Understanding spec sheet

Thread Starter

Crispin

Joined Jul 4, 2011
88
Hi Folks,

I am using an ACS714 to measure current from a battery charger. The spec sheet: (http://www.hobbytronics.co.uk/datasheets/ACS714.pdf)
It has a zero current Vout of 2.5V (VCC / 2). As my netduino has a max of 3.3 on the ADC, I am using a simple voltage divider (2 x 1k) to half the output.

Having a look at the spec sheet, Page 4, line for “Output Resistive Load”
Output Resistive Load RLOAD VIOUT to GND (min) 4.7kΩ (Sorry, formatting is grim)
Does this min that you cannot have a load of less than 4.7k? If so, this means my divider of 2 1k resistors is contributing towards the weird readings I am getting.
Would a divider of 2 10k be better? (any negative effect on that?)


Thanks
Crispin
 

t06afre

Joined May 11, 2009
5,934
As I read the spec sheet. The circuit shall have a load that is bigger than 4.7Kohm. If you plan to use a voltage divider. I would suggest that you put an operational amplifier between the divider and the Arduino ADC input.
 

Thread Starter

Crispin

Joined Jul 4, 2011
88
As I read the spec sheet. The circuit shall have a load that is bigger than 4.7Kohm.
Sorry, that confuses me:confused: The circuit outputs a voltage scaled to the current sense. How could the output part of the hall effect have a load if it never sinks?
 

t06afre

Joined May 11, 2009
5,934
The output stage will source current. At lower loads than 4.7 Kohm. The result will probably for some reason become inaccurate.
 

Thread Starter

Crispin

Joined Jul 4, 2011
88
ok, read and reread what you said and that is how I understand it. "don't put anything less than 4k7 here"

The Netduino's input has a much higher impedance though so surely an opamp would not be needed.

Guessing my divider is causing an issue though
 

t06afre

Joined May 11, 2009
5,934
The Netduino's input has a much higher impedance though so surely an opamp would not be needed
What is the input impedance for pins used as ADC input then. Also if you use 10K resistors in your voltage divider. That will give very high series resistance for your ADC. This will for sure give inaccurate ADC conversions. Due to how the ADC unit work.
 

wmodavis

Joined Oct 23, 2010
739
You don't mention what the "weird readings I am getting" are so it is hard to theorize what is going on.

My reading of the spec agrees that the min resistive load at the output pin should be 4.7k and you have 2k so you are below the minimum. Your voltage divider should be composed of two 2.4k resistors in series or greater to meet the min.

But other issues could be the cause of weird 'readings I am getting'. To eliminate the load as the potential problem you must correct than first then.

As I look at the chip specs, it seems to me that too low of a load likely would have the effect of causing the read output voltage to be somewhat lower than expected but I would expect it to approximately reflect the input current however accuracy would likely compromised.

Cool looking chip though. I've got use for that function.
 

Thread Starter

Crispin

Joined Jul 4, 2011
88
The spec sheet for the Amtel is here:
http://www.atmel.com/dyn/resources/prod_documents/doc6120.pdf
Page 620 seems to allude to the input impedance. Not sure though?

The user can drive ADC input with impedance up to:
• ZOUT
≤ (SHTIM -470) x 10 in 8-bit resolution mode
• ZOUT
≤ (SHTIM -589) x 7.69 in 10-bit resolution mode
with SHTIM (Sample and Hold Time register) expressed in ns and ZOUT
expressed in ohms.
Sorry, I should have made it clear what the “weird” bits were.
Yes, zero current voltage was not 1.25 and it would appear the reason for that would be due to the divider pulling the output down.
The other issue was “jitters” on the ADC but that seems to be more ADC issue than the divider / ACS714. On my Fluke the voltage is steady and only fluctuates by a microvolt. The result of the ADC swings rapidly over 10s of mv. Another topic though.

I’ll get some 1% 4.7k and 10k on the way home and try that.
 

SgtWookie

Joined Jul 17, 2007
22,210
I suggest that you place a 10nF capacitor (poly metal film or ceramic) from the ADC input to ground, and use metal film resistors for the divider. Metal film resistors are more accurate, less noisy, and won't shift resistance as much as carbon or carbon film resistors.

Your ADC needs to have a low-impedance input in order to get an accurate reading. Adding a 10nF cap provides that low impedance. At the same time, you must allow sufficient time between measurements for the voltage on the cap to stabilize, or you'll see a good bit of jitter.
 

t06afre

Joined May 11, 2009
5,934
How often do you read the current value. A trick I often use. Is to base the result on not one reading but sevral readings say an average of 10 readings. Other things that could help is to have a second order filter before the ADC input. If you do as recommended and place a opamp as buffer before the input pin. You can make the filter around the same opamp
 

Thread Starter

Crispin

Joined Jul 4, 2011
88
I put a 10nf cap on this morning and the jitters seemed to have reduced. I did not test much as I had to get going for work. I'll swap the resistors tonight and see.

I've no room on the breadboard for opamp on this version, if it does fail I'll do it in v2.

I normally take 5 readings 10ms apart and average that. I thought to try solve the problem this time by getting the input right. Test version polls acd every 500ms for a single reading.
 

SgtWookie

Joined Jul 17, 2007
22,210
You really only need to wait about 1 millisecond to be so close to the actual voltage that you won't be able to measure the difference, even with a pair of 10k resistors as the divider.
 

Thread Starter

Crispin

Joined Jul 4, 2011
88
This is for measuring charge current for a bank of Pb batteries (810Ah) and measuring the charge current which ranges from 30A to about 1-2 when full so if I loose a bit of change, I'm not fussed.

Thanks for the tip though, I've seen where people say wait before readings to allow for the ADC to settle but not really seen anything about how long.
 

t06afre

Joined May 11, 2009
5,934
Then I test ADC setups. I often use a fresh 1.5 volt battery. And connect that to the ADC with as short leads as possible. A battery is very low noise source. And a fresh battery will also have very low series resistance. With this setup you can get a good estimate. On how many bits noise you have inherent in your system. Given that you do this test in place with not that much electrical noise
 

Thread Starter

Crispin

Joined Jul 4, 2011
88
quick update (and thanks)

Put in two 10k resistors for the divider and it is a log more stable now. Only drifts one or two bits. At the same time, the fluke is jumping by 100uV.

Oddly the mcu still disagrees with fluke but I'll play with that some more. (and confirm with the battery idea)


thanks
 
Top