Flash ADC encoder logic equations

I don't know about noise. The technology changes to fast for my dated knowledge to be useful. AD made a 10-bit (8 would be plenty) 2-quadrant multiplying DAC. The analog audio went on the reference, and the DAC output was taken off the DAC output. The digital input set the gain. They even made a log taper!

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

I should add that multiplying digital signals by a gain factor costs dynamic range, vur the "digital potentiometer" working on audio doesn't. It's much easier to track digital pots in a stereo setup than the usual analog pots. I don't see many other reasons for the scheme.

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

On 2011/04/02 19:12, glen herrmannsfeldt wrote:
> Jerry Avins<jya@ieee.org>  wrote:
>
>> What do you call a digital potentiometer? A DAC with an external
>> bipolar reference has been used for that.
>
> I have wondered about record level controls on digital recorders.
> It seems to me that the better quality recorders use analog
> (real potentiometers), where others use digitial controls
> (up/down buttons, and a memory for the current setting).
>
> It would seem that you could use a DAC, with the input as the
> reference voltage, and so as a digital level control.  Otherwise,
> you use a DAC as input to a voltage controlled amplifier.
>
> Are the digital versions noisier?

It depends upon how many circuit elements are in the path.  A digital 
pot ideally only places one, selected element in the signal path.  The 
AD5204 has a resistor noise of 9 nV/&radic;Hz, which probably compares quite 
well to an analog pot with no digital control.

In contrast, a DAC using the input signal as a reference could place up 
to 24 circuits in the signal path, one for each high bit of the digital 
volume control word.  Typical digital pots only have 8 bit volume 
resolution, so that would be up to 8 circuits.  Of course, it depends 
upon the DAC technology, but most are a ladder with one circuit element 
per bit, and the reference flows through every circuit element that is 
enabled by a '1' bit.  I could see the possibility that a DAC-based 
digital volume control could be noisier than the typical digital pot for 
this reason.

Note that digital pots can be very noisy at the point in time where the 
volume setting is changed, because this can cause a discontinuity in the 
signal.

Brian Willoughby
Sound Consulting

Jerry Avins <jya@ieee.org> wrote:

> I should add that multiplying digital signals by a gain factor 
> costs dynamic range, vur the "digital potentiometer" working 
> on audio doesn't. It's much easier to track digital pots in 
> a stereo setup than the usual analog pots. I don't see many 
> other reasons for the scheme.

It is much easier to make a remote control work with a digital
level control.  I have seen stereo amplifiers with motorized
potentiometers (the knob turns).  If you do it digitally,
it is all electronic.  There can still be a knob that then
goes into a quadrature encoder.  

-- glen

Of course, that applies to analog signal paths. There must be a good way to handle loss of dynamic range at low levels with digital signals, but I don't know it.

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

On Apr 3, 10:42&#4294967295;am, Jerry Avins <j...@ieee.org> wrote:
> Of course, that applies to analog signal paths. There must be a good way to handle loss of dynamic range at low levels with digital signals, but I don't know it.
>

auto-ranging A/D converter, i s'pose.  you can, conceptually, make an
A/D converter that outputs a float.  and a *meaningful* float (not
merely converting the integer value to a float format).

one idea that a friend of mine had was this expensive A/D with
multiple identical (except for input scaling) A/D working in parallel
on the same sample input.  but each of these parallel A/D's had their
analog input scaled (by use of a passive voltage divider) to be twice
the effective V_ref of the adjacent A/D.  say you wanted a float value
with a 5-bit exponent; this would require 32 identical A/D's working
in parallel, and each with twice the voltage range of its neighbor.
all 32 values go into a DSP that looks at all 32 values, makes a
decision about identifying the saturated A/D's so to discard their
value, and since the resistors in the voltage divider cannot have
perfect precision, the DSP would have to do some scale tweaking
between the different A/D's that reconcile their values.  anyway, what
would come out could be a value that has about the same S/N ratio for
*nearly* any range of input.  it's essentially an A/D that needs no
trimpot on the input.

r b-j

For one job I did using a micro had to read three-phase power. The micro had an 8-channel, 8-bit ADC. One can read three-phase power with two wattmeters. (Google "two wattmeter method" if you care.) I used two for voltage and three with different gains -- 4x and 16x -- for each of the current transformers. Fortunately, the ADCs saturated, so finding the lowest gain not in saturation was easy. I had at least 6 bits of resolution over a 12-bit range.

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

robert bristow-johnson <rbj@audioimagination.com> wrote:
(snip)
> auto-ranging A/D converter, i s'pose.  you can, conceptually, make an
> A/D converter that outputs a float.  and a *meaningful* float (not
> merely converting the integer value to a float format).

How about u-Law or A-law coded instead?  It seems to me that it
works better for ADC.
 
> one idea that a friend of mine had was this expensive A/D with
> multiple identical (except for input scaling) A/D working in parallel
> on the same sample input.  but each of these parallel A/D's had their
> analog input scaled (by use of a passive voltage divider) to be twice
> the effective V_ref of the adjacent A/D.  

Calibrating is the hard part, but it sounds good to me.

-- glen

On Sat, 2 Apr 2011 04:55:27 -0700 (PDT), Daku <dakupoto@gmail.com> wrote:

>Thank you for each of your replies. What I am really after is
>the priority encoder at the end of the flash ADC, which
>generates the bits - I would like to verify the Boolean
>equations for these.
>
>
>
>On Mar 31, 9:59 pm, "m...@sushi.com" <m...@sushi.com> wrote:
>> On Mar 30, 9:16 am, Tim Wescott <t...@seemywebsite.com> wrote:
>>
>>
>>
>> > On 03/29/2011 11:54 PM, Daku wrote:
>>
>> > > Could some DSP or electronics guru please direct me ? I am trying =
to
>> > > get hold of a typical (some standard number of resolution bits)  =
flash
>> > > ADC encoder logic equations.  Any hints, pointers would be very
>> > > valuable. Thanks in advance for your help.
>>
>> > There may be some tricks, but basically a flash encoder with N bits =
has
>> > 2^N - 1 comparators, that will go off in a pattern of
>> > [[M ones] [2^N - M - 1 zeros]].  The logic is just a matter of
>> > accurately counting to M, probably while also ignoring patterns that=
 go
>> > [1 1 ... 1 1 0 1 0 0 ... 0 0] or similar.
>>
>> > Have you googled "flash ADC circuit" yet?
>>
>> > --
>>
>> > Tim Wescott
>> > Wescott Design Serviceshttp://www.wescottdesign.com
>>
>> > Do you need to implement control loops in software?
>> > "Applied Control Theory for Embedded Systems" was written for you.
>> > See details athttp://www.wescottdesign.com/actfes/actfes.html
>>
>> That pattern "ignoring" scheme is what they call anti-sparkle logic.
>>
>> http://inst.eecs.berkeley.edu/~ee247/fa04/fa04/lectures/L19_f04.pdf
>>
>> This paper has a thermometer decoder towards the end. I'm not sure if
>> anyone ever published their anti-sparkle logic, but it's not that hard
>> to come up with a few schemes. Then again, many people don't know the
>> inside of a logic gate these days. ;-)
>>
>> Flash converters are notorious for being difficult to drive. A flash
>> converter with the buffer on on-board is worth the money if the specs
>> include the buffer in the testing.

The last time i did a priority encoder was back in the Tango Schematic =
for
DOS days.  Say 1980s?  Not sure if i have any copy of that circuit any
more.  I think the last time a saw one was for the CD3306 or CD3308.
I think they show one in the 8259 programmable interrupt controller as
well.