# Optimally Accurate 28-bit A/D conversion

1. ## Optimally Accurate 28-bit A/D conversion

If I want e.g. 28-bit accurate A/D conversion by multiple conversions and averaging, what is the optimal A/D converter - so that total conversion time is the shortest?

•

2. ## Re: Optimally Accurate 28-bit A/D conversion

Averaging multiple conversions can reduce noise but not inherent ADC errors, particularly non-linearity. To achieve 28 bit accuracy, you need an ADC with 28 bit integral linearity. There are a few ADC in the 24 to 32 bit range available on the market, I presume all of the sigma/delta type. What's the intended sample rate?

3. ## Re: Optimally Accurate 28-bit A/D conversion

Originally Posted by FvM
Averaging multiple conversions can reduce noise but not inherent ADC errors, particularly non-linearity. To achieve 28 bit accuracy, you need an ADC with 28 bit integral linearity. There are a few ADC in the 24 to 32 bit range available on the market, I presume all of the sigma/delta type. What's the intended sample rate?
What is the fastest 28-bit A/D converter?

•

4. ## Re: Optimally Accurate 28-bit A/D conversion

Hi,

More bits doesn't mean better performance.
I've seen a lot of 24 bits ADCs that don't have better performance than 16 but ADCs.

And honestly I've not seen an ADC with true 24 bit performance.
My recommendation: forget about 28 bits.

The other point is, that even if you have an ADC with true 24 bit performance..you need to design the surrounding circuit to keep this performance.
I'm doing high precision electronics for decades now, but I don't think I'm able to design a true 24 bit performance system.
For a 3V range a 24bit resolution means about 180nV/LSB.
Thermocouple effects and even mechanical stress will make 24bits precision impossible.

What is the fastest 28-bit A/D converter?
If I want to know this ... I'd use selection guides and chart from manufacturers and distributors.

Klaus

•

5. ## Re: Optimally Accurate 28-bit A/D conversion

Originally Posted by KlausST
Hi,

More bits doesn't mean better performance.
I've seen a lot of 24 bits ADCs that don't have better performance than 16 but ADCs.

And honestly I've not seen an ADC with true 24 bit performance.
My recommendation: forget about 28 bits.

The other point is, that even if you have an ADC with true 24 bit performance..you need to design the surrounding circuit to keep this performance.
I'm doing high precision electronics for decades now, but I don't think I'm able to design a true 24 bit performance system.
For a 3V range a 24bit resolution means about 180nV/LSB.
Thermocouple effects and even mechanical stress will make 24bits precision impossible.

If I want to know this ... I'd use selection guides and chart from manufacturers and distributors.

Klaus
New production techniques may be necessary - e.g. all Cu ICs including leads, microwelding. For a given A/D converter it is possible to compensate for nonlinearity.

6. ## Re: Optimally Accurate 28-bit A/D conversion

Flash A/Ds anyone?

Although I don't think 28 bit flash A/Ds were commercially produced. The power consumption alone would be humongous.

7. ## Re: Optimally Accurate 28-bit A/D conversion

Hi,

No, there surely is no 28 bit flash ADC.
As already said: it will be a sigma/delta type.

New production techniques may be necessary - e.g. all Cu ICs including leads, microwelding. For a given A/D converter it is possible to compensate for nonlinearity.
I can't think of an application where 28 bit precision is needed ...
It is a dynamics of about 1:300,000,000
It's like measurung the distance between center of earth to cener of moon with a resolution of 1m....while we know the suface of the earth differs by 8,000m ... and the radius of the earth differs by about 20,000m .... and the distance varies in time by about 40,000,000m.

In 99% it makes no sense. And in case you think your application is within the remaining 1%, I recommend to do some error calculations first.
What voltage reference, what Opamp, ....or the most basic: What resistor do you think you can use that maintains the expected stability?

Klaus

•

8. ## Re: Optimally Accurate 28-bit A/D conversion

What is the fastest 28-bit A/D converter?
You can answer the question in two regards.
- What's the maximum sample rate of ADC on the market (or may be experimental devices in literture) with >= 28 bit resolution?
- What's the achievable data rate for noise-free 28 bit resolution? Expect some 100 Hz maximal.

New production techniques may be necessary - e.g. all Cu ICs including leads, microwelding. For a given A/D converter it is possible to compensate for nonlinearity.
Major manufacturers (Analog, TI, Maxim etc.) are working hard for decades to push the limits. Be pretty sure that they already consider all available technologies.

As mentioned previously, the discussion suffers from not referring to an ADC application range and required/desired sampling rate.

9. ## Re: Optimally Accurate 28-bit A/D conversion

Originally Posted by KlausST
Hi,

More bits doesn't mean better performance.
I've seen a lot of 24 bits ADCs that don't have better performance than 16 but ADCs.

And honestly I've not seen an ADC with true 24 bit performance.
My recommendation: forget about 28 bits.

The other point is, that even if you have an ADC with true 24 bit performance..you need to design the surrounding circuit to keep this performance.
I'm doing high precision electronics for decades now, but I don't think I'm able to design a true 24 bit performance system.
For a 3V range a 24bit resolution means about 180nV/LSB.
Thermocouple effects and even mechanical stress will make 24bits precision impossible.

If I want to know this ... I'd use selection guides and chart from manufacturers and distributors.

Klaus
Agree except I'd point out that the way this is handled is with ranging.

It is possible to make a system with 180nV resolution that can also handle 3V full scale....just not at the same time.

--[[ ]]--