Author Topic: 4 wire kelvin to parallel resistors?  (Read 1730 times)

0 Members and 1 Guest are viewing this topic.

Offline jdraughn

  • Regular Contributor
  • *
  • Posts: 106
4 wire kelvin to parallel resistors?
« on: September 11, 2017, 03:53:17 pm »
I am layout a board design for the DRV8704 motor driver and it wants separate traces to the current sensing resistor. I plan on using multiple resistors in parallel to help spread the heat and/or keep the costs down by giving me for flexibility with resistor values since I am not 100% sure what value I am going to use right now.

My question is, would it be better to run the trace right up to the closest resistor, or right between the two resistors?
 

Offline Signal32

  • Regular Contributor
  • *
  • Posts: 234
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #1 on: September 11, 2017, 04:02:45 pm »
What you'll want to do is have the two resistors connect to each other via a trace, then have this trace connect to the rest of the circuit via a single point.
Put the sense trace where this common trace meets the rest of the circuit.
Ex: You don't want to connect the parallel resistors directly to a ground / power plane. Connect them to a trace that will go into the power plane (and sense at the point where the trace meets the power plane).
« Last Edit: September 11, 2017, 04:05:27 pm by Signal32 »
 

Offline Habropoda

  • Regular Contributor
  • *
  • Posts: 56
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #2 on: September 11, 2017, 06:30:47 pm »
This page seems to do a good job of discussing sense resistors:
https://www.electropages.com/2016/01/10-tips-for-designing-with-current-sense-resistors/
See tip #10 for your situation.

Get rid of the thermals, this is a power path and you don't want additional resistance.

I don't get Signal32's suggestions.  For accurate current sensing you need to measure the voltage drop across a know resistance.  His suggestion adds the unknown resistances of the additional traces.

 

Offline Signal32

  • Regular Contributor
  • *
  • Posts: 234
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #3 on: September 11, 2017, 09:12:54 pm »
I don't get Signal32's suggestions.  For accurate current sensing you need to measure the voltage drop across a know resistance.  His suggestion adds the unknown resistances of the additional traces.
Tip #10 there is basically a different interpretation (or more like the broader solution) to the problem. Putting the resistors on an isolated trace is meant to ensure the same thing -- equal current sharing. The resistance of the "additional traces" will also happen for your scenario because even if you're sensing directly on the terminals of resistor #1, the actual resistance will depend on Res1,Res2 ... etc AND the resistance of the traces connecting them.
 

Offline Habropoda

  • Regular Contributor
  • *
  • Posts: 56
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #4 on: September 12, 2017, 12:31:17 am »
I still don't get it.  How about showing a layout using your solution.
 

Offline Signal32

  • Regular Contributor
  • *
  • Posts: 234
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #5 on: September 12, 2017, 01:02:32 am »
I still don't get it.  How about showing a layout using your solution.
It basically comes out to a very similar layout.
Edit: you are correct, in my case there is not 100% exactly equal current sharing, which will affect in a very tiny way the heat up of the resistors, but it will not affect the accuracy / performance of the kelvin sense connection.
« Last Edit: September 12, 2017, 01:10:03 am by Signal32 »
 

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 3359
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #6 on: September 12, 2017, 01:33:04 am »
If your current sense resistors are identical and the layout of the power lines is identical the layout of the sense lines is not critical.  Just need to get to one of the resistors.

If your current sense resistors are identical the layout of the power lines does matter because it can have an impact on current sharing, either through resistance or reactive terms.  The conceptual problem is which resistor represents the current.  Unless you have separate sensing lines and circuitry for each resistor you don't know the total current.

If your current sense resistors are not identical (say 1% tolerance) you run into the same problem again. 

In principal these issues can be resolved by measuring the impedance from the power source through each leg of the circuit, perhaps by installing the resistors one at a time.  You could then compute the way the current is shared and compute total current based on sense lines to any one of the resistors.  Assuming that nothing changes-ever.
 

Offline Habropoda

  • Regular Contributor
  • *
  • Posts: 56
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #7 on: September 12, 2017, 02:52:37 am »
Or the OP could consult the users' guide to the DRV8704 Evaluation Module (DRV8704EVM) which uses dual parallel sense resistors.

The snip below is from that doc.  The arrows points to the sense trace connnections.

OP - note that all solutions shown so far do not have thermals for connections of the pads to the pour.  If you are using Eagle you would look at the properties for the pour and uncheck the box for "Thermals".  It makes it harder to solder but better for the power path.  That should get you going.  The rest of us can continue discussing the fine points so we can get our post count up to the coveted 1000 mark.
 
The following users thanked this post: jdraughn

Offline jdraughn

  • Regular Contributor
  • *
  • Posts: 106
Re: 4 wire kelvin to parallel resistors?
« Reply #8 on: September 12, 2017, 12:13:48 pm »
What is usually done to keep parts from floating around too much as you reflow the board if you don't have any thermals?

 

Online NANDBlog

  • Super Contributor
  • ***
  • Posts: 4485
  • Country: nl
Re: 4 wire kelvin to parallel resistors?
« Reply #9 on: September 12, 2017, 01:01:41 pm »
Nothing. They dont just float away, especially the big parts. Small parts like 0402 might tombstone, but I've yet to see one. I think you should go with just one resistor. Because you are going to pay for the technology, not the value. Just design in a nice 2512 or 1206 metal element resistor.
If we learned anything from Bob Pease, is that spending time on it for trivial things like motor current sensing is not worth it.
http://www.edn.com/design/analog/4422201/Bob-Pease--His-last-challenge--Part-three-The-precision-resistor
 
The following users thanked this post: jdraughn

Offline Habropoda

  • Regular Contributor
  • *
  • Posts: 56
  • Country: us
Re: 4 wire kelvin to parallel resistors?
« Reply #10 on: September 12, 2017, 02:49:14 pm »
In a case like the EVM where both pads are on a pour you would have a problem, but the solder mask keeps them from wandering.
 

Offline jdraughn

  • Regular Contributor
  • *
  • Posts: 106
Re: 4 wire kelvin to parallel resistors?
« Reply #11 on: September 13, 2017, 07:43:42 pm »
Thanks for all the replies, it's all been very useful. I had forgotten about the solder mask, in my mind I was picturing a big huge copper area where the resistors could move around willy nilly as the solder melted.  As far as removing the thermals goes, how will that effect other components on the board? I am mainly worried about the mosfets. I have been designing the board with the "directFet" package in mind, specifically the IRL7486MTRPBFCT. https://www.digikey.com/product-detail/en/infineon-technologies/IRL7486MTRPBF/IRL7486MTRPBFCT-ND/5723312]

[url]https://www.digikey.com/product-detail/en/infineon-technologies/IRL7486MTRPBF/IRL7486MTRPBFCT-ND/5723312
[/url]
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf