AC Line Waveform Distortion

[spiff72]

I hope this thread hasn't died - I would really like to hear others' thoughts on the waveforms (and possible sources of the distortion).

I tried to re-run this FFT on my AC waveform, and got the attached.  I screen grabbed all the settings pages in the MATH menus to show how I had the FFT set up to run.

I also ran a similar pass on a 60 Hz square wave from my function generator for reference.  Unfortunately this scope doesn't seem to provide very granular access to the settings (Hz/div, etc because it seems dependent on the scopes settings for the horizontal timescale), and as a result, I don't have the same Hz/div setting on the AC wave vs the square wave.

I think i successfully got the FFT plot to show DC/0Hz at the left edge of the screen, and it looks like the 3rd and 5th harmonics are obvious high points.  The high peak at about 700Hz seems odd?

The square wave looks reasonable to me with the gradual falloff of odd harmonics.

[bdunham7]

Looks like a pretty good FFT for what you have.  I'll have to try your setup. Your THD from the 3rd and 5th is probably pushing the 5% limit, but that 700Hz spike is more interesting.  It would be about 6Vp-p and that isn't a small amount for a non-harmonic that high up.  And then you have all those smaller spikes at 40Hz intervals starting at 220Hz.  If you can, start turning stuff off and see if you can find a source in your house.

[spiff72]

I might have to make a weekend project of turning off breakers in the house while monitoring the waveform and the FFT plot until it either goes away or at the very least rule out anything within the house (if I can't get rid of it with only one circuit running).

Just need to get everyone out of the house so dad can get some more "mad scientist" time without driving everyone crazy. 

[bdunham7]

Here is my power with as close to your setup as I could get.  I have 10K samples and a full screen,  you have 2K and not a full screen.  I was unable to easily figure out what setting changes that.  You can see my 3rd is way down and 5, 7 and 9 are all much lower than yours as well and reflect less than 1% THD.  The second picture is with the probe disconnected so you can see what the background is--and why the FFT on this scope is more interesting than useful.

[spiff72]

 Bedtime here, but a few question about your settings:

• are u using the same 1054z scope on the latest firmware?
• Is it modded to get 100MHz bandwidth? Mine isn't, and not sure if that makes any difference in this case since frequencies are very low
• Is the timescale on the horizontal what determines the different sample rates (or at least the limits that can be chosen)? I couldn't get the same center and Hz/div when I tried to retest the square wave.

I will try again with your settings tomorrow.

Thanks again for the reply!

[capt bullshot]

What sorts of factors can distort the waveform like this?

This particular waveform (picture from your original post) does look quite flattened.
The usual reason for this is:
- your line has a relative high impedance, either a literally long line or a weak transformer can cause that.
and
- your load consists for large parts of rectifier-input devices

Edit:
power line distortion isn't measured in dB but rather in % of the rms or fundamental voltage, and only the harmonic bins are shown on linear axes.
If you want to do this measurement right, you'll probably end up using a rectangular window over an exact multiple of power line cycles (typically a 200ms window - 10 periods @50Hz / 12 periods @60Hz - is used).

Components at non-harmonic frequencies can be caused by ripple control systems or loads with fluctuating power consumption (e.g. some kind of modulation of a resistive heater, or an arc furnace at large scale).

[spiff72]

Thanks.

I think I am going to retry the measurement again later with the same number of cycles on the waveform displayed on screen as were shown by bdunham7 and see if my plot looks more like his.  I suspect that this is the way that he got 10k sam/sec instead of the 2k/sec that I was getting.

Most of the info i have been seeing only suggest using the "Memory" option for the FFT operation rather than "Trace" to increase the number of datapoints, but in practice i wasn't getting very good results this way.  I think that I just need to fiddle with the settings a bit more to see what changes.

As per capt bullshot's suggestion - I will try to do what he suggested as well using a 200ms window for the trace and see what that shows.

[bdunham7]

I don't think you can do 200ms.  There are 12 divisions and your window seems to be 12X your timebase setting, unless there is some other way to change it.  Still, 120ms works for 50 or 60Hz.

