Correcting a LED response curve?

[John Fields]

Hi,

I am designing a product using a Vactrol (led controlled resistor) to
simulate a RTD. The led intensity will be control by the buffered output of
a D/A chip. It is working, but to be useful, we have to operate the led in
the worst region of it's curve, resulting in very coarse response. Since I
do not have enough resolution on this D/A to correct that from software I am
searching for a method to render the Vactrol response a bit more linear. Any
one can suggest a simple way to create a amplifier using a simple opamp or
transistor to reproduce the reverse curve of the led and correct the Vactrol
response. I tough that maybe using a led in the feedback path of an opamp
may result in a non linear amplifier response and that can be use to drive
the Vactrol. Any suggestions?

 

[Jim Thompson]

Note Figure 1, It has the same curvature problem as a Vactrol.

The OP is trying to get a linear resistance versus control current
curve.

...Jim Thompson

 

[John Popelish]

Hi,

I am designing a product using a Vactrol (led controlled resistor) to
simulate a RTD. The led intensity will be control by the buffered output of
a D/A chip. It is working, but to be useful, we have to operate the led in
the worst region of it's curve, resulting in very coarse response. Since I
do not have enough resolution on this D/A to correct that from software I am
searching for a method to render the Vactrol response a bit more linear. Any
one can suggest a simple way to create a amplifier using a simple opamp or
transistor to reproduce the reverse curve of the led and correct the Vactrol
response. I tough that maybe using a led in the feedback path of an opamp
may result in a non linear amplifier response and that can be use to drive
the Vactrol. Any suggestions?

If I was doing this, my first step would be to add a series
and parallel resistor to the Vactrol so that the network was
just capable of producing the small range of resistance
needed over the full usable range of LED current. Then I
would add a voltage to current converter between the DAC and
the LED.

 

[Joel Koltner]

The OP is trying to get a linear resistance versus control current
curve.

I haven't been watching this thread, so forgive me if this has already been
mentioned, but isn't the usual way to achieve this to try to get another
devices that matches the first one quite closely and run a feedback loop on
the second one (where you can turn the resistance into a voltage or whatever
and measure it directly), driving both with the same output signal from your
error amplifier?

 

[Jim Thompson]

I haven't been watching this thread, so forgive me if this has already been
mentioned, but isn't the usual way to achieve this to try to get another
devices that matches the first one quite closely and run a feedback loop on
the second one (where you can turn the resistance into a voltage or whatever
and measure it directly), driving both with the same output signal from your
error amplifier?

Yes, As I mentioned in...

From: Jim Thompson <[email protected]>
Newsgroups: sci.electronics.design
Subject: Re: Correcting a LED response curve?
Message-ID: <[email protected]>

I just made a Vactrol model, but probably won't have time to
demonstrate the loop method until morning... granddaughter chiming
that she's hungry ;-)

...Jim Thompson

 

[Phil Allison]

"Jacques St-Pierre"

I am designing a product using a Vactrol

** Vactrol is a TRADE NAME - you wanker.


(led controlled resistor) to simulate a RTD.


** Don't use pretentious acronyms without explanation.

The led intensity will be control by the buffered output of a D/A chip. It
is working, but to be useful, we have to operate the led in the worst
region of it's curve, resulting in very coarse response. Since I do not
have enough resolution on this D/A to correct that from software I am
searching for a method to render the Vactrol response a bit more linear.

** Current drive should help.

Or better still, use PWM - the average light output will then be a
linear function of duty cycle.



....... Phil

 

[sycochkn]

More likely because the resistance of this "vactrol" thing does not
vary linearly with light! LEDs are in fact quite linear.

You could try an arrangement with *two* Vactrol devices, provided
their transfer functions are sufficiently similar. Close a loop around
one to get linear resistance control, and feed the LED of the other
with the same current. Hopefully the output resistance of the "other"
will then be more linear, too.

The circuit is a bit too complicated for me to draw out in ASCII art
right now - perhaps someone else will comment. It may be a waste of
time anyway if you can't get matched "Vactrol" devices.

I have just looked them up - you can get dual ones apparently!

(Although I think Spehro is right - you need another approach)

Something like this?

Bob

http://home.earthlink.net/~sycochkn1/rtd.pdf

 

[John Larkin]

I haven't been watching this thread, so forgive me if this has already been
mentioned, but isn't the usual way to achieve this to try to get another
devices that matches the first one quite closely and run a feedback loop on
the second one (where you can turn the resistance into a voltage or whatever
and measure it directly), driving both with the same output signal from your
error amplifier?

Vactrols are notoriously unstable, and the Fairchild thing looks
pretty bad, too, not to mention its tiny linear signal range. An RTD
usually changes about 4000 PPM per degree C, so you need a precise
resistance to simulate to any decent accuracy. I really doubt that
even servoing "matched" Vactrols or Fairchild gadgets would be good
enough.

We have a design for a VME module that's an 8-channel RTD simulator;
one of these days we'll get around to announcing it as a product. It's
brute-force, with each channel fully isolated. The resistance is
simulated using a switched super-precision resistor (each channel has
4 ranges, I recall) whose value is actively scaled by opamps and
mdacs.

