Searching \ for 'RMS to DC conversion' in subject line. ()
Help us get a faster server
FAQ page: techref.massmind.org/techref/index.htm?key=rms+conversion
Search entire site for: 'RMS to DC conversion'.

Truncated match.
'RMS to DC conversion'
1998\08\18@141425 by

I'm playing around with some ideas for RMS to DC conversion and
thought I'd bounce ideas off the list (and yes, the DC output IS to drive
a PIC!).
The typical RMS to DC converter is pretty much an analog
computer, computing the square, the average, the square root, etc.
Another way to determine RMS is to go back to the "heating value"
idea of RMS.  This is used in the old thermocouple RF ammeters (such as
those from Simpson).  The thermocouple seems like an interesting
approach, since it is indeed a differential temperature sensor.  Since a
thermal resistance will "drop" so many degrees per watt, it seems that we
can directly determine the power dissipated in a heater by using the
differential temperature measured by the thermocouple and the
characteristics of the "insulator" or "thermal resistor" developing the
temperature drop.  This SHOULD be independent of ambient temperature,
since it's temperature rise (or drop) we're measuring, not absolute
temperature.
Linear Technology uses a similar idea in the LT1088.  Here,
however, instead of measuring temperature drop, they have a pair of
heaters and a pair of temperature sensors.  An op amp is used to drive
one heater so its temperature matches the other heater.  It's then easy
to figure out the power dissipated in the op-amp driven heater, which
matches the power driving the "unknown" heater.  It seems that there
might be a difficulty due to temperature coefficient of resistance of the
heaters, but as long as they match, I'd guess that the errors cancel.
The LT1088 is an interesting part.  It accurately measures RMS up
to 100 MHz.  However, the "unknown" heater terminals must be within 40
volts of ground.  I'm interested in measuring RMS of an isolated circuit
that is up to several hundred volts from ground.
So...  how about measuring the LIGHT emitted by the heater,
instead of the temperature of the heater?  I recall years ago seeing
optocouplers (I think they were called Sigma Datacells) that included an
incandescent lamp driving a CDS photocell.  It SEEMS that this could be
used to measure the RMS voltage applied to the incandescent lamp.  There
may be repeatability problems and nonlinearity problems, but it MIGHT
work.
CDS cells suffer from "memory" in that they have a "light
perhaps a photodiode or a phototransistor would be a better sensor for
this purpose.  I think we need to stay with the incandescent lamp to get
the RMS conversion.  An LED is just too fast.  The output would just
follow the input.  The lamp filament design could determine the time over
which the RMS value is determined.

Thanks!

Harold

Harold Hallikainen
haroldhallikainen.com
Hallikainen & Friends, Inc.
See the FCC Rules at http://hallikainen.com/FccRules and comments filed
in LPFM proceeding at http://hallikainen.com/lpfm

_____________________________________________________________________
Get completely free e-mail from Juno at http://www.juno.com
Or call Juno at (800) 654-JUNO [654-5866]

Harold speculated:

(much snipped)

So...  how about measuring the LIGHT emitted by the heater,
{Quote hidden}

Interesting idea. One problem is that "nonlinearity" is an understatement.
The light output of a typical incandescent lamp at its rated voltage varies
as something like the twelfth power of voltage.

One classical way to deal with transducers that are sensitive but nonlinear
is to put them inside a feedback loop where they are driven to maintain a
fixed operating point. Then use the external parameter to be measured as an
input that tends to disturb the loop. The voltage/current/whatever that it
takes to keep the loop at the operating point is a measure of the external
parameter. This effectively divides the nonlinearity by the loop gain.

Compensating for temperature sensitivity is often a matter of developing a
differential signal from a pair of more-or-less identical transducers
exposed to the same environment. With light output, you can also use a
chopper to alternately view the test device and a standard.

As you pointed out, the advantage of your idea is that it tends to be
insensitive to frequency and crest factor, and the output is inherently
isolated from the input.

There's a device in my "exotica" junkbox somewhere that I think was
designed for this. It has two separately terminated filaments in a little
glass envelope. My recollection is that it used the nonlinear resistance of
the filaments in a bridge circuit.

For a minimalist design, the TSL235 light sensor from TI is in a TO-92
sized package. It can feed a PIC a digital signal directly. Just count the
number of output pulses in a fixed time window. Compensate for
non-linearity in software. Cheap and simple (everything is simple when you
haven't actually done it...).

Reg Neale

Reginald Neale wrote:

> One classical way to deal with transducers that are sensitive but
> nonlinear is to put them inside a feedback loop where they are driven
> to maintain a fixed operating point.

