Continue to Site

Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

  • Congratulations SSS148 on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

ADC (SAR) Resolution Selection 1

Status
Not open for further replies.

UsmanLula

Electrical
Aug 10, 2005
34
Hi All,

I am new to this forum.

I have selected an SAR ADC for a piece of medical instrumentation. However, I wanted to know how I can select the most suitable resolution ADC. What are the steps in selecting the most suitable resolution e.g. 10 bit?, 12 bit? or 14 bit? or higher?

Does it have anything to do with the Dynamic Range? And if so, how am i supposed to do the calculations?

Cheers

Usman
 
Replies continue below

Recommended for you

You need to know your requirements. The basics are dynamic range and bandwidth. Once you have those you would select on cost and interface (serial vs a parralel or a bus of some sort).

The number of bits comes from the ratio of how much resolution you need in your measurement vs how large a signal might be. If you do not know this you could look at your noise floor and make you step size similar or 4X smaller.
 
For a tutorial on the subject, you may want to consider consulting some application notes from companies such as TI, Linear Tech, Analog Devices and Microchip. With a little bit of web digging, you should be able to find plenty of information on how to select a proper ADC.

The documents will typically describe the different types of converters, accuracy and precision issues, and discuss how to interface to the converter, which is a subject in and of itself.

Armed with this knowledge, you should be able to determine what features you require in a device and then select amongst applicable devices.
 
For instance, you need to measure temperature with a resolution of 0.1F in the range of 10F to 200F.

You choose an LM34 temp sensor which is linear and easy.

It puts out 10mV per degree F.

This means you're A2D needs to cope with a RANGE of:
10mV X 10F = 0.1V
to
10mV X 200F = 2.0V

and since you want 0.1F and the sensor is 10mV/degree you
need a resolution of 10mV/10 = 1mV.

Now if you find an A2D that is scalable 0V to 2.5V (common)
you would need it to resolve 2,500 different values to get your 0.1F.

Well 8-bits is 256 different readings... not enough!
10-bits is 1024 different readings... not enought!
12-bits is 4096 different readings. which is greater than 2,500 so this does it!

You would then search to A2Ds that will get you 12bits at 0-2.5V that interface the *way* you need it to at the *speed* you need it to as UsmanLula stated.

This is the fundamental stuff.

Then you need to look at your drift, and noise figures of the candidate A2Ds to make sure the temperature of your A2D's package environment won't cause it to read different 1mV values for the same sensor output as the A2D's case temperature varies over the environment it's in.

Also the internal noise of the A2D must be low enough to not swamp your readings.

Your reference voltage selected,(2.5V), must not have too much noise or drift either. You check these by adding up everything in a worst case senario and checking the numbers.
 
itsmoked,

Your post reminded me of a bit of information I read once in an ADC app note. It said that for a 12 bit (or better) ADC that you need to use at least a four layer PCB with a solid power and ground plane. The reason, the note stated, is that the electrical noise associate with simpler PCB layouts would become a dominant noise factor.

While I believe that it is may be possible to get a low enough noise design with a two layer board, one would certainly have to be careful. It does bring up the point, though, about how PCB layout techniques are critical to the noise levels in an analog circuit. At the very least, it is necessary to follow the practice of keeping digital lines out of the analog portion of the circuit and properly implementing the analog common.
 
Noway2; Oh yeah, do it ever!

A 12-bit A2D design on a two layer board is often a crapshoot. You are totally living on the edge there. In my opinion the best you should go for on 2 layers is 11-bits. Even 4 layers requires that you follow the rules you mention and the A2D maker's admonitions. Then there's always that 50/60 Hz lurking about on your sense leads..
 
Yes, I agree that a four-layer board is good in most cases. But there are situations where a double layer can do quite well. Especially low-frequency applications. I once had a circuit with low band-width, temperature - around 10 Hz. A few capacitors in strategic places and careful layout saved some dollars for that customer. You know the rule: "There are no easy rules"

Gunnar Englund
 
Sure skogs, just my rules-of-thumb. Also depends on what you are measuring. If your sensor is 0-10V then 12 bit becomes a lot easier to achieve. Same as you say, with something like 1 Hz readings you can filter to death or grossly oversample.
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor