whisky-dave wrote:
On Wednesday, May 1, 2013 2:54:09 PM UTC+1, wrote:
On Wednesday, May 1, 2013 1:54:36 PM UTC+1, whisky-dave wrote:
But you need a dedicated clock you can;t use the average MPUs clock as it's too inaccurate, you need a proper RTC, you do know that computers keep pretty crap time unless they get it from a server.
You wouldn;t use that method anyway as you need to mearuce the actual frequency of teh main not simulate it.
and time the period from local zero crossing to mains zero crossing, and log the latest value to see how far it moves.
Not necessary.


All one need do is

a) xtal control the cpu clock, an xtal with no oven is more than accurate enough to read 50Hz +/- 0.1%.

XTAL DO NOT READ FREQUENCIES and niether do CPUs.

No, but you can use even the cheapest CPU to count clock cycles with a
smidgeon of machine code. Add an external circuit to flag zero crossings
of the mains waveform, and it is then trivial to calculate the mains
frequency. It is also trivial to write a prgram and design a circuit
that will automatically turn off a load if the frequency drops below a
certain level.

A cheap resonator clock is well within the accuracy meeded for this
application.


b) count a divided version of the clock from one mains zero crossing to the next. Problem solved.

That's not the way to do it, you don;t need zerop crossing wither until you decide to switch something even then zero crossing has nothing to do with it.

Wrong. Counting zero crossings and comparing the timing to a refrence
frequency, such as the clock controlling a microprocessor is the easiest
way to work out the frequency of a signal,especially the relativiely
pure sine wave that comes down the mains supply.


The slight frequency spread of the xtals prevents all appliances going off at the same time.

That's not really relible, XTALS have a very high accuracy when made, but temperature and series as well as parellel capacitance play a part.

Wow, you got something right. Even after aging and cooking, cheap
crystal and ceramic resonator frequencies are accurate to within not too
many parts per million. You're trying to check frequency stability of
the supply to within parts per thousand.

If yuo really need to switch things off then measure the frequency of the grid and if it's below what you see as a miniumium then send a single to turn off particular devices that are in yuor home that the smartmeter recognise.


So you're proposing that appliances have 2 way communications,

No they don;t need two way communication, any more than your TV does.

They do, otherwise how does the smart meter know what's switched on to
turn off? Minimum is a "Hello, controller, I am XXXXXXX" when switched
on, and "Goodbye from XXXXXXX" when turned off manually, and the
capability to receive "XXXXXXX Turn off now" and "XXXXXX Turn on now".

and that smart meters talk to each appliance in each house. Doable of course, but its more money, so not attractive.

Why is it more money ?

More complex circuitry. the frequency measuring can be done for about a
pound a unit by the maker, and if there's already a microprocessor in
the unit, it comes down to a few pence once the design costs have been
written off, the communications requires about a fiver.

Be silly to have to measure the frequency of every device.


Either you measure locally or decide centrally and communicate. Do you think 2 way communication's cheaper?


Yes.


I'm struggling to see how enabling 2 way comms and playing with MACs is cheaper than a divider in the chip.

What do you mean by a divider in the chip, which chip are you refering to ?

The cheap, 4 or 8 bit CPU in the appliance. A new use for a Z80 at last....


Measuring low frequecies accuratly is more difficult than you think.
What you have to do is count cycles in one second yuod get perhaps 47 or 53 so you need to average it out especailly in times of high or low demand.
So you need to sample it for at least 10 seconds to giove you 0.1 resoultion and in that time the situation can change.


Its not hard at all, just time the number of clock cycles between 2 successive crossings.

Two crossing of what ?

Zero voltage of the supply in this case.

You could average lots of counts but I don't believe there's a need to, a small percentage of mistimes is immaterial in practice. Simply averaging several yes or nos would be trivially easy.

Yes ot no's to what exactly .

Yes, the frequency is too low, no it isn't.

Anyway it would be cheaper to have just one device measuring frequency rather than every electrical device each reading it. If I were designing such a thing it'd use teh smart meter to measure it rather than install the ciruitry in my kettle and fridge.freeze, TV cooker shower , washing machine, coffee maker storage heaters fires hair dryer etc.....


but you'd still need to install everything in the appliances,

Yes of course unless you turn the whole suply to tteh hosue off.

?\The unit would be installed in the factory, with an option to fit it
to older machinery as well. To retrofit it, the only work needed in the
houseis to unplug the appliance. Your way means installing a new meter,
which *does* require disconnecting the supply.



idf all 2KW+ kettles had a 1/2 power switch then my first try would be to turn those kettles is 1/2 power them, I wouldn't turn off motors or mmoving things such as food blenders.


I'd agree there. There are many things can be turned off or down if its ever worth it. Today its mostly not though.

I agree so why bother ?
As I said before better to turn off scotland with all those deep fat fryers ;-)

You'll have Scotty on you before you know it if you suggest that.


--
Tciao for Now!

John.