I've monitored line frequency with one of my old HP frequency counters. A
filament transformer with a potentiometer across the secondary allows me to
dial a safe voltage for the counter, about 1 V RMS.

My ancient counters are limited in their data collection abilities; I'm
sure your picPET device (whatever that is!) will do a better job.

Jeremy
N6WFO

On Wednesday, April 6, 2016, Jay Grizzard elfchief-timenuts@lupine.org
wrote:

--
Sent from Gmail Mobile

Hi Jay,

Building a mains frequency monitor is a great way to expose yourself to almost everything about precise time & frequency and measurement -- for a few dollars. Working with quartz, rubidium, cesium simply moves the decimal place over a few digits.

Have a look at:

http://leapsecond.com/pages/mains/

http://leapsecond.com/pages/mains-cv/

It turns out that worrying about how to measure the exact zero-crossing of every 50/60 Hz cycle to the microsecond has little to no effect on any mid- to long-term analysis that you do. The low-level microsecond phase noise quickly averages away. You can see that here:

http://leapsecond.com/pages/mains/mains-adev-mdev-gnuplot-g4.png

The main thing you want to avoid is accidentally missing a cycle, or accidentally adding a cycle. But a glaring 16 ms gap is so easy to spot.

Some people go through elaborate electronics and filtering to avoid this. That's fine. Whatever works for you.

I use a picPET through a 5 VAC transformer with no filtering and just measure what shows up. Once in a great while, due to excessive noise, I see a stray pulse. But the cool thing about timestamping counters is that if a pulse shows up when you know it isn't expected you can just delete that data point in s/w and all is well. Similarly, if for whatever reason you miss a cycle, you just interpolate it in s/w. I get glitches like this at the rate of a few a year. There are about 2 billion 60 Hz cycles a year so this level of data repair is fine with me. Over the past 5 years the worst problem is city-wide power failures. But in those cases I just trade data with Hal Murray, who is in the same grid as me, but a different state.

So I think between the two of us we have a complete record of 60 Hz phase going back years. Check the time-nuts archives as this interesting subject of mains monitoring comes up a every year or two.

I logged data every cycle for a while. Then I switched to every second. Even that's more than enough.

Some people use transformers, or opto-isolators, or RC filters, or Schmitt triggers, or even 60 Hz PLL's. Just pick one that you think will work and play with it for a couple of days or weeks and see how you like it. I also run a wall-mount, synchronous motor, kitchen clock to keep me honest. You can see that, compared to a cesium clock, here:

http://leapsecond.com/pages/tec/mains-clock-ani.gif

/tvb

----- Original Message -----
From: "Jay Grizzard" elfchief-timenuts@lupine.org
To: time-nuts@febo.com
Sent: Wednesday, April 06, 2016 6:21 PM
Subject: [time-nuts] Building a mains frequency monitor

Jay wrote:

I posted a very simple circuit that performs very well some time
ago.  It can be found here:

http://www.ko4bb.com/getsimple/index.php?id=download&file=02_GPS_Timing/4_App_Notes_and_Articles/Simple_AC_mains_zero_cross_detector.pdf

This circuit was designed specifically to avoid the problems with
Schmitt triggers and with iffy input circuits.  It is transformer
isolated, but uses the second primary winding of a small dual-voltage
power transformer for input (i.e., 120v), which is immediately
dropped by resistors into a diode clamp.  Why?  Because the higher
the voltage you start with, the faster it slews through the zero
cross and, therefore, the more precise the trigger point will
be.  Using a 120v winding gives you a 10x slew rate advantage
compared to using a 12v winding.  Free gain, free precision!

There was some on-list discussion starting on 12/16/14, with the
Subject: "Simple AC mains zero-cross detector".  (I'd include a link
to the first message in the thread, but febo.com appears to be down
at the moment.  Use the link at the bottom of this message, choose
list archive, then choose December 2014 and display the messages by date.)

Best regards,

Charles

Transformers really are nearly noiseless.  They can pick up magnetic fields
but most of that is the 60Hz you are trying to measure. (the problem is
normally with audio transformers picking up the 60Hz)

A transformer is by far the safest way to go.  I'd even go so far as to use
a split bobbin type so even if the transformer fails you are still safe.
So look for a quality name brand plug-in transformer with AC output and a
"UL" stamp on it.