Nice idea, but in the case of temperature sensing at least, there is
still a problem of compensating the "fixed operating point".  You can
for example, add DC power to RF to stabilise your thermocouple at a
certain temperature, but the required DC power is now dependent on
ambient temperature as well.  You need a second identical thermocouple
to compensate this, or else switch ("chop", albeit slow) your RF input
on and off.

Either you use two matched heater/ thermocouple pairs with the
required isolation built-in (traditional RF meter thermocouples have
*no* isolation; all four wires are connected), or you will have to
switch something using mechanical (relay) switching.  I can't see any
other way to fulfil isolation, linearisation and temperature
compensation.
--
Cheers,
Paul B.

At 02:10 PM 18/08/98 EDT, you wrote:
>        I'm playing around with some ideas for RMS to DC conversion and
>thought I'd bounce ideas off the list (and yes, the DC output IS to drive
>a PIC!).
>

OK, so we already have a device to measure the RMS level, just that it is
not isolated. So why not use a simple Voltage to Frequency converter? Or use
one of those isolated opamps from Burr Brown?
Just some ideas if you need something that works. I see many problems with
the incandescent light source, one of them being repeatability over time.
Non linearities can be fixed up by putting the sensor into a bridge
configuration and good old PIC software.

Dennis

>For a minimalist design, the TSL235 light sensor from TI is in a TO-92
>sized package. It can feed a PIC a digital signal directly. Just count the
>number of output pulses in a fixed time window. Compensate for
>non-linearity in software. Cheap and simple (everything is simple when you
>haven't actually done it...).

I have used the TSL235 quite a bit in a Optical Strain Guage project I have work
ed on.  The TSL is a great product however it has
quite a bit of jitter in it's output.  I seem to remember 10Hz or so.  (I was tr
ying to get 1Hz resolution).  My sample period was a 1
second window and I ended up haveing to average that. Plus don't even think abou
t using it underneath 60Hz flourescents with out
a light shield of some sort.  It's sensitive to IR as well but in this case I gu
ess that's good.

--
Richard A. Smith                         Bitworks, Inc.
rsmithbitworks.com               501.521.3908
Sr. Design Engineer        http://www.bitworks.com

On Wed, 19 Aug 1998 08:35:32 +1000 "Paul B. Webster VK2BZC"
<paulbmidcoast.com.au> writes:
{Quote hidden}

This "fixed operating point" idea is something I've seen in Hall
effect DC ammeters - apply an equal but opposite field to cancel the
"unknown" field.  So, all we need to do is reverse the voltage to our
heater or light bulb!
Well....  How about setting the "fixed operating point" at full
scale and having a SECOND heater (or light bulb, if we are sensing light
instead of heat).  The SUM of the heat or light output of the pair would
be held at full scale by adjusting the power to the "known" heater or
light bulb.  This would get rid of nonlinearity of the sensor, but not
that of the heater and/or light bulb.
The LT 1088 has two heaters and two temperature sensors.  The
"unknown" voltage heats one heater.  The "known" heats the other to the
same temperature.  The RMS voltages then match (assuming identical
heaters, temperature sensors, insulation resistance, and ambient
temperature).
The use of a photodetector is interesting to me in that the
output should now be independent of temperature (until ambient approaches
the filament temperature).  I'd LIKE a simple isolated input RMS to DC
converter that I can throw across a current sense resistor and connect
the other side to a PIC analog input.  There are probably extreme
problems with linearity and repeatability, but I'll mess with it.
Anyone know of an optocoupler with an incandescent lamp instead
of an LED as the light generator?  There used to be a "Sigma Datacel",
but it appears it is no longer manufactured.

Thanks!

Harold

Harold Hallikainen
haroldhallikainen.com
Hallikainen & Friends, Inc.
See the FCC Rules at http://hallikainen.com/FccRules and comments filed
in LPFM proceeding at http://hallikainen.com/lpfm

_____________________________________________________________________
Get completely free e-mail from Juno at http://www.juno.com
Or call Juno at (800) 654-JUNO [654-5866]

>        Anyone know of an optocoupler with an incandescent lamp instead
>of an LED as the light generator?  There used to be a "Sigma Datacel",
>but it appears it is no longer manufactured.
>

C.P. Clare Co. used to have a line of these also. They may still be around.

Reg Neale

Harold wrote:

> Anyone know of an optocoupler with an incandescent lamp instead
> of an LED as the light generator?  There used to be a "Sigma Datacel",
> but it appears it is no longer manufactured.

EG&G Vactec have a whole spectrum of incandescent/CdS opto's in their VTL3Axx li
ne.
They even have a very simple RMS measurement circuit in the Typical Applications
section
of their Optoelectronics Data Book.

