Hi all,

This would probably fit better in another NG, but I know quite a few
people on here who will be able to provide some good ideas, so here goes.

I have an experiment where a voltage of ~250 VDC is applied across a
load, resulting in a current consumption of around 2A. The experiment
runs for around 5 minutes.

The nature of the load is such that the current consumption is very
spiky - it may average to around 2A, but you get very short, sharp
spikes, probably of 10-15A.

I need to be able to measure, with 1% or better accuracy, the total
energy dissipated in the load over the duration of the experimental run.

I don't need instantaneous V or I values - all I need to know is the
total energy consumption over the run.

The only sensible way I can think of doing this is to use a data logger,
and sample V and I at a high rate thoughout the experimental run, then
calculate average power over the run, then multiply by run time.

However, this seems like a very round-about way of achieving my goal.
Anyone have any thoughts on other ways of doing this? I must stress that
it is the electrical energy input to the system that I'm interested in,
so measuring temperature rise of the load etc. is out of the question.


TIA


--
Grunff

"Grunff" wrote in message
...
Hi all,

This would probably fit better in another NG, but I know quite a few
people on here who will be able to provide some good ideas, so here goes.

I have an experiment where a voltage of ~250 VDC is applied across a load,
resulting in a current consumption of around 2A. The experiment runs for
around 5 minutes.

The nature of the load is such that the current consumption is very
spiky - it may average to around 2A, but you get very short, sharp spikes,
probably of 10-15A.

I need to be able to measure, with 1% or better accuracy, the total energy
dissipated in the load over the duration of the experimental run.

I don't need instantaneous V or I values - all I need to know is the total
energy consumption over the run.

The only sensible way I can think of doing this is to use a data logger,
and sample V and I at a high rate thoughout the experimental run, then
calculate average power over the run, then multiply by run time.

However, this seems like a very round-about way of achieving my goal.
Anyone have any thoughts on other ways of doing this? I must stress that
it is the electrical energy input to the system that I'm interested in, so
measuring temperature rise of the load etc. is out of the question.


you need a digital data logger connected to a PC. eg Maplin sell
multimeters with RS232 output which give a serial data o/p. so if the load
has no phase problems you just need to measure the current. I don't have
detailed knowledge of the specs.
http://www.maplin.co.uk/Module.aspx?... ID=&doy=20m2
A serial software interface program like ProCom or VisualBasic Studio is
also needed to sample the readings in the PC. Or GPS navigation software
for NMEA might do the same sampling job at a pinch.

John wrote:

you need a digital data logger connected to a PC. eg Maplin sell
multimeters with RS232 output which give a serial data o/p. so if the load
has no phase problems you just need to measure the current. I don't have
detailed knowledge of the specs.


If I'm going to go down the data logger route, I will probably go for
one of these:
http://www.audon.co.uk/labjack.html

I used one before in a previous project, and found it to be a very nice
piece of kit.

But that wasn't my question really - I can see how to get the
measurement I want by sampling V and I with a data logger - but I can't
help wondering if there isn't a cleverer, simpler way, given that all I
want is total energy.

Another way of asking the question - I can see how to integrate the data
electronically, after collecting a huge number of data points. But I'm
wondering if there isn't another way of doing the integration. If that
makes sense.


--
Grunff

"Grunff" wrote in message
...
John wrote:

you need a digital data logger connected to a PC. eg Maplin sell
multimeters with RS232 output which give a serial data o/p. so if the
load has no phase problems you just need to measure the current. I don't
have detailed knowledge of the specs.


If I'm going to go down the data logger route, I will probably go for one
of these:
http://www.audon.co.uk/labjack.html

I used one before in a previous project, and found it to be a very nice
piece of kit.

But that wasn't my question really - I can see how to get the measurement
I want by sampling V and I with a data logger - but I can't help wondering
if there isn't a cleverer, simpler way, given that all I want is total
energy.

Another way of asking the question - I can see how to integrate the data
electronically, after collecting a huge number of data points. But I'm
wondering if there isn't another way of doing the integration. If that
makes sense.


