Power-company voltage waveform measurement

[DrDeke]

I have been having some unusual power quality (rapid voltage fluctuation) issues at home recently, and I would like to see approximately what the voltage waveform is doing during these events should another one happen.

In an attempt to measure/record the waveform, I connected a transformer to the mains with a rated output of 8 V, 1000 mA. I connected the transformer output to a(n approximately) 10:1 voltage divider consisting of a 22k resistor and a 2.2k resistor. I then connected the voltage divider output to the line-in jack of a computer's sound card and used Audacity to capture and record the signal. I am a bit confused about the results; the recorded waveform appears to be clipping (or something like that), but only on the signal's positive peak voltages, not its negative peaks:



I thought perhaps the clipping might be caused by the voltage still being too high for the sound card's input circuit, so I added another 22k resistor in series with the first one, bringing the ratio to 44k:2.2k. Other than reducing its magnitude, this had no effect on the shape of the recorded waveform.

I think that the quality of this waveform will probably be good enough to answer the questions I would like answers to if the power quality event happens again, but I am still curious what might be causing this apparent distortion on only one half of the AC cycle. I look forward to hearing any ideas you all may come up with.

[wasedadoc]

If you swap over the connections to the transformer secondary does the waveform remain the same or do the flat tops move to the negative half cycles?

[DrDeke]

I tried that just now and the waveform remained the same - the flat tops stayed on the top/positive half cycles.

[TizianoHV]

Could be leakage from the transformer to the PC psu, a battery device (laptop, smartphone) would clear this issue.

What about adding a battery in series to the divider? (to create an offset and see if changes something...).

[Buriedcode]

A soundcard is less than ideal for measuring voltage, its not really for absolute accuracy, plus the maximum input voltage varies by manufacturer, so you really have no idea at what voltage the signal clips.

As wasedadoc points out, by swapping the polarity, if it still clips the top you know its not the signal itself that is clipped, but your setup causing the waveform to be distorted.  If it was the signal looking like that, the bottom would be clipped when you swapped the wires.

Your 8V transformer will likely have a much higher output than 8V since the resistor divider and soundcard present very light loads (~24k input impedance) - so even with your 11:1 voltage divider, and a transformer voltage of say 10V RMS would be 10*1.414 / 11 = 1.29V (thats peak from 0v, so p-p would be double this, at 2.58V p-p).  Thats less than 1V RMS, which most soundcards can handle, but as I said, the transformer could be outputing a much higher voltage than the rated 8.

If you have access to a known amplitude signal, say a 1kHz 1VRMS source, you can use this to check your soundcards input capability, and amplifiying it to see when it starts to clip. 

In the meantime I would try to load your 8V transfortmer somewhat. Say with ~200mA, and then checking it again in Audacity.

[wraper]

Did you try to attach a capaitor in series? I suspect you are driving the input of sound card below GND and it does not have a bipolar power supply.

[wasedadoc]

Did you try to attach a capaitor in series? I suspect you are driving the input of sound card below GND and it does not have a bipolar power supply.

I never encountered an unbalanced line input on a soundcard that did not have its own input capacitor.  Mic inputs are a different kettle of fish as they sometimes provide dc for electret mics.

[Terry Bites]

You want a reasonable load on the XFMR.
2k2 is very low.

I agree, this looks very much like clipping. Make sure you sure you're using the line input not mic. Mic has a DC pullup.
You don't want more than about 150mV on the soundcard line input. I'd go less than that, say 50mV.
As pointed out, you still have over 500mV peak from your divider. (8*1.4/20)V
 Check the gain on your PCs audio mixer control.
Audacity may not be the best thing to use here even with 0dB rec gain.

Get a copy of digilent "waveforms", it has a soundcard option and its free.
digilent.com/reference/software/waveforms/waveforms-3/start

[Simon]

I take it an oscilloscope is not available?

[DrDeke]

Thanks to everyone who has replied so far! I have several updates:

* First, I tried putting a 40 ohm resistance across the transformer secondary, so as to pull about 200 mA. This did nothing at all to change the recorded wave form.

* I confirmed that I am definitely using the sound card's line-level input, not its microphone input.

* I increased the value of R1 in the voltage divider from my already-increased 22k 44k value to 44k 88k. This results in 0.221 V RMS going into the sound card (as measured with an RMS DMM in AC voltage mode), and also fixed the issue! The waveform now appears mostly sinusoidal with only a small amount of "flattening" near the positive and negative peaks.


