you want to calculate the ratio of two signals. Sampling both as closely spaced in time as possible is generally best. Collect both samples, _then_ do any required processing. If you do something like reading multiple times to get an average, interleave the readings of the two channels, don't do all the samples for one then move on to the other.
Don't forget that "calibration" in code can result in missing counts. Offset correction is usually easy and clean. Gain (scale) correction often isn't.
The resistance of the wires used to connect to a bridge device can introduce significant error in some instances. This can often be fixed with four-wire schemes and some good instrumentation amps carefully applied.
@@d614gakadoug9 Good points but at 1.5% accuracy I don't think they mater most of the time.
In a high accuracy situation all relevant sources of errors need to be considered.
A separate clean reference for the sensor and the ADC is a good idea. The point is that you don't want to have two separate voltage references for the sensor bridge and the ADC since errors in the voltage references then directly translates to measurement errors.
Looks to me that you have the 20A version, they have an output of 100mV per amp +- 20mV. So if you take the % accuracy into account, it's about right.
nearly.
Except that he says it's an "05" part and that's also what it is labeled as.
@@jrkorman interesting, I don't have any of the 5 amp ones but do have some 20A versions, they all check ok with the spec sheet.