There are commercial units that just use a heap of resistors and
relays, which works well except for switching transients when you
change value.

We toyed with PWMing a real resistor, which is mighty appealing, but
has too many side problems.

John

 

[John Larkin]

Yes, As I mentioned in...

From: Jim Thompson <[email protected]>
Newsgroups: sci.electronics.design
Subject: Re: Correcting a LED response curve?
Message-ID: <[email protected]>

I just made a Vactrol model, but probably won't have time to
demonstrate the loop method until morning... granddaughter chiming
that she's hungry ;-)

...Jim Thompson

Vacs, or specifically the ldr's inside, have ghastly tempcos, and huge
history effects, and bizarre, complex time constants.

John

 

[John Devereux]

Something like this?

Bob

http://home.earthlink.net/~sycochkn1/rtd.pdf

Yes, looks good. Simpler than I expected since I didn't think of
putting the leds in series for some reason!

Works for the dual vactrol too.

Nice.

 

[John Devereux]

Note Figure 1, It has the same curvature problem as a Vactrol.

The OP is trying to get a linear resistance versus control current
curve.

And yet on the front page it says ">= 99.9% linearity".

At any rate, I think RTDs are around 100 ohms, so the device does not
seem well suited here.

 

[sycochkn]

Yes, looks good. Simpler than I expected since I didn't think of
putting the leds in series for some reason!

Works for the dual vactrol too.

Nice.

Depending on the OP's application it may or may not be close enough.

Bob

 

[sycochkn]

And yet on the front page it says ">= 99.9% linearity".

At any rate, I think RTDs are around 100 ohms, so the device does not
seem well suited here.

Perkin Elmer VTL5C3 1.5 ohms to 30k ohms is $3.79.

Bob

 

[Jacques St-Pierre]

I am designing a product using a Vactrol (led controlled resistor) to

simulate a RTD. The led intensity will be control by the buffered output
of a D/A chip. It is working, but to be useful, we have to operate the led
in the worst region of it's curve, resulting in very coarse response.
Since I do not have enough resolution on this D/A to correct that from
software I am searching for a method to render the Vactrol response a bit
more linear. Any one can suggest a simple way to create a amplifier using
a simple opamp or transistor to reproduce the reverse curve of the led and
correct the Vactrol response. I tough that maybe using a led in the
feedback path of an opamp may result in a non linear amplifier response
and that can be use to drive the Vactrol. Any suggestions?

Sorry if I offend anyone using the Vactrol name for the device. Since I did
not know of any other similar device, I tough using Vactrol should be clear
enough, but yes it's a "LED Controlled Resistor" as mention by Phil.

In fact I am using VTL5C4.

Thanks for your comments, I will experiment on them and post the results.

For now I intend to test this:

http://home.earthlink.net/~sycochkn1/rtd.pdf

It may be what I needed.

Bye
Jacques

 

[Jim Thompson]

Something like this?

Bob

http://home.earthlink.net/~sycochkn1/rtd.pdf

Exactly my thoughts except I was going to use a dual Vactrol, and some
resistance in series with the LED

I don't know about stability, though. The log-log plot on the data
sheet hides the wild curvature, thus extreme loop gain variation.

...Jim Thompson

 

[Jim Thompson]

Vacs, or specifically the ldr's inside, have ghastly tempcos, and huge
history effects, and bizarre, complex time constants.

John

Yep. In practice I seriously doubt getting reasonable stability.

...Jim Thompson

 

[john jardine]

"Fred Bartoli" <" "> wrote in message

[...]

Or use a "HF" pilot so sense the LDR resistance, that will be filtered
out for the actual output resistance.

Works nicely. Used it for a Twin T notch null.

 

[John Larkin]

"Fred Bartoli" <" "> wrote in message

[...]

Or use a "HF" pilot so sense the LDR resistance, that will be filtered
out for the actual output resistance.

Works nicely. Used it for a Twin T notch null.

Or use two LDRs and some analog switches. Alternately use one while
calibrating the other.

Are ldr's analog-voltage linear, in other words ohmic? RTD simulation
would need extreme linearity.

John

 

[Jim Thompson]

"Fred Bartoli" <" "> wrote in message

[...]

Or use a "HF" pilot so sense the LDR resistance, that will be filtered
out for the actual output resistance.

Works nicely. Used it for a Twin T notch null.

Or use two LDRs and some analog switches. Alternately use one while
calibrating the other.

Are ldr's analog-voltage linear, in other words ohmic? RTD simulation
would need extreme linearity.

John

Since it's injected carriers I'd guess it to be nonlinear, like a
MOSFET, for other than small voltage drops.

...Jim Thompson

 

[Spehro Pefhany]

Yep. In practice I seriously doubt getting reasonable stability.

...Jim Thompson

A precious-metal RTD changes less than +0.4%/K. Accuracy of the
*sensor* is typically +/-0.3 to +/-0.03K, so you'd normally want an
instrument to be considerably more accurate and stable than that. An
LDR doesn't have a hope in h-e-double-hockeysticks of getting there
unless this is a particularly sloppy & crappy application. The usual
method is to use an active analog/digital circuit.
Best regards,
Spehro Pefhany