Right, ideally you need to multiply I and V and integrate the product.
Easy if you had an 1960's analogue computer, but these are all in museums
now. The only analogue route I can think of is to use an analogue
multiplier (search in Google) to get the product of V and I. Then construct
an integrator using an OP amp.
http://www.st-andrews.ac.uk/~www_pa/...rt8/Page3.html
You still need to sample the integrator's output, so the integration may as
well be done in the PC rather than in an IC..

Grunff wrote:

However, this seems like a very round-about way of achieving my goal.
Anyone have any thoughts on other ways of doing this? I must stress that
it is the electrical energy input to the system that I'm interested in,
so measuring temperature rise of the load etc. is out of the question.

I know you can't measure the temp rise of the load, but is there any
reason why you couldn't measure the temp rise of an in series resistor
bolted to a known thermal mass?

Been a while since I did physics But shouldn't there be a definite
relationship between total real power and the temp rise, even for spikey
loads?

Lee
--
Email address is valid, but is unlikely to be read.

"Grunff" wrote in message
...
Hi all,

This would probably fit better in another NG, but I know quite a few
people on here who will be able to provide some good ideas, so here goes.

I have an experiment where a voltage of ~250 VDC is applied across a load,
resulting in a current consumption of around 2A. The experiment runs for
around 5 minutes.

The nature of the load is such that the current consumption is very
spiky - it may average to around 2A, but you get very short, sharp spikes,
probably of 10-15A.

I need to be able to measure, with 1% or better accuracy, the total energy
dissipated in the load over the duration of the experimental run.

I don't need instantaneous V or I values - all I need to know is the total
energy consumption over the run.

The only sensible way I can think of doing this is to use a data logger,
and sample V and I at a high rate thoughout the experimental run, then
calculate average power over the run, then multiply by run time.

However, this seems like a very round-about way of achieving my goal.
Anyone have any thoughts on other ways of doing this? I must stress that
it is the electrical energy input to the system that I'm interested in, so
measuring temperature rise of the load etc. is out of the question.


TIA


--
Grunff

Use some voltage to frequency converter chips and count the clicks?

"Grunff" wrote in message
...
Hi all,

This would probably fit better in another NG, but I know quite a few
people on here who will be able to provide some good ideas, so here goes.

I have an experiment where a voltage of ~250 VDC is applied across a
load, resulting in a current consumption of around 2A. The experiment
runs for around 5 minutes.

The nature of the load is such that the current consumption is very
spiky - it may average to around 2A, but you get very short, sharp
spikes, probably of 10-15A.

I need to be able to measure, with 1% or better accuracy, the total
energy dissipated in the load over the duration of the experimental run.

I don't need instantaneous V or I values - all I need to know is the
total energy consumption over the run.

The only sensible way I can think of doing this is to use a data logger,
and sample V and I at a high rate thoughout the experimental run, then
calculate average power over the run, then multiply by run time.

However, this seems like a very round-about way of achieving my goal.
Anyone have any thoughts on other ways of doing this? I must stress that
it is the electrical energy input to the system that I'm interested in,
so measuring temperature rise of the load etc. is out of the question.


TIA


--
Grunff

I may remember incorrectly, but old rms power meters used to work on the
temperature rise
reated by the power passing through, didn't they? The time constant of such
a meter may be long enough to integrate the spikes out. However, the output
will be analogue which isn't convenient
for data analysis if the rmspower varies significantly on a timescale longer
than the thermal time constant of the meter. You could run theoutput to a
pen plotter ( if available ) and integrate visually
the area nder the curve. All very laborious of course. Perhaps apparatus
that is essentially a calorimeter would work ( i.e. pick off some of the
load power with a series resistor which sits inside a calorimeter: wait five
minutes and measure the temperature rise ).

Andy.

Grunff wrote:
Hi all,

This would probably fit better in another NG, but I know quite a few
people on here who will be able to provide some good ideas, so here goes.

I have an experiment where a voltage of ~250 VDC is applied across a
load, resulting in a current consumption of around 2A. The experiment
runs for around 5 minutes.

