Traditional amplifiers vs ADC amplifiers

[loop123]

In the concept of ADC amplifier. You put the signal into the range of the ADC by an initial low amplification Programmable Gain Amplifier. In traditional amplifiers, you amplify the entire signal using large gain and all analog,

What applications when the ADC amplifiers (like 24bit or 32 bit) are used and advantageous?  where do you use these?  Can they beat any traditional amplifiers in terms of noise?

[Zero999]

The question doesn't make any sense.

Just select an op-amp or instrumentation amplifier and connect it to your ADC. Often different gains are required, in which case use a programmable gain amplifier which can be bought as an IC, or made by switching different value resistors into an instrumentation amplifier using an analogue switch.

[SiliconWizard]

I'm not sure I get the question either. It looks as though either it came from some ill-defined homework question, or just some topic to feed language models.
There's one way I sort of make sense of the question, but I don't even really want to answer according to this hypothesis, as it feels like feeding either the troll, or something I don't want to feed anyway.

[loop123]

I mean the resolution of 24 bit or 32 bit ADC is so high that it can even resolve nanoVolts. So instead of using amplifiers to amplify say audio signal. Some just use ADC to resolve the audio millivolts levels without using any amplifiers. So im just asking the pro and con of each and whether you just use ADC over amplifiers in your project. Or why not if you dont.

[langwadt]

I mean the resolution of 24 bit or 32 bit ADC is so high that it can even resolve nanoVolts. So instead of using amplifiers to amplify say audio signal. Some just use ADC to resolve the audio millivolts levels without using any amplifiers. So im just asking the pro and con of each and whether you just use ADC over amplifiers in your project. Or why not if you dont.

that's like asking why you would use a screwdriver instead of a hammer, they do two completely different things

[loop123]

The end result is the same supposed you need to process the signal in a computer. But the journey to the computer can take 2 forms. One is by amplifying the signal to volt level and using separate 16 ADC to digitize the signal. The other is to use 24 bit ADC from the start and use the resolution for amplication without using ang amplifier. I was asking the pro and con of each method. What would be more noisefree for instance?

[NiHaoMike]

It's not just a matter of voltage but also impedance. Many ADCs expect to be driven by a low impedance source, otherwise the performance will not be as specified.

[hans]

I mean the resolution of 24 bit or 32 bit ADC is so high that it can even resolve nanoVolts. So instead of using amplifiers to amplify say audio signal. Some just use ADC to resolve the audio millivolts levels without using any amplifiers. So im just asking the pro and con of each and whether you just use ADC over amplifiers in your project. Or why not if you dont.

Resolution, sample rate and power. Pick 2. They are a trade-off: https://www.researchgate.net/figure/ADC-power-consumption_fig2_220365673

You can run a 10-bit ADC at say 100MSPS. That may consume 1W. However, you may need to preamplify by 1000x to get the required detail. So from my understanding, you're wondering.. why not use a 20-bit ADC @ 1000MSPS? Those 10 extra bits would get us 2^10=1024x greater resolution. However, by that chart such as an ADC would consume 1kW.

So that leaves 999W for us to amplify the signal. In theory that isn't enough as there is no amplifier that doesn't add noise. But also there is no ADC that doesn't "add" noise, such as quantization noise, but also its SNR ceiling. In addition, the figure I linked doesn't show a point-cloud of state-of-the-art ADC designs and exactly how close they get to those theoretical boundary lines. I bet there are a lot less 20-bit 100MSPS designs out there than 10-bit 10GSPS designs, for example.
So in practice, building that 60dB 50MHz amplifier is going to be far more practical.

[soldar]

I believe the OP may be referring to class D audio amps.

[nctnico]

In the concept of ADC amplifier. You put the signal into the range of the ADC by an initial low amplification Programmable Gain Amplifier. In traditional amplifiers, you amplify the entire signal using large gain and all analog,

I guess you mean supplying the signal to the reference input and use the ADC as a multiplier. This is used in some applications and for those some ADCs have specifications for bandwidth of the reference signal. But you'll still need at least an output buffer amplifier. On top of that modern day ADCs don't have a huge voltage swing at their output.

[Kleinstein]

In a few cases it can make sense to use a higher resolution ADC instead of using more effort to scale / amplify the signal to just fit an ADC.  Noise wise the extra amplification often is better or can bet better. With relatively cheap, high resolution SD ADCs the direct way may make sense to simplify a circuit. Also the gain stability may be quite good with the ADC.
An example for this could be a PT1000 sensor and sometimes also thermcouples.  It may be worth using a 24 bit ADC  and save on precision resistors and an amplifier.

[loop123]

In a few cases it can make sense to use a higher resolution ADC instead of using more effort to scale / amplify the signal to just fit an ADC.  Noise wise the extra amplification often is better or can bet better. With relatively cheap, high resolution SD ADCs the direct way may make sense to simplify a circuit. Also the gain stability may be quite good with the ADC.
An example for this could be a PT1000 sensor and sometimes also thermcouples.  It may be worth using a 24 bit ADC  and save on precision resistors and an amplifier.

Can you give a rough list what applications can use a 24bit ADC as amp instead of dedicated amps? Because it seems many people are not even familiar with the concepts. I just want to understand what situations where each is used and the advantages and disadvantages.

[Kleinstein]

The higher resolution SD ADCs often offer quite good zero drift (they often use some chopping internally) and the internal "gain" (especially if not true amplifier, but just more frequent sampling).  So it can make sense to use the input directly without an extra amplifier for a rather simple solution, e.g. where space / part count is important and
 - lowest noise is not important
 - good gain stability wanted (e.g. with the resistive sensor)
 - not too high impedance (e.g. < 2 K)
 - not much protection needed (e.g. no accessible terminals, ESD)

The ADC internal gain steps from more frequent sampling are usually quite good. I don't know for sure but chances are that the Sigilen SDM3045/3055 use the interal gain steps for some ranges. They of cause have some buffer or similar in front.
How linear the ADC is with the extra "gain" depends - there can be a higher INL when using the gain. Having the option to switch the gain is also quite convenient.

[loop123]

This thread stemmed from this passage I read previously:

"Gain by itself is meaningless if you don't know what voltage range it maps to. Modern A/D converters can be quite sensitive, so you don't need much gain to get better resolution than older high-gain systems had."

I wasn't exactly sure what it was saying. Lets use an example. Supposed you had a pressure sensor (with 100mV signal full scale pressure) whose result you want outputted to a computer. You could choose 1 Volt or 5 Volt output for the gain. So would an ADC with 1V or 5V input range would be better? If you don't need as much gain. You can choose just 1V? But what if you used gain to make it 5V? Would the output be better or more noise free? 

[MasterT]

The noise is answer to this question. 24 or 32-bits ADC  should be weighted vs data rate, all of ADC I know have less than 16 noise-free bits above 1 ksps. In other words, last one or two bytes just garbage -not data.
 
 Recently I was digging into 4 channels ultra low noise aquisition sustem, and didn't find any "ready" solution, have to build my own.
 There are some good DS (SD) adc from AD, ad7124, ad7172, some from Cirrus Logic CS5534 - low noise ONLY with low data rate, < 30 sps. More over, all of the single channel - multiplexed. 
 Using high speed 1 msps SAR is not an option, again single channels only, and if try to do multipliplexing - forget about 1 nV. Seems like conspiricy, switches - bugged, there are no "slew rate switching control" on a market.  AZ amplifier - bugged as well, no way to get access to internal switching frequency to have ability synchronize sampling - do glitch reduction.

 So, to get 1 nV noise I have to solder amplifier. 

[ejeffrey]

This thread stemmed from this passage I read previously:

"Gain by itself is meaningless if you don't know what voltage range it maps to. Modern A/D converters can be quite sensitive, so you don't need much gain to get better resolution than older high-gain systems had."

I wasn't exactly sure what it was saying. Lets use an example. Supposed you had a pressure sensor (with 100mV signal full scale pressure) whose result you want outputted to a computer. You could choose 1 Volt or 5 Volt output for the gain. So would an ADC with 1V or 5V input range would be better? If you don't need as much gain. You can choose just 1V? But what if you used gain to make it 5V? Would the output be better or more noise free?

Basically it depends on the noise of the ADC vs the amplifier.  You can usually find the ADC input noise somewhere in the specs, it's an analog noise parameter mostly unrelated to the digital resolution.  Some ADCs have quite low noise, others have more.  If the ADC input noise is low enough you may be able to get by without a lot of gain.

You still may need an amplifier even if it has low gain.  Low noise and high resolution ADCs may have a really unpleasant switching noise *current* at the inputs.  If you don't drive it with a suitably low impedance buffer that can get turned to excess measurement noise.   You might also need to translate the common mode voltage.

[loop123]

The noise is answer to this question. 24 or 32-bits ADC  should be weighted vs data rate, all of ADC I know have less than 16 noise-free bits above 1 ksps. In other words, last one or two bytes just garbage -not data.

Do you know the best references (or even illustrations about this)?  What circuit inperfection caused the 32-bit to have only effective 16-bit data, with rest garbages above 1ksps?  What then is the purpose of 32-bit ADC if half are noises?

Also for 16 bit ADC, the noise-free bits are maybe only 8-bit?

[MasterT]

You can find illustartion in the  data sheets for mention ADC. Tables called "noise vs sampling rate".
It takes some effort to minimize noise below 16-bits, imperfection mostly comming from analog switches - glitches, injection current.
I know how PCM179X series designed, to be the best. Some manufacturers don't know how to do it, others - fulled theirs greediness by Moors law.
 Threre is no sense to have 32-bits - just marketing to  fooling around naive.

[loop123]

You can find illustartion in the  data sheets for mention ADC. Tables called "noise vs sampling rate".
It takes some effort to minimize noise below 16-bits, imperfection mostly comming from analog switches - glitches, injection current.
I know how PCM179X series designed, to be the best. Some manufacturers don't know how to do it, others - fulled theirs greediness by Moors law.
 Threre is no sense to have 32-bits - just marketing to  fooling around naive.

check this out ADS1254. Perhaps it is the only exception to your statement: " all of ADC I know have less than 16 noise-free bits above 1 ksps. In other words, last one or two bytes just garbage -not data."
 

https://www.ti.com/lit/gpn/ADS1254

RMS Noise vs Data Output Rate is same as your "noise vs sampling rate"?  At 1ksps and up, it seems to have higher than 16 noise-free bits above 1 ksps

[MasterT]

Still not 24-bits. If consider peak-to-peak vs RMS noise, tham situation much worse.
 Good luck with ads125x, try if can get real 12-bits - S/H circuits is just an act of sabotage

[loop123]

To aid in understanding. Can you guys give examples of any products that uses solely ADCs to amplify without using any amplifiers? Perhaps there is sound card that only uses ADC, maybe the ones you plugged in the USB like a flash drive? Does it have any amplifiers inside it? Or any radio or ipod like device that only uses ADC, and compare to same class of products that uses amplifiers and ADCs? I want to see if I can detect any differences using hearing and Audacity in the waveforms (to see the dreaded noises produces in such ADC only devices).