Current Sensing circuit issue 3

[Katy1163]

I have recently designed a servo loop circuit in which current is sensed across a 0.1 ohm resistor and fed back into the input. The difference across the resistor is fed into an opamp with a gain of 10. The output of this opamp with the current artwork is approximately 30% higher than expected (a gain of about 12.6). The previous artwork of the same circuit functioned correctly and the only difference that was noticed is that the trace on the low side of the current sensing resistor is thinner than the other side. The way we found to fix it is to supplement the trace with a jumper, which indeed fixed the issue. The problem is that the calculations don't match. The increased resistance due to the smaller trace should not be enough to affect the output by 30%.

A couple notes: The values of the passives in the circuit were measured out of the circuit and found to be correct. The problem is known to be a common-mode gain of about 0.03, but not sure where this gain is coming from (typically impedance mismatch). The opamp is functioning correctly based upon measurements and calculations.

Any suggestions of things to look into would be appreciated!

 

[btrueblood]

"The increased resistance due to the smaller trace should not be enough to affect the output by 30%."

Sure, if the trace is uniform at the dimension measured. But, if it's an etched trace, there may be very small areas (cracks, flaws) where the trace width is actually much smaller than you think it is.

 

[Katy1163]

We had measured the impedance of the trace and it was only about 50 mOhms. The trace is only about an inch long. Also, there were two separate boards with the same artwork and both boards were acting in the same manner, so it was not a random flaw in just that particular board. Thanks for your input!

 

[MacGyverS2000]

So you're using a sense resistor of 100 mohms, and your two connecting traces combined are 100 mohms, as well? Way too much PCB variability possible for my taste. The traces should be an order of magnitude (at least) less resistance than your sense resistor for good repeatability. An over-etched board could easily account for 30% mismatch, or a nick somewhere, etc.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

Milliohms add up pretty quickly. The ostensible solution would be to use a Kelvin, 4-wire measurement across the resistor, i.e., using non-current carrying sense lines.

TTFN

FAQ731-376

 

[Katy1163]

"So you're using a sense resistor of 100 mohms, and your two connecting traces combined are 100 mohms, as well?"

The one 8 mil trace branching off from the sense resistor to the input resistor to the positive side of the opamp is about 50 mOhms. The trace branching off of the other side of the sense resistor is a 20 mil trace. The way we determined that the output is off about 30% is by measuring the voltage across the sensing resistor and calculating what the output should be based on measured values of resistors (out of the circuit). With a trace impedance of 50 mOhms in series with a 0.05% 1K resistor, the trace impedance should not be an issue.

"The ostensible solution would be to use a Kelvin, 4-wire measurement across the resistor, i.e., using non-current carrying sense lines."

Unfortunately, this configuration can not be changed now and also the problem is "fixed." What I am wondering now is exactly what the cause of the problem was and why the jumper fixed it.

 

[Comcokid]

Another question is what kind of current are you sensing with the 100mOhm resistor? A constant DC, or a PWM signal at some base-frequency rate?

When doing switch-mode design with a current-mode controller with a resistor for current sense, care has to be taken in the current sense, and the traces and resistor used for the current sense need to be non-inductive or errors will result. Problems will occur even at switching frequencies of a few tens of kilohertz.

Now, I realize that your doing a motor servo loop of some type and not a switch-mode power supply, but the issues are similar.

 

[IRstuff]

??? your sense resistor is 0.1 ohm or 1K? That's a big difference.

With a 1K, then it's more likely to be some sort of inductive issues as suggested by Comcokid, possibly causing some amps to oscillate, which could create all sorts of anomalous effects.

TTFN

FAQ731-376

 

[MacGyverS2000]

With a trace impedance of 50 mOhms in series with a 0.05% 1K resistor, the trace impedance should not be an issue.

Agreed... but you said the sense resistor is 0.1 ohms. Now you're saying it's 10,000 times larger. Which is it?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[Skogsgurra]

Input resistor >>> sensing resistor. Differential amplifier, I presume.

(">>>" means "much bigger")

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[RajeevCell]

I think, I can visualize Katy's circuit. But a circuit diagram would have helped in discussion.

As far as I understand about this circuit, it is already using "four wire" connection for current sensing. Load current flows through large traces, into the sense resistor (0.1 ohm), and smaller traces bring the "voltage across the resistor" to an OP-amp. The 1 kOhm resistor mentioned is not the sense resistor, it is one of the resistors of the amplifier network. A circuit diagram would indeed have helped.

@Katy, I do think that both of your PCBs had the same etching defects. I cannot think of any other explanation.

 

[Katy1163]

Yes Rajeev, exactly. Unfortunately, I don't believe I am permitted to post a diagram of the circuit, no matter how simple.

I have tried the circuit with a DC current and with a 100Hz sine wave input, since the input signal will not be changing too rapidly. In both cases, it had the same result.

 

[Henneshoe]

I attached a quick sketch of what I assume your circuit looks like. If this is correct, I would check the power connection from your sense resistor to the ground plane. If you have an open via in that leg, all of your current will flow through your sense leg. In this case, the 50mOhms of trace resistance would be added to the impedance of you sense resistor and could cause your error.

Let us know if my sketch is incorrect.

 

[IRstuff]

While it's unlikely that you'd get opens at the same via over and over again, it is extremely likely that you might have a cold joint at that spot, and addition of soldering the jumper made the connection whole again.

This would be particularly true if the resistor's solder pad is not correctly laid out. Usual best practice for components connected to large copper areas is to have the solder pad isolated from the main pad physically, and only connected through normal, short traces, sort of like a Maltese cross in a hole in the large copper pad. This retains sufficiently low connection resistance, but provides a high thermal isolation for proper soldering.

TTFN

FAQ731-376

 

[itsmoked]

Keep in mind that copper has a huge tempco that changes resistance dramatically with temperature. You could be having a thermal(Seebeck)issue that is being stopped by the heatsinking of the added jumper and of course the resistance drop of the jumper further reducing heating.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Henneshoe]

I had a layout once where a via was connected to a trace and went into the ground plane, but for some reason the via did not inherit the net name of the trace. The effect was the ground plane did not connect to the via. Therefore, on every PCB that via looked open.

I am not saying that is the issue here, but it can happen.