Op-amp circuit....

Yes. This is not a simple y = ax + b -- is that intentional? If so, what relationship do you really want?

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Yes it is intentionally a non-linear relationship. If I can get something close to the above it would be great since it does not have to be a perfect match as long as I am close.

-- or implement y = a2 * x^2 + a1 * x + a0. It all depends on what you're after, really. If you only care about those three points and the rest of your input range simply doesn't matter at all then you can do it with a few comparators and resistors, and you won't have to mess around with diodes or multipliers or any stuff like that.

You may also want to consider doing this with a microprocessor, DAC and ADC. Depending on the precision you need and the complexity of the curve you may come to a more economical solution that way. Without knowing your exact requirements between these points I'd consider all four possible solutions.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Yes, provided the input/output relationship is monotonic.

Surf for piece-wise linear curve-fitting.

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

| 1962 | I love to cook with wine. Sometimes I even put it in the food.

if the relationship is monotonic, could you use a ADC, a lookup table implemented by memory and a DAC for it?

See.....

Newsgroups: alt.binaries.schematics.electronic Subject: Piece-Wise Linear Question (from S.E.D) - PWL-Example.pdf Message-ID:

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

| 1962 | I love to cook with wine. Sometimes I even put it in the food.

What about using r-r opamps and letting some of them saturate to define inflection points? Of course, that will have the same glitch problems of most of the ideal-diode circuits.

I've done breakpoints with non-ideal (real) diodes to bend the curves, with some other diodes for tempco. The knees aren't as sharp, but there are no transient spikes.

John

yeah, but you only gave spot points, a complete piecewise function would be nicer.

Are there only three points or is it a continuum between 11.7 to 12.5? Second, what is the precision? I see three significant figures, two significant figures and one significant figure, which is it? Three significant figures implies an accuracy and precision of 0.1% while one significant figure can vary by 10%. This makes a big difference in how you approach the problem and the precision of the resistors and non-linear network required. There are also temperature and drift issues as well.

One poster mentioned using a microprocessor with ADC's and DAC's. It could be a simple look up table to develop the relationship. Beyond that, can the whole problem be done digitally without ever being in the analog, voltage domain? That may be the best and simplest. In other words, where do the voltages come from and how are the results used? It might be easier if those were digital numbers and not voltages from the get go.

Many "single supply" op-amps will swing very close to the (-) rail. If you run the op-amp from +15 to +24 and ground, you can make an op-amp's output stay at ground until you hit the knee in the transfer curve. A second op-amp can combine the output of this op-amp with your original signal to make a transfer function with an abrupt change in gains.

Op-amps like the LM324 come 4 to a package. You can make 3 knees in the transfer function this way.

--
--
kensmith@rahul.net   forging knowledge

Some op-amps have sharper "knees" close to the negative rails than others. I used this in a product about 20 years ago for thermocouple linearization-- the CMOS op-amp I used was nice and sharp. The LMV824 is not so good. WIth the LM324 you may have to keep the current through the output to a few tens of uA to keep it sharp (there's a nominal 50uA pulldown internally).

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Very nice idea.

Is this an expanded scale voltmeter for auto/bike use?

Okay so to answer some of the questions:

1) The voltage has to be continuous between the points; I am not just looking for those points 2) Precision has to be close, but not exact 3) Do not want to implement digital method 4) I am making the circuit to control the response of an RF attenuator that I am using to improve on an older system, and as you can see Analog stuff is not my strong point. The Existing system can change the output by 6dB by varying the control from 11.7-12.5 volts. These numbers cannot change. I am attempting to implement a COTS attenuator to do the same function, but I need to map the existing control voltage to the COTS attenuator voltages. I hope this makes sense.

I am thinking the PWL approach might be a good way to solve my problem

Between the points given, what do you have to be close to?

As I suggested elsewhere, a quad op-amp can give you 3 knees in the curve. That would be a 5 point PWL curve. This was based on the idea that the sense of all of the knees were the same way. If the slope does not monotonically increase, you can get extra knees by using rail to rail op-amps and having them hit each rail.

You can also fit to curves using X^N sort of functions implemented with analog multipliers. There are several analog multipliers on the market and some parts like the LT1228 and some AGCed IF stages that will serve as multipliers. If you can't stand PWL methods erros, maybe combining PWL with some non-linear function such as X^2 will do it for you.

--
--
kensmith@rahul.net   forging knowledge

12.08X² -282X +1649.5

Did everyone miss my piece-wise linear post....

Newsgroups: alt.binaries.schematics.electronic Subject: Piece-Wise Linear Question (from S.E.D) - PWL-Example.pdf Message-ID:

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Let the record show that I posted it 33 minutes before Ken did! So how come it wasn't *my* nice idea?

John

The slope looks increasing. If you have only 3 reference points, maybe a an exponential amplifier could suffice?

(Cf.

, figure 2.)