I think you really do want a zero crossing detector.  The biggest source of
error is going to be the non-constant AC line voltage that might go from
125 volts to 110 volts depending on the time of day and it might even cycle
along with the next door neighbor's air conditions unit or if your
microwave oven is running.

even with zero crossing I think the very best why to do this is to assume
the trigger is not perfect so you try and measure the imperfection then
remove it in post processing.  So perhaps you have two triggers that in
theory should be symmetric.  One measure the raising edge the other the
falling edge.  They should be exactly 180 degrees out of phase and have the
same period.  You compare what should be to what is measured and assume the
difference is some kind of imperfection in the trigger that needs to be
accounted for.

Then you do your best to fit a variable frequency sine wave to your data
points.

Are you only interested in frequency?  What about line transients?  For
that you simply run the AC into a A/D converter and log all the data points
at maybe 48K samples per second.  Sounds like a lot of data but with not
really, only a few hundred megabytes per hour and you can trash the data
after it is processed  You likely be looking at an FFT to get frequency
domain data from your sampled AC line.

On Wed, Apr 6, 2016 at 6:21 PM, Jay Grizzard elfchief-timenuts@lupine.org
wrote:

--

Chris Albertson
Redondo Beach, California

My measurements show the fractional frequency deviation of the power line
at my home is ~25 e-6 with a 1-second measurement and 100 loops through the
calculation. I agree with Bill that a precision frequency source is not
necessary.

Jeremy
N6WFO

On Wednesday, April 6, 2016, Bill Hawkins bill.iaxs@pobox.com wrote:

--
Sent from Gmail Mobile

A good source for what is actually going on with power line frequency is
the web site of the University of Tennessee, which in partnership with
Oak Ridge National Labs has a mains frequency monitoring program at
http://fnetpublic.utk.edu/. The "Table Display" page shows frequency
data for the United States and some other parts of the world. My station
is #853 in the "Western Interconnection."

The frequency bounces around seemingly at random but within ±0.1 Hz
maximum and usually half of that. The random pattern of the frequency
shifts certainly could be used to identify a point in time at which a
recording was made. By the looks of the data, I doubt that small
measurement errors such as those being discussed would affect results.

Jeremy

On 4/6/2016 6:21 PM, Jay Grizzard wrote:

On 4/6/2016 11:34 PM, Nick Sayer via time-nuts wrote:

Would you be so kind as to elaborate how to do this?  Been looking for
such a solution off and on for a while...and I'm not having a lot of
luck with Google search at the moment.

thanks much and 73,
ben

The instructable I wrote about it is at http://www.instructables.com/id/Science-fair-How-accurate-is-the-AC-line-frequency/

There’s code for the Arduino and the Linux side as well as schematics.

It works the same on PC hardware

The PPS causes an interrupt and the handler captures the value of a counter
that is driven by the system clock.  It is typically a nanosecond level
clock that just free runs.  It saves the captured value were a user level
process can read it.  The user level process can do whatever it likes it
typically load the captured counter value.  I would not use the syslog time
stamp as that has more lag but for this purpose maybe it is close enough.

The source for Linux PPS has a test program that does exactly the above.

On Fri, Apr 8, 2016 at 10:32 AM, Ben Hall kd5byb@gmail.com wrote:

--

Chris Albertson
Redondo Beach, California

On 4/8/2016 7:19 PM, Nick Sayer via time-nuts wrote:

Hi Nick,

Awesome, thanks mucho!!!

thanks again,
ben, kd5byb

On 4/9/16 10:20 PM, Bill Hawkins wrote:

Or a series of R/C stages: you don't care about loss.

Alternatively, use a

A clever idea because of the mechanical low pass filtering, but probably
impractical..

A record turntable with a synchronous motor?  That's going to be ancient
and hard to find in this age of digital music players. People like us
would happen to have something in the garage.. but for a science fair
project, unlikely that a 6-12th grader would have such a thing, or even
know where to find one.  I have a lot of junk in the garage, and even
some synchronous motors, but not one that could directly be connected to
the mains.

an AC powered rotary dial electric clock, perhaps? (assuming it's not a
wall wart powering a "quartz movement".)