The nature of the load is such that the current consumption is very
spiky - it may average to around 2A, but you get very short, sharp
spikes, probably of 10-15A.

I need to be able to measure, with 1% or better accuracy, the total
energy dissipated in the load over the duration of the experimental run.

I don't need instantaneous V or I values - all I need to know is the
total energy consumption over the run.

The only sensible way I can think of doing this is to use a data logger,
and sample V and I at a high rate thoughout the experimental run, then
calculate average power over the run, then multiply by run time.

However, this seems like a very round-about way of achieving my goal.
Anyone have any thoughts on other ways of doing this? I must stress that
it is the electrical energy input to the system that I'm interested in,
so measuring temperature rise of the load etc. is out of the question.


TIA



If you want super accurate, you'll need to sample the instantaneous
voltage and current rapidly and record the results for the length
of your experiment.

From that you can work out the power consumption by computing the
instantaneous power, and then integrating that over the period of
the experiment.

For less accuracy, you could:

1. Assume a constant mains supply voltage & frequency
2. Using a True RMS meter with peak avg/min/max record
facility, measure the current.
3. Compute the avg/min/max power from the meter readings
and assumed supply voltage.

For example, a Fluke 87 meter would do this quite happily.

Cheers,
Mark.

mark wrote:

1. Assume a constant mains supply voltage & frequency

It's a regulated DC power supply.

2. Using a True RMS meter with peak avg/min/max record
facility, measure the current.

It's DC.


3. Compute the avg/min/max power from the meter readings
and assumed supply voltage.


But the rest is spot on ;-)


--
Grunff

andrewpreece wrote:

Perhaps apparatus
that is essentially a calorimeter would work ( i.e. pick off some of the
load power with a series resistor which sits inside a calorimeter: wait five
minutes and measure the temperature rise ).


I think this is very cunning (also suggested by Lee) - it may well be
the way to do it. The only thing that concerns me is what kind of big
power resistor to use and guarantee that it will remain sufficiently ohmic.


--
Grunff

On Sun, 20 Feb 2005 17:18:09 -0000, "John"
wrote:

Right, ideally you need to multiply I and V and integrate the product.
Easy if you had an 1960's analogue computer, but these are all in museums
now.

There are still plenty in the Burr-Brown or Analogue Devices
catalogues. This is a common enough task and an analogue multiplier is
still an appropriate solution to it.

Given how cheap accurate and stable op-amps have been for years, it's
not that hard to build your own anyway.

Andy Dingley wrote:

There are still plenty in the Burr-Brown or Analogue Devices
catalogues. This is a common enough task and an analogue multiplier is
still an appropriate solution to it.

Given how cheap accurate and stable op-amps have been for years, it's
not that hard to build your own anyway.


But (and please correct me if I'm wrong) I still need to sample the
output of the multiplier at a reasonably high rate, then integrate the
data over the time of the run, right?

So all the multiplier means is that instead of logging V and I, we log
the power? That means I still have to get a reasonably fast data logger.
If I'm going to do that anyway, then I'd be inclined to skip the
multiplier, grab the raw numbers, and do the calculation afterwards - or
have I missed something?


--
Grunff

On Sun, 20 Feb 2005 21:35:23 +0000, Grunff wrote:

But (and please correct me if I'm wrong) I still need to sample the
output of the multiplier at a reasonably high rate, then integrate the
data over the time of the run, right?

Depends how analogue retro you want to be. You _could_ integrate it in
an analogue stylee, before sampling. Although I wouldn't. I have
done this, but not for about 15 years. A/D and data capture is just
faster these days.


So all the multiplier means is that instead of logging V and I, we log
the power?

The reason for using an analogue multiplier is when the power factor
(i.e. the ratio between the measured values) is all over the place.
This may need to be "sampled" at a high rate, so that appropriate
pairs of values are multiplied together.

Your "spiky" load is probably not that spiky, compared to this
requirement. I'd expect you could easily sample and mulitply
digitally, with a speed well in excess of the experimental variation
(You'll need to actually work this out anyway, to do your error
analysis).

