High side current measurement

[galaxy]

I would recommend a shunt, but you need to be very careful selecting it,
if your Imax = 60A then if you select .001 ohm shunt it will give you full range output of .06V and the reason you need to be careful is the power over that resistor will be 3.6 watts, you need to make sure that your shunt can handle that, you can interface the shunt with any energy measurement IC or use PGA to filter and amplify the signal before reading it.
if your change your resistor value to .01 ohm you will have much higher output (.6v) but your power will be 36 watts  !

Best Regards

[zab]

Hey Max, I know you're not looking for an integrated solution but go ahead and file this one away for when you do.  I've used a lot of these in the past.

Hall Effect  aka good isolation but only reasonable bandwidth/response depending on what your shooting for

50A   100A  and they work well in parallel if you want to get into some serious amps
AC
DC

high side, low side, middle side, who cares,  when the part is dirt cheap and the size of a quarter in real estate, why fiddle around with a dozen or more parts.  Wanna get fancy then use a DAC pin for scaling a VGA gain stage w/ offset on the output and poke it into the ADC.  Sprinkle a little microP code....be happy 

High side, high current monitoring doesn't get any easier than this:

http://www.allegromicro.com/Products/Current-Sensor-ICs/Fifty-To-Two-Hundred-Amp-Integrated-Conductor-Sensor-ICs/ACS756.aspx


H_A

Hi Hrdwr_Adct,

I too have used this device from allegro. But had a bad experience with it. this device is too sensitive especially to sparking that occurs when you connect wires to the terminals.After few times of connection and disconnection the allegro ic ACS756 stop working.So I don't consider it safe.
Have someone  faced this  problem or not .What type of protection these devices required against such problems.

[Mentor]

Hi Hrdwr_Adct,

I too have used this device from allegro. But had a bad experience with it. this device is too sensitive especially to sparking that occurs when you connect wires to the terminals.After few times of connection and disconnection the allegro ic ACS756 stop working.So I don't consider it safe.
Have someone  faced this  problem or not .What type of protection these devices required against such problems.

Sparking? What do stopped to work, just the Vout, the conductor opens, maybe both?

The datasheet says that the conductor can handle an overcurrent of 5 times the nominal current, maybe are you having some inrush current bigger than that?

I was planning to use some of these devices, but the regular distributors never had them in stock. In fact I don't know until now who sells them. 

[zab]

The device  I used has 50A +_rating and having 5 time current would be 250A. No, the limit can not be even touched. The  problem was that output pin got short with 5v. still could not find any other reason of its failure.

[solutions]

"Cannot" is a bad word. Worse than the F word for a designer/troubleshooter.

Instantaneous current, from a zero impedance source into a zero ESR capacitor, is infinity and beyond...

[pickit2]

to infinity and beyond...

[PeterMcMonty]

There is a very clever and cheap solution for high side current monitoring: an Op amp, four resistors (one, R1 in the attached picture, is the shunt) and a NPN transistor (or, if you prefer, a NMOS).

Say R1 = 0,001 ohm (that gives a 60 mV drop voltage at I_load = 60 A with 3,6 W power dissipation)

we can put R2 = 1 ohm (that gives I_R2 = 60 mA at I_load = 60 A)

then R3 = 82 ohm gives Vo = 60 x 10exp(-3) x 82 = 4,92 V full scale reading.

U1 has to be supplied by Vin or higher, it has to be able to stand a common mode input up to +Vcc and (this is a problem, in this case) be able to withstand Vin = 60 V as +Vcc - -Vcc differential supply.

Luckily exists an old device by NSC called Norton amplifier: LM2900 or commercial grade LM3900 (four op-amps in a 14 PDIP case), or LM259 / LM359 (dual op-amps) that have a current mirror input, so you simply put two high values resistors in series with the op-amp inputs and you get a common mode input as large as you want. The IC should be supplied with a 9 to 15 V referred to ground (see datasheets at NSC web site).
Two 10 Mohm resistors should be adequate: the input current should be something between 1 uA and 10 uA.

Q1 should have Vceo > 60 V of course (or Vdsmax > 60 V if a NMOS is used).

There are also SOT-23 IC's that implement such a circuit: they need R1 (shunt resistor) and R3 (output resistor). Op-amp, transistor and resistors R2 and R4 are all within the chip. Unfortunately I was not able to find ones that withstand more than 20 V maximum between input and output pins.



Posted on: July 28, 2013, 11:52:51 23:52 - Automerged
...and here the circuit diagram with Norton Amplifier LM3900

[max]

Thanks PeterMcMonty for the circuit, is it possible to directly measure the mains voltage using lm3900?

Regards

[solutions]

Nice circuit, PeterMcMonty.

Aren't you going to get in trouble with Vgs breakdown if you use NMOS at more than about 20V on the supply? Kind of goes with your argument about using current devices for voltage range.

[max]

see the attached files for the similar circuits extracted from on-semi app notes.

[Pice]

I also need current sensing for a solar MPPT I am working on. I have gone with
an integrated high side and a 0.01 Ohm sense resistor.