I also have some follow-up questions about oscilloscopes and the Digilent Waveforms software:

* While I do not have an oscilloscope available to me at the moment, how would I use one to capture these events if I did? I have very, very little experience using oscilloscopes for making anything but the very simplest of measurements and I am unsure how I would set up a trigger to capture these events. Is it possible to set most oscilloscopes to trigger on a condition along the lines of "the voltage stays below X volts for Y consecutive milliseconds" or something like that? If not, what approach would you recommend?

* I set up the Digilent Waveforms software that Terry Bites mentioned, and it looks much more useful for analyzing one of these events (should I manage to capture one) than Audacity would be. But I am having some trouble getting the software to do what I want. What I would like it to do is continuously capture and record the soundcard channel and write the data to disk for later analysis.

I can capture data using the regular oscilloscope tool in Waveforms, but there appears to be a limit of something like 2M samples per capture. At 8 kHz, this only amounts to 250 seconds. Since I can't predict when an event will occur and don't know how to set up an appropriate trigger, I seem to need a longer recording period than this. There is a 'REC' button in the oscilloscope window which brings up a dialog box with settings to capture data in various formats for extended periods of time, but no matter which file format I tell it to use, it seems to record all values of "0" for each sample instead of the actual value measured by the sound card.

In the mean time, I have gone back to letting Audacity capture sound card audio (having made a note of the capture start time), but if anyone has any suggestions for getting Waveforms to capture the needed data, I'd be very interested to hear them.

[Simon]

Maybe we need to go back to the beginning. You say you have power fluctuations, how do you know? where do you see this? what happens?

[BeBuLamar]

I am not sure if the waveform would have anything to do with the power functuation but I think in this case you have the problem with your sound card. Not so much how you feed it but something is wrong with your sound card. Try to record something from an audio source and see if it's still clip the positive side of the waveform.

[jonpaul]

extreme overload of PC audio input. Checdk soecs of audio input to PC.

Use 100 or 1000:1 atten

Jon

[BeBuLamar]

extreme overload of PC audio input. Checdk soecs of audio input to PC.

Use 100 or 1000:1 atten

Jon

I don't know how to overload only the positive half and only recorded as 50% of max. If I use an opamp and set the offset toward the positive I can clip the positive only but then the amplitude would be at 100% not 50%.

[DrDeke]

Not so much how you feed it but something is wrong with your sound card.

The sound card is working correctly in general, and the soundcard waveform problem was fixed by further increasing the ratio of the voltage divider (R1=88k, R2=2.2k) per Terry Bites' suggestion.

extreme overload of PC audio input.

Indeed, further reducing the voltage being fed to the soundcard fixed the problem.

Maybe we need to go back to the beginning. You say you have power fluctuations, how do you know? where do you see this? what happens?

Ahh, yeah, good point; I had not mentioned what _kind_ of fluctuations I had been having and wanting to measure.

The electric power in my neighborhood is fairly unreliable. This usually manifests itself in frequent prolonged total outages, or in brief (3-5 second) total outages during which a temporary fault presumably trips a protective relay which is then closed after a few seconds by an automatic recloser.

In the past two days, however, I have been seeing a brand-new-to-me type of power problem. The new problem has not occurred again (yet) since setting up the soundcard based monitor, so all I can go by to describe what's happening are my own observations.

What happens is that for a period of somewhere around 4 seconds, the lights in my house flicker deeply and rapidly, at what appears to be a steady pace (maybe somewhere around 8-10 Hz, but again, this is estimated from my visual observation, not measured). At the end of the event, power returns to normal. Further observations:

* This affects lights on multiple branch circuits, on both legs of the 120/240 V, 60 Hz split-phase service

* When I say the lights "flicker deeply", I mean they (visually) appear to turn all or almost-all of the way off during the flicker cycle. This gives a strobe light-like appearance, quite unlike the momentary brightness fluctuation you might see if you turned on a large motor or other non-resistive load in the house. 

* When this happens, it causes both of my double-conversion UPSes to transfer their loads to battery and record "ConditionInputFrequencyDeviation" and "Input Power Supply Fail" messages to their logs. This seems notable because these UPSes are quite tolerant of variations in input voltage and frequency. Their spec sheets indicate that they can accept input voltages between 70-140 V and input frequencies between 40-70 Hz without having to transfer the load to battery.

I have never seen a power problem like this before, and I have no idea what conditions on the distribution system might cause something like this. This is my reason for wanting to measure the utility voltage waveform. If this particular problem happens again, I would like to be able to characterize what is happening in more detail than my "the lights flickered deeply and rapidly" observation above.

