measuring frequency

It would with an infinite sampling rate but in the real world it is necessary to filter noisy signals so that there is as close to nothing above the Nyquist frequency as you can possibly make it.

You would also need a very large digital filter kernel to do the necessary convolution at such a high sampling rate. Whilst there might be ways to do it you are going to struggle more in the digital domain.

It is always better to take the right amount of really good data than take loads of terrible data and then try to fudge it in software.

Sometimes you don't have a choice if you are already working right at the limits of signal to noise when detecting pulsars for instance.

--
Regards, 
Martin Brown

Ten minutes! Yikes. That will affect peoples' lives.

Here, old synchronous clocks and the clocks on ovens and such are AC-line referenced, and they are exact to their resolution between power fails, which happen very briefly maybe a couple of times per year.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Get a synchronous motor running at, say, 1500 RPM. Use a 333333.3333:1 gearbox to make the output shaft rotate at 1000000 RPS second, then use your old counter with a suitable pickup, scaling the numbers as required.

Why do people have to overcomplicate everything?

Cheers

--
Clive

Yup. "Postprocessing is not a substitute for good data." Digitizing the crud wastes dynamic range too (though probably not so much in this case since the signal itself is so large).

In my business you'd be astounded how many software people think that the way to do any optical thing is to point a webcam at something and crunch the daylights out of the resulting crappy video.

Cracked.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

;)

A slightly newer counter that runs in reciprocal or universal mode can do lots of digits in 1 second, as long as you filter first.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Of course a soundcard already has an anti-aliasing filter before the AD. I also do not believe that there will be that much crud above 24kHz behind an iron-core mains transformer that it will be a problem when determining the zero crossings of a sine wave.

There are other fields where unbelievable things are now possible with just an A/D converter and clever software or programmable hardware, like software-defined radio. Connect your antenna (via antialias lowpass filter of course) to a fast A/D and process the data in an FPGA and computer, and it will run circles around your classic superheterodyne radio with crystal filters etc.

That is true _if_ ha A/D converter would be ideal. With only 50 Hz of interest and "24 bit" A/D at 48/96/192 kHz sampling rate, one could consider modern converters quite ideal.

For higher frequencies and lets say 8 bit converters, the situation might be quite different. For proper digitizing of mains waveforms, you should be prepared to set the full scale to equivalent of at least

+/-1000 V to properly handling all repeating peaks that might ride on the sinuous mains waveform. This will quickly use your dynamic range with a low number of bits in the A/D.

If the sampling rate is much higher than the frequency of interest (or bandwidth in SDR, Software Defined Radio), the random crud in analog input acts as dither noise and you can get a few more bits compared to the physical A/D resolution.

You really have to consider on a case by case basis in which domain (analog or digital) the processing is done.

Am 07.03.2018 um 11:10 schrieb Rob:

Here I wrote a quick soundcard line frequency monitor in LabVIEW (MS OS) It used the mic input .. abused a headphone to pick up the hum of a transformer. Used some digital filtering, but the 'trick' is to apply a sine fit or do it right in the freqency domain. (article in in link)

With the second channel I would capture the DCF77 (Germany) or another LF time signal and use that frequency (not the pulses) to calibrate the internal soundcard clock.. or directly just use the frequencies relationchip.... maybe no need to capture two channels :) the hum will it's way into the time ref signal...

Greetings from Germany Henrik

Am 07.03.2018 um 20:34 schrieb Henrik:

Here is a screenshot

I think you really need to look at the actual mains on a scope.

--
Regards, 
Martin Brown

A handy low pass circuit has already been given.

The devil is always in the detail.

You must analogue low pass filter so that there is as close to no content above the Nyquist frequency anyway so why not do it properly?

You will have to work very very hard digitising it well enough and then filtering it down without adding systematic bias from glitches. The crude analogue filter simply does a better job for low pass filtering with a lot less effort. You might be able to make it work by cunning tricks in the digital domain but you will end up working very hard.

Try it and see.

Sorry. I meant a buffer for the waveform of a decent length and a buffer for observed zero crossing events and cycle counts.

I suggest you download Daqarta and use its evaluation period to take a look at the mains waveform in both time domain and frequency domain. I think you will be surprised just how scrappy the waveform actually is.

A lot of traffic lights may also be out by 5 minutes at the moment.

--
Regards, 
Martin Brown

What has happened on mainland Europe is that the Balkans are not pulling their weight and Germans having shut down all their nuclear generating plant after the Japanese tsunami have little to spare.

Chapter and verse on the EU event here. Dropping the Balkans off the synchronisation zone abruptly might help focus minds. They have in effect been stealing power from the rest of Europe without paying.

A very cold snap in the UK last week put us under pressure too with spot gas price almost going through the roof and some industrial bulk users asked to drop off so that CCGT could stay running.

--
Regards, 
Martin Brown

It's been a long time since I took two semisters of Electrical Machinery.

Can't you just drop the field excitation a bit and open the steam valve to get higher frequency and lower voltage, to catch up?

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

You could if you had any generating capacity to spare.

The UK was pretty close to a network fail last week when gas went almost zero availability due to unseasonally cold weather. Relying on CCGT isn't very smart if you don't actually have any gas reserves.

Essentially the problem is that the Balkans are freeloading on the rest of the EU network and so far they have been allowed to get away with it.

There isn't enough capacity in the rest of the system to hold the European mains on target frequency and UK margins are also tight so we can't help much either. Germany shutting down all their nuclear generation plants is a factor in the lack of total capacity.

--
Regards, 
Martin Brown

That is why I would think that sampling at 48 kHz and then looking in a bandwidth of maybe 1 or 2 Hz would be entirely trouble-free.

But hey, maybe I am wrong.

That was my initial thought as well: when the production facilities are overloaded, why do they drop the frequency instead of the voltage?

But, it appears that the type of generator being used always has a phase slip that increases with load, so the frequency is not determined solely from the RPM. Probably quite tricky to get the correct combination of RPM and field to get 50.00 Hz at a lower voltage and still keep everyone in lock.

Thank you, your approach and result depicts what I was expecting to see. Lots of people here are quite sceptical but I would think it is quite easy to filter the 50Hz out of the crud and not have all those problems determining the exact frequency.

Unfortunately I am not proficient with labview, it would likely be much easier to build this than to code it in C. (like it is easier to build a radio using gnuradio than to program it)