Since when is 2 bits worth 3 dB???

The July 2006 "Electronic Products" magazine has the following
alarming adverstisement from Analog Devices on their "high-speed
14-bit ADC" (the AD9259):

  "The part provides a 3-dB improvement in SNR over 12-bit
  multichannel A/D converters available today ...".

Last I checked, 2 bits would be worth 12 dB, not 3 dB. Is this a
12.5 bit part being marketed as 14 bits?

--RY

Randy Yates wrote:
> The July 2006 "Electronic Products" magazine has the following
> alarming adverstisement from Analog Devices on their "high-speed
> 14-bit ADC" (the AD9259):
>
>   "The part provides a 3-dB improvement in SNR over 12-bit
>   multichannel A/D converters available today ...".
>
> Last I checked, 2 bits would be worth 12 dB, not 3 dB. Is this a
> 12.5 bit part being marketed as 14 bits?
> 
> --RY

ENOB vs  data bus width?

Mark

Mark wrote:
> Randy Yates wrote:
> > The July 2006 "Electronic Products" magazine has the following
> > alarming adverstisement from Analog Devices on their "high-speed
> > 14-bit ADC" (the AD9259):
> >
> >   "The part provides a 3-dB improvement in SNR over 12-bit
> >   multichannel A/D converters available today ...".
> >
> > Last I checked, 2 bits would be worth 12 dB, not 3 dB. Is this a
> > 12.5 bit part being marketed as 14 bits?
> > 
> > --RY
> 
> ENOB vs  data bus width?

Why waste the wires?

--RY

Randy Yates wrote:
> Mark wrote:
>> Randy Yates wrote:
>>> The July 2006 "Electronic Products" magazine has the following
>>> alarming adverstisement from Analog Devices on their "high-speed
>>> 14-bit ADC" (the AD9259):
>>>
>>>   "The part provides a 3-dB improvement in SNR over 12-bit
>>>   multichannel A/D converters available today ...".
>>>
>>> Last I checked, 2 bits would be worth 12 dB, not 3 dB. Is this a
>>> 12.5 bit part being marketed as 14 bits?
>>>
>>> --RY
>> ENOB vs  data bus width?
> 
> Why waste the wires?

Why is ENOB ever less than word width? (The E in ENOB stands for 
"effective".)

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:
> Why is ENOB ever less than word width?

I suppose ENOB could be a function of several parameters,
such as sample rate. temperature, etc.

I suppose it could also be the case that there
actually are N bits (N > ENOB) in the underlying
architecture (e.g., a flash converter that had
256 comparators but only 7 ENOB).

But, apart from these, I really don't see why the number
of wires would ever be more than ceil(ENOB).

--RY

Randy Yates wrote:
> The July 2006 "Electronic Products" magazine has the following
> alarming adverstisement from Analog Devices on their "high-speed
> 14-bit ADC" (the AD9259):
>
>   "The part provides a 3-dB improvement in SNR over 12-bit
>   multichannel A/D converters available today ...".
>
> Last I checked, 2 bits would be worth 12 dB, not 3 dB. Is this a
> 12.5 bit part being marketed as 14 bits?
>
> --RY

SNR of an A/D depends on a myriad of items other then the number of
bits it has, look up the SNR of any ten12 bit A/D's, they are all
different. There are many reasons to choose a 14 bit A/D over a 12 bit
A/D even if both have identical SNR's.

steve wrote:
> There are many reasons to choose a 14 bit A/D over a 12 bit
> A/D even if both have identical SNR's.

Please name one or two, Steve.

--RY

Randy Yates wrote:
> steve wrote:
> > There are many reasons to choose a 14 bit A/D over a 12 bit
> > A/D even if both have identical SNR's.
>
> Please name one or two, Steve.

Steve,

I should rephrase the question: Name one
or two reasons that are signal-related. In
other words, reasons like "a smaller
package" or "less power" are not
what I'm talking about.

--Randy

Randy Yates wrote:
> Jerry Avins wrote:
>> Why is ENOB ever less than word width?
> 
> I suppose ENOB could be a function of several parameters,
> such as sample rate. temperature, etc.
> 
> I suppose it could also be the case that there
> actually are N bits (N > ENOB) in the underlying
> architecture (e.g., a flash converter that had
> 256 comparators but only 7 ENOB).
> 
> But, apart from these, I really don't see why the number
> of wires would ever be more than ceil(ENOB).

Alan Herriman, Where are you? If ENOB was always the same as the number 
of output bits, we wouldn't have the term. Look at the SNRs at different 
sample rates at http://www.analog.com/en/prod/0,,AD7760,00.html

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Randy Yates wrote:
> Randy Yates wrote:
>> steve wrote:
>>> There are many reasons to choose a 14 bit A/D over a 12 bit
>>> A/D even if both have identical SNR's.
>> Please name one or two, Steve.
> 
> Steve,
> 
> I should rephrase the question: Name one
> or two reasons that are signal-related. In
> other words, reasons like "a smaller
> package" or "less power" are not
> what I'm talking about.

At high speeds, settling time may not allow the low bits to be 
completely independent of the earlier signal.

One can have 16 bits of resolution but only 12 bits of linearity.

Differential linearity and absolute linearity can be very different.

ENOB can therefore depend on the characteristic being measured.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������