Power Quality Analysis with the QA403: Proof of Concept

Hi @djrix, another board has been released using dual 1206 for the issue you identified. Thanks very much for catching. Other changes include getting rid of the isolated DCDC and relying on a second RAC03-05SK module. So, this is two modules, one for generating 5V for primary, and another for generating the 5V for secondary.

Also, the RAC03-05SK ripple is fairly high (50mV or so) and was showing up in the output via limitations on the AMC1200BDWVR PSRR. And the ripple occurs at the frequency we’re most interesting in measuring. So, there’s now a 4.5V LDO on both primary and secondary so that the supply to the AMC1200 will be ripple free. The noise density on the AMC1200 is around 800nV/rthz, suggesting a good win with the ripple knocked down.

AC Line Isolator REVE.pdf (60.9 KB)

Update on REVE Performance of the Line Isolator:

The no-load condition of the isolator is significantly improved with the LDO addition. The plot below shows the current output from the Line Isolator. Note the 60 and 120 Hz are measuring about -100 dBV, The 1A output level is -21.94 dBV, so this is about 80 dB below 1A, or about 100uA of resolution with 12A full scale (suggesting slightly better than 16-bits). There is some hash out at 1.5 to 2.5 kHz (varies by unit) that looks related to the start/stop of the isolated AC/DC converters. That is, when not delivering the full load, the converter stops operating for a bit, until the output voltage on the output cap drops, and then it starts again. So, as your load changes, this peak moves around. Not ideal.

The point of entry for this hash on the REVE boards for this was due to poor routing and combined grounds on the primary side. If the supply to the AMC1200 on the primary side was disconnected, the peak went away. I tried the best I could to fix the routing with flying wires, but wasn’t able to do much. So, a REVF should hopefully clean this up. The good news in the LDOs used have good ripple rejection around 2 kHz (about 70 dB) and so this 15 mVpp noise coming out of the ACDC should be knocked down to under -100 dBV.

This is a plot with no-load connected. Note that the Irms is shown as 4.3mA. But this is dominated by the hash. If the hash can get fixed, this should drop to well under 1 mA.

Here’s a Tesla Model Y wall charger idle, with charging off:

Here’s the same charging the car at 5A:

And here’s the same charging the car at 12A:

Here’s a Dell laptop charger at idle (nothing connected)

Here’s the same laptop charger with laptop running:

Attached are schematics for REVF (not yet built) and also gerbers and a xls BOM. There are probably two more revisions that will be done before this is considered done.

Recall the aim here was to look into the feasibility of using an audio analyzer to safely measure line-loads up to 12A. The huge dynamic range of the analyzer could then measure down to extremely low-currents without having to go through ranging. The use of direct sensing (instead of current clamps) means the accuracy is much better and offset and drift errors are almost completely eliminated.

There’s still more work to do here, but the end is in sight.

Gerbers, schematic PDF and BOM attached. Again, these are for REVF which hasn’t been built yet.

Line Isolator REVF Design Docs 20231029.zip (197.1 KB)

Hi Matt

Can I build this device? Is this function active in the program or can I activate it?

Merry Christmas for everyone!

Hi @Krisz77,

Yes, I think the REVF is good to build. It’s been subjected to 2.5 KV across the isolation barrier and exhibited >6Gohm. I noted in the another post that I think this will become a product, and CE safety will be pursued (but not UL. CE is generally more strict than UL, and mandatory for sale in EU, while UL isn’t mandatory for sale in the US).

In any case, the software shown above in this post was just for proof of concept. I now think a better route is something I posted elsewhere in the forum, which is a dedicated app for power quality analysis. An early version of that app is shown below. This lets you see time domain waveforms of the measured current and voltage, along with spectrum. The acquisition is continuous time, such that every 200mS a 32Kfft is done on the voltage and current, and from that true power (W) and apparent power (VA) is graphed. Ultimately, this will also let you keep an eye on single cycle excursions for in-rush current, or single-cycle line fluctuations (if you developing a UPS, for example).

I have no idea when hardware might be available, but if you decide to build your own, I’d send you a case once your board is complete and you could run a version of the app below with QA403 hardware.


1 Like

That sounds great Matt! Yes, I want to build one for myself.

I realize I’m super late to the party here, and nobody likes scope creep. But one of the super important things to look at is ground leakage, especially at higher frequencies. Since that leakage current may go off in other directions than through the green wire, the best evaluation is to look at hot current and neutral current at the same time, whether they are individually evaluated, or simply subtracted in a differential amplifier such as the input of a QA40x…

Second, it appears you are looking at voltage between the hot and ground, rather than between hot and neutral. This misses voltage drop on the neutral conductor between the bond and the input.Some devices are sensitive to that and it can tell you a lot about the quality of the source. I’ve designed many technical power systems (separately derived, isolated ground, etc) for AV and G-N noise is an important part of that.

