ADC specs

[percy]

Take a look at these two ADCs. Which one do you think is better suited for audio ADC ?

What I am trying to understand is that even though the DNL/INL of the 18bit ADC is much worse than the 16bit ADC, why is the SNR, SINAD and SFDR better by atleast 3db than the 16bit ADC and the THD still the same as the 16bit ? How is that possible ? What am I missing about interpreting or understand ADC specs ?!
I notice that they have conveniently 'forgotten' to give ENOBs for the 18bit dac.

[poobah]

Higher SINAD and SNR etc... are BETTER numbers.

In particular: SNR is Signal to Noise Ratio: The higher the signal, or the lower the noise... the higher this ratio will be.

SINAD is similar...

Now... DNL and INL are bit more complex... lower is better in this case. Keep in mind that each additional bit of resolution implies 6
dB of range... So comparing INL and DNL directly between 16 & 18 bit units is not apples to apples. The 18 bit converter has a 12 dB head start. You could say as well the 16 bit converter has a 12 dB handicap.

So while the 16 bit unit has .5 lsb xNL... and the 18 bit has 2 lsb xNL; the 18 bit unit effectively has nearly the same noise level as the 16 bit unit...

When all is said and done a difference of 1.5 LSB's corresponds to a "hit" of 9 dB on a 12 dB advantage... why the SNL and SINAD are only about 3 dB better on the 18 bit unit.

Confused yet?

SINAD would be a good factor to compare for audio use...

[percy]

I understand now. I should have realized INL/DNL errors manifest themselves as loss of snr.

Looks like analog to digital conversion hasn't progressed much past 16bits. (delta sigma need not apply). so much for "32 bit DAC"s!

[poobah]

There is some truth to what you say. The fact of matter is... it can be difficult to get a 12 bit ADC to behave.

That doesn't mean that the extra resolution is a farce. It simply means that sample averaging (filtering), and dithering (usually happens for you automatically) are required to extract the finer measurements.

One of the central issues with ADC design lies in the fact that the same piece of silicon doing the conversion is also an inherently noisy place to be... it is digital after all.

Look at some of the other ADC TI & BB have to offer... the PCM series come to mind. These have much higher SNR's etc...

Be prepared to stand on your head and "arf like a seal" if you wish to achieve the full capabilities of these devices... it ain't easy. You won't do it with alligator leads or a radio shack proto board.

In the case of a 24 bit converter with a 5 volt differential input: 1 count equals 0.6 uV.

[percy]

in the 80's when the first recordings on CD came out, I wonder what ADCs were used for the master recordings(24bit). If 24bit ADCs are a struggle even now, how were they 'then' ?

[poobah]

I had a few of the ADCs back then... BURR BROWN was still the big kid on the block.

They were laser trimmed hybrid construction... and about $1K...

The new ADC techniques... shoving little packets of current around allows much better noise performance.

[flshzug]

AD was the king of the r2r's as far I read

Someone has PCM75 datasheet for me ?

apparently you can reach 18 bit with those - additional residual signal tinkering.

[rfbrw]

Originally posted by poobah

They were laser trimmed hybrid construction... and about $1K...

I imagine that would be fine for the defence and oil industries but I'd have thought most mere mortals would have stuck to the likes of the AD1879 and the PCM1750.

[poobah]

This may have been about 1983... at the time... they were the best I could find. And NO... I did not pay the 1K$... got sampled on 2 of them...

[flshzug]

I born in 83

What about soft clipping on SAR adc vs. d-s ?
I read on prosoundweb , some rec.engineer wrote that each & every track he gets from clients is soft limited ...