Solar inverter making digital clock run slightly fast

[Circlotron]

A friend has recently got solar panels + inverter installed. Now he tells me that 3 different digital clocks in the house gain time, one of them at the rate of about 1.5 hours per week. These clocks would sense the zero crossings of the 50Hz mains, and a bit of noise on the mains from the PWM of the inverter would mess this up. I don't want to dive into the insides of the clocks to add a bit of extra low pass filtering to the zero crossing detector cct, so has anyone seen a line filter that I could plug into the power outlet and then the clock into the filter? Something with a bit of L&C, not just a MOV. This is Australia BTW, celebrating 30 years of electricity.

[edpalmer42]

Are you sure that noise on the zero crossings is the problem?  My first thought is that the frequency of the inverter isn't precisely 50 Hz - or maybe it varies with load.  1.5 hours per week is less than 1%.  Could be a reasonable specification for an inverter.  This type of clock is about the only thing that would notice that kind of error.

If that's the problem, there's no practical way to get the clocks working properly.  They'll have to be replaced.

Ed

[Circlotron]

It's a grid-tied inverter so the actual mains dictates the inverter frequency.

[DanielS]

A MOV would not help you with zero-crossing detection anyway.

It would help to see what the event leading to the extraneous counts looks like. Maybe they are filterable glitches but it could also be lower frequency stuff if the inverter is having trouble matching the grid.

[edpalmer42]

Many plug-in "surge suppressors" include both LC filtering and MOVs for transient suppression.  They can be rather pricey.  If you can scavenge a filtered IEC socket you could make your own.  If you've got two or three sockets, put them in series.

Ed

[Phoenix]

Now he tells me that 3 different digital clocks in the house gain time, one of them at the rate of about 1.5 hours per week.

How old are these clocks that they need to use the grid to measure time, are they some sort of antique or just a clock? Get new clocks, with crystals, probably easier and cheaper than EMI filters...

[coppice]

Now he tells me that 3 different digital clocks in the house gain time, one of them at the rate of about 1.5 hours per week.

How old are these clocks that they need to use the grid to measure time, are they some sort of antique or just a clock? Get new clocks, with crystals, probably easier and cheaper than EMI filters...

Many clocks are still mains powered, because they are in devices with other functions. Many of these are timed from the mains, which is far more accurate than a crystal in much of the world. Timing something mains powered from a crystal makes sense if you have backup power (battery, supercap, etc.) to ride through a power cut, but its not the obvious choice in places where the power only fails once a decade or more.

[mzzj]

Many clocks are still mains powered, because they are in devices with other functions. Many of these are timed from the mains, which is far more accurate than a crystal in much of the world.

AFAIK grid frequency is only better in long run. Scandinavian grid is specified to nominal maximum error of  +-30 seconds.
At this moment Fingrid website shows that there is +10.07 second error on "grid time" 
For continental Europe: http://www.mainsfrequency.com/gridtime.htm

[Circlotron]

How old are these clocks that they need to use the grid to measure time, are they some sort of antique or just a clock?

Actually they are a digital alarm clock + radio. The type that a 14 year old "invents".

[Seekonk]

I have a cheap HF inverter at my camp that is 10 years old.  Frequency is set by a pot and I was amazed how accurate it was after all this time. I made a RC snubber in a box with an outlet to power electronic devices, a couple of MKP capacitors totaling about 1uF and a 10 ohm resistor in series.  Outlet is parallel with capacitor.  This inverter powers a refrigerator so it is on off a lot and the intent was to absorb turn off spikes and any line ringing.   I would give this a try.  Common LC line filters wouldn't do much good at lower frequencies.

[Lightages]

If it is truly noise that is causing the problem, then perhaps a simple isolation transformer could kill the problem.

[Kalvin]

And, if the inverter's frequency is drifting, a GPSDI would be a nice project

[SeanB]

Open up the clock and there is a small transformer inside, with a secondary of 9-0-9VAC to drive the clock. Add a film capacitor of around 0.47uF 63V across each secondary winding. That should, along with the transformer inductance, reduce the noise on your supply line.

[Circlotron]

I'm thinking that might be the smartest thing do do at this point. Hopefully they come apart okay without the tuning dial cord making life difficult.

[SeanB]

You do not need to remove the board. The transformer will have a mains lead going in the one side, and either 3 or 5 wires coming out of it. There will be 3 going to the clock board, typically 2 white and one red, or vice versa.  Simply strip a little of the wire near the board where there is some space and solder the capacitors into the bared copper core.  Then use some hot glue to stick the caps down where they will not interfere with the case top or the rest, there normally is enough space for the 2 green mylar capacitors. Spot of hot glue over the solder joints to insulate them and put the cover back on.

[LabSpokane]

OK, time for troubleshooting 101. 

Step 1:  Measure the mains frequency without the inverter.

Step 2:  Take a look at the mains waveform without the inverter.

Steps 3 and 4:  Do 1 and 2 with the inverter operating. 

Don't apply a solution until you understand the problem.

By the way, I believe the mains frequency is supposed to be between 49.85 and 50.15 Hz, so it may well be that the inverter is actually running out of tolerance and needs to be looked at, rather than fixing the clocks to match the inverter.  Without the data, you will never know for sure....

Also, it may not be a "noisy" inverter.  The inverter may be outputting a stepped approximation to a sine wave and that is what is causing the issues. 

[Circlotron]

It is a grid connected inverter so I can't see how it could shift the mains frequency. I live about 3 miles away so it would affect the clocks in my house too if it did.

Of course, me being unable to see how something could happen is not a good basis for rejecting it out of hand.   

[Circlotron]

Here we go.
Dug out a pair of 4u7 multi layer ceramics with very low impedance at HF. (probably worth as much as the entire clock, but got them for nothing 15 years ago). About 12mm square. Connected each one from transformer secondary centre tap to each end. Now to see if it does the trick.

[LabSpokane]

It is a grid connected inverter so I can't see how it could shift the mains frequency. I live about 3 miles away so it would affect the clocks in my house too if it did.

Of course, me being unable to see how something could happen is not a good basis for rejecting it out of hand.   

You are correct that it can't shift the mains frequency, but it could superimpose a signal over the top of the mains frequency if it were operating incorrectly. Most of the interference will remain localized due to the fact that your friend's system has two transformers and a fair bit of transmission line separating you.

This type of noise likely isn't good for any of the electronics in his home, so really would be best dealt with by dealing with the source.

[NiHaoMike]

Is the inverter some off brand known for breaking conducted EMI limits? Other than that or an inverter with a bad output filter, maybe the clock is really EMI sensitive. Is the AM radio much noisier during the day?