24 bits should be enough for anybody, right?

[Mark R]

I was idly looking at what sort of ADCs you can get these days (looking for something to interface to a mid-range MCU, that was better than the built in ADCs), and I noticed that TI have a new range of 31 bit ADCs.

Are these devices genuinely useful? The dynamic range and SNR of 24 bit devices is already ridiculous compared to real signals, and real analog front ends. What sort of signals would really benefit from more than 120 dB of dynamic range? I can't think of any? The datasheet states 'energy exploration' - but that means nothing to me.

The 31 bit ADCs are supposed to offer about 140 dB of SNR - although, looking at the datasheet, it appears that manufacturer wasn't able fully to characterise the performance, because the devices out-performed their signal generator.

I was tempted to buy one for a laugh - then I realised that the chip cost $200.

 

[PottedMeat]

I was tempted to buy one for a laugh - then I realised that the chip cost $200.

You talking about the ADS1282-HT? Skimming their site I think the HT's are cherry picked and put into more rugged ceramic DIPs. There are free samples of the ADS1282 (TSSOP packaging).

 

[Cogman]

Perhaps it is more of an "The ADC is not holding the system back" sort of notion rather then "do we really need it." Better over engineered then under

For those that don't know/are too lazy to google it. ADC stands for analag to digital converter. I have a simplistic view of them (I haven't don't too much) however, the basic idea is that you set two pins to an upper and lower bound voltage and a third pin takes the voltage to be measured. The output pins tell you essentially the position of the voltage relative to the two range pins, IE if you had 1V and 3V and the output was 0, then you would say that the pin has, at most, one volt on it.... just so you know

What frequency are these things rated at? I know you said the signal generator couldn't keep up, so how fast was the signal generator going? Does the SNR measurement give you an idea of the frequency these things are capable of operating at (We didn't cover signal to noise ratio stuff when building our joysticks . That, and 8 bits was more then enough for use, I shaved off a bit for better wiring and had some of the best performance in the class ).

 

[CycloWizard]

Is it a 32-bit processor? If so, it's probably not costing any extra and could even save you some cycles to use data in this format. Other than that, I'm not sure what the benefit might be.

 

[bobsmith1492]

Above 24-bit you'd easily swamp any extra resolution in the noise floor, not to mention the ridiculous dynamic range available at that point.

Case in point: your 31-bit A/D. A 31-bit quantized signal has a theoretical maximum of 1.76+6.02Q dB SNR. So, the 31-bit A/D could have up to 188.4dB but naturally it won't; since it probably runs only on 3.3V or so, at the low end the junction noise from the transistors will cause a noise floor at a certain voltage. 31 bits would be steps of 1.5 nanovolts!

On the other hand, even a 24-bit A/D could have up to 146dB SNR, which is on a similar level as that claimed by the 31-bit A/D. The difference between the cheaper 24-bit and expensive 31-bit is probably in the analog circuitry surrounding the A/D. With modern delta-sigma converters the extra bits aren't hard to do.

 

[JSt0rm]

the reasons you want more then 24 bits is when summing large numbers of audio tracks digitally. 64bits is desired in that situation. For a end user 24bits represents 144db of dynamic range. plenty.

 

[William Gaatjes]

the reasons you want more then 24 bits is when summing large numbers of audio tracks digitally. 64bits is desired in that situation. For a end user 24bits represents 144db of dynamic range. plenty.

Unfortunately, the ADS1282-HT will not be of much use in the audio range.
The samplerate is 4KSPS. This means 4000 samples per second. For proper audio sound you really need a minimum of 2 times the maximum frequency in your audio signal. that would be 40,000 samples er second for a 20,000 Hz range. I agree when listening to current cd's with today's style music the quality is horrible, it could very well have been a 4KSPS ADC . But coming back to the ADS1282. I think this is used for very accurate sampling of signals that do not change fast. Although looking at that range, it sounds more marketing gimmick. A quick and dirty look would say 5V input range divided by 31bit. Approximation : 5 / 2^31 = 2,3 nano Volts. I am just awake and i think i did not have enough coffe yet. Because this looks wrong.


Then again, the datasheet mention seismic measurements as application.



http://www.ti.com/lit/gpn/ads1282-ht


http://focus.ti.com/docs/prod/folders/print/ads1282-ht.html

 

[Modelworks]

For audio there are 32 bit ADC if you look around the pro audio field, but they are not cheap. The one below is 32 bit ADC, 120db, 200Khz sample rate
http://www.esstech.com/index.php?p=products_ADC

And some impressive DAC to go with it.
http://www.cdrlabs.com/News/oppo-di...or-bdp-83-special-edition-blu-ray-player.html

The ES9016 Sabre32 Ultra DAC is the latest addition to the world’s highest performance and critically acclaimed Sabre32 Reference DAC (ES9018) introduced in early 2009. Unlike conventional sigma-delta (Σ&#916 DACs, the ES9016 Sabre32 Ultra DAC incorporates innovative patented circuits to deliver spectacular music with an unsurpassed sound stage, with up to 128dB dynamic range and 0.0003% (-110dB) total harmonic distortion, and free from clock jitter common in digital audio systems. The patented circuits in the Sabre32 DAC include the 32-bit HyperStream™ modulator, capable of 100% modulation and unconditional stability, the Revolver™ Dynamic Element Matching which ensures the highest performance over an extremely wide audio dynamic range, and a Time Domain Jitter Eliminator to remove the digital jitter that causes distortion.

The ES9016 Sabre32 Ultra Audio DAC supports either 8-channel or stereo conversion from a digital audio input (up to 500kHz sampling rate) in PCM, DSD or SPDIF formats, features click-free soft mute, volume control, de-emphasis, programmable zero detect, per-channel customizable roll-off filter characteristics and consumes less than 100mW.

 

[William Gaatjes]

Well, maybe the OPA211 could be used with the 31bit ADC .

http://focus.ti.com/docs/prod/folders/print/opa211.html