My problem is the following. Being an outdoor product, the device will
be exposed to extreme temperature fluctuations. In addition, during hot weather the
current increases dramatically, so the power dissipated across the resistor compounds
the heating issue.

For the MPPT function, all I care about is the relative power. However, I would like to able to get
accurate absolute readings as well.

The only solutions I can come up with are to use a resistor with a massive current rating (50W instead of 5W) to minimize
the heating and then compensate in software using the onboard temp sensor. Alternately I could mount the resistor
on a heatsink together with a discreet temp sensor.

Is there a cheap(er) current sense that can handle 20A with accuracy over -10 + 70 Celsius
range?

[Gallymimu]

I also need current sensing for a solar MPPT I am working on. I have gone with
an integrated high side and a 0.01 Ohm sense resistor.

My problem is the following. Being an outdoor product, the device will
be exposed to extreme temperature fluctuations. In addition, during hot weather the
current increases dramatically, so the power dissipated across the resistor compounds
the heating issue.

For the MPPT function, all I care about is the relative power. However, I would like to able to get
accurate absolute readings as well.

The only solutions I can come up with are to use a resistor with a massive current rating (50W instead of 5W) to minimize
the heating and then compensate in software using the onboard temp sensor. Alternately I could mount the resistor
on a heatsink together with a discreet temp sensor.

Is there a cheap(er) current sense that can handle 20A with accuracy over -10 + 70 Celsius
range?

might need to use a Manganin shunt for precision over temperature.  In a nice current measurement device you'll see a thru hole piece of Manganin that has been ground down to give an exact resistance (or you could just calibrate your measurement system).  I use some 500A Manganin shunts for precision high current measurements but they are bulky!  They cost about $30 and are 50mV at 500A.

http://www.murata-ps.com/data/meters/dpm_shunts.pdf

You should be able to find something in the 20A range.

LEM also makes some powered DC current transformers I think but they are pricey.

[pablo2048]

I'm using ACS712 in my Arduino project without problem - so You can try...

[Pice]

Thanks. I hadn't heard of Manganin shunts before. I did a quick search and came up with
the Murata 20A 3020-01098-0 on Digikey for $20.

Unfortunately the Allegro ACS712 is sensitive to temperature so I didn't even try sourcing it.

I think I going to have to just go with temp. monitoring and SW compensation to keep price down.

Any ideas on how to attach an LM35 to a power resistor ?

[Gallymimu]

Thanks. I hadn't heard of Manganin shunts before. I did a quick search and came up with
the Murata 20A 3020-01098-0 on Digikey for $20.

Unfortunately the Allegro ACS712 is sensitive to temperature so I didn't even try sourcing it.

I think I going to have to just go with temp. monitoring and SW compensation to keep price down.

Any ideas on how to attach an LM35 to a power resistor ?

GLUE!

I'd probably try using an SOT23 temp sensor right underneath a thru hole power resistor or maybe between a couple in parallel to try and keep the heating uniform.  As long as your thermal load isn't very high and your temps aren't changing too fast you can use just about any glue or thermal compound to help block convection from causing problems.

LM35s are pretty pricey these days for what they do.

How accurate do you need to be, there may be lots of other less complicated options, but you didn't include enough information for anyone to help you since you just said "accurate from -10 to +70"  Personally I think 1% is pretty accurate but you might need 0.0001% Who would know!

[Pice]

Thanks Gallymimu

The thru hole over sot23 sounds like an elegant solution.

I'm am still in the early design stages. I'm shooting for 0.1% over the temperature range. If it's
not economically feasible, I'll have to settle for less. The more impressive it sounds, the better.

The temp sensor accuracy is much less critical. A 2^o C error will only give a 400ppm error
on the reading.

I just know the LM35 because that's what I used to use for discreet measuring. I haven't used one
in years. I'll take a look at some of the newer ones.

[Gallymimu]

Microchip makes some sot23 temp sensors that are good to 1 or 2 degrees and they are cheap.

There are current sense resistors out there that are 0.25% and once calibrated might be awfully close to 0.1% and pretty good over temperature.  For instance if you got a 100ppm current sense resistor that would get you 0.01

Take a look at these, $1, 5W, 5mOhm or 10mOhm, 1% initial tol, 20ppm... not quite there but getting pretty good.
http://www.digikey.com/product-detail/en/OAR5R010FLF/989-1097-ND/2407724
http://www.digikey.com/product-detail/en/PWR4412-2SDR0100F/PWR4412-2SDR0100F-ND/1242812
http://www.digikey.com/product-detail/en/OAR5R005FLF/989-1096-ND/2407723

[Pice]

I just thought of another possibility. Since it is for a solar regulator, current
and heat should be quite closely related ( Hotter day = more current & more current = hotter resistor).
Therefore I could probably estimate the temp error based on the current/voltage reading alone.

