Analog Attenuation

Hello. I need to attenuate 5V analog signals down to 3.3V so they can be read into a 3.3V DSP's ADC. It has to be super SUPER accurate since I will be double-integrating the voltage signal with time and any errors will be hugely magnified.

There seems to be no prexisting attenuation ICs so I was looking at precision volttage dividers. But that I need to use large resistances for less power consumption but this causes problems the output impedence to be too high and input impedence to be too low (since they are potentially unknown this is a general purpose signal conditioner board). As a result, I considering at buffering the voltage divider with an op-amp or integrating the voltage divider with an op-amp for superior input resistance (see diagram).

But there are strange little non-idealities about op-amps such as input bias currents, offset voltages, slew rate, as well as another curious parameter called "Minimum Stable Closed-Loop Gain". This last one concerns me because my circuit will have a gain <1 and all the op-amps I have seen are all have minimum stable gains of 1 or higher.

Can anyone shed some light on this minimum stable gain and other op-amp no-idealities? The offset of 1mV on the output (cannot calibrate for it) causes huge errors to build up in my circuit...

Hmm, it might be best to calculate values, but fine tune the design using a multiturn pot on breadboard, that way you'll know it works perfectly in practise.

You're making this FAR more complicated than it is! - all you need is TWO RESISTORS - a simple potential divider. You are correct in that a non-inverting opamp has a minimum gain of ONE, so couldn't be used (except as a buffer for the above mentioned potential divider).

dknguyen said:

my circuit will have a gain <1 and all the op-amps I have seen are all have minimum stable gains of 1 or higher.

Click to expand...

The gain of the op amp is regardless of the voltage divider and it's set by the feed-back resistors. Considering a voltage follower, the gain is always 1.

The minimum stable gain is a condition for frequency stability. Be aware that this parameter refers to the voltage gain between the output and the input of the op amp only for the basic non-inverting configuration.
Most op amps are compensated for unity gain.

dknguyen said:

The offset of 1mV on the output (cannot calibrate for it) causes huge errors to build up in my circuit...

Click to expand...

Some op amps have offset nulling pins.

I would have to agree with Nigel on this one a simple voltage divider is all that is needed.

if you get some high-precision ones as well you will not get any extra offset (wich will cause yr integrators to give false-outputs)

adding in OPAMPS will just add offsets which unless you really need to buffer the signal try to stear away from

The only problem I have with using a simple divider is the potential of loading the input or output too much. I would like to use big resistors to minimize power consumption and increase the input impedence, but wouldn't it load down the output if I did that? (It's happened to me before)...regardless...a simple divider looks really REALLY inviting since I need to cram a bunch of them onto the same board. Its just the output impedence thing...I am not quite sure what the input impedence is for an IC.

dknguyen said:

The only problem I have with using a simple divider is the potential of loading the input or output too much. I would like to use big resistors to minimize power consumption and increase the input impedence, but wouldn't it load down the output if I did that? (It's happened to me before)...regardless...a simple divider looks really REALLY inviting since I need to cram a bunch of them onto the same board. Its just the output impedence thing...I am not quite sure what the input impedence is for an IC.

Click to expand...

Resistive attenuator followed by a unity gain opamp buffer, as I suggested before it depends on the impedances involved - which you've never mentioned?.

What EXACTLY are you trying to do?.

I am trying to make a general purpose PCB to convert 5V analog sensor signals so that they can be read into the DSP's ADC (which is only 3.3V tolerant). So in general, I neither know the sensor's output impedence or the DSP's input impedence.

Nigel Goodwin said:

Resistive attenuator followed by a unity gain opamp buffer

Click to expand...

I agree. The low output impedance of the op amp assures proper operation of the ADC.

It is what we do in our products. The joy is that once layed out, resistor values can be chosen for any desired attenuation factor.

Again, WHAT SENSOR WHAT ADC WHAT CONNECTION?
Do you need high absolute accuracy, or just consistency? Resistors aren't super-accurate until you calibrate their values in the microcontroller code. But they're generally the same one moment to the next.
If you have a differential sensor, chances are a resistive divider is not a good idea.
In fact I'm already betting your sensor could be driven just as well with 3.3v, making this question silly.

dknguyen said:

I am trying to make a general purpose PCB to convert 5V analog sensor signals so that they can be read into the DSP's ADC (which is only 3.3V tolerant). So in general, I neither know the sensor's output impedence or the DSP's input impedence.

Click to expand...

In which case your question has no specific answer, as it's not a specific question. But it's already been answered fully, it's simply a question of designing it accordingly when you know the specs you need.

Yeah, thanks. I think I got it. I forgot that I was outputting through a filter chip with a known input impedence...so I should be able to optimize for that with no problems. I'm just going to stick the buffer at the input so it can deal with the unknown output impedences of a wider range of sensors better.

I wish the sensors could be driven with 3.3V! But nooooo, it has to cut out at 4.75V would make things sooo much easier.