The other option is of course a pair of synchronised analogue
sample-and-holds (dead easy), then digitising their values at leisure.

In article ,
Grunff wrote:

[snip]
The nature of the load is such that the current consumption is
very spiky - it may average to around 2A, but you get very
short, sharp spikes, probably of 10-15A.

Do you have an idea of the timing, of the shortest
duration of a current spike?

I need to be able to measure, with 1% or better accuracy, the
total energy dissipated in the load over the duration of the
experimental run.
[snip]
The only sensible way I can think of doing this is to use a data
logger, and sample V and I at a high rate thoughout the
experimental run, then calculate average power over the run,
then multiply by run time.

I'm not sure that Average*Time will get the total
energy. Don't you have Integrate the instantaneous
power over duration of the run time?

Take V+I readings at fixed time intervals (significantly
shorter than the shortest spike), multiply V*I, and
accumulate.

The multiply and accumulate could either be done in
real time, or the raw V+I data log could be processed
off-line, say with something like MathCad.

--
Tony Williams.

"Grunff" wrote in message
...
andrewpreece wrote:

Perhaps apparatus
that is essentially a calorimeter would work ( i.e. pick off some of the
load power with a series resistor which sits inside a calorimeter: wait
five
minutes and measure the temperature rise ).


I think this is very cunning (also suggested by Lee) - it may well be
the way to do it. The only thing that concerns me is what kind of big
power resistor to use and guarantee that it will remain sufficiently
ohmic.


--
Grunff

A 25W wirewound would be good to better than 1% probably - the blocky gold
finish type with solder terminals, RWR### I think they might be called. I
have no idea of what
type of power you are trying to measure. Anyway, you need only sample a
fraction of it, then
multiply up the result to get the power consumed by the load. Might be
better to do it over a longer period than 5 minutes though so temperatures
in a calorimeter have a chance to equalise.
The Fluke 87 idea sounds like a lot less hassle, is that the one that
measures 'true rms'? I'd
check that it has a sampling time fast enough to record your glitches
though.

Andy.

Grunff wrote:
mark wrote:

1. Assume a constant mains supply voltage & frequency


It's a regulated DC power supply.

OK. Depends on how good the regulation is as to whether
this assumption holds. Does it sag a bit?


2. Using a True RMS meter with peak avg/min/max record
facility, measure the current.


It's DC.


The current isn't static is it? It's a spikey load.
There will still be a min/max/avg figure even for DC.


3. Compute the avg/min/max power from the meter readings
and assumed supply voltage.



But the rest is spot on ;-)



thanks
Mark.

mark wrote:

OK. Depends on how good the regulation is as to whether
this assumption holds. Does it sag a bit?


It doesn't under normal load, but I'm sure that under the high peaks it
does, yes.


The current isn't static is it? It's a spikey load.
There will still be a min/max/avg figure even for DC.

Right, OK - I thought you'd misread as AC. I see where you're coming
from now.


--
Grunff

andrewpreece wrote:

A 25W wirewound would be good to better than 1% probably - the blocky gold
finish type with solder terminals, RWR### I think they might be called.

Yeah, I know the ones - I have a bunch of them somewhere.



Anyway, you need only sample a
fraction of it, then
multiply up the result to get the power consumed by the load. Might be
better to do it over a longer period than 5 minutes though so temperatures
in a calorimeter have a chance to equalise.

I think I can set up a reasonably accurate calorimeter, and a few
calibration runs should allow me to compensate for its characteristics
quite nicely.



The Fluke 87 idea sounds like a lot less hassle, is that the one that
measures 'true rms'? I'd
check that it has a sampling time fast enough to record your glitches
though.

I will look into this further, thanks.


--
Grunff

Andy Dingley wrote:

The reason for using an analogue multiplier is when the power factor
(i.e. the ratio between the measured values) is all over the place.
This may need to be "sampled" at a high rate, so that appropriate
pairs of values are multiplied together.

Ah, ok, gotcha.


Your "spiky" load is probably not that spiky, compared to this
requirement. I'd expect you could easily sample and mulitply
digitally, with a speed well in excess of the experimental variation
(You'll need to actually work this out anyway, to do your error
analysis).

I think based on all the good advice so far, I'm going to go for a two
pronged approach. I'll set up a calorimiter, and use the resistor method
discussed elsewhere in the thread, and I'll also get a data logger (as
fast as I can afford) and do the measurement that way as well. Should be
interesting to compare the results.


--
Grunff

Grunff wrote:

I think based on all the good advice so far, I'm going to go for a
two
pronged approach. I'll set up a calorimiter, and use the resistor
method
discussed elsewhere in the thread, and I'll also get a data logger
(as
fast as I can afford) and do the measurement that way as well. Should
be
interesting to compare the results.

I dont rmember you saying how quickly the spikes repeat. If it many
times per second, all you need is an analogue meter, problem solved. MC
meters read average current.

If thats no good I'd use a small series R and feed the signal to an
opamp integrator.


NT

In article ,
Grunff wrote:

I think based on all the good advice so far, I'm going to go for
a two pronged approach. I'll set up a calorimiter, and use the
resistor method discussed elsewhere in the thread, and I'll also
get a data logger (as fast as I can afford) and do the
measurement that way as well. Should be interesting to compare
the results.

Might be useful to look at the LEM Hall Effect Current
Sensors. This would do isolated current sensing with
minimum voltage drop. eg, Radiospares 257-414 is 10V
dc out for 200 Ampere-Turns full scale input. Simply
pass 20-turns (of insulated flexible wire) through the
central hole and it is 10V out for 20A in. The voltage
drop caused by the sensor is just the resistance of 20
turns of wire.

There are other HE sensors from LEM which give a precise
current ratio. For example one of these could be used
to provide a an exact 20A:20mA ratio, with the 0-20mA
going sideways into a calorimeter, or to a (digital)
integrator, or to a data logger.

--
Tony Williams.

wrote:

I dont rmember you saying how quickly the spikes repeat. If it many
times per second, all you need is an analogue meter, problem solved. MC
meters read average current.


It varies from a few hundred Hz to a few 10s of kHz throughout the
experiment. The analogue meter is a fine idea, but would require manual
number-taking, which I'm trying to avoid (average I doesn't stay
constant, and I need to calculate total energy).


--
Grunff

Tony Williams wrote:

Might be useful to look at the LEM Hall Effect Current
Sensors. This would do isolated current sensing with
minimum voltage drop. eg, Radiospares 257-414 is 10V
dc out for 200 Ampere-Turns full scale input. Simply
pass 20-turns (of insulated flexible wire) through the
central hole and it is 10V out for 20A in. The voltage
drop caused by the sensor is just the resistance of 20
turns of wire.


That's a great idea, thanks. 257-414 is discontinued, but there are
quite a few other suitable ones.


--
Grunff

Grunff wrote:

If I'm going to go down the data logger route, I will probably go for

one of these:
http://www.audon.co.uk/labjack.html

I used one before in a previous project, and found it to be a very
nice
piece of kit.

But that wasn't my question really - I can see how to get the
measurement I want by sampling V and I with a data logger - but I
can't
help wondering if there isn't a cleverer, simpler way, given that all
I
want is total energy.

Another way of asking the question - I can see how to integrate the
data
electronically, after collecting a huge number of data points. But
I'm
wondering if there isn't another way of doing the integration. If
that
makes sense.


--
Grunff

there is, use an opamp integrator. No sampling, along with the errors
that introduces.

NT

In article ,
Grunff wrote:

That's a great idea, thanks. 257-414 is discontinued, but there
are quite a few other suitable ones.

I use quite a number of the LA55-P (RS 180-7357).
16-turns of ptfe insulated wire to sense 10Adc
full scale being pulled from a 347Vdc supply.
15-0-15V supplies, with following opamp to provide
the Zero and Gain adjustments.

