2
\$\begingroup\$

I am building a digital wattmeter intended for measurements on AC power lines. I have went through a lot of application notes and example designs, and stumbled upon a slightly confusing motif recurring in a number of designs from Texas Instruments' TI Designs.

They employ a similar frontend for voltage measurement, where the mains voltage is first scaled down using a voltage divider. The scaled-down voltage signal is then taken as differential and passed through some low pass filter before being fed to a differential-input amplifier or directly to a high-resolution differential-input ADC.


Question 1

What is the purpose of the connection from Neutral to the reference ground, if it results in an impaired differential signal?

schematic 1 (simplified diagram of the input section from SimpleLink™ Wi-Fi® CC3200 Smart Plug Design Guide)

After running a simulation of this input circuit, I have found that the ground connection on the Neutral line renders V_SENSE- signal thousands of times smaller than the V_SENSE+ signal, and causes a non-180° phase difference between them, resulting in a suboptimal differential signal. On the contrary, if the connection is removed, V_SENSE+ and V_SENSE- signals are of the same magnitude and 180° relative to each other, creating a perfect differential signal.

What is the purpose of this connection from Neutral to the reference ground?


Question 2

What is the purpose of using a smaller value for R5? Why is it that the previous configuration use the same value for both R5 and R6? How do I know when to use which?

schematic 2 (simplified diagram of the input circuit from Smart Plug with Remote Disconnect and Wi-Fi Connectivity)

This is largely the same as the previous configuration. Apart from the differences in component values, the most intriguing difference is in R5, which is 100Ω as opposed to 1kΩ for R6. According to the document, the justification for the smaller value is to make up for the much larger impedance at V_SENSE+. However, simulation results show that the outputs are almost identical to those of the previous configuration. Moreover, the magnitude mismatch and phase angle problems still exist.

What is the purpose of using a smaller value for R5? Why is it that the previous configuration use the same value for both R5 and R6? How do I know when to use which?

\$\endgroup\$
  • 1
    \$\begingroup\$ Have you tried contacting TI? \$\endgroup\$ – Andy aka Aug 16 '16 at 17:50
  • \$\begingroup\$ Yes. In fact, I've posted the same question on TI's forum 3 days ago. \$\endgroup\$ – user120905 Aug 16 '16 at 18:04
  • \$\begingroup\$ Here is the link to the question: link But it has rarely been read. This one of the many TI forums is so rarely visited that I fear I might not have the question answered any time soon. Also I notice that the TI technicians are more likely to answer questions directly related to their ICs and products. \$\endgroup\$ – user120905 Aug 16 '16 at 18:10
  • \$\begingroup\$ What chip to the V_Sence wires connect to? \$\endgroup\$ – Andy aka Aug 16 '16 at 18:22
  • \$\begingroup\$ It could be connected to a differential-input ADC (VSENSE+ to Positve input and VSENSE- to Negative input) to directly measure the voltage. Example: ti.com/lit/df/tidrb17/tidrb17.pdf Or it could be fed to an instrumentation amplifier to produce an amplified single-ended output, which similarly will be read by an ADC, but this time without the need of a differential input. Example: ti.com/lit/df/tidrfq8/tidrfq8.pdf \$\endgroup\$ – user120905 Aug 16 '16 at 18:29
2
\$\begingroup\$

Despite the galvanically isolated supply, you cannot have ground floating around wrt the line- it has to be connected (more or less) to one side of the power, specifically the side with the current shunt connected to it.

The sense voltage at the low end of the voltage divider would be expected to be almost zero.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ Thanks for answering my first question. :) But if VSENSEP and VSENSEN are perfectly differential, why would it matter if any of the subsequent signals after the filter are not referenced to the neutral? Also, if VSENSEN is almost zero, what is the purpose of the LPF formed by C7 and R8? Would you mind giving me a more detailed answer, coz I'm a newbie. Thank you. \$\endgroup\$ – user120905 Aug 16 '16 at 18:19
  • \$\begingroup\$ There's a limited common mode range in a real differential amplifier- a few volts at most. The LPF is shorted out by the two GNDs so the reason is to filter voltages that appear across the PCB traces (presumably why they are using a diffamp rather than single ended in the first place) \$\endgroup\$ – Spehro Pefhany Aug 16 '16 at 18:29
  • \$\begingroup\$ Thank you. I now understand that the reference ground prevents an excessive common-mode voltage. But I don't understand "to filter voltages that appear across the PCB traces". Do you mean the voltage developed across the two GND_S (one at the neutral and the other between C5 and C7)? Also what would be the primary cause for such a noise voltage? Is it due to the large current nearby (in the shunt)? \$\endgroup\$ – user120905 Aug 16 '16 at 18:38
  • \$\begingroup\$ Yes, current in the shunt could be one reason but if you are trying to measure very accurately, even a few mA through a trace or across a ground plane can affect the measurement. 150V into the divider is ~230mV at the resistor and 0.1% of that is 230uV. Only 5mA through a 1" 10 mil 1-oz trace will cause that. \$\endgroup\$ – Spehro Pefhany Aug 16 '16 at 19:17
  • \$\begingroup\$ Thank you for your answers. I have one more question regarding this design from TI. According to the full schematic the LMV324's are powered from the +3.3V (Vcc) and -3.3V supplies, which result in a total of 6.6V across the LMV324's supply pins. However, the datasheet of LMV324 specifies the absolute maximum supply voltage to be 5.5V. The reference design couldn't be wrong (or could it?). What have I missed here? \$\endgroup\$ – user120905 Aug 18 '16 at 18:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.