They are at:
http://www.egginc.com/bin/webmate/egg/page/egg/index

Best of luck,

Michael

*************************************************************************
When the way of the Tao is forgotten, kindness and ethics must be taught.
Men must learn to pretend to be wise and good.  --  Lao Tzu
*************************************************************************

On Wed, 19 Aug 1998 02:04:08 -0700 Mike Ghormley <mikegBLACKSAND.COM>
writes:
>Harold wrote:
>
>> Anyone know of an optocoupler with an incandescent lamp instead
>> of an LED as the light generator?  There used to be a "Sigma
>Datacel",
>> but it appears it is no longer manufactured.
>
>EG&G Vactec have a whole spectrum of incandescent/CdS opto's in their
>VTL3Axx line.
>They even have a very simple RMS measurement circuit in the Typical
>Applications section
>of their Optoelectronics Data Book.
>
>They are at:
>        http://www.egginc.com/bin/webmate/egg/page/egg/index
>

I've had trouble getting data out of EG&G Vactec.  It took
several tries to get the Optoelectronics databook.  I'd request it and
they would send me their "short form catalog" which tells almost nothing
about the products.  Now, I DO have the Optoelectronics Databook (dated
November 1997).  It lists ONLY LED optocouplers (no incandescent).  Also,
the part numbers are all VTL5..., no VTL3 series.  Checking the web site
you gave, I again find VTL5's with LEDs.
Is the incandescent VTL3 a current product?  Do you have a full
URL to a web page on this product?

Thanks!!!

Harold

Harold Hallikainen
haroldhallikainen.com
Hallikainen & Friends, Inc.
See the FCC Rules at http://hallikainen.com/FccRules and comments filed
in LPFM proceeding at http://hallikainen.com/lpfm

_____________________________________________________________________
Get completely free e-mail from Juno at http://www.juno.com
Or call Juno at (800) 654-JUNO [654-5866]

Harold Hallikainen wrote:

> This "fixed operating point" idea is something I've seen in Hall
> effect DC ammeters - apply an equal but opposite field to cancel the
> "unknown" field.  So, all we need to do is reverse the voltage to our
> heater or light bulb!

Ha!  The single most elegant application using this approach I think
is the precision weighing balance.  The balance position is set so the
design of the solenoid is non-critical, the sensor can be a standard
opto-chopper and accurately controlling current is nowadays, quite easy
and precise.  The mechanical engineering of the balance becomes
relatively easy to do at home (I'm not looking for an argument here!).

> Well....  How about setting the "fixed operating point" at full
> scale and having a SECOND heater (or light bulb, if we are sensing
> light instead of heat).  The SUM of the heat or light output of the
> pair would be held at full scale by adjusting the power to the "known"
> heater or light bulb.  This would get rid of nonlinearity of the
> sensor, but not that of the heater and/or light bulb.

An interesting variant on what I said and you quoted.  If you have a
dual coupler with good matching at least within a specified range, but
not linearity, you can make a decently accurate bolometer.  If in
addition you have matching over an extended range and good isolation in
each coupler, you can have an isolated and accurate system.  Yours seems
to be a cute variant on the "isolation" theme.

If you had isolation *and* linearity, you wouldn't need matching.
Sort of "you can't have it all" effect.
--
Cheers,
Paul B.

Harold, I have'nt followed all of this thread but you might just use
a light bulb and a CdS cell. Jameco has several CdS cells. Try:

http://www.jameco.com

- Tom

At 01:03 PM 8/19/98 EDT, Harold Hallikainen wrote:
{Quote hidden}

On Fri, 21 Aug 1998 08:57:28 -0700 Tom Handley <thandleyTELEPORT.COM>
writes:
>   Harold, I have'nt followed all of this thread but you might just
>use
>a light bulb and a CdS cell. Jameco has several CdS cells. Try:
>
>      http://www.jameco.com
>
>   - Tom
>

I know I started a deluge of mail on this, but I think I just
MIGHT have enough time left on the PIC to do the RMS calcualtion in
software.  I'll use a current transofrmer driving a bridge rectifier
driving a terminating resistor.  The PIC A/D reads the voltage across the
terminating resistor.
To compute the RMS, I'm thinking of doing something I did in an
inventory system 15 years ago or so, and that is to use a "weighted
average".  It would go something like this...

total=total-(total/64)          ; Subtract out a typical
sample, assuming we
; are going to average 64 samples.  64
used since
; it's so easy to divide by.  Could use
128, 256, etc.
GetNewSample
Square(NewSample)       ; perhaps using a lookup table?
total=total+sample^2
MeanSquared=total/64    ; Get the new average
RootMeanSquared=sqrt(MeanSquared)  ; Get new RMS

So... where was that square root routine archived???

Harold

Harold Hallikainen
haroldhallikainen.com
Hallikainen & Friends, Inc.
See the FCC Rules at http://hallikainen.com/FccRules and comments filed
in LPFM proceeding at http://hallikainen.com/lpfm

_____________________________________________________________________