PIC 16f877 A/D conversion Allegro ACS712

[DavyJones]

Hello all... I posted some threads a while back and got some help with the 16F822 and it's A/D conversion process. I worked through the math with a pressure transducer and had thought I worked out the math with the Allegro current sensor ACS712 but I'm pretty sure I was originially wrong on that and now just want to get a second opinion or 2 to see if I finally got it. If anyone has any comments please let me know.

I did find one thread dealing with what I am pretty sure is the ACS712 perhaps a typo on the name or it was called that before I started to use them. In this thread they do math on this device but I think the outcome of this math is going to be a little bit off if they are truly talking about the ACS712. It turns out the company I work for does allot of business with Allegro and the were more then happy to help me out. I did review the data sheet on the Acs712 but initially didn't spot the output voltage until just yesterday. I emailed them and they confirmed the output range of the 712 is 2Vdc which is 2.5Vdc to 4.5Vdc

So my question/observation is using an 8bit number by my calculations the multiplier of the value returned by the pic would be .02272. From the PIC for a 0 to 30 amp reading from the 712 I could expect to see 128 to 216 which gives me 88 to work with divide 2/88 and get .02272 thus my calculation is
amps = reading * 2272
amps = DIV32 1000

Did I get this right or is my math still off? Hope thats not to long winded.
Thanks
David

[Darrel Taylor]

DavyJones ...
Sorry ... but that one gets sent to the "Locker".

Yup, we were talking about an ACS712 from Allegro in that thread.
But, hell_pk had also specified 10-bit A/D with a 2.61V VREF-.
You seem to be using 8-bit A/D without a reference, so yes, things are going to be different ...

With VREF+ = 5V, and VREF- = 0V, at 8-bit ... each A/D step is 0.0196V. (5 / 255)

The ACS712 idles at 2.5V, and with 8-bit resolution that's an A/D reading of 127.

And assuming you are using the default 66mv/A sensitivity, the formula looks like this ...

(ADvalue - 127) * 196 / 66

Which should give 1 decimal of precision ... 10.0amps = 100
With quite a bit of integer truncation due to the 8-bit resolution.<hr>
ADD:
That only works with positive currents, so to make sure it never sees negative numbers, you can add this before the formula to limit the A/D values ...

ADvalue = ADvalue MAX 127

hth,

[mackrackit]

Here is a longer way of looking at what is happening. Not as nice as Darrel's but it may help you understand a bit. This works for sensors that do not start on zero.

The old formula of
y = mx-b
y = output 'amps in your case
m = slope
x = input units
b = offset

You are wanting to read form 0 to 30 amps.

Start off by looking at the units as being volts. You have a 2 volt span. This means that every 1 volt input will equal 15 amps. Slope will equal 30 / 2 = 15..
m = 15

The offset. Starting at 2.5 volts. 2.5 volts = 0 amps. the offset is 2.5 * 15 = 37.5
b = 37.5

Plug all this into the formula and lets say the input is 4.5 volts.
y = (15 * 4.5) - 37.5
y = 30 amps

Now convert this to an 8 bit PIC resolution.
Like Darrel did I will assume a 0 to 5 volt reference.

2.5 volts = 127
At 8 bit resolution each volt is 51 steps.
Spanning 2 volts or 102 steps will have 4.5 volts equaling 229.

New value for m is 30 / 102 = 0.2941
m = 0.2941

New offset value is 127 * 0.2941 = 37.35
b = 37.35

y = (ADC * 0.2941) - 37.35
ADC = 229
y = 29.99 amps

Possible code...

ADC = 229
z = ADC * 2941  ' equals 673489
z1 = DIV32 10000  ' z1 = 67
y = z1 - 37       ' Did you want precision???

I am sure something is wrong but it may help you understand.

[DavyJones]

Darry and Mack thanks very much for your response. I undersatand the 2 different ways you came to your solution. Either my ACS712 is fried or I am doing something wrong. I am passing AC through the 712. No matter which math I use the current reading seems to high but 2 to 3 amps. That is according to my clamp meter I am using taking a reading at the same time and the device I am using to test this. I am running a heat gun which states on the side it's 1200 watt 10amp heat gun.

Any suggestions? I can post my code it's a very simple program.

I appreciate the help, I've got to sweep up this big pile of hair I just yanked from my head

Thanks
David

[ukemigrant]

I drew this diagram to illustrate the different units involved.

For a numeric value of 166 read from the analog-to-digital converter, Darrin's equation brings back 11.58 and Mackrackit's brings back 11.47.

They are close enough given the low resolution arithmetic taking place especially if you just round up to no decimal places.

I can attest to the annoyance of one meter showing 9 Amps when the equations bring back the numbers above.

http://www.picbasic.co.uk/forum/albu...0&pictureid=25

[Darrel Taylor]

I am passing AC through the 712. No matter which math I use the current reading seems to high but 2 to 3 amps

What Dave and I both "assumed" was that you were measuring DC currents.
The part about an A/C signal would have helped in the beginning.

Being that the heat gun is mainly a resistive load, you could get a much closer reading by finding the highest (peak) voltage and multiply by 0.707, which gives the R-M-S value of a perfect sine wave. Then convert to current.

The heat gun also has a fan, which is an inductive load, so you'll never get a perfectly "accurate" reading with the 0.707 multiplier. But since the fan is only a small portion of the load it would be closer.

But I doubt the eventual use of your project will be measuring the current from heat guns. And the type of load makes a HUGE difference.

If you'll be primarily measuring motor currents, the wave forms will NOT be sine, and .707 will not even come close. In that case, there is only one way to go.

Run the signal from the current sensor into an R-M-S converter. Don't bother trying to do it in software. I like the LTC1966, but it's a really tiny chip and isn't great for hobby use cause they're too hard to solder.

There are several other chips available that do the same thing. But for sure you will need an R-M-S (Root Mean Square) reading for any inductive loads.
<br>