From: Winfield Hill
Subject: Re: Watt Meter Project....Attn Win
Date: 24 Dec 2002 10:31:47 -0800
Organization: Rowland Institute
References: <3E04C4D7.firstname.lastname@example.org> <email@example.com>
X-Newsreader: Direct Read News 4.20
> I copied the following from the National AN 222 pdf.
> [snip useful paragraph - thanks Mike!] The last sentence says;
> "This circuit measures power on negative cycles only,
> and so cannot be used on rectifying loads."
> How about a fullwave bridge rectifier with capacitor filter.
> Wouldn't both halves of the cycle be identical?
There are several problems with this idea, for example to work
correctly during the single ac half cycle, the circuit still
must operate correctly with both positive and negative current.
Perhaps if a second PNP transistor pair were employed some way
could be found to combine the two outputs into a single meter.
By "rectifiying loads," they refer to a single diode in series
with the load, which is a rather rare requirement. But if a
hobbiest wanted to take such a measurement, the two ac lines
to the instrument could be swapped for a second measurement,
and the two measurements added together. :>)
> Also fullwave bridges have large short current pulses, would
> this be accurate for this type of signal?
This is a good question, Mike. To first order it'll work as
desired. However in some measurements, such for some computer
power supplies, the peak-to-rms ratio can get very high, even
over 10x, with all of the current taken at the very tippy-top
of the ac cycle. In this case the circuit's nonlinearity can
cause a considerable error.
>> . | _ |
>> . +-|_|-+ meter
>> . | |
>> . V ~ Iac c c
>> . ,------ b b --,
>> . | e--+--e |
>> . ---+--/\/\----- | -------+----
>> . |
>> . AC I ~ Vac AC
>> . LINE | LOAD
>> . ----------------+-------------
We can predict the error from the equation for a long-tail pair
differential amplifier, i_out/Ie = tanh(v_in/52mV), where the
52mV comes from 2kT/q.
For example, 5mV yields 9.58% and 15mV (full scale) yields 28.07%,
which is an acceptable -2% drop from the maximum gain. So the
circuit linearity is OK for nice waveforms up to full scale.
By contrast, a 5x full scale input, or 75mV, yields 89.4%, which
is only 3.18x out rather than 5x, for a large -36% error. Also
note, 89% is nearly to clipping, so the performance deteriorates
rapidly for signals beyond 5x. Sheesh!
This shows that there's a good market for wattmeters that employ
real multiplier ICs or digital conversion and calculation. :>)