Question about ADC, 6dB/bit and spurious free level

It's available from Rochester, which suggests EOL. Digikey's link to the TI data sheet doesn't work.

But there are probably lots of cheap video ADCs around.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

o acquire a signal up to 25MHz, preferable with high dynamic range in the r ange of 80 dB

times 6dB per bit)

6fa.pdf

oise -100dB down

ce could do:

to -90dB

ssing?

Don't know that I can help you, but maybe you can explain something to me. You say you need a 14 bit converter, then mention 90 dB and 16 times 6 dB. That's three different things that are not equal and yet you seem to be t alking aboout them as equivalent. I'm confused.

Did you mean 16 bits and the 16 bit device has a 90 dB noise floor?

Various specs typically don't match the 6 dB per bit, so 90 dB goes well wi th a 16 bit converter, but I don't get the 14 bit comment.

--

  Rick C. 

  - Get 1,000 miles of free Supercharging 
  - Tesla referral code - https://ts.la/richard11209

Hard to say but maybe not. The OP may have meant a 16 bit converter or may not have given mathematically accurate figures, just near enough.

So for an 8 bit converter the dynamic range is 48 dB = 48/8 = 6 dB / bit For a 14 bit converter 20*log10(2^14) = 84 dB = 84/14 = 6 dB / bit for a 16 bit converter 20*log10(2^16) = 96 dB = 96/16 = 6 dB / bit But that's all theoretical and ADC specs will give figures which take other device specific parameters into account.

It costs 15 USD, but has power spectrum on page 19 with a nice peak and

Page 9 shows the output power spectrum, shows a nice peak and noise down

Specsmanship again.

There are two cases for considering digitizer carrier-to-noise ratio: unipolar and bipolar. Either way, an ideal N-bit ADC (i.e. noiseless and having zero DNL and INL) contributes quantization noise. In general this is hard to treat, but as Widrow showed BITD, as long as your signal is at least a few LSBs in amplitude, the quantization noise can be accurately described as uniformly-distributed additive white noise of amplitude (Vref/2**N)/sqrt(12). The sqrt(12) comes from computing the RMS deviation of a staircase from a straight line--you can derive it in a few lines of algebra.

In the unipolar case, the signal can have an amplitude right up to Vref, so the maximum CNR is

CNRmax(dB) = -20*log(2**-N/sqrt(12)) = 6.02 N + 10.79.

In the bipolar case, where the peak-to-peak range of a sine wave has to fit in [0, Vref), we lose a factor of sqrt(8), so

CNRmax(dB) = -20*log(2**-N * sqrt(8/12)) = 6.02 N + 1.76.

Lower-resolution ADCs sometimes approach these values.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

It's available from 5 other suppliers also:

But, yes, I need to consider longevity

:

to acquire a signal up to 25MHz, preferable with high dynamic range in the range of 80 dB

16 times 6dB per bit) 916fa.pdf

noise -100dB down

vice could do:

n to -90dB

missing?

. You say you need a 14 bit converter, then mention 90 dB and 16 times 6 dB . That's three different things that are not equal and yet you seem to be t alking aboout them as equivalent. I'm confused.

with a 16 bit converter, but I don't get the 14 bit comment.

Sorry for that, I was not exact in the OP

It's a comparison between the 10bit and the 14bit. So theoretically 60dB vs 84dB

I was just surprised about that the plots from the datasheet showed 90dB no ise floor for the 10bit ADC and 100dB for the 14bit. So number don't add up ?

Cheers

Klaus

I've seen things in high volume production sold by Rochester. I don't know that everything they make has been abandoned by others. In fact, octopart often reports the lowest prices at Rochester.

--

  Rick C. 

  - Get 1,000 miles of free Supercharging 
  - Tesla referral code - https://ts.la/richard11209

te:

DC to acquire a signal up to 25MHz, preferable with high dynamic range in t he range of 80 dB

(16 times 6dB per bit)

25916fa.pdf

nd noise -100dB down

device could do:

own to -90dB

m missing?

me. You say you need a 14 bit converter, then mention 90 dB and 16 times 6 dB. That's three different things that are not equal and yet you seem to be talking aboout them as equivalent. I'm confused.

l with a 16 bit converter, but I don't get the 14 bit comment.

vs 84dB

noise floor for the 10bit ADC and 100dB for the 14bit. So number don't add up?

Are you sure it showed the numbers you state? I've seen such graphs where the 0dB point was not where you'd expect it.

--

  Rick C. 

  + Get 1,000 miles of free Supercharging 
  + Tesla referral code - https://ts.la/richard11209

At least with audio ADCs and DACs it seems to be common to quote the SNR on the 0-20 kHz bandwidth only even with 192 kHz sampling rate. The noise shaping helps to get nice numbers for the specification :-).

te:

DC to acquire a signal up to 25MHz, preferable with high dynamic range in t he range of 80 dB

(16 times 6dB per bit)

25916fa.pdf

nd noise -100dB down

device could do:

own to -90dB

m missing?

me. You say you need a 14 bit converter, then mention 90 dB and 16 times 6 dB. That's three different things that are not equal and yet you seem to be talking aboout them as equivalent. I'm confused.

l with a 16 bit converter, but I don't get the 14 bit comment.

vs 84dB

noise floor for the 10bit ADC and 100dB for the 14bit. So number don't add up?

Noise spectrum plots show the noise spectral density. To get the signal to noise ration you need to add up the noise components in each frequency bin in the correct way (which generally relies on the assumption that the noise in each bin is uncorrelated with that in other bins). The "noise floor" on such plots depends on the measurement bandwidth of the frequency analyser (often an FFT) used to generate the plot. John

Rochester seems to be expanding into 'life cycle management', i.e. functioning effectively as a second source for parts that are still active but may be getting old enough that you'd worry about continued supply. A smart idea if so. One good thing about them is that unlike some of the grey-market outfits, they buy enough stock that they don't need to jack up the price as the supply shrinks.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

A lot of Digikey and Mouser parts are from Rochester lately. I hope that doesn't tag EOL for those parts.

They may be in the reel-breaking business, which is fine.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

[snip]

Thanks, that made me dig out John Watkinson's Introduction to Digital Audio. It's on page 46.

Their minimum order is $250 per item IIRC. (It might be $250 total.) That reduces the small-quantity nuisance.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

That may be because if they spec'd at the higher bandwidth, the noise would be worse.

Audio equipment specs kind of misleading these days and most pro audio specs seem to be A-weighted to get better numbers.

Didn't used to be like that before around 1995 or so ?

Oh dear..... unfortunately a common misunderstanding.

The dynamic range is the (maximum coded analog signal)/(minimum coded analog signal).

The minimum coded signal is *1/2* lsb, not 1 lsb. That is any, analog signal over 1/2 lsb is a valid code that is actually detected as a 1 lsb. Any signal below 1/2 lsb is ignored. This means that the dynamic range of an 8 bit converter is:

DR = log(255/(1/2)) = 54.1 dB

The dynamic range should not be confused with the resolution or error.

-- Kevin Aylward

- SuperSpice

6.02 * numbits:

I don't know how you managed to escape my plonk file, but back you go again.

PLONK

--
Science teaches us to trust. - sw