Zeroing in on the Precision Op Amp Your System Needs: A Graphical Approach
PETER LIU: Hi, this is Peter Liu of National Semiconductor Precision Systems Business Unit. I manage the applications group. We're going to spend the next 20 minutes or so going over a graphical approach to help you zero in on the Precision Op Amp that your system needs.
Objectives
The objectives of this seminar are two-fold. First we're going to show you six parameters from which you can start your selection process and we're going to describe how they impact the sensitivity of your system and help you identify when and where these affect portable and distributed sensing systems. We're going to close out by showing the four graphs on National's Precision Amplifiers homepage and how to use them to accomplish the above. And then, of course, you are going to have to delve into the data sheets for your final selection.
Why a graphical approach?
So to get started, why a graphical approach? Well the front-end amplifier of a system is arguably the most important part of your system because it determines the sensitivity for your entire system and, therefore, you have to consider a lot of parameters. With over 15 parameters for each of 30+ amplifiers available to you becomes a very daunting task. We realize your time is money and, therefore, we have developed a set of graphs to help you visualize the relationship between different amplifiers so you can make these tradeoffs quicker.
Modern precision systems and 4 basic parameters
So front-end amplifiers in precision systems are almost always, if not always, high gain and that's because of best signal to noise ratio, or SNR, is achieved by having as much of the gain in the first stage because each subsequent stage adds noise and also gains up the noise of the previous stage and none of this noise can be taken away. This hasn't changed since the dawn of electronics. The signal to noise ratio of the amplified signal and the resolution of the ATD converter that will digitize that signal combine together to determine the system sensitivity. Because of this high gain, all the imperfections of the Operational Amplifier, such as input offset voltage and the input voltage noise all get gained up as well. And particularly in today's portable and distributed systems, you tend to find that input bias current and current noise start playing a bigger role more often. And that's because in these systems the total energy consumed must be monitored very carefully and, therefore, in design budgets higher valued resistors are used in feedback and input networks and sometimes even the sensors as well.
Voltage and Current Noise in two common topologies
And so to provide context for our subsequent graphs, we're going to look at two circuits which are commonly encountered in sensing systems. The first one is the non-inverting gain configuration and the second one would be the transimpedance configuration. In the non-inverting configuration current noise on the negative pin of the amplifier gets converted to voltage noise by the parallel combination of R2 and R1 and that gets gained up to the output. Similarly, the current noise on the plus pin of the Op Amp gets converted to a voltage by the sensor's resistance and whatever impedances that hang out here and also get gained up. And lastly, we have the voltage noise of the amplifier getting gained up as well. Because these three terms sum up in quadrature tends to accentuate the most largest component which happens to often be the voltage noise component. In the transimpedance amplifier the current that normally flows through the diode gets gained up by the feedback resistor, RF, to become an output voltage. And if you notice here, the current noise on the negative pin follows the same path and gets gain up as well. However, the voltage noise just shows up with a gain of one and, therefore, for most systems it's the current noise times the feedback impedance in this case that determines the total noise. Two things to keep in mind are that noise cannot be calibrated out and that less noise means a higher signal to noise ratio, which results in higher system sensitivity which is what we're all after.
Graph: Current vs. Voltage Noise
So here we plot voltage noise on the X axis and current to noise on the Y axis. From here we can see that there're plenty of amplifiers available for designs that are 16 bits or better. The current noise starts being a significant contributor when the system impedances approach 10Kohms because 1 pico-amp per root Hertz times 10Kohms starts giving you 10 nano-Volt per root Hertz of equivalent noise. And, therefore, amplifiers in this region are usually better starting points.
Graph: Current vs. Voltage Noise
\In a transimpedance application, the feedback resister is even larger, in the Mohm to 10Mohm, sometimes even 100Mohm regions. And there 10 femtoamps gets translated to 10 nanoVolt per root Hertz and these amplifiers here become good choices. For extreme cases of very high feedback resistors, of course, the LMC 6081 and 662 becomes the preferred choice for their very, very, very low current densities.
Input Offset
The next graph we want to show you related to the input offset. An input offset affect the circuit in the same way as input noise sources. What we can do is just replace a voltage noise source here with VOS, the current noise source here with IB. And, in general, amplifiers with lower VOS and IB allow for more of the output range to be allocated to the signal which, again, gives you better SNR and better system resolution.
Graph: Input bias current vs. offset voltage
And that is why we plot VOS versus I-Bias. On the X axis we plot VOS on the Y axis, I-Bias and we can see, again, our same group of amplifiers in their relative positions. It's important to note that for system impedances greater than 100Kohms you start having to pay attention to the input bias current. And again, such systems are commonly found in portable and distributed sensing systems. So 10 nano-Amps of I-Bias translates into 1mV of equivalent offset voltage. Now National guarantees a maximum VOS and IB through rigorous testing. And what this means to you is that you can bound your error budget with these max numbers instead of typical numbers.
Graph: Noise vs. Supply Current
The third graph in our collection is one of input voltage noise density versus supply current. Low supply currents are critical for portable and distributed sensing applications, particularly. Because we often have to ask the question, how quiet of a system does my power budget afford. And what we can see here is in plotting voltage noise on the Y axis versus supply current on the X axis, we can quickly make tradeoffs saying that if I have 1mA available, this is how much of a quiet system I can afford. If I have more current available for my battery than I can get lower noise. So this green line helps us calibrate that anything along this green line represents state of the art for achieving best balance between supply current and noise.
Graph: Noise vs. Operating Voltage
The last graph in our collection is one plotting voltage noise versus operating voltage. And low operating voltages are vital for applications powered either from a battery or from harvested energy because low sources just don't provide very high operating voltages. Historically, low noise has only been associated with amplifiers requiring high power supplies. But that all changed with National's VIP50 process, which enables this class of amplifiers down here at low operating voltages while achieving excellent noise performance. So this brings us to the highlight of our webinar here, using all four graphs together that we covered previously to understand how all these amplifiers relate to each other,
Precision Amp Homepage: Use all 4 graphs together to save time.
And how we can tradeoff between them. So we can see that LMP7715 is quite a well-balance amplifier in terms of noise performance for both current and voltage. It's also very well-balanced for VOS and IB. Its operating voltage range is from 1.8 to 5 and it is consuming just above a milli-Amp to achieve better than 6 nanoVolt per root Hertz of noise. Now when we find after our initial system design that we can't quite afford that much current for the supplies and we have to back off a bit, what is our next best choice? The graph shows that LMP7701 and LMV771 along with LMV851 are the next best candidates. From the noise graphs we can see that they're only slightly higher for noise values and if offset voltage is absolutely imperative, the LMP7701 is just close to the 7715. Whereas the other two are out here. If we find that we can afford more supply current and we need less noise in our system then, of course, we can pick the 7731 and perhaps even the LMH6624. They reside in this quadrant of the noise graphs. So I hope I have shown you quite an effective way of using these four graphs which can be found on the Precision Amplifier homepage of National Semiconductor.
For Additional Information
For more information, please consult these additional sources.
Digg
Del.icio.us
