| Electronics > Projects, Designs, and Technical Stuff |
| Reading high currents on PCB: Hall effect sensor vs shunt resistor |
| << < (4/7) > >> |
| superKris:
--- Quote from: fcb on May 06, 2020, 05:53:00 pm ---ACSxxx sensors aren't actually very good - you'll probably get within your few% though. Zero drift will be your #1 problem to solve with anything hall based. I'm assuming your application is DC and requires isolation - looking at your PNG then it looks like a a low voltage system (mention of accessories, chargers etc..) - in that case you could easily use current shunt shunt resistors and float the differential measurement? Either way, I would defintley look at using shunt resistors rather than hall sensors if you want a reliable ZERO. Texas make some IC's worth looking at if your system is lower voltage DC: LMP8480, INA282 range. --- End quote --- I'm already using the INA219 for external current shunt and voltage measurements. Incredible little chip. However, like I wrote i fear about the dissipation that will come with that. I'm hoping to run 40A (with peaks of 50) trough a PCB with 2oz copper "traces" on both sides. Below you can find a drawing of what it currently looks like. Its a automotive application (RV) so temperatures can get high easily. If i take for example a INA219, the lowest shunt voltage setting is 40mV. Even that is 2W at 50A. I have 2 of those + 2x 20A shunts. While i doubt the total disipation will go over 3W often, i have my doubts about adding that heat to a PCB that can barely hold these currents in the first place. I'm aso VERY unfamiliar to using these kinds of shunts, and i have no idea about their accuracy, and the way to mount them. |
| ejeffrey:
--- Quote from: superKris on May 05, 2020, 09:19:07 pm --- --- Quote from: ejeffrey on May 05, 2020, 07:03:14 pm ---This ADC has a differential input. Make a 2:1 resistor divider from Vcc and connect that to the (-) input of the ADC. Your chosen ADC is buffered and has a high input impedance so you should basically be good to go. --- End quote --- If i'm not mistaking a voltage divider would need to be extremely accurate. Exactly the same 0,5 * VCC as the Hall effect sensor. At a difference of 3mV (1 LSB) I would need to do a analog adjustment of the Wheatstone bridge or do a digital adjustment. However, once set the reading should be fine even with a changing VCC. --- End quote --- Well, .1% resistors or even .01% are not that expensive, but that is unlikely to be the most important factor. The Vcc/2 offset on chip is probably not that accurate either and even if it was the offset and drift of the sensor itself are going to be relatively large. A voltage divider is going to get you about as good as you can do with that sensor. Zero calibration will help with the intrinsic offset, but fundamentally you aren't going to get a great zero value on a hall sensor. Even a unipolar sensor won't help much: it gets rid of the Vcc/2 divider accuracy problem but doesn't really address the offset and drift of the sensor itself. Can you do periodic zero calibration? Either deliberately stop the current periodically or just rely on times of known zero current such as at startup. |
| fourtytwo42:
--- Quote from: David Hess on May 06, 2020, 03:25:35 pm --- --- Quote from: fourtytwo42 on May 06, 2020, 08:52:41 am ---Your problem is using a ratiometric sensor with an absolute ADC. Simple solution use the internal Arduino ratiometric ADC's that you dont like for the current sensors and use the absolute ADC for whatever you are currently using the internal Arduino ADC's for. --- End quote --- The hall effect sensor he is using is not ratiometric; its gain is fixed. What changes is its offset which is always 1/2 of the supply voltage. --- End quote --- Oopps yes sorry David your right, it's been a while since I used the things! Actually I much prefer the LEM, much more stable and less noise sensitive, as I use it unipolar I just follow it with a precision diff amp to remove the offset (the LEM has a reference output at what it thinks is 1/2vcc so it tracks nicely). |
| superKris:
I'm afraid any periodic calibrations will be difficult. I really have many doubts if hall sensors remains the way to go. They are awesome at not dissipating any real heat, they are getting to be a pain in the ass. Meanwhile i have been searching a lot for suitable shunt resistors but i can only find them at ridiculous prices of € 5 to 20 per shunt. And basically I have not been able to find any with a suitable rating: 750uOhm or 50A/40mV. I was only able to find this SMD type, but i dont really understand how this would work on the suggested 20oz board at 50A. Even with a shitload of vias... Wil it hold the current? http://www.farnell.com/datasheets/2710720.pdf?_ga=2.148078839.1947980846.1588791907-1152630265.1588791907 |
| Vovk_Z:
--- Quote from: superKris on May 06, 2020, 06:57:13 pm ---However, like I wrote i fear about the dissipation that will come with that. I'm hoping to run 40A (with peaks of 50) trough a PCB with 2oz copper "traces" on both sides. Below you can find a drawing of what it currently looks like. Its a automotive application (RV) so temperatures can get high easily. --- End quote --- --- Quote from: superKris on May 06, 2020, 06:57:13 pm ---I was only able to find this SMD type, but i dont really understand how this would work on the suggested 20oz board at 50A. Even with a shitload of vias... Wil it hold the current? --- End quote --- You may divide one shunt into several parallels. But is there a strong need to put shunts onto the PCB? 40-50 Amps looks good for external shunts. |
| Navigation |
| Message Index |
| Next page |
| Previous page |