I think that the key, especially for the putative science fair project,
if I were the judge in senior division, would be good software to look
for anomalies and excising oddball transients.

have a simple data logger that logs the time of zero crossings (or
similar), and then post process to reject zero crossings that aren't
within some "expected band" and which tolerates "missing pulses".

That would be a winning project. Assuming it weren't copied from
somewhere - and that's why they want people like us as judges.

The International Science and Engineering Fair is in May in Phoenix this
year, and they're always looking for judges. Always the week after
(U.S.) Mother's day.

https://student.societyforscience.org/grand-award-judges

Ignore the formal qualifications listing - if you have the chops and
experience, they'll take you.

On Sun, 10 Apr 2016 12:03:51 -0700
jimlux jimlux@earthlink.net wrote:

Why all these complicated filtering systems? As Jim wrote, we live in
a digital world. One can easily sample the 60Hz with an ADC, 200sps is more
than enough, the resolution doesn't need to be good either, 8bit would be sufficient. Do some filtering in the digital domain with some narrow band
FIR or IIR filter. No need to worry about temperature stability or whether
there will be any spikes. Time stamping is also easy as the zero crossing
"detection" is just a simple bit of math.

All this can be done in a 16bit uC.. or use one of the many cortex-M0/M3 out
there... or if you want to use floats for simplicity, use an M4.

If you want to go all out and do a luxury solution, how about sampling
at 100ksps (something which most modern 32bit uC support with their
internal ADC), so a simple 200Hz or so RC filter would be enough to get
rid of all harmonics, spikes and other stuff that would cause aliasing.
Filter in the digital domain, decimate, filter again, decimate, filter
until you're down at a couple of Hz of bandwidth. Then mix the signal
down to DC with an 60Hz generated from an NCO and detect the phase.
Send phase value to PC using USB or ethernet interface.

		Attila kinali

--
Reading can seriously damage your ignorance.
-- unknown

I invite you to perform the same experiment with and without the extra filtering and report exactly how much benefit the extra filtering conveys. Otherwise, it’s just Monday morning quarterbacking.

So, the “science fair folks” is me. It wasn’t a real science fair entry, per se. I titled it that way because it seemed to me to be a great idea for someone to use in a science fair if they wanted.

In any event, unless you’re suggesting that the data is wrong, then I would assert that the precision was sufficient. Perfect is the enemy of good.

Really? Did you miss the 3rd and 4th pages? The ones with the Arduino sketch and the Linux monitor program? In C? That code?

I just finished similar project recently. I am using OCXO 9,830400MHZ as
the reference. My MCU driven by this crystal (No PLL). And I was using
opto-coupler as ZCD.
At the begining I was using every crossing (which 120 times per second
in case of 60Hz). However that design was not good. Long live to tom
(/tvb) who did analyses of raw data I provide and gave me very valuable
advises how to improve my project.

The main purpose of my project was to drive the Telechron clock. The
Main measurement was added just for fun. My project has three clocks
inside. Its RTC, driven my simple watch crystal, it is Main clock,
driven my ZDC and it is MCU clock, driven by OCXO.

The MCU still capturing every ZDC events (120 events per second in case
of 60Hz). But I concentrate on full cycles. Also, I implement the switch
how often do the records to the log file. Its either every 64 ZDC events
or once per 8192 cycles. Basically it recording the time difference
between number of ZDC events.

The log looks like this:

Uptime:      139 hours

RTC time:    18:00:00

MCU time:    18:00:00

MAIN time:    18:00:03

[Time Stamp]            [Period]              [1/9830400]

16-04-11 18:00:17.253 [+] 0.01665924072269      163767  -4
16-04-11 18:01:25.500 [+] 0.01666137695315      163788  -21
16-04-11 18:02:33.765 [+] 0.01666656494141      163839  -51
16-04-11 18:03:42.035 [-] 0.01666798909505      163853  -14
16-04-11 18:04:50.300 [+] 0.01666564941407      163830  +23
16-04-11 18:05:58.570 [-] 0.01666748046875      163848  -18
16-04-11 18:07:06.847 [-] 0.01666961669921      163869  -21
16-04-11 18:08:15.097 [+] 0.01666239420575      163798  +71

Time stamp, then "sign" which means if we are bellow or above the
"ethalon", then period, then raw value of capturing timer and then delta
between two readings.
Works great for me.

