Question regarding burden resistors on Current Transformers...

Hi everyone,

I want to confirm an observation... just to be sure I get it right. Specifically, the use of burden resistors on current transformers.

For energy monitoring applications, many ADC datasheets show the use of a current transformer with a burden resistor and a RC circuit to keep the output voltage within the limits of the ADC inputs and to filter out higher harmonics. What I find interesting is that some datasheets show the burden resistor simply connecting across the outputs of the current transformer. This configuration is used by the open-energy monitor group, for example. That said, applications like the openenergy monitor use a single-ended ADC without a bipolar input capability.

However, for fully differential bipolar ADCs, energy metering application notes for chips like the AD7753 (p.4) as well as the MCP3911 (p.20) show the burden resistor split into two, with both of them being attached to a respective current transformer output as well as analog ground.

I presume that the split burden resistor would help bias the common-mode voltage of the differential ADC to be zero? However, what confuses me slightly is that the sample circuit datasheet for the AD7753 on p.15 shows a single burden resistor on a current transformer. Is that just a simplification given that their application note for an actual power meter shows a split burden resistor configuration?

So what drives the decision to split or not to split the burden resistor on a current transformer when using a differential ADC like the MCP3911 or the AD7753?

Constantin:
So what drives the decision to split or not to split the burden resistor on a current transformer when using a differential ADC like the MCP3911 or the AD7753?

I think it depends on the design of the current transformer. For most current transformers, I would expect the split burden resistor arrangement to give best rejection of transient voltages on the primary, which may be capacitively coupled to the secondary. OTOH if the current transformer has a shield between windings, with one of the secondary terminals designated as ground and the other as output, then use a single burden resistor. But this isn't my area of expertise, so I'm prepared to be corrected by someone with more experience of current transformers.

Thanks DC42!

I presume that the common mode issue may also explain why Analog attaches Neutral directly to one of their ADC inputs (see p. 5 of the application note for the AD7753) for the purpose of line voltage measurements. Another explanation is the benefit of fewer drop resistors being required...

My voltage sensing is done through a tiny 0.08VA EE20 transformer. Do you reckon that I should ground one of its outputs per the Analog recommendation? Currently, I have two parallel paths to the analog inputs, both featuring a 36k resistor in series and a RC filter (1k and 100nF). The Analog application note suggests dropping the second line and just keeping the two RC networks. Interesting! Presumably, I'd have to double the resistance of the remaining resistor under the circumstances.

Another option could be to use a transformer with a center tap on the secondary for analog ground. Then run two lines to the ADC as before, but this time with 2x resistance since the output voltage effectively doubled (2x6V vs. 1x6V output).

Constantin:
I presume that the common mode issue may also explain why Analog attaches Neutral directly to one of their ADC inputs (see p. 5 of the application note for the AD7753) for the purpose of line voltage measurements. Another explanation is the benefit of fewer drop resistors being required...

My voltage sensing is done through a tiny transformer. Do you reckon that I should ground one of its outputs per the Analog recommendation and then use the (much smaller) drop-down resistors for the input signal? Currently, I have two parallel paths to the analog inputs, both featuring a 36k resistor in series and a RC filter (1k and 100nF). The analog application note suggests dropping the second line and just keeping the two RC networks. Interesting!

Another option could be to use a transformer with a center tap on the secondary for analog ground. Then run two lines to the ADC as before, but this time with 2x resistance since the output voltage effectively doubled (2x6V vs. 1x6V output).

They connect the neutral to V2N because it needs to measure the voltage between line and neutral, and neutral is near ground potential. Your current arrangement is OK assuming that the bottom ends of the 1K resistor and capacitor are connected to the chip ground, and provides better protection than capacitively-coupled transients than with one side grounded. But having one side grounded would probably be satisfactory too. I wouldn't bother wish a centre tapped secondary, the 2 voltage dividers to ground have a similar effect already.

You will have some phase shift in the transformer, due to the finite primary inductance and primary resistance. However, you should be able to cancel this out by using larger resistors or capacitors in the RC filter.

