60 Hz Power Frequency Accuracy

tomcrawford

I was wondering how closely a clock that uses the 60 Hz power line would keep time. Of course, over the long run, it is quite good; we all have electric clocks that only need resetting for Daylight Saving and when power failures occur. But what about over the short term (day to day)? Supposing I set a clock to the exact time. How accurate would it be? The wikipedia article says the Western US grid is correct to +/- 2 seconds.

Using a propellor mini, I compared the line to a 1 PPS pulse from a GPS receiver.

   waitpeq (1<< PPSPin, 1<<PPSPin, 0)
   GPSCycles += 60
   CurrentDelta = GPSCycles - GPSCycles

and in another cog,

   waitpeq(1<<Hz60Pin, 1<<Hz60Pin, 0)
   Hz60Cycles++

I collected data for about 10 days in April 2016. Seven days (Thursday through Wednesday) are shown in the appended graph. The blue line is time of day; the red line is the error. The error does not start at zero because the run was begun a few days before the data starts. It can be seen the the peak-to-peak error is about nine seconds over the run. I plan to collect data for a few months to see if this is typical.

Heater.

Back in 1971 I built a digital clock with TTL and Nixie tubes. It counted mains cycles for time keeping. Sadly I have no idea how well it kept time because almost every other clock around, with various motor driven mechanical displays was synched to the mains as well.

It did drift a bit compared to my grandfather's pendulum timed, well, grandfather clock. But I suspect that grandfather clock was more accurate than the mains.

My understanding at the time was that the mains frequency in Blighty would drift throughout the day but they tried to get the count of cycles in a day correct.

Now a days it's so much easier to get accurate timing from GPS, or an oven controlled XTAL clock, or even a rubidium clock off ebay.

Phil Pilgrim (PhiPi)

Heater,

Of course you realize the mains frequency in Blighty is consistently slow by 10 Hz, right?

-Phil

Heater.

Ha, I do indeed. It's that "consistently" thing that's important.

We also consistently have bigger gallons than the US. Which means that beer comes in consistently nice big pints.

Bean

That's an interesting graph.
What are the units ? You say it is 9 seconds peak-to-peak, but the units go from about +/- 300 ???

Bean

Bean

Oh, I guess the units are "cycles" of the 60 Hz mains.

Bean

tomcrawford

Bean wrote: »

That's an interesting graph.
What are the units ? You say it is 9 seconds peak-to-peak, but the units go from about +/- 300 ???

Bean

Sorry, I should have made that more clear. The units *are* cycles. I was just thrilled to get the grid lines at intervals of 60.

Incidentally, I once had a Heathkit digital clock at a place where the mains were so noisey that I had to put a capacitor to filter it to keep from gaining several seconds a day.

Chris Savage

Some of my Z80-based controllers had the 60 Hz from the mains picked up via Opto and piped into the NMI line to run the clock without an RTC chip. A trick I "borrowed" from Commodore.

Heater.

tomcrawford,

Ha, ten years I was involved in rolling out a new system where 200 nodes in a network had to be in sync with the mains. Don't ask why, it's historical.

When it was up and running I was called out to investigate why many nodes were counting time twice as fast as reality. At the week end. A flight from Helsinki to Oslo. God dammit.

Turned out they had designed a new mains filtering/conditioning circuit that fed 50Hz pulses into out digital stuff.

Turned out, after I ripped one of the units apart and looked at the circuit, that it was acting like a high pass filter. Plus they had a lot of noise on the mains. It was counting the harmonics!

I could have done better with a low pass RC circuit. But I'm not the electronics guy in that particular situation.

After they fixed that, redesigned and replaced all those mains conditioners, I was called out again. Another flight. Another day in the rain in the streets of Oslo. Turned out the installers had all the connections of all those units to our digital stuff reversed.

Sometimes, all I want to do is write code....

Sapphire

Tom,

The WECC (Western Electricity Coordinating Council) has gone to a time error correction at ±30 seconds. This means you should see deviations up to 1 minute over a long period of time.

I have been monitoring this continuously since before the discussions of eliminating time error correction. What use to be ±5 seconds to trigger a correction is now at ±25 seconds to initiate a correction.

That's too bad, but was done to reduce the workload of the Balancing Authorities in synchronizing to other parts of the US-grid. I have a lot of synchronous motor clocks, that while still accurate over the long term, they are sometimes off by ±30 seconds due to the relaxed time error correction.

As a historical note, it was the synchronous motor clock, as invented by Henry Warren, that forced the early electricity companies to stabilize their frequency. He demonstrated in Boston in 1916 that the Edison Electric Illuminating Company had horrible frequency control, and his inventions lead to the ability to use the power lines for timekeeping. The name of the company he founded, Telechron, means "time from a distance".

jmg

Sapphire wrote: »

.... What use to be ±5 seconds to trigger a correction is now at ±25 seconds to initiate a correction.

That's too bad, but was done to reduce the workload of the Balancing Authorities in synchronizing to other parts of the US-grid.

Wow.

What 'workload' ?! - surely this is automated, so the numbers become arbitrary, and one would expect these to improve over time, not lurch backwards on a bureaucratic reflex, as seems to be the case here.


What exactly does '±25 seconds to initiate a correction' mean ? - how quickly do they then apply that correction ?

Sapphire

The workload is $. When a Balancing Authority is out of tolerance with another on an adjacent grid, they are charged for the error, and have to act to correct it. This could involve running more generators, switching on or off hydroelectric power, etc. With the old setting of ±5 seconds, this was a daily occurrence, often many times a day as load changes. Going to ±25 seconds gives them more room to run without correcting. At one point a few years ago, NERC (North American Electric Reliability Corporation) wanted to completely do away with time error correction, but the outcry from industry was to much to bear, and they settled on this relaxed scheme.

The ±25 seconds means that once the time (as determined by line frequency) has drifted away from the standard amount by 25 seconds, they begin initiating a correction, speeding up or slowing down the line frequency. The goal is to stay within ±30 seconds of the established time. Right now, the WECC is at +18 seconds, and we have been positive for the past week.

jmg

Certainly sounds antiquated in the extreme, and I'm amazed bits of paper are still involved.
You would think Grid frequency balancing (on average) came for free, as part of Grid load balancing.

Ah, bless the Swiss, of course, their Grid info has to include the time !

http://www.swissgrid.ch/content/swissgrid/en/home/experts/topics/frequency.html

Currently shows 3.572 s

Also says they correct via +/- 10mHz which means a sustained error of 200ppm could be present for some hours, while they apply the correction.

Sapphire

jmg,

Interesting. They adjust at ±20 seconds.

Here is the site for ERCOT, the Electric Reliability Council of Texas:

http://www.ercot.com/content/cdr/html/real_time_system_conditions.html

Note they are right at the threshold for starting a correction (as of the time of this post).

Heater.

Actually I have always wondered how this works.

My naive assumption is that in a good old national electric grid there are dozens of different power stations of different sorts all connected in parallel to a common grid.

Clearly they all have to be in phase. Surely if some station were to connect whilst out of phase something really bad happens as all the others drive against it!

Is it so that a bunch of AC generators wired in parallel naturally tend to lock together in phase? After all generators are motors and motors are generators so they can all push and pull against each other.

But then how do you get any control over the frequency the whole grid runs at?

And the, we end up connecting national grids together to trade power, who is in control then?

Sapphire

They have to be in phase at the moment of connection or disconnection, but they don't have to be at the same time error point as long as the phase matches. So a grid at -8 seconds can interconnect with one at +12 seconds if it is done when both are in phase and running at the same frequency. However, once interconnected, they need to stay in phase, or there are tremendous system losses as the generators fight to keep on frequency. That's why each BA is supposed to keep the drift to a minimum. Having the same or close time error lets them drift together without one getting out of tolerance, which would force a disconnection.

Interconnection is also done with DC ties, which eliminates this problem. Power can be passed between grids without respect to phase, and each grid feeds it into or out of their own synchronous converters.

Electrodude

According to what the tour guide at the Hoover Dam told me, they start generators by first running them as if they were motors to get them up to speed and in phase, and then they start pushing them (in this case by running water by them) to generate power.

The Swiss Grid website jmg linked to says that the frequency is dictated by supply and demand. If there's more load per generator, the generators will spin slower because they will be harder to turn. If the frequency is too low, they bring more generators online to compensate.

localroger

When I was in high school in the 1970's there was a then-old book in the school library of Scientific American projects for the amateur scientist. I remember to this day that one of those projects was an electronic clock timebase which was necessary for astronomers because of the drift of then-ubiquitous mains clocked clocks. Back in the 1960's the article explained that mains frequency was synchronized to deliver exactly the right number of cycles per day, but not necessarily the right number in any given hour, the next hour perhaps making up for a shortage or excess of the previous one.

In the day the solution involved a ridiculous number of vacuum tubes dividing down a 100 KHz crystal oscillator with a lot of fiddling because of divide-by-10 stages preferring to divide by 9 or 11 if not tuned properly. The end result was as complicated as a high-end shortwave radio and existed mainly to provide enough really accurate 60 Hz power to drive a mechanical clock.