Calibration can be done off-line, with (say) a
20Vdc source, 2 ohm resistor, and digital ammeter.
Once set up it has a repeatability within 0.25%
(of fsd).

AFAIK, the sum to be done will be something like.

T*(P1+P2+P3.......Pn)
Total Energy = --------------------- Joules.
N

P1(etc) are a series of discrete Power calculations,
taken at time intervals T/N, where T/N is small
compared with T. T/N must also be small compared
to the shortest transient P.

For a 5 minute run, T= 300 seconds, and T/N might
typically be 100uS. This gives N= 3000000, which
is the number of (pairs of?) data points to log.

--
Tony Williams.

wrote:

there is, use an opamp integrator. No sampling, along with the errors
that introduces.

Now that does seem like a good idea - I could integrate both V and I
using opamps, then multiply to get total power - any problem with that
thinking?


--
Grunff

Grunff wrote:

Now that does seem like a good idea - I could integrate both V and I
using opamps, then multiply to get total power - any problem with that
thinking?

In general yes there is because the product of the means is not the same
as the mean of the product. You need Int[v*i]dt and _not_
Int[v]dt * Int[i]dt.

However in your case v is constant (IIRC) so they are the same, more or
less - depending on how good your stabilised supply is and whether
voltage drop in the wiring is an issue.

Drift is always a potential problem with analogue integrators though;
I'd go with your original data logging idea (provided you can sample
fast enough) and then number crunch in the digital domain.
--
Andy

wrote:
Grunff wrote:


If I'm going to go down the data logger route, I will probably go for


one of these:
http://www.audon.co.uk/labjack.html

I used one before in a previous project, and found it to be a very

nice

piece of kit.

But that wasn't my question really - I can see how to get the
measurement I want by sampling V and I with a data logger - but I

can't

help wondering if there isn't a cleverer, simpler way, given that all

I

want is total energy.

Another way of asking the question - I can see how to integrate the

data

electronically, after collecting a huge number of data points. But

I'm

wondering if there isn't another way of doing the integration. If

that

makes sense.


--
Grunff


there is, use an opamp integrator. No sampling, along with the errors
that introduces.

NT


haha, yes, nice one. You think that an analog integrator
isn't going to have errors? Especially accumulated over
minutes?

The data logger is the best idea, just record the I&V
waveforms over the period under question then do everything
else numerically.

Cheers,
Mark.

mark wrote:
wrote:
Grunff wrote:

I'm
wondering if there isn't another way of doing the integration. If
that makes sense.


there is, use an opamp integrator. No sampling, along with the
errors
that introduces.


haha, yes, nice one. You think that an analog integrator
isn't going to have errors? Especially accumulated over
minutes?

every method on the planet has errors. Try not to be dumb. We just need
to get in below 1%.

Digital sampling is going to be especially problematic with a spikey
load, as the digital sampling has to be so fast it maps the spikes out
accurately, ie a great number of samples per spike so the spike's shape
is followed faithfully throughout. This is necessary to come in at
below 1%.

It is difficult to achieve this unless you know the highest possible
frequency component of said spikes. In practice, if you dont, you have
a mathematical problem to pick the sampling frequency needed to keep
total error budget below 1%.

We know the spikes occur at upto 10kHz, but that is the repetition
rate, not the highest spike frequency component, which inevitably will
be orders of magnitude higher.

Analogue integration has issues like every method, but its relatively
easy, simple, and low cost, and coming in at 1% is practical.


The data logger is the best idea, just record the I&V
waveforms over the period under question then do everything
else numerically.

There might be a bit more to it.

BTW v and i need to be multiplied before the power product is
integrated. Since you need 1% accuracy you will need to measure v, you
wont do it just measuring and integrating i.


I always tell people not to ask electronics qusetions in uk.d-i-y, we
have a remarkable amount of expertise on diy, building etc, but this
can lull some into imagining this is a group with expertise on tronics.
While we do have some, most do not, and the truth does not fall out
clearly like it does with other topics. Only if you know the answers
will you recognise in most cases. sci.electronics.design is the place
to go for this stuff.


NT