Wifi Module effect on Power IC measurement

hello

I have a strange problem. We use in our application the Power IC ADE7953 , and it gets its supply from a smps DC supply (1 A rated). It gives good measurement and it has no problem when it is connected alone, but when connecting wifi module of our project to the same DC source, the Power IC get blind in low current ranges (currents less than 20 mA), while it is still relatively good in high ranges. After several trials, we became sure that wifi makes some unknown behaviour that destroys the measurement of Power IC in low ranges. I know that we should have used smps supply with isolated channels, but sadly we have already bought the components. Are not there any suggestions about how to eliminate or at least reduce this effect?

Hi,

This sounds like:
* either a noise problem in the supply voltage
* or an EMC problem
* or other PCB layout problem

We can not verify this with the given informations.
* complete schematic with detailed device informations
* PCB layout
* scope pictures of supply voltages
* test conditions, reference measurement values, expected values, measured values
* maybe the IC performs a calibration routine, then we need to know the process flow and the calibration values.

Klaus

What do you mean by "blind"?

RF noise, or pulsations in power draw, could bother
the regulation loop or the measurement of it.

When you say it "destroys the measurement" it
makes me think you are only looking at results and
should be looking for causes with non-ATE
instruments. Like a high bandwidth 'scope that
is referred to measurement ground, looking at
ADE7953 ground, power, output, feedback and
other controls - including IC pins that may have
been left to float - while performing the tests
on the platform that is giving you trouble.

What does "destroys" mean, anyway - just an
unacceptable (per some arbitrary limit) repeatability,
a gross malfunction (like repeatably 3V out instead
of 1.8), ???

dick_freebird said:

What do you mean by "blind"?

RF noise, or pulsations in power draw, could bother
the regulation loop or the measurement of it.

When you say it "destroys the measurement" it
makes me think you are only looking at results and
should be looking for causes with non-ATE
instruments. Like a high bandwidth 'scope that
is referred to measurement ground, looking at
ADE7953 ground, power, output, feedback and
other controls - including IC pins that may have
been left to float - while performing the tests
on the platform that is giving you trouble.

What does "destroys" mean, anyway - just an
unacceptable (per some arbitrary limit) repeatability,
a gross malfunction (like repeatably 3V out instead
of 1.8), ???

Click to expand...

sorry for the vague word "blind"
what happens exactly, is that I put a load that draws current about 2.5 mA. If we disconnect the wifi, the calibrated power IC shows on serial terminal that the current drawn
is near this value (it shows something like 2.3 or 2.4 mA) , and this is works as needed. Also in other loads like a 0.86 A filament lamp, a 0.11 A florescent lamp with, the measurements works very well. The problem appears when wifi is connected, and the measurements of loads of florescent and filament lamp becomes nearly the same, while the 2.5mA load gets "blind".
blind here means, that if I put the 2.5 mA, the power ic measures something that looks like biased noise, it shows 0.11 mA, 0.35 mA , 8mA, 16mA etc... , also this behavior keeps happening even if I disconnected the load (left the measuring channel floating), and does not change also even if I increase the 2.5 mA load to 5mA, or 10mA , all of these loads appears as the same random noise. We tried to change the supply type, nothing changed. we tried to give the circuit a 12 A supply to be sure that this is not supply current problem, also nothing changed.

As said, this looks like a RF interference problem. Let me add for consideration that it's hardly possible to trace interfering 2.4 GHz RF with an oscilloscope probe.

ADE7953 reference schematic has RC low pass filters at the analog inputs and sufficient power supply bypass capacitors. But are they present in your deigns and are they placed effectively?

I'm talking about a continuous ground plane (or at least an almost continuous copper pour) below the energy meter IC and each bypass and filter capacitor connecting directly to it. If necessary also series ferrite beads for external connections that might pick up RF.

The datasheet claims 0-1% - 0.2% (of full scale) accuracy -
but this presumes you set up your application as nicely as
ADI sets up their evaluation test systems - a rather poor
bet.

What is your full scale range as you have things set up?
1A full scale would say no way are you getting clean data
at anything below a milliamp. Even without board design
and interference nonidealities.

You might play with the WiFi module's operating modes
and see whether this is TX activity, data link activity,
happens in only some state or other (perhaps states
are cycled through and you can see an interferer on
the current reading or current-related pins if you went
and plotted samples strip-chart style - although it looks
like ksps data rate in my skim of the datasheet and this
might really "decimate" the real goings-on. Still some
'scope poking might show -something- at some frequency
that can be traced back to its source (like TX / RX cycling
or serial bus packet bursts or whatever).

I remember a while back looking at current consumption for a WiFi module with a high frequency current probe.

I remember being surprised that, although the average current consumption was low, its crest factor was something like 5:1 or 7:1 and fairly quick slew rate!
That was causing the power supply to droop momentarily. To make a long story short.....the problem was eventually solved by replacing the power supply with another from a different brand.
We surmised that the loop bandwidth of the offending supply was slow. Many power supply designers, in an effort to provide nice phase-margin specs, lower the crossover frequency too much. This is fine if your load is not changing rapidly and/or if the load will tolerate some drooping.

Perhaps something similar is happening to you.