At the cost of nearly doubling the parts count, my wish list would be to have a hot current and neutral current output on two BNCs, so by simple patching you can look at each, both, or difference them. I also want voltage between hot and ground on one jack and between neutral and ground on another. Likewise you could analyze either, both, or difference them.

Complete power analysis involves evaluating what is being supplied by the source and what is being done with the power by the load DUT. As presently implemented it really only looks at part of the picture.

Thanks, and Merry Christmas,

Dale Shirk

Hi @daleshirk, thanks for this very valuable input! It does feel like there’s a more complete product somewhere in here. Maybe take something like an 8 channel audio ADC, make the product USB powered (for the secondary) and keep the primary powered as before. And then add in a host of isolated opamps for the various measurements needed. And then, there’s not a QA403 at all in the picture.

I just got one of the commonly available “power station” products because I was curious what modern inverter outputs looked like. These products are usually a 100 WH or larger battery, and an AC plug. And you charge the battery via solar (or line), and then you can run USB plugs and an AC product or two when the power’s gone. In any case, I took a look at the AC it generated and the THD was very good, with harmonics below 40 dBc. the output voltage was nominally 110V at no load, and dropped 5V under moderate loads (40W). So, something else interesting to know here would be battery voltage and current, and then inverter efficiency could be known.

Do you have any insight into the frequencies where neutral currents can leak onto ground? For example, it’s pretty easy to verify there’s no path at DC from neutral to ground. But once X/Y caps are added, it gets a little harder to measure and so the path you outline becomes very important.

I think the learning from this effort thus far is as follows: The isolated opamps have plenty of dynamic range for the signals involved, bandwidths all seem sufficient, the primary/secondary protections are suitable, etc. You are right, there are a lot of directions this could go. I think as-is, it can give some guidance to those looking to understand how their product consumes line power. But it’s not very helpful in terms of revealing DUT attributes that would be interesting from a safety perspective.

And Merry Christmas to you too!


To take your points in order. You’re quite right that power analysis does not require the ADC quality of the QA40x. A rather pedestrian consumer quality 8- channel audio ADC with proper gain staging and isolation could result in a more capable power analyzer. To test at the supply panel level it also needs to handle Clip-on current sensing. If one wants to compete with Fluke and others on that, that’s another large project.

Analyzing inverter efficiency is another big project, involving DC voltage and current measurement as you suggested. For one or two units, one could just measure the DC side with VOMs and do the calcs manually.

RE frequencies of interest; both capacitive and inductive coupling increase with frequency, therefore pretty much the whole audio band is in play. When I started in pro audio in the 80’s we fought “dimmer buzz” which was most audible in the several KHz region because of the peak in the ear’s sensitivity, but on the RTA, it could run out the top of the audio spectrum. It was actually voltage induced onto the ground wire by unequal magnetic coupling of the equal-but-opposite fast rise time currents on the hot and neutral. We ignorantly blamed proximity of “large AC fields” while it was the power cable geometry the whole time. The ground differences became audio because of unbalanced interfaces and Pin-1 problems. (Equipment that is sensitive to current flow on the signal Shield/ground terminal) Today “Dimmer buzz” isn’t just dimmers, but every cheap switch mode power supply, but we have cleaned up Pin-1 sensitivity, and have largely moved to digital signal transport.

You mentioned measuring DC isolation Neutral to ground. To do a full analysis you’ would also measure capacitance neutral to ground and hot to ground. Then you would also need to know about the inductive coupling neutral to ground and hot to ground. To measure that directly and in isolation you need to drive current through each, which means access to both ends and a separated return path. I’ve done those measurements in my study of power wiring geometry and it’s not trivial to keep noise and RF out. That’s why I’m interested in simultaneously measuring the current on both legs going by a particular point and comparing them to see if any current is escaping. UL requires, if memory serves, less than 50 microamps leakage for ungrounded equipment and 7 ma for grounded equipment.

Full safety testing includes Hi-Pot testing, which is a whole 'nother kettle of fish. There are tools for that. I’m more interested testing the actual leakage of grounded equipment that passes UL but may have design flaws or problems introduced by repairs.

To that end adding a second voltage and a second current sensor shouldn’t involve extensive design work. The only significant issue is to make sure that the hot side current sensor can safely and correctly float away the 120Volt common mode that rides on it. Maybe easier for me (not a circuit level designer) to say than for you to do.


Thanks, @daleshirk, all very helpful. I think the market for full-blown power quality analysis is well covered. For leakage, there is of course the hi-pot tester which is usually non-destructive. But, there are also low-cost meggers that will let you step from 250V up to 2500V that can achieve almost the same. Tie LINE + NEUTRAL together, connect the hot side of megger to that, and then check impedance at 250, 500…2500. That won’t get you coverage at 50 or 60 Hz, but it will probably shake out the most common faults.

hello Matt,
i want to build this power analyzer but before to start i need to know if the pc app is available somewhere for download?

Hi @DATIR, let me see about getting this posted. It’s a really useful tool for the QA403 for sure.

Hi @matt,
yes, it is a nice tool and can be very helpful together with qa403