What's All This Best Amplifier Stuff, Anyhow?

This article is in Bob Pease on Analog Volume 2 in the Analog section of the Electronic Design Library. 

Members can download the PDF ebook.

After I wrote up that amplifier scheme (see “What’s All This Bridge Amplifier Stuff, Anyhow?”), I was thinking about how to get better common-mode (CM) range.

That June 11 amplifier had excellent CM accuracy, but darned little CM range, about +1.9 V to perhaps +2.6 V, which was adequate only for the 4.4-V bridge shown there. What if you want a CM range of ±2 or 4 or 7 or 10 V? That amplifier was hopeless, but let’s consider the CM extender circuits (see the figure).

As the CM voltage goes up and down, A1 acts as a simple follower/buffer and bootstraps the powersupply voltage of the next stages. So, A2 and A3 have basically a constant CM voltage. They have to worry very little about changes in V_CM.

Thus, the common-mode rejection ratio (CMRR) can be very good. A6 is set up to bring the outputs of A2 and A3 down to ground. A6 will almost always need a CMRR trim, because the resistors R3, R4 aren’t accurate enough. So, P1 is shown.

The Preferred Embodiment

Let’s choose R1 = 100 and R2, R3 = 10k, so the first stage has a gain of 201. This can give very low dc offset and drift. The two zeners can be any 2.5-V zener, such as the LM4040-2.5, to give the first amplifier a 5-V total supply voltage. The total CM range is about +12 V to –10 V.

But actually, while the LMP2022s provide excellent dc drift, this composite amplifier doesn’t have very low noise for low source impedances. I mean, you can put in almost any type of op amp as A2 and A3. So, your “preferred embodiment” could put in a couple LMH6624s and get very low voltage noise, such as 1.4 nV/Hz, total for the whole circuit, assuming your R_S is lower than 200 .

Specifically, I can never tell you what is the “best” low-noise op amp for your application, until you tell me your R_S and your desired bandwidth and your signal size. As in any amplifier, you have to engineer it for best results, planning for the R_S and those other factors.

So if you want to go to high impedance, you could choose an LMC662, which has 5 fA of IB and ZIN better than 10¹⁴ . Maybe10¹⁶ The noise voltage is ~22 nV/Hz per amplifier. Or, the LMV651 offers only slightly worse IB but 6.5 nVHz. In other words, you can use almost any kind of amplifier, and you choose suitable types—as you always have to!

Other Considerations

High bandwidth: You might want to take a gain of more like 21 in the first stage and 10 in the later stage. Use fast amplifiers.   

• High gain: maybe a gain of 201 × 100?  
• High CMRR: You will normally need to trim because the output resistors R3, R4 aren’t perfect. 1% resistors will prevent you from getting more than about 130 dB (referred to input) with your best trim. Using 0.1% resistors for R3, R4 can get your trim range smaller so you can get 150 dB. Anything more than that, and you have to engineer it. You may be able to get up to 120 or 130 dB without a pot. Refer to my LB-46 (see www.national.com/an/LB/LB-46.pdf#page=1).  
• Fast CM signals? A1, A4 and A5 will have to have high slew rate, and you’ll have to test for that.  
• High output drive? Put a buffer inside the output loop of A6.  
• Large input signals, larger than a volt? You may need the floating supply pushed up from ±2.5 to ±5 V or more and use higher-voltage amplifiers for A2 and A3.  
• Strange CM ranges? Well, you could run the whole amplifier on +20 and –10 V, or +5 V and –25 V, to get an asymmetrical CM range.

Have I built this? Well, mostly in my head, as a paper study. The road to Santa Clara is paved with good intentions. When I realized I would have to build several circuits to evaluate, I just got intimidated and built none of them. But it’s still a good framework for almost any strange or wild set of bridge-amplifier features you may want. Have a ball! /rap