Regards,
Vlad

On 2016-04-06 21:21, Jay Grizzard wrote:

--
WBW,

V.P.

A more modern name for a synchronous motor is a permanent
magnet stepper motor.  Any PM stepper, and a couple of
microfarad capacitor becomes a synchronous motor when
connected to the power line.

-Chuck Harris

jimlux wrote:

On 4/10/16 3:19 PM, Attila Kinali wrote:

Excellent idea.. a $19 teensy can handle 200 ksps 16 bit ADC samples and
filter them to 4 Hz BW with no problem.

If you want to put a narrow band 60 Hz filter in there, and a zero
crossing interpolator, it would be easy.

In a mixture of fixed and floating point.

If you don't want to digitally log the frequency or want .00000hz accuracy then use one of the "tuning fork"? indicators. They have a row of reeds (12 or 20, long time since I have seen one) that resonate at differing frequencies from a few hz.below nominal to nominal frequency and then to a few hz. above nominal.

It gives a quick at a glance showing of the frequency. Ones I saw were used to adjust the phase of generator to AC mains before switching back to AC mains.

Cheers
Will
ZL1TAO

It gives a quick at a glance showing of the frequency. Ones I saw were used to adjust the phase of generator to AC mains before switching back to AC mains.

On Wed, 6 Apr 2016 18:21:43 -0700, you wrote:

Given the low power line frequency and high noise content, this method
provides plenty of resolution.

High frequency noise on the 60 Hz waveform itself will dominate
sources within the circuit.  Phase controlled light dimmers are
especially noisy but intermittent sources of noise from heavy loads
starting and stopping like motors and microwave ovens are also an
issue.

Take a look at page 12 of Linear Technology application note 31 for an
example of a simple Schmitt trigger based injection locked 60 Hz
oscillator intended for exactly this application:

http://www.linear.com/docs/4124

Schmitt triggers are great for preventing false triggering but are
less useful where low jitter is desired since they amplify high
frequency noise.  The old alternative I have sometimes seen involves
an analog phase locked loop but it would be pretty easy to implement a
modern DSP version in a simple microcontroller.

A transformer is one of the better ways because it provides galvanic
isolation; it is safe, low noise, and easy to get right.  Modern DSOs
use an optocoupler for their line trigger.  High value (and voltage)
resistors to neutral and ground will work but the high impedance will
pick up more ambient noise.  Capacitive coupling would work better
than resistors and Y type safety capacitors could be used for better
safety.

Hi Jay,

Ah! if it "feels naive" then I've accomplished my goal for the picPET: a simple solution that gives superb results.

Is there a better way? Well, another way to measure mains frequency is record the duration of each cycle directly (e.g., 16.6666 ms). It works, but be careful:

1. Period, or especially frequency readings are susceptible to numerical round-off errors.
2. Depending on how the interval is measured there may be deadtime effects, or timing windows.
3. With period or frequency measurements, if you lose even a single reading, you lose track of phase (timekeeping).
4. Certain types of noise create erratic or too-short periods which are hard or impossible to repair in the data.
5. Double wide cycles are detectable but missed cycles are not.

Here are the advantages of the timestamping method:

1. You get accurate period data by just subtracting adjacent timestamps.
2. There is zero deadtime; and never rounding errors.
3. You get perfect long-term phase tracking, even if there is noise or glitches or lost or corrupted data.
4. If you want, you get a count of expected or unexpected events.
5. Extra or missing cycles are easy to detect and repair with no loss of phase information.
6. picPET output is directly readable by TimeLab via serial/USB.
7. Unwrapped timestamp data can be decimated to any level for frequency(tau) averaging, simultaneously.

There is some noise but it was too low for me to worry about. Look at the ADEV plot to understand why.