Hi DC42 and thanks again.

I measured the phase lag to be about 1.32ms between the current phase and the voltage phase using my oscilloscope. However, that is using a light, unitary PF load (i.e. 40W incandescent lamp). Under higher load (i.e. 1340W unitary PF toaster), the phase shifts decreases to 0.7ms. As mentioned over in the science section, these results have baffled me somewhat. Pito suggested this may have to do with the mains, something I need to follow up on by testing at the 400A power panel vs. in the kitchen.

Either way, either lag cannot be fixed using just the phase register in the MCP3911 since the lag is well outside the realm of the ADC phase register range (+/- 0.3ms with the current settings). As I see it, I have two options: adjust capacitors and resistors as you suggest (not that easy on a 0805 footprint) or accept the lag and simply use offset pairs of readings in the calculations (the ADC is sampling at about 1418 sps).

If my oscilloscope readings are correct, it might make sense to simply pair ampere reading (x-2) with voltage reading (x) to achieve a ~1.4ms correction and then use the phase register to correct the remaining 0.08ms. That is, assuming the light load phase lag is correct.

I would expect the voltage transformer to introduce a small voltage lead, and the current transformer to introduce a current lead. Are you using the recommended burden resistor on the current transformer? What happens if you reduce the burden resistor? (put another resistor in parallel with the one(s) you already have if they are difficult to change).

Adding a resistor in series with the voltage transformer primary will, I think, increase the voltage lead and therefore could be used to correct the phase error. Look up or measure the primary resistance of the transformer (probably a small number of kohms for a small transformer), then try adding the same resistance again, or more.

0805 resistors are not hard to remove, if you use a soldering iron bit large enough to cover both pads and a solder sucker. But you don't want to be changing them too often, or you may damage the pcb pads.

Hi and thanks for the suggestion... I may simply remove the burden resistor (single, connecting the two outputs) and start by splitting the burden resistor into two, using through hole resistors instead, terminating them on AGND. I did note a higher offset error on the current line than on the voltage input (i.e. less than one LSB on the voltage line vs. multiple on the current line). I presume that 'anchoring' the bias on AGND via a split burden resistor will help with that bias.

AFAIK, I did calculate the burden correctly - this is a CR8348-2500-N and the burden I have on it now is 63 Ohms. The max input into the ADC is +/- 0.6Vpeak, so I sized the max. output to be 0.6V assuming a 18Arms load. That corresponds to an R of 59 Ohms and the closest precision resistor I had on hand was a 63.4 Ohm model. The output seems to be OK vis-a-vis currents measured vs. output voltage.

A smaller burden resistor for the current transformer is a possibility since I have multiple gain settings to play with. The additional noise of Gain =2 vs. unitary gain on the input side is pretty negligible. I'll measure the primary resistance on the transformer and try your suggestion. That will require a higher-voltage resistor like a 1W flameproof model, right?

Constantin:
I'll measure the primary resistance on the transformer and try your suggestion. That will require a higher-voltage resistor like a 1W flameproof model, right?

The actual power dissipation in the resistor will be quite low (should be less than 0.5W, probably a lot less), and I don't think you will need to drop more than a few tens of volts across the resistor, Nevertheless, I suggest you choose a resistor with at least 100V voltage rating. Most types of through-hole metal film resistors should be OK.

Hi DC and thanks again!

For some reason I thought I would have to have each resistor be able to handle a 250v working voltage safely - so they tended to be 1w models.

Fwiw, the output from the transformer is pretty good, despite its 0.08va size. I need to compare the sinusoidal curves to references in the scope but so far it looks like I could use constant plus a multiplier to predict voltage to an R^2 of 0,9992 when just reading averages off my fluke. I plan on sampling data with a yokogawa wt310 later this week to see how well it does when comparing sampling sets with each other. I expect the Yoko to be a step above the fluke or the kilawatt.

I would be very happy if I can continue to use the voltage transformer since i would then not have to worry about using a adum5xxx series chip to isolate the adc from the rest of the circuit.