[TriEmbed] High Current Sensing

Rick nm3g at triad.rr.com
Sat Jun 14 13:15:39 CDT 2014

Good afternoon Adam,

I've been there ... had 8 current transducers (600 amps each), feeding 
into 12-bit ADCs with the transducers providing the isolation.

The problem is the current transducers can be a bit expensive (Phidgets 
lists a 100 amp transducer for $40.00 
(http://www.phidgets.com/products.php?product_id=3588), but only shows a 
200 mA resolution).

If you want, take a look at Caddock's current measuring (Kelvin) 
(http://www.caddock.com/Online_catalog/current_Sense/current.html), but 
they limit out at around 60 amps and will require a heatsink. You don't 
want to mount these on a PC board, but rather inline with your high 
current leads. This type of measurement will require isolation ... look 
at isolation amplifiers, but be prepared for sticker shock (The Analog 
Devices AD202 is over $50.00 each ... hmmm price dropped since the last 
time I bought them). An alternative is known as a "poor man's isolation 
I've used the AD629, which is a very high impedance isolated opamp that 
provides great, but not perfect isolation.

A third approach is to purchase a surplus current shunt 
(http://ctrsurplus.com/newdccurrentshunt100amp50mv.aspx), as new shunts 
are as expensive as current transducers. Here again you will have to 
provide isolation (I'd build a board and mount it adjacent to the shunt 
with the instrument amplifier and ADC right there, then transport the 
information via I2C or what ever protocol you desire).

Accuracy is important, but remember you will be limited by the 
resolution of your ADC. I highly recommend a 16-bit ADC if you need fine 
current monitoring (1.5 mA for 100 amp shunt with a 16-bit).

To reduce your noise, put a low pass filter on each ADC input line with 
the time constant equal to the sample rate (I sampled every 100 mS, thus 
I installed a 100 mS low pass on each line) ... this cuts your random 
noise down. I also read each ADC 8 times ... throwing away the first 
reading and averaging the remaining 7 ... and was getting very 
repeatable, very accurate measurements on non-sine wave current waveforms.

Hope this helps.



On 6/14/2014 11:35 AM, Adam Haile wrote:
> As those that saw my projects at the Maker Faire know, I work with a 
> lot of high current LED projects. So I'm always looking for higher 
> current supplies and my variable bench supply definitely doesn't cut 
> it so I generally use supplies like the 5V/40A supply I used for my 
> 24x24 matrix. But that is even a too little for some of the projects I 
> have planned. I'm thinking about getting a 500 - 700W (@5V) supply 
> (something like this: http://www.ebay.com/itm/370888481354) and 
> building it out with a few more features.
> Mainly, I want to add built in current sensing that would be displayed 
> on an LCD screen (likely all Arduino controlled). The supply itself 
> would be unmodified - I'm not crazy. I would just add all that between 
> the supply output and connectors on the outside of an enclosure I 
> would build.
> So, my main question is if anyone has any advice for how I could go 
> about sensing this much current? I'm thinking about using something 
> like the TI INA219 (http://www.ti.com/product/ina219) which, with the 
> right shunt, should be able to handle it and would have a nice I2C 
> interface. I'd also like to have 2 - 4 separate outputs, each with 
> their own separate sensor... so each would need to handle up to 100A 
> or so.
> I realize that running the current through any PCB trace to the shunt 
> would probably require insanely wide traces, so is there a good, high 
> precision shunt that would work off PCB?
> Basically, I don't know what I don't know, so I'd love any direction 
> that could be given.
> _______________________________________________
> Triangle, NC Embedded Computing mailing list
> TriEmbed at triembed.org
> http://mail.triembed.org/mailman/listinfo/triembed_triembed.org
> TriEmbed web site: http://TriEmbed.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.triembed.org/pipermail/triembed_triembed.org/attachments/20140614/c5a1ac65/attachment.html>

More information about the TriEmbed mailing list