At the seconds level, even at the cycle level, mains wanders around so much that jitter at the microsecond level is uninteresting from a timekeeping perspective. If you're interested in a full phase noise spectrum or THD and the like, then a zero-crossing detector and counter is not the right instrument; go with high-speed ADC sampling instead.

I'd say see if you have a problem before you worry about improving it. I originally had plans for a fancy zero-crossing detector or robust mains signal filtering, as others have proposed. But when I tried a plain 6.3VAC signal (via 1kR) on the PIC's Schmitt trigger input pin I got such clean data for days and months at a time that I never went back and made it more complicated. I now log once a second (instead of once a cycle).

/tvb

----- Original Message -----
From: "Jay Grizzard" elfchief-timenuts@lupine.org
To: time-nuts@febo.com
Sent: Wednesday, April 06, 2016 6:21 PM
Subject: [time-nuts] Building a mains frequency monitor

Hi

If you do want to go with the fancy ADC approach, there are MCU’s that are built specifically for
power meter applications. They have cool stuff like 24 bit ADC’s and DSP capability. Since the
target it high volume, the cost is sub $10 …

That’s still an IF by the way. It’s still not at all clear to me that the DSP stuff really adds much to the
process if the ultimate result will be highly decimated.

Bob

This one line item twigged my interest.

Dedicating my 53220A to certain long running tasks that don’t require its full capabilities irks me a little. If can design my own gizmo to provide time interval sampling/averaging, what’s the best route to connecting it up to TimeLab?

Let’s start with the assumption that I can generate serial data and get that into a Raspberry Pi that’s on the network running ser2net. So I have a TCP listening socket that will spit out… what? Time interval values in ASCII one per line in seconds in scientific notation (3.14159E-3)? If I can do that, what driver (that is, acquisition source type) do I use in TimeLab to gather it up?

I’m strongly inclined to use networking for this because I run TimeLab in a VM on my laptop, and for long-running tasks, it’s stationed rather far removed from my workbench.

Nick wrote:

Several things to note about that front end circuit, from a time-nut
perspective (the circuit was apparently created as a science project,
and it may be fine for that):

1. The LM358 makes a very poor comparator, even for a 60Hz ZCD, with
   rise and fall times of tens of microseconds.  Also, its output
   doesn't pull closer than about 1.5v to the positive supply.

2. With the non-inverting input biased to +2.5v, the switching
   threshold is over 3 volts positive from the zero cross of the AC
   mains voltage, which guarantees that mains voltage variations will
   create timing offsets.  Ideally, the non-inverting input would be
   biased one diode drop below ground so the actual switching threshold
   would be near 0v.  However, that is not within the input voltage
   range of the 358 running on a single supply, so ground would be the
   closest workable choice (the 358 is a "single supply" op-amp, so its
   input common-mode range includes the negative supply -- ground, in
   this case).  Just remove R2 to implement this change.

3. The unused section of the LM358 has its noninverting and
   inverting inputs grounded, with the output left floating.  This is
   not a good way to connect an unused op-amp.  Generally, one should
   connect the noninverting input to a potential that is within both the
   input common-mode range and the output voltage range (here, from
   ground to about 3.5v), and connect the output to the inverting input
   (making it a unity-gain follower).

A much better solution is to use a real comparator with the threshold
at 0v.  Instead of a series diode on the input, use a series resistor
and clamp diodes as necessary to keep the comparator's noninverting
input within its allowable voltage range.

I didn't review the code, but anyone building the circuit should
check it carefully to see if there are any similar issues on the software end.

Best regards,

Charles

At one point, I did try an LM393 instead of a 358. The result was that noise caused excessive false triggering. The 358, so far as I can tell, when acting as a comparator lacked sufficient bandwidth and/or speed to keep up with the noise. The result was that the per-second cycle offsets reported were almost always ±1, and were not every second. When I plotted my results, the scale of the “cycle debt,” as I termed it, was so broad that I concluded that whatever inaccuracy there was was being swamped by the signal being measured. My results also seemed to be on a par with the published results of other similar investigations (particularly those of tvb).