I only got decent results at these low frequencies using "Trace" and either the Blackman or Hanning windows.  For those of you following along but without a DS1054Z, the noise at -60-70db is high, but not as obscuring as you might think.  In real time, the noise is flickering while the harmonic humps hold pretty steady.  You can easily see them and pick them out, even on "clean" power like mine.  I think we can conclude that for all its faults, the FFT can readily distinguish between clean power (mine) and marginal or not clean power (spiff72).

This capture is my last, best attempt to illustrate and what we can see and measure.  This is using a 4kHz LPF, although there wasn't much difference between that and the scopes own 20MHz LPF.

[spiff72]

I think I misspoke there - my intent was to try to replicate your full-width FFT by experimenting with the timebase by getting the number of cycles visible in the trace to be around 12 periods (20ms/div or 25 would get me close - whichever the scope will let me do).

I am still trying to understand the sample rate shown for the FFT, as well.  You showed 10k/sec in your plot, while I only had 2k/sec.  I am thinking this is a function of me having about 30 periods visible when I did it, and you had fewer.  I have to play around with this some more, too.

[bdunham7]

• are u using the same 1054z scope on the latest firmware?
• Is it modded to get 100MHz bandwidth? Mine isn't, and not sure if that makes any difference in this case since frequencies are very low
• Is the timescale on the horizontal what determines the different sample rates (or at least the limits that can be chosen)? I couldn't get the same center and Hz/div when I tried to retest the square wave.

1054z, but not latest firmware (I'm one behind)

100MHz and 500uV options installed

I'm not sure about the FFT sample rates or even if the terminology is correct here.  The display in the FFT window says "KSa/S".  I would think it would just be "KSa", or the number samples in each window.  Thus 2 KSa would limit the bandwidth to 800Hz, thus the point where your graph cut off.  My 10 KSa would allow 4KHz and thus my display went to the end of the window, or 1.2KHz.

I struggled to get back to 600Hz center and 100Hz/div as it wasn't giving me a 100Hz option at 20ms/div.  I had to use the 10ms timebase.  I used 1.2M memory points as well, but I'm not sure it was necessary.

[spiff72]

I took another stab at this tonight.  This is the cleanest looking FFT I could get and fill the screen with a reasonable Hz/div value (100).  In order to get this, I actually used "Memory" instead of "Trace" for the FFT, but to get a reasonable horizontal scale, I had to drop the memory depth to 12K points.  Timescale is 50ms/div.

I noticed that this one did NOT have the odd non-harmonic 700Hz peak.

I then reloaded the settings from yesterday (via a saved settings file), and this is the second image.

[spiff72]

Well, I did a test over the weekend and plugged the scope into an outlet near my panel and set up my probe to test the line voltage.

I then shut off every breaker in the house except for one that feeds the circuit I was plugged into.  The result was that there was no difference in the signal. so I have concluded that one of the following two statements is true:

- The source of the distortion is outside the house (already present on the incoming feed)
- The source of the distortion is on the same circuit that I had powered up to check

I didn't get a chance, but I might re-run this test one more time to rule out a couple of additional doubts that came to mind after the test.

- I was using the "averaging" method of the scope during the test (under the acquire menu).  This could have possibly slowed the response time of the scope, but I doubt it.  I might repeat in high-res or normal mode.
- Plug scope into a different circuit to rule out that last circuit as the source of the distortion.
- I also want to re-capture the waveform and FFT plot for the "before" and "after" shutting off all breakers.  I did this, but lost track of which was which because the scope doesn't have a RTC to put a proper timestamp on the captured files.

At the end of the day, if the source is not in the house, do i have any leverage with the power company to request they clean up the power coming into my house?  Is it even anything to be concerned about?

Thanks again!

[Ragnarok]

3rd harmonics is generally from unbalanced 3 phase load. The transformer in the street is usually 3 phase (with households tapping different Phase + Neutral)

5th/7th/11th/13th harmonics are generally from rectifiers. Rectifiers will generally have harmonics generated on the (6n +/- 1) fundamental, where N is based on the topology of the rectifier design. Usually these harmonics are probably stray and seeps through the network, possibly from industrial customers connected on the local grid.

[soldar]

This is what mine looks like: