INA333: How about one with differential *output*

Hello Rap,

Unfortunately we do not currently have a differential output INA like you have mentioned. This is something we are hoping to address soon.

Typically, you want to use an instrumentation amplifier for a bridge application because of the high input impedance. However, INAs are typically not very good at driving ADCs, particularly in high gain configurations, because they don't have the necessary bandwidth. Because all of the fully-differential amplifiers (FDA) in our portfolio are high speed, they tend to be well suited to driving ADCs with high accuracy. Due to their increased bandwidth, FDAs tend to require lower values of resistors around them, unless you compensate the amplifier appropriately. Thus you wont tend to see FDAs in bridge applications.

There are a few different approaches you could take:

1) If the input common-mode is already at the desired level and remains relatively constant, you could build up a discrete instrumentation front end (the input buffer stages without the difference amplifier) using a device such as the OPA2320.

2) You could use an INA as the bridge amplifier and follow this with a fully differential amplifier in low gain configured for a single-ended to differential conversion to drive the ADC. This will be the highest accuracy solution and you should easily be able to get 16 bits of resolution doing this.

3) There are also some ADC designed to interface directly with bridge sensors if this is something you'd like to explore. The ADS1131 is one example.

Zak I thank you very much for your reply. I was sort of dumbfounded when I looked around thinking - "well this has GOT to be something that already exists". Leave it to me to find the only pot hole to step in. It seems so natural to draw this in your mind or on paper -----

1. Bridge on left of your design paper/screen. (Differentially supplying signal)
2. INAMP in the middle of your design paper/screen. (Differentially clobbering noise off lines and supplying needed GAIN)
3. Differential in ADC on right of your design paper/screen. (WAITING ON DIFF OUTPUT FROM THE NONEXISTING DIFF OUTPUT INAMP.... Doh! )


I'm going to "line by line" this since you make great points and I can comment on each.

"Typically, you want to use an instrumentation amplifier for a bridge application because of the high input impedance."
Yes. No loading the sensor. even if it is a collection of 4 (full bridge) low impedance(resistance) strain gauges.

"However, INAs are typically not very good at driving ADCs, particularly in high gain configurations, because they don't have the necessary bandwidth."
Yes if you need the BANDWIDTH. I assume the "bandwidth" you mention is INPUT SIGNAL BANDWIDTH which allows digitizing a relatively FAST input signal. I DON'T NEED REALLY EXTRAVAGANT BANDWIDTH. Just digitizing human input forces which may get sampled at around 1KHZ.

"Because all of the fully-differential amplifiers (FDA) in our portfolio are high speed, they tend to be well suited to driving ADCs with high accuracy. Due to their increased bandwidth, FDAs tend to require lower values of resistors around them, unless you compensate the amplifier appropriately. Thus you wont tend to see FDAs in bridge applications."
Yes a FDA with a HIGH gain + HIGH BANDWIDTH looks like it might get a little unpredictable. Usually turning into a TRANSMITTER of some sort.

Your #1.
One of your FDA datasheet suggested making an INAMP from 3 of the FDAs. A little more cumbersome but doable I guess.

Your #2.
Yeah but more parts might equal more noise. I'm considering this one. While POUND resolution probably is good enough for our application - I’d like to get OUNCE resolution in 300lb’s.That’s 300 / (1/16) or about 4800 ouncecounts. (ouncecounts?) 12-13 bit sitting still in my reading with 3 or 4 bits wiggling.
3 volt system. Means 350 – 700 uV per count.

Your #3. As is - my ADC is in my micro. Someone thought it would be a good idea to toss some strain gauges in our system and send them communicating through that micro. I do have SPI though. I could plop an external ADC on a PCB close to the gauge and then I'd have a 4 foot trip with SPI back to my micro to worry about.

I've never be so conflustered as to which way to go before. I just worry about noise - the RESOLUTION KILLER...

Make sense?

Hm.... Apparently TI 's presentation group thought the idea that popped into my head wasn't too far off. In THIS, they show the exact thing I was thinking at slide 19 but no mention of a suitable FDA :

www.ti.com/.../slyp163.pdf

Rap,

The benefit of bandwidth in driving an ADC is actually more about the decrease in load transient response time due to increased loop gain rather than input signal bandwidth. Even if you bandlimit your input to the same cutoff frequency, a device with a higher gain bandwidth product will tend to perform better from an output settling perspective.

When working with FDAs, it is generally more straightforward to run them with gain, because this avoids many of the interactions that contribute to instability.

I haven't seen the instrumentation amplifier with FDAs, but this makes sense and while a bit cumbersome, I imagine would also work well. You could also build a differential INA by paralleling two INAs, but you have to be careful to remain within the linear operating region of each device.

Generally speaking, it is best to put the gain towards the front end of your signal chain so the noise of the first stage is the only component that sees significant gain. If you limit the bandwidth of the FDA after the INA, its noise contribution should not be very significant. Hope this helps!