Choosing an ADC

[Tris20]

I've spent the better part of an hour looking for a simple table of ADCs like:

Type                                         Bit Range                       SPS range             Other info

Sigma delta                                14-24                           10 - 100k
Sigma delta(audio specific)        12-22                            5k - 1M
SAR                                            8-20                              10k - 10M
Pipeline                                      8-18                              1M - 1G
Flash                                             ??                                   ??                 High power consumption

I can't find a simple bloody table so I'm trying to make my own based on vague diagrams and I'm not sure with the values.

Anyone able to chip in on this?

Thanks in advance.

[dannyf]

I can't find a simple bloody table

Maybe because such a table makes no sense or has no value?

[Tris20]

Why not? Surely it's useful to compare ADC types based on bit range and sample rate quickly?

[nctnico]

Why not? Surely it's useful to compare ADC types based on bit range and sample rate quickly?

Why would that be useful? When I look for an ADC I go to Farnell's website and punch in some quick specs (sample rate, interface and resolution) to get a list with devices. Sort by price (low to high) and start at the top looking for a suitable ADC.

[Tris20]

I concede that the above would be a more practical approach and what I will do after graduation. However, as an attempt to give us an understanding of the types of each ADC, we need to choose a type of ADC during an exam without access to a list of components or suppliers. To me this feels a bit backwards but I guess it's to make us familiar with the types of ADCs and there common applications. Hence I'm trying to make a table to give myself (and anyone else who may find it useful) a better understanding of the comparisons between types.

[Tris20]

I know. Test questions like this are starting to remind me why I left education in the first place. 

[strangelovemd12]

My first thought when reading dannyf's reply was "It certainly has educational value to me", and I don't find what you've descried to be a backwards method at all.  I just wanted to put on record that I'd be interested in what comes of this.  I.E. if you find/make a table, please post it.

[Tris20]

Yeah I definitely intend to expand the table in the top of this thread over the next few weeks. I'll quote my sources as well. Again, if anyone else would be kind enough to chip in and add to this it'd be really helpful and greatly appreciated.

[tszaboo]

You have cascaded, bit per stage, dual slope, single scope architectures, also from SAR there is at least 2-3 different type with different characteristics. And you better include typical input signal, latency, typical noise free resolution, typical digital interface and other parameters in the table.
Most design engineers have the table in their head, and there is no point writing it down, because it is usually very obvious which type of ADC to choose for which application.

[strangelovemd12]

Most design engineers have the table in their head, and there is no point writing it down, because it is usually very obvious which type of ADC to choose for which application.

The point of writing it down is to facilitate its transfer into other heads, to create more engineers.

[peufeu]

There are other types of ADCs like single and dual slopes, etc.

Basically you need to list what you need :

- number of bits
- samples per second

That's the abstract stuff, but there is also the real (analog) stuff :

- DC offset
- Gain error
- INL, DNL (google those or read ADC datasheets)
- Noise
- Settling time
- Analog bandwidth
- Bandwidth of interest
- Power draw
- Cost
- etc

Selection of the proper ADC depends a lot on these.

For example, say you want to make a multimeter with way too many digits to make Dave drool.

You'll want excellent DC precision, offset, gain error, INL, DNL. You don't care too much about speed, say 30 updates/s will do. But you'd like the ADC to have a settling time of say, a few samples max, the less the better. Bandwidth isn't important (a few Hz) and noise should not exceed the last digit on the display flipping +/-1 or something.

From these specs you'll likely select a sigma-delta these days.

It is easy and cheap to get excellent precision on :
- measuring time with quartz.
- calculating ratios of time digitally

It is hard and expensive to do the same with analog values like voltage.

That's why all converters that are suitable for your multimeter actually :
- convert voltage into time and measure the time (dual slope converter, used in all multimeters before sigma delta)
- convert digital signals (ie, ones and zeros) plus time into voltage and compare it to the value to measure, correcting the digital values to get closer and closer to the analog value (feedback) : that's the sigma delta

Those are extremely accurate but slow. They also allow easy implementation of digital filters (like rejection of 50 Hz, etc)

If you want to measure RMS in your multimeter cheaply you'll need a faster ADC to do the RMS calculation digitally. Fortunately sigma deltas allow a speed/precision tradeoff. You can get 100x more samples, at 100x less precision. So, for example you could do 10ksps (for RMS) at 0.1%, or 10sps at 0.0001%. Neat !

A SAR ADC is faster but less linear and relies on matching of tight tolerances on-chip resistors, introducing errors.

A Flash ADC is more of the same : much (much much) faster, much more error, much more expensive since it includes 2^N precision comparators.

[codeboy2k]

Whenever I need an ADC I usually have a few parameters that are most important for the application at hand.
Then to find an ADC and narrow the search I just use the parametric search capability of the manufacturers' websites. 

Thus, I don't see the need for such a table as the OP suggests, since the manufacturer has already done it for me and it's searchable.

[cloudscapes]

audio ADCs don't need to go up to 1M. top limit of 48k, 96k or 192K is pretty common. more than that and you're probably venturing into lab ADCs, not consumer stuff.

[krenzo]

Here's a chart that contains what you want: http://www.analog.com/library/analogDialogue/archives/39-06/architecture.html

[Tris20]

True about audio ADCs. I actually got most of my values from the chart in the link above and it looked like it extended to 1M. Again, yes I would like to refer to a website to pick an ADC in future, yet this is part of an assignment to familiarise us with basic ADC types and their approximate ranges. I won't have access to google in the exam unfortunately...

Also, big thanks to peufeu for his/her contribution to this thread. A lot of useful information in there which isn't given out in lectures (we are only second years after all).

[Alexei.Polkhanov]

Analog Devices has a book online entitled "MIXED SIGNAL AND DSP DESIGN TECHNIQUES". Section 3 covers everything that you are looking for including description of how different ADCs work:
http://www.analog.com/en/content/mixed_signal_dsp_design_book/fca.html

[casinada]

Some ADC's have been in production for 20 years. I used the MAX132 20 years ago and it is still in production. Same with others. So pick a mature one with proven record.
Most of the manufacturers gice examples of the best use of their products, of course, you can use them for other things. Most of the designs don't just come out of the vacuum, I'm sure some big company ask the ADC manufacturer that they need a few millions of chips with certain specs and that's how chips are born.

[peufeu]

Also, big thanks to peufeu for his/her contribution to this thread. A lot of useful information in there which isn't given out in lectures (we are only second years after all).

you're welcome

it's useful to know the various ways the different breed of chips work internally to pick the right one for the job...

if you want more details you can read some datasheets and compare different technologies, especially LT datasheets are very well written.