In my work designing the HP3478A (I see one on your shelf!) I learned that copper has a terrible temperature coefficient. On our first prototype, the sense wires were connected to the shunt in such a way that the effective resistance of the shunt included some copper trace. Put in a high current, and the measured value slowly changed, out of spec. Maybe that's not as significant if you're not designing a 5 1/2 digit multimeter. Let's see, Google tells me that copper tempco is .00393 /k, that's 3930 ppm per degree c? I think any design with multiple resistors is going to have copper tempco problems, so you need to keep the resistance of the copper way below the resistance of the shunt to minimize it.
9:13 circuit with terminals on each side is best for multiple resistors is best since it is symmetrical, but can be thermally influenced by nearby heat sources
2:00 1 RON = 0.23 USD (random exchange rate website evaluation for June 9th 2020/doesn't matter at all, just a curiosity) Thanks for the insights. I learned about metal film temp-co the hard way. ..."what the ?! My 0402 10mR shunt can't handle a piddly 10 amps!" ...maybe not that bad...but... What is the best practices consensus for traces under a surface mounted package? Thanks for the upload. -Jake
It's quite a general question so the answer wont be very specific, in general I just consider available clearance and possible noise coupling into sensitive inputs.
I have a good question to you - why in different DMMs, manufactures always use a proper shunt resistors (usually a thick "П" shape wire ~20 mm length) but in serial conjunction with a PCB trace, which is a cooper? Usually ratio is that ~95% of resistance (and voltage drop) is on the proper shunt, and ~5% is on PCB traces. As a result, even not cheap DMMs (imagine UT61E) have a current measurement drift in time, on bigger current. For example the UT61E (20A rated) with 6A, after 10 minutes of current passing through, will increase reading by 15mA, while other current measurement devices in the same circuit keep current exactly the same and stable, up to 1mA precition. I tried to google and could not find a good answer for that. To compensate non-zero TCR of the proper shunt? Then why it's worse than other not-DMM type of devices, where only shunts are used current sensing? Help please to get this point.
Another great video Voltlog! I learned a lot of this material the hard way; designing and redesigning the Joulescope front-end. As @Tiago Ferreira mentioned, the parasitic effects really start getting difficult to manage as frequency increases. Inductance is particularly problematic. Larger shunt resistor = larger inductance = lower bandwidth. But smaller resistor = worst power dissipation = more temperature dependency.
Matching the length of the sensing pads seems wrong to me. This just makes the total path longer which decreases the voltage at the sensor. Or am I missing something?
The idea is to keep those lines balanced, because any voltage difference is mistaken as being caused by the passing current through the shunt, thus increasing the measurement error.
Hi since you're talking about shunts and current measurements, then may i suggest you to make a video about switch mode current regulator? I am trying to design one which in "theory" would but probably won't and i can't find any good circuits for that. So i would appreciate if you make a video about a switch mode current regulator without a uC/MCU.
About a million times more important than the matched sense paths is controlling thermocouple effects. Make sure the shunts are thermally balanced, so that temperature difference between the two sense points is minimized. And of course, also minimize any thermal gradient across your sense amp...
In my work designing the HP3478A (I see one on your shelf!) I learned that copper has a terrible temperature coefficient. On our first prototype, the sense wires were connected to the shunt in such a way that the effective resistance of the shunt included some copper trace. Put in a high current, and the measured value slowly changed, out of spec. Maybe that's not as significant if you're not designing a 5 1/2 digit multimeter.
Let's see, Google tells me that copper tempco is .00393 /k, that's 3930 ppm per degree c?
I think any design with multiple resistors is going to have copper tempco problems, so you need to keep the resistance of the copper way below the resistance of the shunt to minimize it.
Thanks for your work on the 3478A, Rich! I own several of them. 👍
Thank you for sharing this and for your work on the HP3478A!
This was quite interesting. I definitely learned a lot. Thanks for sharing this.
Thank you!
Excellent! Learned lots
Great to hear!
9:13 circuit with terminals on each side is best for multiple resistors is best since it is symmetrical, but can be thermally influenced by nearby heat sources
Great walkthrough
Good practice
Thanks for sharing 👍😀
Thanks for the visit
Very helpful video 👍
Nice, thank you for sharing!
Thanks for watching!
2:00
1 RON = 0.23 USD
(random exchange rate website evaluation for June 9th 2020/doesn't matter at all, just a curiosity)
Thanks for the insights. I learned about metal film temp-co the hard way.
..."what the ?! My 0402 10mR shunt can't handle a piddly 10 amps!"
...maybe not that bad...but...
What is the best practices consensus for traces under a surface mounted package?
Thanks for the upload.
-Jake
It's quite a general question so the answer wont be very specific, in general I just consider available clearance and possible noise coupling into sensitive inputs.
Good Information. Thanks.
Glad it was helpful!
I have a good question to you - why in different DMMs, manufactures always use a proper shunt resistors (usually a thick "П" shape wire ~20 mm length) but in serial conjunction with a PCB trace, which is a cooper?
Usually ratio is that ~95% of resistance (and voltage drop) is on the proper shunt, and ~5% is on PCB traces. As a result, even not cheap DMMs (imagine UT61E) have a current measurement drift in time, on bigger current.
For example the UT61E (20A rated) with 6A, after 10 minutes of current passing through, will increase reading by 15mA, while other current measurement devices in the same circuit keep current exactly the same and stable, up to 1mA precition.
I tried to google and could not find a good answer for that.
To compensate non-zero TCR of the proper shunt? Then why it's worse than other not-DMM type of devices, where only shunts are used current sensing?
Help please to get this point.
Another great video Voltlog! I learned a lot of this material the hard way; designing and redesigning the Joulescope front-end. As @Tiago Ferreira
mentioned, the parasitic effects really start getting difficult to manage as frequency increases. Inductance is particularly problematic. Larger shunt resistor = larger inductance = lower bandwidth. But smaller resistor = worst power dissipation = more temperature dependency.
Thanks for the feedback!
Matching the length of the sensing pads seems wrong to me. This just makes the total path longer which decreases the voltage at the sensor. Or am I missing something?
The idea is to keep those lines balanced, because any voltage difference is mistaken as being caused by the passing current through the shunt, thus increasing the measurement error.
Hi since you're talking about shunts and current measurements, then may i suggest you to make a video about switch mode current regulator?
I am trying to design one which in "theory" would but probably won't and i can't find any good circuits for that.
So i would appreciate if you make a video about a switch mode current regulator without a uC/MCU.
Okay, I've add your suggestion to my list and will consider it for a future video. Thanks!
@@voltlog Thanks. Would appreciate more if you don't use dedicated ics for that and even if you use then forget to explain it.
Check out the TH-cam channel KatKimShow, great primer on switch mode power supplies and then read some TI materials. That's what I did.
About a million times more important than the matched sense paths is controlling thermocouple effects. Make sure the shunts are thermally balanced, so that temperature difference between the two sense points is minimized. And of course, also minimize any thermal gradient across your sense amp...
thank you for sharing this 👍
Love all the ideas.
Take a look here for more improvemets: www.analog.com/en/analog-dialogue/articles/optimize-high-current-sensing-accuracy.html#