how to take 0.5-1.8V signal and change it to 0-4.096V signal. Op-amp difference amplifier maybe?

You need a subtractor with a bit of gain. An opamp + 4 resistors.

R1 goes from the 0.5 reference voltage to the inverting input. R2 goes to the output from there.

R3 goes from the signal input to the non-inverting input, and R4 goes from there to ground.

R1 = R3, and R2 = R4.

The gain of the thing is R2/R1

This will subtract 0.5V from the 0.5-1.8 volt input, and multiply the result by R2/R1.

Using a divider with 1.8k and 200, you'll get 0.5V out of a 5V rail. This is your reference.

Using 24k for R1 and R3, and 100k for R2 and R4 gives you the right amount of gain.

A rail-to-rail opamp, like the microchip MCP6294, will keep the voltage output near the rails accurate.

The main issue is input impedance, which is going to be equal to R1. This can cause inaccuracy if your sensor has high output impedance. If that's a problem, use a buffer between the sensor and R1 (input to + input of opamp, - input to output of opamp)

--
Regards,
  Bob Monsen

If a little knowledge is dangerous, where is the man who has
so much as to be out of danger?
                                  Thomas Henry Huxley, 1877

Joerg wrote in news:UtVCe.740 $ snipped-for-privacy@newssvr21.news.prodigy.com:

All three channels can be on the same chip. Honestly the op-amp is the least of my worries though - I'm worried about the board being covered in pots (I think I need something like 9 pots)

-Michael

If possible, I would design the amplifier to keep the output between 0 and 4.096 for the worst case tolerance of the input, and digitally correct the zero and span for all other cases.

I don't think you are going to find any ganged pots, so if you want to adjust both zero and full scale, you will need two pots per channel. But remember that opamps are not much larger than resistors, so you may be able to use a multi opamp amplifier for each channel and reduce the other parts counts. And as long as you can stand a little interaction between the zero and span adjustments, you don't need a full subtracter. A non inverting amplifier (input signal to +, output to feedback resistor to end of span pot, wiper to -, other end of span pot to fixed resistor to wiper of zero pot across .6 volts or so and ground.) will not load your signal pot, and the resistor values can be chosen to produce only a tiny interaction between the two adjustments. 1 opamp, two fixed resistors and two pots per channel. Use a dual or quad amp.

"Mook Johnson" wrote in news:%3XCe.24565$ snipped-for-privacy@tornado.texas.rr.com:

Well - it's a 10b ADC with a 4.096V vref (I could probabaly lower that to

2.048 if it would help anything). I want to get about 360 steps between .5 and 1.8v. So - I could do something like just run the input signal through a non-inverting amplifier which would bring it up to a max voltage of 4.096V, so I'd get ADC readings of about 284 - 1023 (=range of 739) which would give me about double the range that I need, so it'd probabaly be OK. But I just think it would be a little more clean if I could use the full range of the ADC.

-Michael

Hello Michael,

Don't know what your supplies are but have a look at ye olde LM324A (not the non-A). This one has 3mV max offset while the regular version is

7mV. I believe the A also comes in TSSOP which should be small enough. And it is cheap.

If they must be scattered look for a SOT23-5 opamp such as the TLV2211.

Regards, Joerg

Hello Michael,

Know what you mean. Too many pots used to be called "rubber engineering". I guess you'd have to think about an alternative solution that either doesn't need all this adjustment or where you can do it digitally. There are lots of multi-channel DACs.

Then there is the measure-and-add-resistor alignment technique.

Regards, Joerg

Do you really need the full range of resolution from the A2D?

How small of a change are you looking to detect from a pot (noisy)? A 16 bit 0-4.096 A2D can detect down to 62uV of change per bit. With your given voltage range, you'll have 20K increments to play with (over 14 bits of resolution).

You can do all the scaling inside the micro.

Hello Michael,

All that effort for a gain of half a bit? If you really need that 1/2 bit why not offset-shift and sample again? Then you'd have 10 1/2 bits.

Regards, Joerg

If you're range is changing like you mentioned earlier, it would me a much more versitile design if you just used a protion of the A2Ds range and allowed the input to be whatever range they needed to be. As long as you have the resolution, you should be fine even with 10 bits. Be sure to tightly couple and filter the signal so you redice the LSB flicker to a minimum and you should be fine.

Could you use a multi-input 12 or 16 bit A2D? It will likely take less space than all those pots and be significantly less trouble as well (noise, drift, vibration sensitivity...etc). (I don't like using pots to set a stable voltage reference)

Because he'd only be using 26% of the ADC's range, and lose two bits of resolution.

Cheers! Rich

Op amp. easy fast and cheap. all you need is a gain of about 2. What are you driving?

My first guess would be an A/D, uC, D/A.

But when you want to A/D why go to this trouble at all? Just A/D the original signal and do the rest in software!

Wouter van Ooijen

-- ------------------------------------

Webshop for PICs and other electronics Teacher electronics and informatics

- does he realy need those two bits?

- feed the A/D a lower reference

- get an A/D with two more bits

- get two bits less per sample but add some noise and sample more

Wouter van Ooijen

-- ------------------------------------

Webshop for PICs and other electronics Teacher electronics and informatics