Nick wrote:

I'm on record as being in favor of pre-filtering to separate true
grid phase artifacts from local noise, as much as
practicable.  However, it would be serendipitous indeed (more like
miraculous, actually) if the slew-rate limit of an LM358 just
happened to be exactly the right degree of filtering.  Close enough
for a science-fair project, perhaps, but not a time-nuts-level solution.

Tom uses a Schmitt trigger input (which, as I pointed out yesterday,
guarantees that the "non-zero-cross detector" [or,
"zero-cross-by-proxy detector"] will have AM to PM conversion), and
(last I knew) he does not filter the input (other than the incidental
interaction of the input resistance of his pickup with the input
capacitance of the gate).  A proper comparator with a 0v threshold
and a few mV of hysteresis, preceded by carefully designed filtering,
can generate a ZCD output with substantially lower cycle-to-cycle and
second-to-second phase errors due to local noise that is a
significantly more faithful representation of the actual grid phase
and frequency.  If what we're interested in is measuring the grid
phase and frequency, rather than the incidental local noise that has
nothing to do with the grid phase and frequency, this is the clearly
better approach, IMO.

Best regards,

Charles

just look what is happening in one amplifier which happened to be a bit
to slow for for the passing trough signal. The part of the spectrum --
which  can't make it -- although will not show up at the output, will
not disappear, but --if the amplitude is large enough --  overdrive
parts of the amplifier, causing recovery effects and delays. the delay
will influence the propagation of the desired signals too..... That is
particularly important if the desired signals  position on the time
range is important.
Also, by comparing the structure of a comparator and an amplifier it is
very good visible, that a comparator, has much less stages which have
all relative high bandwidth and lower gain, and the dominant pole --
which at the mentioned amplifier [LM358 ] is not much higher than 5
[five] Hz, yes you could see it here:
http://www.onsemi.com/pub_link/Collateral/LM358-D.PDF
It could happen that the signal coming out from a miss-used amplifier
look very nice, but it is very likely, that it is not exactly correct,
and there is a good reason, why the solid-state industry produce not
just amplifiers but also compactors, which would need pre-filtering, but
provides the correct output signal
73
KJ6UHN
Alex
who used to be corporate applications engineer at ON Semiconductor,
Intersil Corporation and Elantec

On 4/12/2016 2:00 PM, Charles Steinmetz wrote:

Attila,

On 04/11/2016 12:19 AM, Attila Kinali wrote:

The professional equipment does it this way. It samples, filters and
decimate the data. For the professional use the absolute phase is
relevant, so group delay needs to be flat, known an
calibrated/compensated. It also produces frequency and ROCOF (Rate of
Change of Frequency) according to standardized estimators. The phasor is
normalized to that of the reference time and the frequency (50 or 60 Hz)
setting. Phasor measurements up to 120 Hz is seen in installations,
typically on voltages of all three phases, but also currents and
sometimes on both.

NIST in Gaithersburg work on calibration of these PMUs. It was
interesting when I could show them some of the T&F tricks to measure
phase of limited slew-rate, proposing calibrator designs etc. so promote
absolute phase measurements etc.

Cheers,
Magnus

God middag Magnus,

On Thu, 14 Apr 2016 03:38:38 +0200
Magnus Danielson magnus@rubidium.dyndns.org wrote:

Jupp, I am aware of those. But I didn't expect many people here
to know of them :-) I have quite a few friends who work in the
energy distribution busines and its related problems (like network
stability, oscillations of dynamic systems, complexity theory etc)
hence I have a cursory interest in the equipment to monitor power
networks as well. It is intersting to see, that modern power networks
rely more and more on precise timing as well. Even more interesting
is that the primary source for time is GPS these days, with all its
problems. I am not sure whether the desginers of these networks are
fully aware of the security and reliability implications that come
with GPS (or GNSS in general). Maybe i should dig a little bit around
and see what's current common practice in europe.

		Attila Kinali

--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson

Guten nacht Attila,

On 04/14/2016 12:24 PM, Attila Kinali wrote:

Indeed. I've even written an article with researchers relating to PMUs.

See my and Alison's presentation at the US PNT advisory board.

Using Alison's words, we need to do some hygiene cleanup.

Cheers,
Magnus