At present, I am using Audacity to record the waveform on an ongoing basis after you all helped me correct my soundcard input circuit yesterday. But if I could get the Digilent Waveform software to do the needed type of capture, I could almost certainly get a better analysis out of the captured waveform if the problem happens again in that software than I can in Audacity.

Thanks again for everyone's input and help; I truly appreciate it.

[Jeroen3]

What you are describing sounds like line fault conditions upstream. Trees shorting or stuff.
There is no fix except put the cable underground? But that's not free of trouble either, the very old cable here arced out last week at multiple places due to the ground being saturated with water.

But it can also be a fault in your own installation. Did you have that checked yet?

Even if you measure the event, there probably isn't much you can do about it, brownouts happen.

[DrDeke]

Even if you measure the event, there probably isn't much you can do about it, brownouts happen.

I fully agree; I do not expect this to be something I can fix. But it doesn't exactly seem like a normal brownout either, and I am very curious to learn what exactly is happening to the voltage during these events.

[james_s]

What happens is that for a period of somewhere around 4 seconds, the lights in my house flicker deeply and rapidly, at what appears to be a steady pace (maybe somewhere around 8-10 Hz, but again, this is estimated from my visual observation, not measured). At the end of the event, power returns to normal. Further observations:

How long does it last? Makes me wonder if one of your neighbors has a large air compressor that cycles on periodically.

[DrDeke]

How long does it last? Makes me wonder if one of your neighbors has a large air compressor that cycles on periodically.

I would say about 4 seconds, but one of the major points of trying to monitor the voltage waveform is to enable me to give more fact-based answers to these kinds of questions than my recollection-based "about 4 seconds" .

Regarding the air compressor (or other nearby motor/inductive/non-resistive load), I would expect that to produce more of a one-off voltage dip than the oscillation that seems to be happening here.

[james_s]

Yeah 4 seconds is too short for the compressor idea, I've never measured mine but I assume it draws current in pulses matched to the compression stroke which I would estimate from the sound to be around 10Hz on my own compressor.

[DrDeke]

Yeah 4 seconds is too short for the compressor idea, I've never measured mine but I assume it draws current in pulses matched to the compression stroke which I would estimate from the sound to be around 10Hz on my own compressor.

Ooh, ok, now I see where you're coming from with this idea! Although yeah, a compressor would probably run for more than 4 seconds at a time. Interesting idea though.

[floobydust]

It could be a tap changer, usually they are automated for voltage regulation (AVR).
Changing an (on load) tap, it takes a long time and the switches/carbon brushes arc. I've seen these on the distribution side (around 13-25kV).
The tend to get slower when old or cold, until they eventually bag. They do blow up really good. Flickering lights is one of their hallmarks with sloppy switching.

I would keep an eye on mains voltage to see if it is moving around for long periods.

For your soundcard, I would give it say 1/10 the signal amplitude you have now (i.e. add 10k  + potentiometer) and confirm the distorted waveform is due to overload. Cheap wall wart transformers do compress the sinewave as they are run at fairly high flux density to keep costs low. So they can make a poor accuracy PT (potential transformer).

[Simon]

It sounds like you want to video the effects on things in your house and report to the power company.

[tggzzz]

Maybe we need to go back to the beginning. You say you have power fluctuations, how do you know? where do you see this? what happens?

Ahh, yeah, good point; I had not mentioned what _kind_ of fluctuations I had been having and wanting to measure.

You missed another important point as well, best summed up by a valuable generic question: "so what?" In other words, suppose you use a scope/whatever to capture waveforms, what can you do with that data?

I doubt a power company would know what to do with your measurements. A video of the effects would be of more use to them. Alternatively, hire a tool that is designed to log the voltages and present "interesting things" in a form that the power company can understand - preferably is related to their service specification and guarantees.

As for the 4s you mention; that might be a large electric motor startup drawing a high current. Normal domestic freezers sometimes draw 10* their normal operating current (>1.2kW vs 100W) for more than a second.

My house is fed from 240V overhead lines. I did experience "chatter" in winds, which was traced to an imperfect connection at the pole in the street. Once tightened up (by someone on a fibreglass ladder ) the problem disappeared.

[BILLPOD]

My local utility company takes 'Power Quality' seriously, as they can receive large fines by state government monitoring agencies.  They will come to your home and install monitoring equipment in your meter box,
to try to identify the causes of your 'flicker'.   Your Utility may have a similar program.  Give them a call.