The resistors I was planning on using are 200ppm/^oC
If I use the 20ppm ones you suggest instead, total error over the temp range
is 0.16%. If I use the crude temp correction above, assuming I can
estimate to within 20^o I can reduce the temp error to 0.04% over
the entire range.

Once temp/current has been accounted for and the resistor has been calibrated, Shouldn't I be able
to get much better than 0.1% ? What else could cause the resistor to drift ?

[Gallymimu]

I just thought of another possibility. Since it is for a solar regulator, current
and heat should be quite closely related ( Hotter day = more current & more current = hotter resistor).
Therefore I could probably estimate the temp error based on the current/voltage reading alone.

The resistors I was planning on using are 200ppm/^oC
If I use the 20ppm ones you suggest instead, total error over the temp range
is 0.16%. If I use the crude temp correction above, assuming I can
estimate to within 20^o I can reduce the temp error to 0.04% over
the entire range.

Once temp/current has been accounted for and the resistor has been calibrated, Shouldn't I be able
to get much better than 0.1% ? What else could cause the resistor to drift ?

If you study the data sheets for all of those shunts you will see specs for a variety of parameters that can cause accuracy drift such as time, temp cycling etc.  So it isn't a guarantee that if you calibrate out the initial accuracy (which will help immensely) and compensate for temp coef (which is different from the permanent change that temp cycling can cause) that you will be SUPER accurate.  The thing that's hard is that some of these items that will drift probably won't occur until it's been in the field for months/years so you may never know it drifted.  Accelerated life testing would be one way to validate (and that's mostly a test of operation at greatly elevated operating temperature).

Posted on: August 08, 2013, 05:16:54 05:16 - Automerged
oh, I'd personally not make the assumption that current is related to temp other than an approximation of self heating of the shunt.  Ambient temp should be considered unknown.  For instance when it's really hot and there is a lot of sun, if a cloud comes over the temp may only drop a little (depending on the size of the cloud) but the current will drop dramatically.

[Pice]

oh, I'd personally not make the assumption that current is related to temp other than an approximation of self heating of the shunt.  Ambient temp should be considered unknown.  For instance when it's really hot and there is a lot of sun, if a cloud comes over the temp may only drop a little (depending on the size of the cloud) but the current will drop dramatically.

Good point, I wasn't thinking. I suppose I'll just go back to a cheap temp sense, resistor & glue. It should do the job.

If I use the shunts you mentioned above ( a flat bar bent like a square "U" ) , a to-92 package is probably easier to attach than a sot23.
I'll just place it next to the shunt with the flat side parallel to the side of the shunt, with a drop of thermal grease in between.

[Gallymimu]

I think a TO-92 will be a little big, you measured temp will be a combination of ambient convection across most of the epoxy package of the case and then some from the connection to the shunt.  Would you consider a thermistor?  A little encapsulation would help prevent ambient temp effects.

Since there isn't a large heat load thermal grease doesn't do a lot for you, just use a very thin layer of glue.

[Mentor]

I've already installed some Pt100 sensors and thermistors with this product:
http://www.electrolube.com/docs/thermalmain.asp?id=36

That's a bi-component "epoxy bonding system", and I think that works for temperatures below 130ºC, but there are some alternatives that work with higher temperatures. I got a sample from the local distributor, probably you can ask for one too. Even the small syringe can be used to install a lot of sensors.

[Gallymimu]

I've already installed some Pt100 sensors and thermistors with this product:
http://www.electrolube.com/docs/thermalmain.asp?id=36

That's a bi-component "epoxy bonding system", and I think that works for temperatures below 130ºC, but there are some alternatives that work with higher temperatures. I got a sample from the local distributor, probably you can ask for one too. Even the small syringe can be used to install a lot of sensors.

again since there isn't a heat load he probably doesn't need anything thermal.  RTV is probably fine if it get's really hot *silicone RTV handles rather high temps), and honestly besides the interface you don't want thermal conduction so a slight encapsulation around the sensor with a NON-thermally conductive compound is best to prevent convective cooling of the non-bonded surfaces of the sensor.

[solutions]

You need to be careful with RTV. It'll really mess up some sensors.

O2 sensor for automotive comes immediately to mind, which needs to run HOT

[PeterMcMonty]

Thanks PeterMcMonty for the circuit, is it possible to directly measure the mains voltage using lm3900?

Regards

I think it's possible, but not very straightforward.

Posted on: August 16, 2013, 02:04:00 02:04 - Automerged

Nice circuit, PeterMcMonty.

Aren't you going to get in trouble with Vgs breakdown if you use NMOS at more than about 20V on the supply? Kind of goes with your argument about using current devices for voltage range.

Thanks for your appreciation, solutions, I dont' think there are troubles with Vgs, since the device operate in a closed loop: when Vgs goes over threshold the MOS starts conduction and the op amp output tends to go lower and so Vgs. Furthermore, the LM3900 should be supplied with no more than 15V, then its output cannot go higher than 15V too.

Sorry for my english, but I don't understand your last phrase: " Kind of goes with your argument about using current devices for voltage range." may you please explain me in some other way?
Thank you