Continue to Site

Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

  • Congratulations GregLocock on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

Analog Linearization

Status
Not open for further replies.

didymus7

Electrical
Dec 16, 2003
4
A couple of months ago, some guy with his own company approached my company with a circuit that consisted (he said) of 7 components (one op-amp, resistors and caps) that could improve linearization. We never got to see the circuit (He believed he could get rich on this...my company didn't bite). While we never had a black box from him, his simulations did show a large improvement in linearization. Normally we use a log amp that or just L/C matching (this is eddy-current sensing)to get some linearity, but depending on range/offset, we may end up with up to 10% non-linearity. Would anyone have any idea as to how he was configuring this circuit to linearize? Again, it was a single active component, an op-amp, with several passive components.

Thanks....Steve
 
Replies continue below

Recommended for you

Traditionally you can shape a transfer characteristic using a diode in series with a resistor across the feedback resistor in say an inverting opamp-based amplifier. Now all this is doing is creating an adjustable non-linear stage. This could compensate out a known non-linearity elsewhere. There is nothing very clever about that.

Or did you mean that the whole circuit responded less non-linearly than another whole circuit?
 
If you E-mail me the output vs input in data table format
and the desired specs, I will see what I can do...

<nbucska@pcperipherals.com>
 
logbook--As far as I know, the only active component is the opamp. Again, I'm only going by what he said.

The only other clue I have is from plotting the non-linearity versus distance. Normal curve shape is the classic 'S' curve. After his circuit, that 'S' curved looked like it had been full-wave rectified--two humps with a notch about mid-range.

Now that my company is not going to buy his idea, I was just wondering if he really discovered something wonderful.

didymus7
 
I can’t quite follow the curve shapes you are talking about, but I would like to. A picture (JPG/GIF etc) would help if you could put on a link to your own website or a site to which you can upload the picture.

It is debatable whether or not an ordinary silicon diode is an active component. Certainly a Tunnel diode and a Gunn diode would be considered active as they have transient power gain. But a 1N4148? I’m not so sure.

S-correction curves are done on TV scan linearisation circuits and consist of magnetically biased inductors. You can also get voltage dependant resistors. Rectification is another matter though.
 
I'm confused by your comment about the &quot;two humps.&quot;

These two humps are riding on top of what should be a straight line for the gain curve?

A zero-crossing notch is usually a really bad thing for amplifiers, since it introduces tons of distortion, particularly at low signal levels.

TTFN
 
Thanks for the responses guys. I'll have to post up on my personal website what I'm talking about. Later today.

didymus7
 
Several cheap methods of linearizing your signal:

1. To use a A/D with the output going to an eprom/eeprom (with the linearization table) output to a D/A (also needs a clock circuit)

2. Analog input to a single chip micro with Anaog in, programmed with linearization table, PWM output with R/C network to obtain a voltage signal.

3. A variation of No. 2 above is to use a V/F circuit to convert the signal to a digital form and use a micro without analog input.

All three of these are in used with some smart(?) sensors.
 
I'm afraid I'm a bit in the dark here. Linearization of what signal under what circumstances?

John
 
The point of this guys was that the signal did not matter. The case in point is the output of eddy current sensor electronics where the change in phase/amplitude is demodulated to a DC voltage. This DC voltage varies non-linearly with the change in phase/amplitude of the bridge to the distance between the sensor and metal. Different metals give different responses. This guy who approached us could take any signal and make it more linear with, he said, seven components, specifically, one opamp and 6 resistors and capacitors. Sorry, haven't gotten any info up on the web yet.

And Rich2001, we do have different models that incorporate all of what you have mentioned. I'm looking for an 'oddball' solution....;-)

didymus7
 
The oddball solution must be something nonlinear: if only
OA+R-s then it must be biased close to input limit or
current feedback amplifier or some similar non-standard OP-AMP.



<nbucska@pcperipherals.com>
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor